· This article was downloaded by: [Department Of Fisheries] On: 15 July 2013, At: 23:15...

This article was downloaded by: [Department Of Fisheries]On: 15 July 2013, At: 23:15Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries ManagementPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/ujfm20

Temperature-Driven Decline of a Cisco Population inMille Lacs Lake, MinnesotaRajeev Kumar a , Steven J. Martell a , Tony J. Pitcher a & Divya A. Varkey aa Fisheries Center, University of British Columbia , 2202 Main Mall, Vancouver , BritishColumbia , V6T 1Z4 , CanadaPublished online: 18 Jun 2013.

To cite this article: Rajeev Kumar , Steven J. Martell , Tony J. Pitcher & Divya A. Varkey (2013) Temperature-Driven Declineof a Cisco Population in Mille Lacs Lake, Minnesota, North American Journal of Fisheries Management, 33:4, 669-681, DOI:10.1080/02755947.2013.785992

To link to this article: http://dx.doi.org/10.1080/02755947.2013.785992

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/loi/ujfm20

http://www.tandfonline.com/action/showCitFormats?doi=10.1080/02755947.2013.785992

http://dx.doi.org/10.1080/02755947.2013.785992

http://www.tandfonline.com/page/terms-and-conditions


North American Journal of Fisheries Management 33:669–681, 2013C© American Fisheries Society 2013ISSN: 0275-5947 print / 1548-8675 onlineDOI: 10.1080/02755947.2013.785992

ARTICLE

Temperature-Driven Decline of a Cisco Population in MilleLacs Lake, Minnesota

Rajeev Kumar,* Steven J. Martell, Tony J. Pitcher, and Divya A. VarkeyFisheries Center, University of British Columbia, 2202 Main Mall, Vancouver,British Columbia V6T 1Z4, Canada

AbstractMille Lacs Lake, Minnesota, has experienced a decline in the population of Cisco Coregonus artedi since the

1980s. Cisco is a coldwater stenotherm, and the population decline is often attributed to a general increase in the laketemperature. However, there also has been a fishery for this species during the last 20 years. To investigate the influenceof temperature on this decline, three versions of a surplus production model (SPM) were formulated: (1) SPM withobservation error only, (2) SPM with observation error and a maximum temperature anomaly, and (3) state-spaceSPM with a maximum temperature anomaly and random effects process error. Data on CPUE of sampling gill netsfrom 1985 to 2007 were used as an index of fish abundance for the lake population. The model parameters wereestimated by fitting the predicted CPUE to the observed CPUE. The analysis indicated that temperature explained36% (model 2) and 40% (model 3) of the change in Cisco abundance. Temperature-influenced, time-varying carryingcapacity and maximum sustainable yield (MSY) were also estimated. We concluded that the causes of Cisco declinewere a combination of temperature and fishing pressure.

Cisco Coregonus artedi is a forage salmonid species widelydistributed in central and northern Minnesota. This coldwaterstenotherm is vulnerable to environmental stress, and its survivalis often related to the temperature–oxygen profile of a lake(Jacobson et al. 2008). Frey (1955, as cited in Colby and Brooke1969) defined habitat suitable for Cisco, termed as the “ciscolayer,” based on their lethal temperature–oxygen tolerance levelas a portion of the water column at temperatures less than orequal to 20◦C with dissolved oxygen (DO) levels higher than orequal to 3 mg/L. Younger Cisco can survive higher temperaturelevels than adults; Edsall and Colby (1970) reported an upperlethal temperature for age-0 Cisco as 26◦C.

Mille Lacs Lake is the second largest lake within Minnesotaand is located in Aitkin, Crow Wing, and Mille Lacs countiesin the east-central part of the state (Figure 1). It is a glacial lakewith an area of nearly 537 km2 and drains about 1,080 km2 ofwatershed (Heiskary et al. 1994). The lake is shallow with amean depth of 8.8 m and a maximum depth of 12.8 m. Sam-pling data from the lake suggest a substantial decline of Cisco

*Corresponding author: [email protected] April 26, 2012; accepted March 12, 2013

abundance has occurred over the last two decades. An increasein temperature over the years is hypothesized as a causativefactor for this decline (Colby and Brooke 1969; Jacobson et al.2008). However, this species was also fished over the same timeperiod: Cisco catch varied from ∼7 metric tons in 1985 to ∼0.5metric tons in 2007 with a peak of ∼30 metric tons in the mid-1990s (Figure 2e). The average summer temperature of the lakeis over 16◦C and rises to around 23◦C (MDNR 1995). Jacobsonet al. (2008) recorded a maximum temperature 26.1◦C in year2006.

Several Minnesota lakes, including Mille Lacs Lake, experi-ence mass mortality of Ciscoes especially during summer whenthe “cisco layer” is relatively small. Pelagic Ciscoes, mostlyadults, move to cooler deeper water (hypolimnetic zone) insummer (Scott and Crossman 1973); mortality occurs whenthe prevailing hypoxic condition in the hypolimnion forces Cis-coes to move up in the water column into the zone where thetemperature is lethal for the species. Since Mille Lacs Lake isquite shallow, wind-driven mixing of water occurs freely from

669

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3

670 KUMAR ET AL.

FIGURE 1. Mille Lacs Lake, Minnesota. Inset map of the lake also shows thelocation of the city of Garrison on the northwestern side. (Source: backgroundU.S. map was generated using SAS/GRAPH map data sets; the inset for MilleLacs Lake was obtained from Minnesota Department of Natural Resources).

top to bottom; consequently, Ciscoes do not get enough thermalrefuge in the deeper parts of the lake. Our objective through thismodeling exercise was to investigate and quantify the influenceof temperature on Cisco carrying capacity and abundance inMille Lacs Lake.

METHODS

Data for the ModelWe used time-series data of CPUE, temperature, and catch

from 1985 through 2007 (Figure 2). The data were obtainedfrom the Minnesota Department of Natural Resources (MDNR).State-licensed anglers and tribal–band fishers (netters, spearers,and anglers) are the two groups that engage in fisheries on thelake (Figure 2e). The CPUE data are from standard experimentalgill-netting conducted by MDNR to monitor the fish populationsin Mille Lacs Lake. Every year, usually in the last week ofSeptember, gill nets are set overnight at 32 inshore locationswidely spread around the lake (Jones 2006), and CPUE foreach year is the average fish catch per gill net. Since the sametypes of nets are set at the same location and time every year, weassume that capture probabilities are relatively consistent acrossyears and should not bias CPUE as an index of biomass. TheCPUE data show a decline of the Cisco population in the lake(Figure 2a); decline was observed in fish age 2 and older butnot in the number of age-1 recruits suggesting that the declinein CPUE has followed the decline in adult Cisco (Figure 2b, c).

Air temperature data were recorded at Garrison, Minnesota,(Figure 1) for the time period 1985–2007 (Melissa Drake,MDNR, personal communication). Water temperature data fromMille Lacs Lake were not available for the 23-year time periodfor which CPUE information was available. For the few years

FIGURE 2. Time-series data from 1985 to 2007 for (a) average observedCPUE data (pounds per gill-net lift) from 32 experimental gill nets, (b) averageCPUE (numbers per gill-net lift) of age-1 fish, (c) average CPUE (numbers pergill-net lift) of age-2 + fish, (d) temperature anomaly for maximum tempera-tures in months July and August, and (e) total catch in metric tons.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3

DECLINE OF A CISCO POPULATION 671

for which we had water temperature data, we checked if themonthly average air temperature data at Garrison were corre-lated with the monthly average water temperature data fromMille Lacs Lake. We found that the two temperature series werehighly correlated except during the winter months with ice coveron the lake. In our analysis, we used records only from July andAugust (correlation coefficient = 0.88 for temperature recordsin July and August; see Appendix 2 for more details).

Also, we used the air temperature series to calculate ananomaly factor, thus took into account only the trend andnot the absolute temperature records. We calculated the sum-mer temperature anomaly based on the maximum temperatures(Figure 2d) observed every year in July–August for the 23-yearperiod. For any data series, the anomaly denotes the deviationfrom the mean, in other words, it is the distance of data pointsfrom the mean relative to the standard deviation. The temper-ature anomaly (Ta) was calculated by standardizing the timeseries of maximum temperature using equation (1),

Tat = (Tt − μ)

σ, (1)

where Tat denotes the temperature anomaly in year t, Tt is themaximum temperature for year t, and μ and σ represent the meanand SD of the temperature–time series. Maximum temperaturesincreased from 1985 to 1987 then declined till 1997. From 1997,the temperatures maintained an increasing trend. Additionally,as a sensitivity analysis, we calculated the temperature anomalybased on accumulated degree-days in July–August for the 23-year period. We used this alternate description of temperatureanomaly in the temperature-dependent surplus production mod-els and explored the results. The additional data and models areprovided in Appendix 2.

Model FormulationThree model formulations were used to examine the effect of

temperature on Cisco population decline: (1) model 1: surplusproduction model (SPM) with observation error; (2) model 2:SPM with a temperature anomaly and observation error; (3)model 3: SPM with a temperature anomaly, observation error,and random effects process error (state-space model).

Model 1: SPM with observation error.—The surplus produc-tion refers to “the difference between increase in biomass due togrowth and recruitment to the fishable population, and the lossin biomass due to natural mortality” (Punt and Smith 2001). Inthe presence of a fishery, if catch is equal to surplus production,stock biomass is maintained at equilibrium; higher catch leadsto decrease in stock biomass, and catch lower than surplus pro-duction leads to an increase in stock biomass (Jacobson et al.2001). The Schaefer SPM (Schaefer 1954) is based on a lo-gistic population growth model, and it has density dependencesuch that growth and survival are faster at lower populationbiomasses than at levels close to the unfished biomass (Puntand Smith 2001). When fisheries catch is incorporated into the

logistic population growth, the model can be modified for nextyear’s biomass as in equation (2) (Hilborn and Mangel 1997),

Bt+1 = Bt + r Bt

(1 − Bt

K

)− Ct , (2)

where K is the carrying capacity of the population, r is theintrinsic rate of population increase, Bt describes the biomass(B) at any point in time t, and Ct is the catch at time t. Equa-tion (2) was reformulated as

Pt+1 = Pt + r Pt (1 − Pt ) − Ct

K, (3)

where P is the biomass expressed as a proportion of the carry-ing capacity. The subscripts t and t + 1 denote time and time1 year later, respectively. This formulation of the model enableseasier control of the biomass falling below an unrealistic “zero”value during the model-fitting process. Since direct annual es-timates of fish biomass are not available, a compromise is tostudy the trends in the population biomass by conducting regu-lar (usually annual) sampling surveys. Sampling-survey CPUEsare commonly used as indices of abundances or biomass. Theobservation model assumes that any change in biomass will bereflected in CPUE as follows:

CPUE = q Bt , (4)

where parameter q is the catchability of the sampling net. Catch-ability expresses the index of abundance (CPUE) as a proportionof the biomass. Thus, if we know the values of parameters r, K,and q and have data on catch trends, then equations (3) and (4)allow us to predict the CPUE for the same years. The CPUE pre-dicted by the model was fitted to the observed CPUE obtainedfrom assessment gill-net surveys.

Carrying capacity K was estimated as loge(K) in the model-fitting so that the estimated parameters had similar magnitude,and the log scale also keeps the estimate positive. If fisheriesdata were available from the beginning or very early historyof the fisheries, then it would have been credible to assumethe first-year biomass at or close to the carrying capacity, butin many instances data are collected only after fisheries haveexisted for some length of time. Since it was difficult to makevalid assumptions on the biomass of Cisco in the first year ofthe time series (1985), the model also estimated P0, the biomassin the first year of simulation expressed as a fraction of K. Thefinal parameter that the model estimated was the SD (σ) in themeasurement error of CPUE.

Model 2: SPM with a temperature anomaly and observationerror.—To incorporate the influence of temperature, carryingcapacity is allowed to vary every year as a function of the tem-perature anomaly according to the relationship

Kt = K · exp (λ · Tat ) , (5)

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3

672 KUMAR ET AL.

where Kt is time-varying carrying capacity, Tat denotes the tem-perature anomaly in year t, and λ is an additional parameter to beestimated. The parameter λ describes the relationship betweenKt and temperature anomaly. If λ = 0, then the multiplier oncarrying capacity is equal to 1. A value of λ < 0 has a nega-tive influence on Kt (carrying capacity in year t is higher thanthe average carrying capacity K), while a value of λ > 0 has apositive influence on Kt.

The production function equation (3) in the SPM was modi-fied to incorporate Kt as follows:

Pt+1 = Pt + r Pt (1 − Pt ) − Ct

Kt. (6)

Model 3: State-space SPM with temperature anomaly.—Processerror was incorporated into model 2 as a random effect on thePt as follows:

Pt = Pt · exp (Xt · τ) , (7)

where Xt is the random effect added to the Pt estimated fromequation (6), and parameter tau (τ) is the SD of the randomeffects process error.

Objective FunctionsThe criteria to fit the models was to minimize the negative

log-likelihood (NLL) of the predicted versus observed CPUEvalues. The NLL for models 1 and 2 was calculated as

NLL = 1

2log (2π) + nlog(σ) +

n∑t=1

ε2t

2σ2, (8)

where NLL is summed over all the years (n) for which CPUEdata were available. Epsilon (ε) is the difference between theobserved and predicted CPUE. The SD of the observation erroris denoted by the parameter σ.

The NLL for model 3 was the sum of equations (8) and (9).Equation (9) describes the NLL for the process error component:

NLL = 1

2log (2π) + nlog(τ) +

n∑t=1

[log (Xt )

]2

2τ2, (9)

where, τ is the standard deviation of the process error.

Model PriorsPrior on r for Cisco.—This prior was calculated using a

demographic method suggested in McAllister et al. (2001) andlater updated in Stanley et al. (2009). The demographic methoduses inputs of natural mortality, stock–recruitment steepness,and growth parameters for establishing a stable age structure andthen numerically solves the Euler–Lotka equation to estimate r.The methodology is detailed in Appendix 1. The density of theestimated prior closely resembled a normal probability densityfunction with a mean of 0.40 and SD of 0.09 (Figure 3). Hence,

a normal distribution with the same mean and SD was used todescribe the prior.

Prior on λ.—A rescaled beta function was used to provide anuninformative uniform prior on λ to confine its value between–1 and + 1 (–1 ≤ λ ≤ 1); the value 1.1 was chosen for bothshape parameters of the beta distribution to keep the distributionshape nearly uniform (equation 10).

0.5(λ + 1) = Beta(1.1, 1.1) (10)

Prior on σ and τ.—A fairly uninformative gamma prior (con-jugate prior on precision for normal distribution) was placed onthe precision of observation and process error (equation 11):

1/σ2 = Gamma(3.01, 0.51) (11)

Since the estimates of q were converging in model 1 and model2, a prior for q was not required in the formulation of these mod-els. In model 3, a penalty, –loge (q), was added to the likelihoodfunction to prevent the q from approaching very small magni-tudes.The models were written in AD Model Builder software(Fournier et al. 2009). The model-fitting procedure dependedon taking derivatives of the negative log-likelihood across theparameter space.

Management Variable: MSYMaximum sustainable yield was calculated as a function of

intrinsic rate of population growth (r) and carrying capacity ofthe system (K). For model 1, MSY was calculated as

MSY = rK/4. (12)

For models 2 and 3, the estimates of carrying capacity equation(5) and MSY equation (13) varied depending on the annualtemperature anomaly:

MSYt = r · Kt

4, (13)

where, Kt and MSYt are the time-varying carrying capacity andthe MSY, respectively.

RESULTSUnder all SPM formulations, the posterior mean for the pa-

rameter r was estimated to be lower than the prior mean of r(Figure 3; Table 1). Parameter r consistently approached thelower bound to capture the population decline when the priorwas not used; the informative prior used here rectified this prob-lem. Since r and K are correlated, providing a prior for r also,to some extent, addressed the problem related to scale of thebiomass estimates.

If the mean of λ was 0, it would indicate no correlation oftemperature with carrying capacity. Posterior mean λ was esti-mated as −0.36 for model 2 and −0.40 for model 3 indicating

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3


FIGURE 3. Parameter densities from all surplus production model formulations. The prior and posterior densities of parameters from model 1, model 2, andmodel 3 are superimposed in each panel. [Figure available in color online.]

that approximately 36–40% of the change in carrying capacitywas a result of change in temperature (Table 1). The posteriordensities of λ estimated by models 2 and 3 almost overlappedeach other (Figure 3) and supported a negative correlation oftemperature with carrying capacity. Correlation plots of the pa-rameters are presented in Appendix 2.

The model fits to the CPUE data were very similar in theperiod 1995–2007, but in the period before 1995 the fits ofmodels 2 and 3 seemed to follow the trends in CPUE better

than in model 1 (Figure 4a) and, compared with models 2 and3, model 1 estimated a lower biomass in 1985 (Figure 4b).Depending on model formulations, estimated biomass of Ciscoranged from 82–147 metric tons in 1985 to 4–10 metric tons in2007. Carrying capacity estimates varied by 37% between themodels; model 2 had the lowest estimates while model 1 hadhighest estimates. The estimated time-varying carrying capacityfor model 2 and model 3 varied as a function of the temperatureanomaly. The estimates from model 3 were consistently higher

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3

674 KUMAR ET AL.

TABLE 1. Parameter estimates across the three models with 95% confidencelimits used to assess decline of Ciscoes in Milles Lacs Lake. r = intrinsic rate ofpopulation growth, K = carrying capacity, q = catchability, λ = correlation co-efficient of temperature with carrying capacity, σ2 = observation error variance,and τ2 = process error variance.

95% confidencelimits

Parameters Models Estimates Low High

r 1 0.30 0.12 0.492 0.32 0.13 0.503 0.38 0.26 0.54

K (metric tons) 1 133.24 106.80 203.702 96.97 72.01 155.533 124.22 94.59 233.35

q 1 0.03 0.01 0.052 0.03 0.02 0.043 0.02 0.01 0.04

λ 2 −0.36 −0.70 −0.103 −0.40 −0.65 0.02

σ2 1 0.64 0.41 1.442 0.53 0.36 1.223 0.25 0.12 0.67

τ2 3 0.28 0.11 0.87

than the estimates from model 2, but on a relative scale thepatterns were very similar and both the models showed morethan a three-fold difference in carrying capacity between lowand high temperature years (Figure 4c).

We compared the model predictions for stock status andfishery status and found that all three models gave similar pre-dictions about current depletion and the trajectory of fishingmortality (Figure 5). All the models suggested that the harvestrate was higher than MSY in several years after 1995 (Figure 5).However, the models estimated substantially different base-year(1985) biomasses. Model 1 estimated the fraction of biomassto carrying capacity at base year (P0) as 0.61 while models 2and 3 estimated it to be nearly at carrying capacity (Figure 5),suggesting a stronger decline in biomass caused by temperaturechange.

Based on the Akaike’s information criterion (AIC) statistic,which measures the goodness of fit while adjusting for the ad-dition of parameters, model 2 was judged to best describe thegiven data (Table 2).

DISCUSSION

Model FormulationIt is not uncommon to incorporate environmental variables

to assess the climatic effects on population dynamics in vari-ous fields. In a study of the Pacific Sardine Sardinops sagax,Jacobson et al. (2005) investigated the environmental effects on

FIGURE 4. Time series of (a) fitted CPUE, (b) estimated biomass, and(c) temperature-influenced, time-varying carrying capacity estimates. Dottedlines in panel c indicate 95% CI of the estimates for respective models. [Figureavailable in color online.]

fish productivity (productivity was correlated with variation inhabitat area related to El Nino) using an “environmentally de-pendent surplus model (EDSP)”. A study on the predicted effectof a temperature increase on a songbird, Cinclus cinclus, popu-lation in Norway related net recruitment rate to temperature ef-fect and estimated the change in carrying capacity (Sæther et al.2000). The response of an Atlantic Cod Gadus morhua popu-lation under various climate change scenarios was explored byrelating recruitment to temperature variables (Clark et al. 2003).In our model, the parameter λ was used to define a correlationbetween carrying capacity and temperature. Since summerkillwas observed for Cisco, we hypothesized that Cisco survival inMille Lacs Lake decreased in years of high temperature, thusaffecting the carrying capacity in any given year. The currentmodel formulation could be altered to include parameter λ asa correlation with intrinsic rate of growth r if it were expectedthat juvenile survival rate was associated with the environmentalcovariate. The alternate parameterization would lead to differ-ent parameter estimates, but the resulting population biomassdynamics predictions would be very similar because r and Kparameters are correlated with each other.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3


FIGURE 5. Biomass relative to carrying capacity (K), and fishing effort (F) relative to FMSY across all the models’ formulations. Dotted lines show 95% CI forrespective models.

Including process error in the model did not alter the es-timate of the temperature decline parameter, and thus showedthat process error in the model could not explain the decline inthe Cisco population. We explored additional model formula-tions, including alternate descriptions of temperature anomalyusing accumulated degree-days, an asymmetric surplus pro-duction model (modified Fox model), and process error onlymodels. Model 2 provided a better fit to the data than any ofthe additional models explored. For the purpose of clarity, theadditional models are not included here, but are provided inAppendix 2.

We consider that our modeling approach is the first step to-ward quantifying the effect of temperature on Cisco. Given thedata limitations, we were not able to explore an age-structuredmodel to capture the effect of temperature on age-specific co-horts of the population. Since several lakes have exhibited Cisco

mortality (Colby and Brooke 1969; Jacobson et al. 2008), it maybe possible to undertake a large-scale meta-analytic modelingof lake-specific and age-specific effects of temperature on Ciscopopulations.

Temperature, Management, and Ecosystem ImplicationsMany coldwater fish species move to cooler, deeper wa-

ters of lakes in the summer months. Deeper waters, however,might present constraints associated with lower oxygen con-centrations. Jacobson et al. (2008) described that Cisco mor-tality was generally observed in lakes with high temperaturesin the epilimnion and hypoxic conditions in the metalimnionand hypolimnion. In Amisk Lake, Alberta, Ciscoes were foundin deeper waters after lake aeration (Aku et al. 1997), whichincreased the available habitat when the oxygen concentration

TABLE 2. Comparison of the three models used to assess decline of Ciscoes in Milles Lacs Lake based on AIC statistics.

Objective function Number ofModels (−loge-likelihood) Loge-likelihood parameters AIC �AIC

Model 1 30.55 −30.55 5 71 4Model 2 27.54 −27.54 6 67 0Model 3 35.60 −35.60 29 129 62

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3

676 KUMAR ET AL.

improved in the hypolimnion. Mille Lacs Lake is located at thesouthern extent of Cisco distribution and is a relatively shal-low lake. As stated previously, a thermocline does not developin this lake, and Ciscoes do not have an opportunity to find athermal refuge in deeper cooler waters. This further establishesthe importance of temperature as a driving factor for Ciscomortality.

All three models were able to explain the Cisco populationdecline (B2007/K < 0.1). It is not possible to rule out whetherthe relationship of the stock size with temperature is causal oraccidental. Previous observations of summerkill in the lake andthe results from models 2 and 3 indicate that Cisco decline isa combined effect of temperature and fisheries. Temperatureswere high in the period from 1987 to 1990, and this periodexperienced lower-than-average carrying capacity (Figure 4c);however, when temperatures decreased in the following years(1990–1997), the carrying capacity increased and biomass in-creased until 1995. Catch decreased after 1997 but the fishingmortality (F) considerably exceeded the FMSY in the years after1995 (Figure 5). Following the high harvest rate in high temper-ature year 2001, the biomass fell to less than 10% of the carryingcapacity. Temperatures in 2002 were similar to the temperaturesin 1991–1992, but biomass levels were much lower. The mainreason was that in years preceding 1991, fishing mortality waslower than FMSY, while in years preceding 2002, fishing mor-tality was higher than FMSY. Thus, the increase in temperaturewas a major cause in decline of Cisco, but fishing (F/FMSY >

2) was also an important contributor to the decline.Results of models 2 and 3 suggested that an increase in tem-

perature in future years would lead to increased stress on MilleLacs Lake Ciscoes and could lead to further population decline.Catch in 2006 was low (<5 metric tons compared with historiccatch levels of >20 metric tons), and the fishing mortality waslower than FMSY. As a result, the biomass in 2007 showed a 2%increase. In order to ensure species recovery, the study suggeststhat strong restrictions on Cisco catch be made in warmer years,especially in light of the current depletion level.

In this research, we did not explore change in predatory pres-sure as a possible cause for the decline of Ciscoes. The abun-dance of Walleye Sander vitreus, the major predator speciesof Cisco, has remained consistent over the study period, andanother forage species, Yellow Perch Perca flavescens, con-tributes to more than 75% of the Walleye’s diet. However, wecannot ignore the possibility of Walleyes being a driver in thedecline of Ciscoes, and this question could be explored furtherin ecosystem-wide models of Mille Lacs Lake.

ACKNOWLEDGMENTSWe thank Minnesota Department of Natural Resources for

providing data and funding for the research. We especially thankstaff at the Fisheries Section, MDNR, in Aitkin for their consis-tent cooperation.

REFERENCESAku, P. M. K., L. G. Rudstam, and W. M. Tonn. 1997. Impact of hypolimnetic

oxygenation on the vertical distribution of cisco (Coregonus artedi) in AmiskLake, Alberta. Canadian Journal of Fisheries and Aquatic Sciences 54:2182–2195.

Clark, R. A., C. J. Fox, D. Viner, and M. Livermore. 2003. North Sea cod andclimate change: modelling the effects of temperature on population dynamics.Global Change Biology 9:1669–1680.

Colby, P. J., and L. T. Brooke. 1969. Cisco (Coregonus artedii) mortalities in asouthern Michigan lake, July 1968. Limnology and Oceanography 14:958–960.

Edsall, T. A., and P. J. Colby. 1970. Temperature tolerance of young-of-the-yearcisco, Coregonus artedii. Transactions of the American Fisheries Society99:526–531.

Fournier, D. A., H. J. Skaug, J. Ancheta, J. Ianelli, A. Magnusson, M. N.Maunder, A. Nielsen, and J. Sibert. 2009. AD model builder: automaticdifferentiation model builder. ADMB Project. Available: admb-project.org.(April 2012).

Heiskary, S., J. Koser, and J. Hodgson. 1994. Lake Mille Lacs 1992 cleanlakes study (314a) water quality report. Minnesota Pollution Control Agency,Division of Water Quality, St. Paul.

Hilborn, R., and M. Mangel. 1997. The ecological detective: confronting modelswith data. Princeton University Press, Princeton, New Jersey.

Jacobson, L. D., S. J. Bograd, R. H. Parrish, R. Mendelssohn, and F. B. Schwing.2005. An ecosystem-based hypothesis for climatic effects on surplus produc-tion in California sardine (Sardinops sagax) and environmentally dependentsurplus production models. Canadian Journal of Fisheries and Aquatic Sci-ences 62:1782–1796.

Jacobson, L. D., J. A. A. De Oliveira, M. Barange, M. A. Cisneros-Mata, R.Felix-Uraga, J. R. Hunter, J. Y. Kim, Y. Matsuura, M. Niquen, C. Porteiro,B. Rothschild, R. P. Sanchez, R. Serra, A. Uriarte, and T. Wada. 2001. Sur-plus production, variability, and climate change in the great sardine and an-chovy fisheries. Canadian Journal of Fisheries and Aquatic Sciences 58:1891–1903.

Jacobson, P. C., T. S. Jones, P. Rivers, and D. L. Pereira. 2008. Field estimationof a lethal oxythermal niche boundary for adult ciscoes in Minnesota lakes.Transactions of the American Fisheries Society 137:1464–1474.

Jones, T. 2006. Mille Lacs Lake creel survey and population assessment sum-maries. Minnesota Department of Natural Resources, Section of Fisheries,St. Paul.

McAllister, M. K., E. K. Pikitch, and E. A. Babcock. 2001. Using demographicmethods to construct Bayesian priors for the intrinsic rate of increase in theSchaefer model and implications for stock rebuilding. Canadian Journal ofFisheries and Aquatic Sciences 58:1871–1890.

MDNR (Minnesota Department of Natural Resources). 1995. Methods andestimates of harvestable fish production in the 1837 treaty area. MDNR,Section of Fisheries, St. Paul.

Punt, A. E., and A. D. M. Smith. 2001. The gospel of maximum sustainable yieldin fisheries management: birth, crucifixion and reincarnation. Pages 41–66in J. D. Reynolds, G. M. Mace, K. H. Redford, and J. G. Robinson, editors.Conservation of exploited species. Cambridge University Press, Cambridge,UK.

Sæther, B. E., J. Tufto, S. Engen, K. Jerstad, O. W. Røstad, and J. E. Skatan.2000. Population dynamical consequences of climate change for a smalltemperate songbird. Science 287:854–856.

Schaefer, M. B. 1954. Some aspects of the dynamics of populations important tothe management of the commercial marine fisheries. Inter-American TropicalTuna Commission Bulletin 1:27–56.

Scott, W. B., and E. J. Crossman. 1973. Freshwater fishes of Canada. FisheriesResearch Board of Canada Bulletin 184.

Stanley, R. D., M. McAllister, P. Starr, and N. Olsen. 2009. Stock assessmentfor Bocaccio (Sebastes paucispinis) in British Columbia waters. CanadianScience Advisory Secretariat Research Document 2009/055.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3


APPENDIX 1: SPECIFICATION OF PRIORSThe demographic method (McAllister et al. 2001; Stanley

et al. 2009) numerically solves the Euler–Lotka equation to es-timate r using natural mortality, stock–recruitment steepness,and growth parameters as inputs. Equation (A.1.1) below is amodification of the original Euler–Lotka equation in which theintegration is performed from age 0; when there is no reproduc-tion in the age-0 class of individuals, the integration starting atage 1 is equivalent (Stanley et al. 2009). The initial guess forr in the calculation was 0.4, based on Jensen’s (1984) surplusproduction model for Ciscoes in Lake Superior. The equation is

T∑t=1

lt mt e−tr = 1, (A.1.1)

where t denotes age, T is the maximum age of the species,lt denotes survivorship at age, and mt is the number of age-1recruits expected to be produced by females of age t. For furtherdetails on the calculation of lt and mt, please see Stanley et al.(2009).

Monte Carlo sampling of 10,000 combinations from the dis-tributions of natural mortality and stock–recruitment steepnesswas done to obtain a distribution around the estimate of the priorfor r. The distribution of M was uniform with a lower boundof 0.30 and an upper bound of 0.65, based on mortality profilesand the resulting M estimates for Cisco year-classes from 1998to 2004 in Mille Lacs Lake (Tom Jones, MNDNR, personalcommunication).

Steepness information was not available for the Mille LacsLake Cisco population; the steepness used (normal distributionwith a mean of 0.53 and a standard deviation of 0.06) wasbased on estimates of steepness for salmonid species from Myerset al.’s (1999) meta-analysis of stock–recruitment relationships.The starting values of r were also varied during the estimation.The density of the estimated prior closely resembled a normalprobability density function, so a normal distribution with amean of 0.40 and a standard deviation of 0.09 was used todescribe the prior in fitting the surplus production model.

APPENDIX 2: SUPPLEMENTAL INFORMATION ONMODEL DEVELOPMENT

Supplementary Figures: Parameter Correlation Plots forModels 1, 2 and 3

FIGURE A.2.1. Parameter correlation plots for model 1 used to assess declineof Ciscoes in Milles Lacs Lake.


Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3

678 KUMAR ET AL.


Additional Data and Models

Temperature DataCorrelation of Air Temperature and Water Temperature Data

Water temperature data from Mille Lacs lake were not avail-able for the 23-year time period for which CPUE informationwas available. For the few years for which we had water temper-ature data, we checked if the monthly average air temperaturedata at Garrison were correlated with the monthly average watertemperature data from Mille Lacs Lake. We found that the twotemperature series were in complete agreement except duringthe winter months with ice cover on the lake. In our analysis,we used temperatures only from July and August where thetemperatures were in agreement. Additionally, we used theair temperature series to calculate the anomaly, thus takinginto account only the trend and not the absolute temperaturerecords.

TABLE A.2.1. Temperature data and corresponding anomaly estimates. Column 2 shows the maximum air temperatures attained at Garrison, Minnesota, in theJuly–August months every year. Column 3 shows the anomaly calculated based on the temperature in Column 2. This is the anomaly data used in the main text.Garrison air temperature data were also used to calculate accumulated degree-days (ADD; column 4). The ADD data were used in additional models presentedlater in Appendix 2.

Jul–Aug Anomaly based on Accumulated Anomaly basedYear maximum temperature maximum temperatures degree-days on ADD

1985 88 −0.75 439.55 −0.811986 90 −0.24 449.05 −0.511987 94 0.77 492.65 0.891988 99 2.03 529.25 2.061989 94 0.77 507.80 1.371990 96 1.27 459.55 −0.171991 91 0.01 453.05 −0.381992 91 0.01 437.30 −0.881993 86 −1.25 431.05 −1.081994 86 −1.25 441.55 −0.751995 88 −0.75 442.30 −0.721996 87 −1.00 427.90 −1.181997 84 −1.76 434.05 −0.991998 89 −0.49 439.05 −0.831999 92 0.26 477.05 0.392000 86 −1.25 435.80 −0.932001 94 0.77 518.55 1.712002 91 0.01 456.80 −0.262003 89 −0.49 467.55 0.082004 91 0.01 442.30 −0.722005 91 0.01 487.55 0.722006 99 2.03 517.05 1.672007 93 0.52 495.05 0.96

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3


FIGURE A.2.4. Time series of air temperature (◦F) at Garrison and Mille LacsLake (MLL) water temperature (◦F) from 2000 to 2005. The number after theyear in the horizontal axis denotes the month.

CPUE Data

TABLE A.2.2. CPUE data from Milles Lacs Lake with CV (for years whenavailable).

Mean weightYear (lb/net) CV

1985 5.201986 7.101987 8.201988 5.001989 8.601990 4.301991 2.101992 2.501993 4.711994 5.701995 2.371996 6.981997 2.551998 2.201999 2.00 2.212000 2.30 1.862001 0.84 2.322002 0.40 2.472003 1.23 1.082004 0.93 1.382005 3.21 1.752006 0.09 2.39

Additional Models

Seven other models, listed as follows, were explored in ad-dition to the three models presented in the main text:

(1) Degree-days model 2(2) Degree-days model 3(3) Process error only model 1(4) Process error only model 2(5) Fox surplus production model 1(6) Fox surplus production model 2(7) Fox surplus production model 3

The performance of models, including degree-days, was verysimilar to the models presented in the main text. The valuesof the AIC statistic were slightly higher than the models basedon maximum temperature anomaly. The estimate of parameterλ was less certain with the models based on degree-days. Theobjective functions for the process error only models indicatedthat these models’ performance approached the other models,but the standard deviations on the parameter estimates for Kwere very high, up to a few thousand metric tons. The modifiedFox versions of the surplus production models did not performas well as the Schaefer surplus production models. However,the parameter estimates were very similar.

Model Parameter Densities

In Figure A.2.5 for parameter densities, densities in bluerepresent alterations to model 1, densities in red representalterations to model 2, and densities in black represent alter-ations to model 3.

Model Details

Degree-days-based models.—The additional models,degree-days model 2 and degree-days model 3, simply replacedthe temperature anomaly data in model 2 and model 3 with thetemperature anomaly calculated based on ADD.

Process error only models.—We do not think it is appropriateto apply a process error only version of any fisheries model as itis never possible to obtain perfect data for abundances. However,for exploratory purposes only, we developed alternate versionsof process error only models 1 and 2. The model formulationsare based on Hilborn and Mangel (1997). Here we assumedthat the CPUE observations were perfect representations of thepopulation biomass and all the error observed was a result ofmodel process error. The details are as follows:

Fox Model 1

Bt+1 = 1

qCPUEt + r

1

qCPUEt

(1 − 1

K q

)CPUEt − Ct .

For model 2, the above equation was modified to include thelambda (λ) parameter:

Bt+1 = 1

qCPUEt + r

1

qCPUEt

(1 − 1

K · exp (−λ · Tat ) · q

)

× CPUEt − Ct .

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3

680 KUMAR ET AL.

TABLE A.2.3. Model performance compared based on the AIC statistic.

Objective function Log– Number of AICModels (negative log-likelihood) likelihood parameters score �AIC

Model 1 30.55 –30.55 5 71 4Model 2 27.54 –27.54 6 67 0Model 3 35.60 –35.60 29 129 62Degree-days model 2 28.83 –28.83 6 70 3Degree-days model 3 37.20 –37.20 29 132 65Process error only model 1 32.76 –32.76 5 76 8Process error only model 2 29.92 –29.92 6 72 5Fox surplus production model 1 35.69 –35.69 5 81 14Fox surplus production model 2 32.39 –32.39 6 77 10Fox surplus production model 3 36.14 –36.14 29 130 63

Fox Version of the Surplus Production Model

The Schaefer surplus production (SP) is represented as

SP = r Bt

(1 − Bt

K

).

The Fox model surplus production is represented as

SP = ln (K )r Bt

[1 − ln (Bt )

ln (K )

].

The Pella and Tomlinson surplus production is representedas

SP = r

pBt

[1 −

(Bt

K

)p].

Haddon (2010) suggests that if the “p” term in the above equa-tion is close to zero, then the parameters from the Fox modelSP equation and the Pella–Tomlinson model SP equation aredirectly comparable. We used the Pella–Tomlinson model equa-tion with p fixed at 0.01 as the Fox surplus production equation.

FIGURE A.2.5. Parameter density plots for additional models described in Appendix 2.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3


APPENDIX REFERENCESHaddon, M. 2010. Modelling and quantitative methods in fisheries. Chapman

and Hall/CRC Press, Boca Raton, Florida.Hilborn, R., and M. Mangel. 1997. The ecological detective: confronting mod-

els with data (MPB-28), volume 28. Princeton University Press, Princeton,New Jersey.

Jensen, A. L. 1984. Dynamics of fisheries that affect the population growth ratecoefficient. Environmental Management 8:135–140.

McAllister, M. K., E. K. Pikitch, and E. A. Babcock. 2001. Using demographicmethods to construct Bayesian priors for the intrinsic rate of increase in the

Schaefer model and implications for stock rebuilding. Canadian Journal ofFisheries and Aquatic Sciences 58:1871–1890.

Myers, R. A., K. G. Bowen, and N. J. Barrowman. 1999. Maximum reproductiverate of fish at low population sizes. Canadian Journal of Fisheries and AquaticSciences 56:2404–2419.

Stanley, R. D., M. McAllister, P. Starr, and N. Olsen. 2009. Stock assessmentfor bocaccio (Sebastes paucispinis) in British Columbia waters. CanadianScience Advisory Secretariat Research Document 2009/055.200.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:15

15

July

201

3



Influences of Riparian Vegetation on Trout StreamTemperatures in Central WisconsinBenjamin K. Cross a , Michael A. Bozek b & Matthew G. Mitro ca School of the Environment , Washington State University , Post Office Box 646410,Pullman , Washington , 99164 , USAb National Park Service, Inventory and Monitoring Division , 12795 West Alameda Parkway,Lakewood , Colorado , 80225 , USAc Wisconsin Department of Natural Resources , Science Operations Center , 2801 ProgressRoad, Madison , Wisconsin , 53716 , USAPublished online: 20 Jun 2013.

To cite this article: Benjamin K. Cross , Michael A. Bozek & Matthew G. Mitro (2013) Influences of Riparian Vegetation onTrout Stream Temperatures in Central Wisconsin, North American Journal of Fisheries Management, 33:4, 682-692, DOI:10.1080/02755947.2013.785989







http://dx.doi.org/10.1080/02755947.2013.785989




ARTICLE

Influences of Riparian Vegetation on Trout StreamTemperatures in Central Wisconsin

Benjamin K. Cross*School of the Environment, Washington State University, Post Office Box 646410, Pullman,Washington 99164, USA

Michael A. BozekNational Park Service, Inventory and Monitoring Division, 12795 West Alameda Parkway, Lakewood,Colorado 80225, USA

Matthew G. MitroWisconsin Department of Natural Resources, Science Operations Center, 2801 Progress Road, Madison,Wisconsin 53716, USA

AbstractSummer stream temperatures limit the distribution of Brook Trout Salvelinus fontinalis and are affected by riparian

vegetation. We used riparian and instream habitat surveys along with stream temperature loggers placed throughoutstreams to determine the potential for riparian vegetation shading to increase the length of stream that is thermallysuitable for Brook Trout. Twelve streams located throughout central Wisconsin were evaluated in the summers of2007 and 2008. Across all streams, nonparametric ANCOVA modeling was used to identify spatial temperaturepatterns within a year for individual stream segments. Riparian tree-vegetated segments had a significantly lowermean change in stream temperature per kilometer of stream compared with grass-vegetated segments during theperiods of maximum daily and weekly average temperatures, when we accounted for upstream temperature. Ripariangrass-vegetated segments increased on average 1.19◦C/km (SE, 0.44) during the maximum daily average temperatureperiod and 0.93◦C/km (SE, 0.39) during the maximum weekly average temperature period, whereas tree-vegetatedsegments decreased 0.48◦C/km (SE, 0.39) and 0.30◦C/km (SE, 0.25) during those respective time periods. Maximumweekly average temperatures were also modeled with different shading levels using a heat budget temperature model,U.S. Fish and Wildlife Service’s Stream Segment Temperature Model. Across 11 study streams (one stream modelcould not be calibrated), modeled stream temperatures in equilibrium with their environmental conditions rangingfrom 23.2◦C to 28.3◦C at 0% shading could be reduced to 18.8–23.5◦C with 75% shading. Modeled increases in shadeup to 75% from the current average of 34% increased the length of surveyed stream thermally suitable to BrookTrout by 4.9 km on Sucker Creek. We conclude that riparian forests are important for maintaining thermal conditionssuitable for Brook Trout in central Wisconsin streams and can be managed to increase the amount of stream habitatthermally suitable for Brook Trout.

Before early European settlement of central Wisconsin,stream riparian areas were dominated by hardwood forests (Cur-tis 1959). During this period, many of these forested streamscontained abundant Brook Trout Salvelinus fontinalis popula-tions (Becker 1983). Land use activities after early settlement

*Corresponding author: [email protected] February 10, 2012; accepted March 11, 2013

in Wisconsin altered forested riparian areas by the removal oftrees for timber harvest and to accommodate agriculture (Curtis1959; Wang et al. 1997). Such land use changes contributedto increased stream temperatures and the extirpation of BrookTrout from streams in which thermal refugia were unavailable

682

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3

RIPARIAN VEGETATION INFLUENCES ON STREAM TEMPERATURE 683

(Becker 1983; Hayes et al. 1998; Baird and Krueger 2003).Given that Brook Trout distribution is strongly linked to streamtemperatures (e.g., Bozek and Hubert 1992; Wehrly et al. 2007;Mitro et al. 2010) and stream temperatures can be linked toriparian shading (e.g., Brown 1970; Poole and Berman 2001;Rutherford et al. 2004), quantifying the relationship between ri-parian vegetation and stream temperatures in central Wisconsinstreams will help determine the importance of managing ripar-ian vegetation to protect or restore thermal habitat for BrookTrout.

Thermal regimes of streams control the abundance anddistribution of Brook Trout throughout a length of stream(Bozek and Hubert 1992; Stoneman and Jones 2000; Wehrlyet al. 2003). Brook Trout generally survive at temperaturesbelow 22.3–23.3◦C (Eaton et al. 1995; Wehrly et al. 2007).Water temperatures can negatively affect reproduction, diseaseand parasite susceptibility, predation vulnerability, interspecificcompetition, feeding, growth, and movement, some of whichare affected by the influence of temperature on metabolism(Brett 1971; Hokanson et al. 1973; Elliott 1981; Taniguchi et al.1998). Behavioral thermoregulation is a strategy Brook Troutuse to cope with physiologically unsuitable water temperaturesand involves the migration of these fish to stream reaches wheretemperatures are more preferable for survival (Meisner 1990;Hayes et al. 1998; Baird and Krueger 2003).

Brook Trout studies have identified optimum temperaturesfor growth to be between 13◦C and 16.1◦C (Hokanson et al.1973; Dwyer et al. 1983), preferred temperatures to be between10◦C and 19◦C (Hokanson et al. 1973), and the upper limit ofthermal tolerance to be between 23.5◦C and 25.5◦C (Fry et al.1946; Cherry et al. 1977). Field observations have identifiedBrook Trout distributions to be limited by maximum weeklyaverage temperatures (MWAT) in excess of 22.3◦C (Eaton et al.1995), 23.3◦C (Wehrly et al. 2007), and 24◦C (Meisner 1990),and limited by maximum daily average temperatures (MDAT) inexcess of 24◦C (Barton et al. 1985; Picard et al. 2003). However,Brook Trout are found at greater abundances when temperaturesare closer to their preferred levels. In fact, Wehrly et al. (1999)found July 3-week average stream temperatures between 15◦Cand 19◦C had the highest densities of Brook Trout, Brown TroutSalmo trutta, and Rainbow Trout Oncorhynchus mykiss in lowerMichigan rivers.

With many factors influencing the quantity of heat energyin a stream, stream thermal regimes can be quite complex andhave very dynamic energy budgets (Bartholow 1989; Poole andBerman 2001; Moore et al. 2005). Thermally influential factorsinclude stream discharge and initial temperature, solar radiation,streambed conduction, air temperature, wind speed, stream mor-phology, and groundwater inputs (Bartholow 1989; Poole andBerman 2001). Different types and amounts of riparian vege-tation can influence the thermal regime of a stream by alter-ing solar inputs, creating various microclimates, and varyinggroundwater inputs. Forest microclimates are cooler than non-forested areas in the daytime and slightly higher at nighttime,

leading to lower air temperatures, decreased air temperaturefluctuations, and more stable and cooler water temperatures(Moore et al. 2005). Solar radiation received by a stream isone of the most influential factors affecting stream temperatures(Brown 1970; Beschta et al. 1987; Johnson 2004). Riparian veg-etation canopies provide shade thereby reducing solar radiationreceived by a stream leading to lower overall summer watertemperatures and a reduction in stream temperature fluctuations(Brown 1970; Barton et al. 1985; Johnson 2004). The effects ofriparian vegetation on stream temperatures becomes evident inareas where riparian vegetation has been clear-cut (Barton et al.1985; Beschta et al. 1987; Bartholow 2000; Mellina et al. 2002).

Management of riparian areas and streams is in need ofintegration and science-based guidance so that shade can bemaintained or restored where it is needed to regulate watertemperatures and possibly increase habitat availability suitablefor Brook Trout. In this study, we quantify the relationshipbetween riparian vegetation and stream temperature in order toassess the potential for gains in thermally suitable Brook Trouthabitat by managing for forested riparian areas. The objectivesof this study were to (1) develop empirical thermal profiles forselected trout streams in central Wisconsin to identify transi-tions from suitable to unsuitable thermal conditions for BrookTrout, (2) quantify how stream temperature responds to riparianvegetation type, and (3) utilize a stream temperature model topredict the effects of riparian shading on stream temperature.

METHODSStudy sites.—We examined 12 streams in our study, six in

2007 and six in 2008 (Figure 1). Stream selection was basedon three criteria. First, streams had to transition from upstreamareas known to support year-to-year survival and reproductionof Brook Trout to downstream areas, before a confluence with alarger river, not known to support year-to-year survival or repro-duction of Brook Trout (WDNR 2002). Downstream reaches inthese streams may become too warm for trout survival in thesummer and were targeted for this reason. Thus, with known nat-ural reproduction in the upstream reaches, increased wild troutdistribution during summer may be realized by extending appro-priate water temperatures farther downstream. Second, streamshad to be less than 6 m in width to practically demonstrate theinfluences of shading by higher levels of riparian vegetationon stream temperatures, because wider streams are not shadedto the same degree as are narrower streams under similar ri-parian vegetation (DeWalle 2010). Third, study streams had tobe located in the North Central Hardwoods Forest Ecoregion ofWisconsin (Omernik and Gallant 1988), a region formerly dom-inated by forested riparian areas (Curtis 1959), which wouldmake the restoration of forested riparian areas more logical.

Data collection.—Streams were divided into segments basedon riparian vegetation type in order to relate stream temperaturechanges to vegetation type and characteristics (e.g., percentshading), which could influence thermal characteristics of

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3

684 CROSS ET AL.

FIGURE 1. Study stream locations in Wisconsin’s North Central HardwoodForest Ecoregion (Omernik and Gallant 1988) with stream identification num-bers. Streams sampled in 2007: (1) Unnamed Creek 17–5/Cunningham Creek,(2) Blake Creek, (3) Magdanz/Hatton Creek, (4) Little Wolf River (NorthBranch), (5) Sucker Creek, and (6) West Branch Shioc River. Streams sam-pled in 2008: (7) Bronken Creek, (8) Gillespie Creek, (9) Onemile Creek,(10) Spring/South Branch Pine Creek, (11) Walla Walla Creek, and (12) Web-ster Creek.

streams. Vegetation types were divided into four categories:(1) grass (≥75% stream shading provided by grasses), (2) shrub(≥75% stream shading provided by shrubs, woody plants <5 mtall), (3) tree (≥75% stream shading provided by woody plants≥5 m tall), and (4) transitional vegetation, which was anyvegetation type not categorized by the three previous categorytypes.

Stream temperature data loggers were placed at distancesno greater than 1,000 m apart throughout the entire surveyedlength of each stream during 2007 and additionally at the up-stream and downstream end of each stream segment in 2008.In one stream, West Branch Shioc River, measurement errors insegment length and the loss of several temperature loggers re-sulted in the largest interval of 1,950 m between data loggers. Inall surveyed streams, temperature data loggers were also placedabove tributaries and in tributaries far enough upstream to avoidany mixing at the confluence of the two streams, which allowedfor discrete measurements of tributary temperature influences.Stream temperatures were monitored from June 1 to August31 to ensure that maximum summer stream temperatures wererecorded. However, in the summer of 2007 some temperaturedata loggers were set as late as June 20. The later date of datalogger placement in 2007 was not thought to have influenced therecording of MWAT or MDAT because of the higher flows andlower air temperatures that occurred prior to June 20 relative tothe rest of the summer. Onsest HOBO Pro v2 water tempera-ture data loggers (accurate to ± 0.21◦C at 0–50◦C) were usedto record stream temperatures, and stream temperatures wererecorded at 65 sites in 2007 and at 101 sites in 2008. Prior touse, loggers were checked for accuracy and placed in PVC tub-ing that covered their entire length to avoid direct sunlight butwhich allowed water to flow through the tubing. Loggers were

placed near the thalweg in naturally shaded areas with no directsunlight and recorded stream temperature at 30-min intervals.

All streams were surveyed to obtain data for use in streamtemperature prediction models while using common surveytechniques (Platts et al. 1987; Bartholow 2004). Parametermeasurements for stream temperature modeling occurred atsegment-, transect-, and site-specific scales. Physical habitatwithin each stream segment was surveyed using cross-sectionaltransects. Transects were randomly stratified throughout thelength of each segment in order to obtain a minimum of fivetransects per segment. Measurements were taken as close tosummer base flow (July–August) as possible because this iswhen stream temperatures are usually highest in the region.

To characterize the stream segment morphology for use inthermal models, stream width, depth, flow, slope, and segmentlength were measured. Average depth, stream width, and per-cent shade were measured at each cross-sectional transect us-ing methods described by Platts et al. (1987). Cross-sectionaltransect variables were averaged for each segment to assignan overall quantitative value. Stream segment length and slopewere measured at the segment scale using GIS. Discharge wasmeasured according to Platts et al. (1983) at the beginning andend of each study stream, at the mouth of tributaries, in the studystream just above the confluence with tributaries, and at severaleasily accessible locations throughout the stream to obtain amore precise understanding of groundwater dynamics.

Variables quantifying riparian vegetation shading and solarinputs included percent vegetative shading, segment elevation,and segment latitude. Percent vegetative shading was measuredalong the cross-sectional transect using a modified sphericaldensiometer according to Platts et al. (1987) and was then aver-aged using a technique modified by Lawson (2005) in order tobe assessed as a percentage. Elevation and latitude were averagevalues calculated for each segment using GIS.

Recorded meteorological variables included air temperature,percent possible sun, relative humidity, and wind speed. Airtemperature values were averaged across each modeling periodusing the closest weather station or by averaging values whenweather stations occurred at similar distances from the studystream (Bartholow 2004). Possible sun, relative humidity, andwind speed were obtained from local climatological data reportsfrom weather stations at major airports because these data arenot reported by smaller weather stations (Bartholow 2004).

Data analysis.—We created longitudinal thermal profiles forall study streams using empirical stream temperature data forthe period coinciding with MWAT. The temporal thermal pro-files were used to identify transitions from suitable to unsuitablethermal conditions for Brook Trout and make comparisons be-tween actual and modeled stream temperatures under variousshading scenarios.

To assess how differences in vegetation type affected changesin empirical stream temperature per segment during this study,nonparametric ANCOVA was used. The nonparametric test wasperformed by running an ANCOVA on ranked data. From the 12

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3


study streams located throughout central Wisconsin 59 streamsegments were analyzed. Vegetation types (grasses, shrubs, andtrees) were the main effects analyzed during the MWAT andMDAT periods. Upstream temperatures were ranked and usedas the covariate. The dependent variable in the ANCOVA, that is,the change in stream temperature per kilometer for modeled timeperiods, was also ranked. Homogeneity of regression slopes foreach factor was tested. Fisher’s least-significant-difference testwas used to determine significant differences between streamtemperature change and individual riparian vegetation types.All analyses were performed using SAS with alpha set at P ≤0.05 (SAS 2010).

We also demonstrated the effects of various levels of riparianvegetation shading on stream thermal regimes using the StreamSegment Temperature Model (SSTEMP: Bartholow 1989). TheSSTEMP is a steady-state physics-based model that includesstream geometry, hydrology, meteorology, and shade inputs, allof which are averaged for each modeling segment (Bartholow1989). The model computes the heat exchange for a volume ofwater with a known initial temperature by simulating heat fluxprocesses occurring throughout a stream segment to calculatethe water temperature at the downstream end of a segment. At-mospheric radiation, convection, conduction, evaporation, solarradiation, stream back radiation, and vegetative radiation wereheat flux processes included in the model. Several assumptionswere made when using SSTEMP and are discussed in detail byBartholow (1989, 2004).

Models for all study streams during the MWAT periodwere created and calibrated to represent actual stream ther-mal regimes. The SSTEMP outputs included stream segmentmean outflow temperature and mean equilibrium temperatures(the temperature that a stream approaches if all environmentalconditions remain under a steady state) for the modeled MWATperiod. Stream temperatures were modeled from the locationof the temperature data logger farthest upstream with sufficientflows for model input (some streams had data loggers upstreamof perennial flows) to the temperature data logger farthest down-stream. The SSTEMP was then individually calibrated for eachstream segment specifically for the MWAT time period to min-imize additive error, which could confound the estimates of thefarthest downstream segment. Calibrations of discharge, lateralinflow temperature, and shading variables were performed onmodels to represent measured stream temperatures accordingto Bartholow (1989, 2004). Discharge calibration for individ-ual stream segments was necessary because measurements werenot made directly at the starting and ending points for all seg-ments. To account for the nonconstant rates of lateral flows,modeled input and output discharge values for each segmentwere calibrated to best represent observed stream temperatures.Discharge calibrations were performed randomly, within rea-son, to coincide with values at points of measured discharge.Lateral inflow temperatures were also modified during calibra-tion. Originally, lateral inflow temperatures were estimated tobe 1.5◦C above mean annual air temperature, representing an

estimate of groundwater temperature (Collins 1925, cited byBlann et al. 2002). However, shallow groundwater and sur-face flows were observed in several segments, and lateral in-flow temperature was increased by 0.79◦C on average (SE,0.12) to account for elevated temperatures of surface waterinputs.

Models were used to predict how various levels of shadingwould affect temperature profiles; simulated shading levels were0, 25, 50, and 75%. One hundred percent shading, simulatingcomplete blockage of all solar radiation inputs, was not modeledbecause it would not occur under natural conditions. Moreover,75% stream shading was deemed appropriate for the purposeof demonstrating the effects of natural higher levels of streamshading on stream temperatures. In fact, for streams less than6 m wide, stream shade modeling suggests that 80% shade canbe achieved by managing for 12-m-wide riparian buffer stripswith 30-m-tall and dense (leaf area index, ∼6) vegetation in mid-latitude areas such as central Wisconsin (DeWalle 2010). UsingBrook Trout thermal criteria of 22.3◦C for the MWAT (Eatonet al. 1995), the length of stream predicted to be thermally suit-able for Brook Trout was recorded for each study stream atthe end of each simulation. Modeled increases or decreases instream length suitable for Brook Trout were assessed to quantifythe potential gains or losses from various stream-shading sce-narios when compared with actual stream temperature profiles.We also reported the predicted stream equilibrium temperatures,that is, the temperature achieved if the stream temperature wasin equilibrium with its environment, which can occur over var-ious distances based on the stream environment and the steadystate of the environmental conditions present.

RESULTSOf the 12 streams 5 crossed the thermal suitability threshold

for Brook Trout at the downstream end of their surveyed lengthduring the MWAT period. In 2007, six study stream lengthsranged from 6.1 to 13.7 km, and the length of thermally suitablewater for Brook Trout ranged from 3.9 to 11.2 km (Table 1;Figure 2). In 2008, the length of six additional study streamsranged from 4.1 to 15.0 km (Table 1; Figure 3). The entirelengths of all the streams monitored in 2008 were found to bethermally suitable for Brook Trout during the MWAT period.Streams from both study years had MWAT temperatures rang-ing from 13.3◦C to 23.6◦C and MDAT temperatures rangingfrom 13.3◦C to 25.4◦C. The change in stream temperaturesmeasured per kilometer of stream during the MWAT periodranged from −3.1 to 3.4◦C/km with a mean value of 0.28◦C/kmamong segments. During the MDAT period, the change instream temperatures measured per kilometer of stream rangedfrom −4.6 to 3.9◦C/km with a mean value of 0.39◦C/km. TheMDAT and MWAT periods did not occur over the same timeperiod within all streams and were different between moststreams.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3

686 CROSS ET AL.

TABLE 1. Actual surveyed length of the study streams that were thermally suitable for Brook Trout (<22.3◦C) during the period of MWAT compared withcalibrated SSTEMP model predictions under 0, 25, 50, and 75% shade conditions for the surveyed length of stream with respective predicted equilibriumtemperatures (Teq).

Actual surveyed 0% Shade 25% Shade 50% Shade 75% Shade

Stream nameLength(km)

Suitablelength(km)

Suitablelength(km)

Teq(◦C)

Suitablelength(km)

Teq(◦C)

Suitablelength(km)

Teq(◦C)

Suitablelength(km)

Teq(◦C)

2007Unnamed

17–5/CunninghamCreek

6.1 4.7 0.4 26.1 2.1 24.4 6.1 22.5 6.1 20.5

Blake Creek 11.2 11.2 6.1 25.7 10.6 24.3 11.2 22.9 11.2 21.4Magdanz/Hatton

Creek6.9 5.3 3.1 26.0 5.6 23.2 6.9 23.2 6.9 21.8

Little Wolf River(North Branch)

13.7 10.5 3.2 26.4 8.1 24.9 13.7 23.3 13.7 21.5

Sucker Creek 8.8 3.9 3.5 26.5 4.2 25.0 6.0 23.4 8.8 21.7West Branch Shioc

River6.3 6.2 2.3 27.0 4.1 23.4 6.3 23.4 6.3 21.3

2008Bronken Creek 8.2 8.2 8.2 23.2 8.2 22.0 8.2 20.8 8.2 19.6Gillespie Creek 4.1 4.1Onemile Creek 12.5 12.5 12.5 28.3 12.5 26.8 12.5 25.2 12.5 23.5Spring/South Branch

Pine Creek11.8 11.8 11.8 23.4 11.8 22.0 11.8 20.5 11.8 18.8

Walla Walla Creek 15.0 15.0 9.0 24.8 15.0 23.1 15.0 21.3 15.0 19.4Webster Creek 10.9 10.9 10.9 25.5 10.9 23.9 10.9 22.2 10.9 20.5

Longitudinal warming occurred under natural conditions in11 of 12 study streams with the exception of Walla Walla Creek,which was the only stream fed by a headwater lake, resultingin warmer initial stream temperatures compared with the otherspring-fed study streams. Walla Walla Creek received large in-puts of groundwater as it progressed downstream, transitioningto colder stream temperatures (Figure 2).

Stream Temperature Response to RiparianVegetation Type

Stream segments with grass-vegetated riparian areas showedgreater increases in stream temperature per kilometer duringthe MDAT and MWAT period compared with tree-vegetatedsegments, which in many cases provided cooling reaches (Fig-ure 4). During the MDAT period, significant differences werepresent in stream temperature change per kilometer among thestreams segments with three riparian vegetation types: grasses,shrubs, and trees (F = 5.95, P = 0.0046). Segments with grass-vegetated riparian areas had significantly higher changes instream temperature per kilometer compared with those withtree-vegetated riparian areas (P = 0.0017), while stream seg-

ments with shrub-vegetated riparian areas did not differ signif-icantly from those with grass- or tree-vegetated areas duringthe MDAT period. On average across all stream segments dur-ing the MDAT period, grass-vegetated segments increased by1.19◦C/km (SE, 0.44) and tree-vegetated segments decreased by0.48◦C/km (SE, 0.39). During the MWAT period, the changes instream temperature per kilometer of stream among the segmentswith three riparian vegetation types were also significantly dif-ferent (F = 6.46, P = 0.0030). Stream segments with tree- andshrub-vegetated riparian areas did not differ significantly fromeach other during the MWAT period, but both were significantlylower than those with grass-vegetated riparian segments (P =0.0023 and P = 0.0224, respectively). An average increase of0.93◦C/km (SE, 0.39) in grass-vegetated segments and decreaseof 0.30◦C/km (SE, 0.25) in tree-vegetated segments were ob-served in all stream segments during the MWAT period. Despitedifferences in the effects of vegetation type on changes in streamtemperatures, there was not a significant relationship betweenthe change in stream temperature per kilometer of stream andthe covariate, upstream temperature, during the MDAT (F =1.59, P = 0.2131) or MWAT period (F = 2.46, P = 0.1224). As

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3


FIGURE 2. Longitudinal thermal profiles of the 2007 study streams during the period of MWAT compared with modeled stream temperatures under variouslevels of shading with the suitability threshold for Brook Trout set at 22.3◦C. The x-axis represents distance downstream (km) and the y-axis represents streamtemperature (◦C). Note the variance in both scales between graphs. Vertical lines represent locations of tributaries entering the stream.

a result, when adjusted for upstream temperature in the analysis,the change in stream temperature per kilometer was affected toa greater degree by the vegetation type compared with the initialupstream temperature.

SSTEMP Modeling of Shading EffectsThe predicted MWAT period temperatures from the noncal-

ibrated SSTEMP models deviated from measured stream tem-peratures in several segments. The greatest deviation was 3.2◦C

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3

688 CROSS ET AL.

FIGURE 3. Longitudinal thermal profiles of the 2008 study streams during the period of MWAT compared with modeled stream temperatures under variouslevels of shading with the suitability threshold for Brook Trout set at 22.3◦C. The x-axis represents distance downstream (km) and the y-axis represents streamtemperature (◦C). Note the variance in both scales between graphs. Vertical lines represent locations of tributaries entering the stream.

for Spring /South Branch Pine Creek. The average maximumdeviation was 2.2◦C for the MWAT period. Using the initial data,all stream models were successfully calibrated except for Gille-spie Creek, for which temperatures on average were 1.69◦C

cooler than our heat budget model could explain (Figure 3).All calibrated stream temperature models deviated from actualrecorded temperatures by 0–1.2◦C per segment with an averagedeviation of 0.1◦C (see Supplementary Figure 1 in the online

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3


FIGURE 4. Change in stream temperature (◦C) per kilometer of stream versusupstream temperature at 59 stream segments (grass, n = 37; shrub, n = 8; andtree, n = 14) across central Wisconsin during the (a) MDAT and (b) MWATperiods.

version of this article). Calibrated stream temperature modelspredicted the farthest downstream temperature with an R2 valueof 0.9995 compared with 0.9246 for noncalibrated SSTEMPmodels.

Varying levels of modeled riparian vegetation shadingdemonstrated the influence of shade on stream thermal regimes(Table 1; Figures 2, 3). Shading modeled at the 0% level led toa decrease in the length of stream thermally suitable for BrookTrout in 7 of 11 streams. Five streams remained suitable un-der the 0% shading scenario because of the abundance of coldgroundwater inputs throughout the modeled segments of thestreams. For the streams predicted to lose thermal habitat under0% shading, Unnamed 17-5/Cunningham Creek had the largestloss in thermally suitable stream length with a predicted lossof 91% (4.3 km) of its current 4.7 km of suitable Brook Trouthabitat. When 25% shade was modeled, only 4 of the 11 streamswere predicted to have decreases in the length of suitable water

for Brook Trout with Unnamed 17-5/Cunningham Creek pre-dicted to decrease by 55% (2.6 km) of its current suitable length.At 25% shade, 2 of the 11 streams were predicted to have slightincreases in suitable water for Brook Trout, with a 6% (0.3 km)increase on Magdanz/Hatton Creek and an 8% (0.3 km) increaseon Sucker Creek. Modeling streams at 50% shade also increasedthe length of surveyed stream suitable for Brook Trout in 5 ofthe 11 streams by as much as 54% (2.1 km) on Sucker Creekand 30% (3.2 km) on the Little Wolf River (North Branch); nodecreases in thermally suitable stream length were predicted.Shading modeled at 75% led to the amount of surveyed streamsuitable to Brook Trout lengthening by 126% (4.9 km) on SuckerCreek.

The effects of increased shading were also evident by the re-sponse of stream equilibrium temperature (temperature achievedif the stream is in equilibrium with its environment). Progres-sive increases in shading led to progressive decreases in equilib-rium temperatures. None of the 11 study streams modeled hadequilibrium temperatures less than 22.3◦C when 0% shade wasmodeled. However, when 25% shade was modeled equilibriumtemperatures for two streams were lower than the suitabilitythreshold. Four streams were lower than the suitability thresholdwith 50% shade, and 10 of the 11 study streams had equilib-rium temperatures less than 22.3◦C with 75% shade modeled.The single stream above the threshold, Onemile Creek, had anequilibrium temperature of 23.5◦C (Table 1). However, OnemileCreek and other creeks remained below suitable threshold tem-peratures for Brook Trout for some distance despite high equilib-rium temperatures because of relatively large, cold groundwaterinputs over short distances.

DISCUSSIONModeling relations between riparian vegetation and stream

temperature can be insightful for management agencies inter-ested in protecting or increasing water habitat that is thermallysuitable for Brook Trout because they are restricted to a nar-row thermal range of colder water temperatures (Elliott 1981;Bozek and Hubert 1992; Stoneman and Jones 2000; Wehrly et al.2003). In this study, we showed that riparian vegetation influ-enced stream temperatures by identifying differences betweenriparian vegetation types and the changes in stream temperatureand using heat budget model-shading simulations. Stream tem-peratures, in our study, were cooler under riparian forests andwarmer under riparian grasses. In southeast Minnesota, Blannet al. (2002) also found, in general, that stream temperaturesdecreased along forested reaches and increased slightly in suc-cessional and grazed reaches. For the streams we surveyed, ourfindings indicated temperatures would not surpass temperaturethresholds suitable for Brook Trout if riparian areas were com-pletely forested, provided initial suitable stream temperatureswere present and all other factors stayed the same (e.g., dis-charge, stream size, and average air temperatures). Moreover,SSTEMP modeling of increases in riparian vegetation shading

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3

690 CROSS ET AL.

demonstrated the opportunity to increase the amount of streamlength that was thermally suitable for Brook Trout.

Because the management of riparian vegetation managementcan influence the amount of solar radiation affecting a stream,trout stream temperatures can be maintained or improved. Theprimary source of heat energy inputs into streams comes fromdirect solar radiation (Brown 1970; Beschta et al. 1987; Johnson2004). Using heat budget analyses similar to SSTEMP, Johnson(2004) found solar inputs to a nonshaded stream reach in theOregon Cascade Range at 1200 hours accounted for 840 W/m2

of the incoming energy fluxes compared with the next high-est heat energy input of 18 W/m2 from convection, with a netenergy flux of 580 W/m2 entering the stream. Conversely, un-der complete shading, the net energy flux was 149 W/m2 withonly 4 W/m2 coming from solar inputs. We predicted streamsto lose as much as 7.3 km (Little Wolf River, North Branch) ofwater suitable for Brook Trout if shading was not present (0%shade). Alternatively, streams that were found to be unsuitablefor Brook Trout in some reaches were predicted to gain thermalhabitat (e.g., as much as 4.9 km in Sucker Creek if continuouslyshaded at the 75% level).

One stream, Gillespie Creek, contained cooling stream seg-ments that could not be explained within our heat budget model.Other studies have identified colder-than-predicted stream tem-peratures caused by the hyporheic flow exchanges between shal-low groundwater and the water in the stream channel (Poole andBerman 2001; Story et al. 2003). Hyporheic flow exchange pro-cesses are unaccounted for in SSTEMP and may have causedthe observed cooling effects in Gillespie Creek.

Equilibrium stream temperatures decreased with increasinglevels of shading using SSTEMP in our study. Gaffield et al.(2005) also predicted equilibrium temperatures in southwestWisconsin streams to lower with increased stream shading. Ourstudy predicted the average equilibrium temperature to lowerby 4.8◦C as stream shading increased from 0% to 75%, andGaffield et al. (2005) predicted average equilibrium tempera-tures to lower by 4.5◦C as stream shading increased from 0% to80%. Moreover, Blann et al. (2002), using similar heat budgetanalysis in southeastern Minnesota, estimated weekly averagetemperatures can be as much as 2.5◦C higher for streams passingthrough nonshaded reaches compared with shaded reaches.

For riparian vegetative management to affect stream temper-atures, additional factors need to be considered. The length ofstream necessary to reach equilibrium temperature for any givenriparian condition is dependent upon stream size (discharge andwidth) because the high specific heat of water and water volumeaffect the rate of change in water temperature (Poole and Berman2001). Rutherford et al. (2004) predicted that it takes 1,200 mfor streams 1–2 m in width to reach equilibrium temperatures inAustralia and New Zealand and that streams with a width lessthan 2.5 m were provided the same amount of shade regard-less of vegetation type. Moreover, once stream temperatures inreaches of nonshaded stream exceed equilibrium temperaturesin shaded reaches, only then can shade have a cooling effect onstream temperatures. If stream temperatures are below the equi-

librium temperature under certain shade conditions, then shadeallows the stream to warm at a slower rate. Under conditionsbelow equilibrium temperature, decreases in stream tempera-ture can only occur as a result of groundwater inputs, but shadeplays a role in the magnitude of the cooling (Bartholow 2000).

Management RecommendationsEfforts aimed at expanding Brook Trout distribution and in-

creasing angling opportunities can focus on identifying BrookTrout streams that are restricted by temperature and manag-ing them for more optimal thermal regimes to improve overalltrout population health. Examination of temporal thermal pro-files can be used to identify stream areas where appropriatemanagement of riparian vegetation may be used to address ther-mal limitations to Brook Trout. Using current large-scale streamtemperature prediction models (i.e., Wehrly et al. 1998; Stewartet al., in press) to identify trout streams with possible thermallimitations during summer can minimize the need for creatingdetailed thermal profiles for all streams region wide by focus-ing on reaches most likely to benefit from improved riparianmanagement. Management agencies can identify target streamtemperatures for Brook Trout; when streams approach or exceedthese temperature thresholds, management actions can be usedto maintain or enhance colder thermal regimes.

Fisheries managers are increasingly concerned about the po-tential impacts of climate change on coldwater resources. Airtemperature plays an important role in determining stream tem-perature (Pilgrim et al. 1998; Isaak et al. 2012). Wisconsin hasbecome warmer on average over the past 60 years, and thiswarming trend is predicted to continue; the statewide averageair temperature will increase by up to 3–4◦C by the year 2050(WICCI 2011). In order to adapt to such changes in climate, fish-eries management may need to focus limited available resourcesto those streams most likely to benefit from active management,such as managing for forested riparian areas to reduce solarinput to streams so that appropriate thermal regimes can be pro-tected or enhanced to benefit native species such as Brook Trout.Stream temperatures can vary widely in limited geographic ar-eas in Wisconsin, which points to the fact that variables inaddition to air temperature play a key role in defining thermalregimes. Our study suggests that managing for forested riparianareas is one option that resource managers can use to managefor coldwater thermal regimes in groundwater-fed streams incentral Wisconsin threatened by warming changes in climate.

ACKNOWLEDGMENTSThis project was supported by Federal Aid in Sport Fishery

Restoration Program, Project F-95-P, through the WisconsinDepartment of Natural Resources. Special thanks go to fieldtechnicians Preston Debele, Thomas Rennicke, and MatthewKohler, who were an instrumental part of this project. Additionalthanks are extended to graduate and undergraduate studentsthat contributed to this project as well as the three anonymousreviewers. The use of trade names or products does not constituteendorsement by the U.S. Government.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3


REFERENCESBaird, O. E., and C. C. Krueger. 2003. Behavioral thermoregulation of Brook

and Rainbow trout: comparison of summer habitat use in an Adirondackriver, New York. Transactions of the American Fisheries Society 132:1194–1206.

Bartholow, J. M. 1989. Stream temperature investigations: field and analyticmethods. U.S. Fish and Wildlife Service Biological Report 89(17).

Bartholow, J. M. 2000. Estimating cumulative effects of clearcutting on streamtemperatures. Rivers 7:284–297.

Bartholow, J. M. 2004. Stream segment temperature model (SSTEMP) for Win-dows, version 2.0. U.S. Geological Survey, Fort Collins, Colorado. Available:www.fort.usgs.gov/Products/Publications/pub abstract.asp?PubID=10016.(April 2009).

Barton, D. R., W. D. Taylor, and R. M. Biette. 1985. Dimensions of riparianbuffer strips required to maintain trout habitat in southern Ontario streams.North American Journal of Fisheries Management 5:364–378.

Becker, G. C. 1983. Fishes of Wisconsin. University of Wisconsin Press,Madison.

Beschta, R. L., R. E. Bilby, G. W. Brown, L. B. Holtby, and T. D. Hofstra. 1987.Stream temperature and aquatic habitat: fisheries and forestry interactions.University of Washington Institute of Forest Resources Contribution 57:191–232.

Blann, K., J. F. Nerbonne, and B. Vondracek. 2002. Relationship of riparianbuffer type to water temperature in the driftless area ecoregion of Minnesota.North American Journal of Fisheries Management 22:441–451.

Bozek, M. A., and W. A. Hubert. 1992. Segregation of resident trout in streams aspredicted by three habitat dimensions. Canadian Journal of Zoology 70:886–890.

Brett, J. R. 1971. Energetic responses of salmon to temperature: a study ofsome thermal relations in the physiology and freshwater ecology of SockeyeSalmon (Oncorhynchus nerka). American Zoologist 11:99–113.

Brown, G. W. 1970. Predicting the effect of clearcutting on stream temperature.Journal of Soil and Water Conservation 25(1):11–13.

Cherry, D. S., K. L. Dickson, J. Cairns Jr., and J. R. Stauffer. 1977. Preferred,avoided, and lethal temperatures of fish during rising temperature conditions.Journal of the Fisheries Research Board of Canada 34:239–246.

Curtis, J. T. 1959. The vegetation of Wisconsin: an ordination of plant commu-nities. University of Wisconsin Press, Madison.

DeWalle, D. R. 2010. Modeling stream shade: riparian buffer height and den-sity as important as buffer width. Journal of the American Water ResourcesAssociation 46:323–333.

Dwyer, W. P., R. G. Piper, and C. E. Smith. 1983. Brook Trout growth efficiencyas affected by temperature. Progressive Fish-Culturist 45:161–163.

Eaton, J. G., J. H. McCormick, B. E. Goodno, D. G. O’Brien, H. G. Stefany,M. Hondzo, and R. M. Scheller. 1995. A field information-based system forestimating fish temperature tolerances. Fisheries 20(4):10–18.

Elliott, J. M. 1981. Some aspects of thermal stress on freshwater teleosts. Pages209–245 in A. D. Pickering, editor. Stress and fish. Academic Press, NewYork.

Fry, F. E. J., J. S. Hart, and K. F. Walker. 1946. Lethal temperature relationsfor a sample of young speckled trout (Salvelinus fontinalis). University ofToronto Press, Toronto.

Gaffield, S. J., K. W. Potter, and L. Wang. 2005. Predicting the summer temper-ature of small streams in southwestern Wisconsin. Journal of the AmericanWater Resources Association 41:25–36.

Hayes, D. B., W. W. Taylor, M. T. Drake, S. M. Marod, and G. E. Whelan.1998. The value of headwaters to Brook Trout (Salvelinus fontinalis) in theFord River, Michigan, USA. Pages 175–185 in M. J. Haigh, J. Krecek, G. S.Rajwar, and M. P. Kilmartin, editors. Headwaters: water resources and soilconservation. A. A. Balkeman, Rotterdam, The Netherlands.

Hokanson, K. E. F., J. H. McCormick, B. R. Jones, and J. H. Tucker. 1973.Thermal requirements for maturation, spawning, and embryo survival of theBrook Trout, Salvelinus fontinalis. Journal of the Fisheries Research Boardof Canada 30:975–984.

Isaak, D. J., S. Wollrab, D. Horan, and G. Chandler. 2012. Climate change effectson stream and river temperatures across the Northwest U.S. from 1980–2009and implications for salmonid fishes. Climatic Change 113:499–524.

Johnson, S. L. 2004. Factors influencing stream temperatures in small streams:substrate effects and a shading experiment. Canadian Journal of Fisheries andAquatic Sciences 61:913–923.

Lawson, L., editor. 2005. A manual of procedures for the sampling of surfacewaters. Arizona Department of Environmental Quality, Report TM05–01,Phoenix.

Meisner, J. D. 1990. Potential loss of thermal habitat for Brook Trout, dueto climatic warming, in two southern Ontario streams. Transactions of theAmerican Fisheries Society 119:282–291.

Mellina, E., R. D. Moore, S. G. Hinch, J. S. Macdonald, and G. Pearson. 2002.Stream temperature responses to clearcut logging in British Columbia: themoderating influences of groundwater and headwater lakes. Canadian Journalof Fisheries and Aquatic Sciences 59:1886–1900.

Mitro, M. G., J. D. Lyons, and J. S. Stewart. 2010. Predicted effects of climatechange on the distribution of wild Brook Trout and Brown Trout in Wisconsinstreams. Pages 69–76 in R. F. Carline and C. LoSapio, editors. Conservingwild trout: proceedings of wild trout X. Wild Trout Symposium, Bozeman,Montana.

Moore, R. D., D. L. Spittlehouse, and A. Story. 2005. Riparian microclimateand stream temperature response to forest harvesting: a review. Journal of theAmerican Water Resources Association 41:813–834.

Omernik, J. M., and A. L. Gallant. 1988. Ecoregions of the upper Midwest states.U.S. Environmental Protection Agency, Environmental Research Laboratory,EPA/600/3–88–037, Corvallis, Oregon.

Picard, C. R., M. A. Bozek, and W. T. Momot. 2003. Effectiveness of us-ing summer thermal indices to classify and protect Brook Trout streams innorthern Ontario. North American Journal of Fisheries Management 23:206–215.

Pilgrim, J. M., X. Fang, and H. G. Stefan. 1998. Stream temperature correlationswith air temperatures in Minnesota: implications for climate warming. Journalof the American Water Resources Association 34:1109–1121.

Platts, W. S., C. Armour, G. D. Booth, M. Bryant, J. L. Bufford, P. Cuplin, S.Jensen, G. W. Lienkaemper, G. W. Minshall, S. B. Monsen, R. L. Nelson,J. R. Sedell, and J. S. Tuhy. 1987. Methods for evaluating riparian habitatswith applications to management. U.S. Forest Service General TechnicalReport INT-221.

Platts, W. S., W. F. Megahan, and G. W. Minshall. 1983. Methods for evaluatingstream, riparian, and biotic conditions. U.S. Forest Service General TechnicalReport INT-138.

Poole, G. C., and C. H. Berman. 2001. An ecological perspective on in-streamtemperature: natural heat dynamics and mechanisms of human-caused ther-mal degradation. Environmental Management 27:787–802.

Rutherford, J. C., N. A. Marsh, P. M. Davies, and S. E. Bunn. 2004. Effectsof patchy shade on stream water temperature: how quickly do small streamsheat and cool? Marine and Freshwater Research 55:737–748.

SAS (Statistical Analysis Systems). 2010. SAS version 9.3. SAS Institute, Cary,North Carolina.

Stewart, J. S., S. M. Westenbroek, and M. G. Mitro. In press. A model for eval-uating stream temperature response to climate change in Wisconsin. USGSScientific Investigation Report.

Stoneman, C. L., and M. L. Jones. 2000. The influence of habitat features onthe biomass and distribution of three species of southern Ontario streamsalmonines. Transactions of the American Fisheries Society 129:639–657.

Story, A., R. D. Moore, and J. S. Macdonald. 2003. Stream temperatures intwo shaded reaches below cutblocks and logging roads: downstream cool-ing linked to subsurface hydrology. Canadian Journal of Forest Research33:1383–1396.

Taniguchi, Y., F. J. Rahel, D. C. Novinger, and K. G. Gerow. 1998. Temperaturemediation of competitive interactions among three fish species that replaceeach other along longitudinal stream gradients. Canadian Journal of Fisheriesand Aquatic Sciences 55:1894–1901.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3

692 CROSS ET AL.

Wang, L., J. Lyons, P. Kanehl, and R. Gatti. 1997. Influences of watershed landuse on habitat quality and biotic integrity in Wisconsin streams. Fisheries22(6):6–12.

WDNR (Wisconsin Department of Natural Resources). 2002. Wisconsin troutstreams. WDNR, Pub-FH-806, Madison.

Wehrly, K. E., L. Wang, and M. Mitro. 2007. Field-based estimates of ther-mal tolerance limits for trout: incorporating exposure time and temperaturefluctuation. Transactions of the American Fisheries Society 136:365–374.

Wehrly, K. E., M. J. Wiley, and P. W. Seelbach. 1998. Landscape-based modelsthat predict July thermal characteristics of lower Michigan rivers. Michi-gan Department of Natural Resources, Fisheries Research Report 2037, AnnArbor.

Wehrly, K. E., M. J. Wiley, and P. W. Seelbach. 1999. A thermal habitat classifi-cation for lower Michigan rivers. Michigan Department of Natural Resources,Fisheries Research Report 2038, Ann Arbor.

Wehrly, K. E., M. J. Wiley, and P. W. Seelbach. 2003. Classifying regional varia-tion in thermal regime based on stream fish community patterns. Transactionsof the American Fisheries Society 132:18–38.

WICCI (Wisconsin Initiative on Climate Change Impacts). 2011. Wisconsin’schanging climate: impacts and adaptation. Nelson Institute for EnvironmentalStudies, University of Wisconsin–Madison and the Wisconsin Departmentof Natural Resources, Madison. Available: www.wicci.wisc.edu. (February2012).

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:16

15

July

201

3



Genetic Monitoring of Threatened Chinook SalmonPopulations: Estimating Introgression of NonnativeHatchery Stocks and Temporal Genetic ChangesDonald M. Van Doornik a , Debra L. Eddy b , Robin S. Waples c , Stephen J. Boe d , Timothy L.Hoffnagle b , Ewann A. Berntson a & Paul Moran ca Northwest Fisheries Science Center, Manchester Research Station , Post Office Box 130,Manchester , Washington , 98353 , USAb Oregon Department of Fish and Wildlife, Northeast-Central Fish Research and Monitoring ,Eastern Oregon University, 203 Badgley Hall, One University Boulevard , La Grande ,Oregon , 97850 , USAc National Marine Fisheries Service, Northwest Fisheries Science Center , 2725 MontlakeBoulevard East, Seattle , Washington , 98112 , USAd Confederated Tribes of the Umatilla Indian Reservation, Grande Ronde Monitoring andEvaluation, Agricultural Services Center , 10507 North McAlister Road, Room 2, Island City ,Oregon , 97850 , USAPublished online: 03 Jul 2013.

To cite this article: Donald M. Van Doornik , Debra L. Eddy , Robin S. Waples , Stephen J. Boe , Timothy L. Hoffnagle , EwannA. Berntson & Paul Moran (2013) Genetic Monitoring of Threatened Chinook Salmon Populations: Estimating Introgression ofNonnative Hatchery Stocks and Temporal Genetic Changes, North American Journal of Fisheries Management, 33:4, 693-706,DOI: 10.1080/02755947.2013.790861







http://dx.doi.org/10.1080/02755947.2013.790861




ARTICLE

Genetic Monitoring of Threatened Chinook SalmonPopulations: Estimating Introgression of NonnativeHatchery Stocks and Temporal Genetic Changes

Donald M. Van Doornik*Northwest Fisheries Science Center, Manchester Research Station, Post Office Box 130, Manchester,Washington 98353, USA

Debra L. EddyOregon Department of Fish and Wildlife, Northeast-Central Fish Research and Monitoring,Eastern Oregon University, 203 Badgley Hall, One University Boulevard, La Grande, Oregon 97850,USA

Robin S. WaplesNational Marine Fisheries Service, Northwest Fisheries Science Center, 2725 Montlake Boulevard East,Seattle, Washington 98112, USA

Stephen J. BoeConfederated Tribes of the Umatilla Indian Reservation, Grande Ronde Monitoring and Evaluation,Agricultural Services Center, 10507 North McAlister Road, Room 2, Island City, Oregon 97850, USA

Timothy L. HoffnagleOregon Department of Fish and Wildlife, Northeast-Central Fish Research and Monitoring,Eastern Oregon University, 203 Badgley Hall, One University Boulevard, La Grande, Oregon 97850,USA

Ewann A. BerntsonNorthwest Fisheries Science Center, Manchester Research Station, Post Office Box 130, Manchester,Washington 98353, USA

Paul MoranNational Marine Fisheries Service, Northwest Fisheries Science Center, 2725 Montlake Boulevard East,Seattle, Washington 98112, USA

AbstractConservation efforts aimed at Pacific salmon (Oncorhynchus spp.) populations have frequently utilized artificial

propagation in an attempt to increase fish abundance. However, this approach carries the risk of unwanted changesin the genetic characteristics of the target population and perhaps others that might incidentally be affected. We usedgenetic monitoring techniques to estimate the amount of introgression that has occurred from nonnative hatcherystocks into native populations and to determine the extent of genetic changes that have occurred in associationwith supplementation efforts over the past 20–50 years in Snake River Chinook Salmon O. tshawytscha populationsfrom northeastern Oregon. A total of 4,178 fish from 13 populations were genotyped for 12 microsatellite DNA

*Corresponding author: [email protected] October 9, 2012; accepted March 25, 2013

693

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3

694 VAN DOORNIK ET AL.

loci. Expected heterozygosity values for each sample ranged from 0.707 to 0.868. Estimates of the effective numberof breeders per year in the naturally spawning populations ranged from 20.6 to 459.1, whereas in the hatcherypopulations they ranged from 33.8 to 1,118.8. We found that introgression from the Rapid River Hatchery stockwas particularly noticeable in the early 1990s but that it appears to have had a substantial effect on only two of thenative populations (Lookingglass Creek and the upper Grande Ronde River) despite the ample opportunities forintrogression to occur. All seven of the native populations sampled have maintained their levels of within-populationgenetic diversity throughout the sampling period. Overall, this region’s supplementation efforts appear to have hada minimal effect on the genetic diversity of its Chinook Salmon populations.

The history of Pacific salmon Oncorhynchus spp. conser-vation efforts is replete with attempts to mitigate habitat lossand increase their abundance by supplementation with artifi-cial propagation. Such efforts have continued for more than100 years, often with poor documentation (Lichatowich 1999).Here we define artificial propagation as the spawning and rear-ing of fish in captivity (hatcheries) throughout a portion of theirlife until they are released into the wild. Conventional artificialpropagation programs have long been used successfully to com-pensate for the loss of naturally produced salmon by providingincreased numbers of fish available for harvest. More recently,artificial propagation programs have increasingly been used tosupplement natural populations. Supplementation is a specialcase of artificial propagation defined as

the use of artificial propagation in an attempt to maintain or increasenatural production, while maintaining the long-term fitness of thetarget population and keeping the ecological and genetic impactson nontarget populations within specified biological limits (RASP1992).

Releasing hatchery-reared fish into wild populations carriesthe risk of creating unwanted changes in the genetic character-istics of those populations. For example, amplification of onlya small portion of a population through hatchery rearing, theso-called Ryman-Laikre effect, can lead to a decline in a popu-lation’s level of genetic diversity and effective population size,particularly in the face of environmental change (Ryman andLaikre 1991). Any such changes would clearly be detrimentalto conservation efforts aimed at those populations. Additionalconcerns arise if the hatchery-reared fish do not share ancestrywith the populations targeted for supplementation. Nonnativehatchery fish have the potential to supplant the native popula-tion or to interbreed with them, resulting in introgression thatcan reduce the level of genetic divergence between the two pop-ulations, cause a loss of locally adapted traits within the wildpopulation, and reduce the fitness of the introgressed populationthrough outbreeding (Waples 1991; Ford 2001; Reisenbichler2004; Araki et al. 2007; Hansen et al. 2010). The exclusive useof native fish to develop a supplementation stock still carriesthe risk of Ryman-Laikre effects but may alleviate some of theother concerns. While the degree of benefits from using nativefish for supplementation programs is still uncertain (Hulett et al.2004), evidence is mounting that hatchery-reared fish of a nativeorigin have greater reproductive fitness when spawning in the

wild than hatchery-reared fish of a nonnative origin (see Arakiet al. 2008 for a review).

Molecular genetics provides techniques that are able to mea-sure the changes in a population’s genetic parameters resultingfrom natural processes and anthropogenic activities (Schwartzet al. 2007). Such techniques typically measure temporalchanges in the allele frequencies of genetic markers. Several pre-vious genetic monitoring efforts of salmonid supplementationprograms failed to find reductions in genetic diversity or effec-tive population size that could be attributed to supplementation(Hedrick et al. 2000; Eldridge and Killebrew 2008; Small et al.2009; Van Doornik et al. 2011). However, other studies did findsignificant changes in genetic measures that could be attributedto the introgression of hatchery fish into wild populations(Tessier et al. 1997; Heggenes et al. 2006; Hansen et al. 2009;Christie et al. 2012). Each supplementation program has itsown unique set of conditions and practices. Thus, every supple-mentation program should have a genetic monitoring programdesigned to measure the effects it may have on the target popula-tions as well as populations that are not intended to be affected.

The Columbia River basin, located in the Pacific Northwest,historically produced several million adult Chinook Salmon O.tshawytscha annually (Chapman 1986). However, by 1950 over300 dams had been built in the basin, most of which providedno means for upstream fish passage, effectively blocking manymiles of spawning area and contributing to the decline of Chi-nook Salmon populations in the basin (Laythe 1950). Hatch-ery programs in the Snake River, a tributary of the ColumbiaRiver, began in the 1960s, with the goal of using artificialpropagation to mitigate salmon losses associated with the con-struction of four lower Snake River dams (Carmichael andMessmer 1995). The primary goal of these programs was toincrease the number of returning adult fish rather than to re-store the natural populations. Spring- and summer-run ChinookSalmon in the Snake River were listed as threatened under theU.S. Endangered Species Act (ESA) in 1992 (NMFS 1992)due to declines in abundance and other factors (Matthews andWaples 1991). When populations were listed for protection un-der the ESA, hatchery goals expanded to include supplemen-tation of these depressed populations. This created a manage-ment dilemma, as the Lower Snake River Compensation Planincluded a legislative mandate for artificial propagation pro-grams to mitigate the losses caused by hydropower develop-ment (Marshall 2010), whereas the ESA limited the affect any

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3

GENETIC MONITORING OF CHINOOK SALMON 695

such programs could have on the threatened populations (NMFS1992).

Two northeastern Oregon river basins within the Snake Riversystem, the Grande Ronde and Imnaha basins, experiencedpropagation efforts in the 1980s and 1990s that were not assuccessful as hoped in sustainably boosting natural production(McClure et al. 2008). Although the release of millions ofhatchery-reared fish into the Snake River basin (Myers et al.1998) has helped provide increased opportunities for tribal andsport fisheries (Hesse et al. 2006), the natural populations havefailed to reach their stated recovery goals for abundance, as mea-sured by the number of naturally produced spawners observedeach year (HSRG 2009). Artificial propagation programs for theGrande Ronde basin have included either a mix of native andnonnative stocks (stocks with ancestry external to the GrandeRonde basin) or solely native stocks. The main hatchery usedin the Grande Ronde basin is at Lookingglass Creek, whichprimarily made use of two nonnative Chinook Salmon stocks.After construction of the hatchery, the major concern for man-agers was mitigation for lost fish rather than the preservationof endemic stocks, thus justifying the use of nonnative stocks(Crateau 1997). Fish from the Carson National Fish Hatchery(on the lower Columbia River) were transferred to Looking-glass Hatchery and released into the Grande Ronde basin from1982 to 1987 (Olsen et al. 1992). Fish were released directlyfrom the hatchery into Lookingglass Creek, and other GrandeRonde basin rivers (Catherine Creek and the upper GrandeRonde River) with the intent that they would return as adultsand spawn naturally. The Carson stock originated from a mix ofseveral potential Columbia River spring Chinook Salmon stockscollected at Bonneville Dam (Pastor 2004). However, this stockperformed poorly in the Grande Ronde basin, producing lownumbers of returning adult spawners with high straying rates(Myers et al. 1998). Straying occurs when adult salmon return tospawn in a stream different from their natal stream, or in the caseof hatchery-reared fish, a stream different from the one whichthey were released. The Carson stock was replaced by fish fromRapid River Hatchery beginning in 1987 (Olsen et al. 1992). TheRapid River Hatchery stock was founded from fish captured atHells Canyon Dam on the Snake River above the Grande Rondeand Imnaha basins in 1964–1969 (Myers et al. 1998). Thus, theRapid River Hatchery stock represented populations external tothe Grande Ronde basin yet still within the Snake River basin.In addition to releases from the hatchery into LookingglassCreek, juvenile fish from the Rapid River Hatchery stock werereleased directly into Catherine Creek, Lookingglass Creek, andthe Grande Ronde River. However, this hatchery stock also hadlow numbers of returning adult spawners and strayed into otherGrande Ronde basin rivers (the Lostine, Minam and Wenaharivers) at such high levels that in some years they represented35–93% of the fish on the spawning grounds (Crateau 1997).Releases of Rapid River Hatchery stock into the Grande Rondebasin were discontinued after 2000, and recent supplementationefforts have made use of fish naturally returning to LookingglassCreek as well as fish trapped in Catherine Creek (Hesse et al.

2006), a nearby Grande Ronde River population. This repre-sented a shift in the goal of the artificial propagation program,from mitigating for reduced population abundance by producingincreased numbers of fish to supplementing natural populationsto aid in their recovery and conservation (Hesse et al. 2006). Incontrast to the Grande Ronde basin, the Imnaha basin has onlyreceived hatchery-reared fish derived from naturally produced,native Imnaha River fish (Carmichael and Messmer 1995; Hesseet al. 2006).

While only some populations in the Grande Ronde basin re-ceived direct plantings of nonnative fish, all of them potentiallyexperienced introgression from nonnative stocks due to the highstraying rates of the Carson National Fish Hatchery and RapidRiver Hatchery fish. Strays can have immediate impacts on thereproductive success of native fish by competing with them formates and spawning locations; however, their genes will be in-trogressed into the native population only if they successfullyreproduce and produce viable offspring. Despite the fact thathatchery-reared fish can have lower reproductive success thanwild fish when spawning in the wild (Leider et al. 1990; Reisen-bichler and Rubin 1999; Kostow et al. 2003; Araki et al. 2007;Berntson et al. 2011; Theriault et al. 2011; Hess et al. 2012),they can still have a significant effect on the native population,especially if they represent a substantial portion of the spawningpopulation (Goodman 2004; Reisenbichler 2004; but see Hesset al. 2012). In order to successfully conserve the genetic legacyof native populations, it is important to understand the influencethat nonnative stocks have had on those populations.

The present study used tissue samples from Chinook Salmonpopulations collected for genetic monitoring when a renewedeffort began to recover threatened populations in the GrandeRonde and Imnaha basins in the late 1980s (Waples et al. 1991).The availability of samples from the two primary nonnativehatchery stocks released into the study area and archived scalesamples for one Grande Ronde basin population from a time be-fore it was subjected to any artificial propagation or supplemen-tation efforts led us to the following objectives for this study:(1) to estimate the amount of introgression that has occurredfrom nonnative hatchery stocks into the native populations and(2) to determine the extent of any genetic changes in these pop-ulations over the past 20–50 years.

METHODSSample collection.—Tissue samples were collected from

spring- and summer-run Chinook Salmon at eight locations inthe Grande Ronde and Imnaha basins (Figure 1). In addition,samples were collected from the Carson National Fish Hatcheryon the Columbia River in Washington State and the Rapid RiverHatchery on the Little Salmon River in Idaho. The samplescollected represent six hatchery populations, five populationsthat have been supplemented with hatchery-raised fish, and twononsupplemented populations that have not received any directplanting of hatchery-raised fish (but may have been affected byunintended strays; Table 1). All populations were sampled in

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


FIGURE 1. Map showing the locations at which samples were obtained (R = river, H = hatchery). Note that the Grande Ronde, Lostine, and Imnaha hatcherylocations identified are only trapping sites; the offspring of fish collected at those sites are raised at Lookingglass Hatchery.

multiple years, resulting in a sample set that represented 4–13brood years for each population.

Tissue samples were collected from the captured fish in avariety of ways. Most samples were obtained by nonlethal sam-pling of fin tissues, which were preserved in 95% ethanol andstored at ambient temperature. Fish were captured as parr inrearing areas by electrofishing or seining or from hatcheries bydipnetting to sample all raceways in a given facility. Earlier parrsamples were originally collected for allozyme analyses, andthus entire fish were kept frozen at −80◦C until fin tissue wasremoved from each fish and preserved in 95% ethanol. A few ofthe collections were from adult fish, sampled either as fin clipsor, in the case of the early Lookingglass Creek samples (collec-tion years 1964 and 1972), as scales stored on gummed scalecards. All collecting was done so as to ensure that the sampleswere random and as broad a representation of the population aspossible.

Sample analyses.—A total of 4,178 fish were genotyped for12 microsatellite DNA loci (Ogo2, Ogo4, Oki100, Omm1080,Ots201b, Ots208b, Ots211, Ots212, Ots213, Ots3M, Ots9, andSsa408) using the methods described by Van Doornik et al.(2011). Because the Lookingglass Creek scale samples wereover 40 years old, we subjected them to a polymerase chainreaction (PCR), preamplification procedure (Piggot et al. 2004).All genotypic data were checked for any instances of duplicatemultilocus genotypes by GenAlEx (Peakall and Smouse 2006).Three individuals were found to have identical genotypes toother fish from the same collection and thus were assumed tobe duplicate tissue samples and were removed from any furtheranalyses.

Statistical analyses.—Departures from Hardy–Weinbergequilibrium were examined for each sample in GENEPOP(Rousset 2008) using a simulated Fisher’s exact test (Guoand Thompson 1992). To test for nonrandom association of

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


TABLE 1. Brood years sampled and the supplementation history for each population used in this study.

Population Brood years sampledPopulation type andsupplementation origin

Stock origin orsupplementation stocks used

Carson National FishHatchery

1994, 1998–2000a Hatchery: nonnative Spring Chinook Salmon stocksabove Bonneville Dam

Grande Ronde Hatchery 2007, 2008 Hatchery: native Grande Ronde RiverImnaha Hatchery 1988, 1992, 1997, 2001,

2002, 2006–2008Hatchery: native Imnaha River

Lookingglass Hatchery 1992, 1993, 1997, 2007 Hatchery: nonnative Carson and Rapid Riverhatcheries

Lostine Hatchery 2005 Hatchery: native Lostine RiverRapid River Hatchery 1988, 1992, 2000, 2007 Hatchery: nonnative Spring Chinook Salmon stocks

above Hells Canyon DamCatherine Creek 1991–1993, 2000, 2003,

2007Supplemented: native and

nonnativeLookingglass Hatchery

Imnaha River 1988, 1992, 1997, 2001,2003, 2005, 2007, 2008

Supplemented: native Imnaha River

Lookingglass Creek 1959–1961,a 1967–1969a,1992–1994, 1997,2003–2006a

Supplemented: native andnonnative

Lookingglass Hatchery

Lostine River 1988, 1991, 1993, 2001,2005, 2006

Supplemented: native Lostine River

Upper Grande Ronde River 1991–1993, 2003, 2008 Supplemented: native andnonnative

Lookingglass Hatchery, UpperGrande Ronde River

Minam River 1989, 1993, 1997, 2001,2002, 2005, 2006

Nonsupplemented

Wenaha River 1991, 1997, 2003, 2006,2007

Nonsupplemented

aEstimate of brood years present in adult samples assuming that the spawning population consisted of 3-, 4-, and 5-year-olds, the age range of almost all Snake River ChinookSalmon spawners (Myers et al. 1998).

genotypes between loci, we performed gametic disequilibriumtests, also using GENEPOP. The Benjamini and Yekutieli falsediscovery rate method was used to correct the critical valuefor multiple tests for departures from Hardy–Weinberg expecta-tions and for gametic disequilibrium (Benjamini and Yekutieli2001; Narum 2006). For each sample, we used the program ML-RELATE (Kalinowski et al. 2006) to estimate the percent ofcomparisons between pairs of individuals that represented full-sibling relationships. The presence of large numbers of relatedindividuals in a sample can violate the assumption of random-ness. We ran the program for 1,000 permutations and report thepercentage of pairwise comparisons that most likely representfull-sibling relationships with P > 0.95; FST values betweenall possible sample pairs were calculated using the programGENETIX (Belkhir et al. 2003), and the data were randomlypermutated 1,000 times to determine the associated P-values.

We used two methods to estimate introgression from thenonnative hatchery stocks into the natural populations. First, wecalculated Cavalli-Sforza and Edwards (1967) chord distancesover 1,000 bootstrap replicates and constructed a consensusneighbor-joining tree with PHYLIP (Felsenstein 2005). The re-sulting tree was depicted with PhyloDraw (Choi et al. 2000),

and the tree structure was used to visualize the pattern of ge-netic relationships among samples.

We also estimated relative migration rates (m) between thetwo nonnative hatchery stocks and each native population. Weused the Bayesian estimator BIMr (Faubet and Gaggiotti 2008)to estimate m between each population throughout our samplingperiod, and our earliest samples from the Rapid River Hatch-ery (brood year [BY] 1988) and Carson National Fish Hatchery(BY 1994). Within a given population sample, BIMr estimatesthe proportion of individuals that have immigrated into thatpopulation from another specified population in the previousgeneration. For each estimate we ran the program for 220,000iterations, the first 20,000 of which were pilot runs to determinethe optimal tree-swapping algorithms, followed by 100,000 iter-ations that were discarded as burn in. The 95% highest posteriordensity interval (a Bayesian analog to a confidence interval)was also computed for each estimate of m. Ten independentestimates were made for each brood year, and the one with thelowest deviance was reported (Faubet et al. 2007).

Possible temporal changes in genetic diversity were exam-ined by comparing several genetic parameters for those popu-lations that we were able to sample over several generations.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


Within-sample variability was measured by calculating the ex-pected level of heterozygosity (HE) using Arlequin (Excoffieret al. 2005) and allelic richness using HP-RARE (Kalinowski2005). Wilcoxon, matched-pairs, signed-rank tests (Zar 1984)were used to evaluate the changes in HE and allelic richness be-tween the earliest and most recent samples for each population(VassarStats; Lowry 2012). We also used FSTAT’s exact testto compare the mean levels of HE and allelic richness amongsamples grouped by their supplementation history (hatchery,supplemented, or nonsupplemented).

For each year in each sample, we calculated the effectivenumber of breeders per year (Nb) using two different methods.The variable Nb is an analog of effective population size pergeneration (Ne) that is more appropriate for semelparous speciessuch as salmon that have variable age at maturity (Waples andTeel 1990; Waples 2005). The first method of estimating Nb

employed the program LDNE, which measures linkage dise-quilibrium at presumably unlinked loci (Waples and Do 2008).As recommended by the program’s authors, only alleles withfrequencies ≥0.02 were used for this calculation, and jackknif-ing was used to calculate 95% confidence intervals. When thedisequilibrium observed in a sample can be attributed solely tosampling error, estimates of Nb become negative. A negativeestimate indicates that the confidence interval around the Nb

estimate includes infinity (∞). The second method employedthe program SALMONNb, which compares allele frequenciesacross three or more temporal samples (e.g., brood years) toestimate Nb for each brood year (Waples et al. 2007). Again,only alleles with frequencies ≥0.02 were used for these calcu-lations, and the proportions of adults returning to each river atages 3, 4, and 5 were estimated from spawning ground surveysand hatchery records (CBFWA 2008; NMFS 2012; Oregon De-partment of Fish and Wildlife, unpublished data). SALMONNbrequires a minimum of three temporal samples from a popula-tion, so we could not make estimates using this method for allpopulations. We also could not estimate Nb from adult samplesusing SALMONNb because the individuals in those samplescould not be separated by brood year. The estimates from LDNEand SALMONNb were combined by calculating their harmonicmeans. Ideally, this would be a weighted harmonic mean thatreflects the relative precision of each estimator, in which casethis strategy should increase the precision of the overall estimate(Waples and Do 2010). However, because SALMONNb doesnot provide a variance associated with its estimates, we com-puted an unweighted harmonic mean. The resulting harmonicmeans were plotted along with estimates of the adult (census)population size (N) for each population over time. When cal-culating the harmonic mean in cases in which one of the Nb

estimates was ∞, a value of zero (1/∞ = 0) was substitutedinto the harmonic mean equation for the value 1/Nb. Estimatesof N were derived from spawning ground surveys and reddcounts that were compiled from several sources (CBFWA 2008;NMFS 2012; Oregon Department of Fish and Wildlife, unpub-lished data). The Nb estimates for the few adult samples we

had are not directly comparable to those for single-cohort, ju-venile samples. The adult samples, which contain individualsfrom multiple brood years, provide an estimate in LDNE thatis influenced by Nb in a number of previous years, whereas thejuvenile samples from a single brood year provide an estimateof Nb in their parental generation (Waples 2005).

RESULTS

Sample AnalysesHigh genetic variability was observed within the populations,

as evidenced by the HE and allelic richness values (Table 2).Tests for departures from Hardy–Weinberg equilibrium foundthat 22.5% of locus pairs showed evidence of significant dis-equilibrium. Slightly more of the significant tests were due tohomozygote excess (12.8%) than were due to heterozygote ex-cess (9.7%). After adjusting the critical value for multiple tests,only 11.5% of the tests were significant. These Hardy–Weinbergequilibrium departures were particularly prevalent in the upperGrande Ronde River and Grande Ronde River Hatchery sam-ples (Table 2). Those same samples also tended to have thegreatest percentages of locus pairs with significant gametic dis-equilibrium and low estimates of Nb. This suggests that thesepopulations have experienced a high degree of nonrandom mat-ing, inbreeding, or introgression from another population. In allsamples fewer than 5% of pairwise comparisons were identifiedas full-sibling relationships (Table 2), suggesting that our resultswere not affected by the presence of large numbers of siblinggroups.

Pairwise FST values ranged from 0.000 to 0.070 (see Supple-mentary Table 1 in the online version of this article). Only 41(1.9%) of the 2,145 pairwise values were not significantly dif-ferent (P > 0.05), and most of these (N = 37) were comparisonsof temporal samples from the same location.

Migration EstimatesEstimates of migration, and thus introgression, from the

Carson National Fish Hatchery and Rapid River Hatchery stocksinto the natural populations varied among populations andthrough time (Figure 2). Given that our samples violate theassumption of BIMr that they consist of nonoverlapping gener-ations (Faubet and Gaggiotti 2008) and the fact that almost allof the estimates had very large 95% highest posterior density in-tervals the values should be interpreted with caution. However,they are still very useful for comparing the relative amount of in-trogression that each of the populations in our study has receivedfrom the hatchery stocks over time. Introgression from the Car-son National Fish Hatchery stock was most notable in CatherineCreek and the upper Grande Ronde River in the early 1990s, butoverall was considerably lower than the estimated amount ofintrogression from the Rapid River Hatchery stock for all pop-ulations. The greatest estimates of introgression from the RapidRiver Hatchery stock also occurred in the 1990s, especially intothe Catherine Creek, Lookingglass Creek, and upper Grande

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


FIGURE 2. Estimates of m, the proportion of individuals in natural Chinook Salmon populations in the Grande Ronde and Imnaha River basins that areimmigrants from Carson and Rapid River hatchery stocks. The Carson National Fish Hatchery stock values are slightly offset on the x-axis so that the 95% highestposterior density interval can be clearly seen.

Ronde River populations. Estimates of m are not shown for theearliest Lookingglass Creek samples because supplementationhad not yet begun when those samples were collected, but, asexpected, they were not significantly different from zero. Thepopulations that have not received any direct supplementation ofCarson National Fish Hatchery or Rapid River Hatchery stockhad varying results. The Imnaha and Lostine rivers showed amuch lower level of introgression than the Minam River, whichshowed a steadily increasing amount of introgression from theRapid River Hatchery stock for the first 4 years of samples,reaching a high in BY 2001, after which time it decreased, andthe Wenaha River, which had a spike up in BY 1992, after whichtime the estimates have been fairly consistent.

Temporal StabilityNone of the populations showed a significant decline in either

heterozygosity or allelic richness (Table 2). In fact, Looking-

glass Creek, the upper Grande Ronde River, and Wenaha Riverhad significant increases in these measures between the earli-est and most recent samples. No significant differences weredetected among samples grouped by their supplementation his-tory for either HE or allelic richness. The Nb estimates show thateffective population sizes do not appear to be declining, and theysomewhat mimic the trends in total population size (Figure 3).The one exception is Lookingglass Creek, which showed a re-duction in more recent Nb estimates compared with the earliersamples from that population, which were taken before supple-mentation began in that system.

The population structure revealed by the dendrogram of ge-netic distances is relatively stable over time for most populations(Figure 4). All temporal samples from the Imnaha, Lostine, Mi-nam, and Wenaha rivers form distinct clusters separate from thesamples from all other populations. In addition, hatchery sam-ples from the Imnaha and Lostine rivers clustered within the

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


TABLE 2. Results for each brood year (BY) sampled, including the number of individuals successfully genotyped, the number of loci (out of 12) that were not inHardy–Weinberg equilibrium (HWE), the percentage of locus pairs that showed significant gametic disequilibrium (GD), the percentage of pairwise comparisonsbetween individuals that revealed a full-sibling relationship (FS), estimates of expected heterozygosity (HE) and allelic richness (AR), and the effective number ofbreeders (Nb) from two different methods and their harmonic mean (HM); na = not available.

Population BYNo. fishtyped

No. locinot in HWE

Sig. GD(%)

FS(%) HE AR

SALMON-Nb Nb

estimateLDNE Nb

estimateLDNE 95%

CIHM Nb

estimate

Carson National 1994 79 3 12.1 0.5 0.849 10.0 naa 120.0 99.5–148.9 naFish Hatchery 1998–2000 52 1 7.6 0.2 0.868 11.1 nab 180.6 134.4–268.7 na

Catherine Creek 1991 60 1 24.2 0.9 0.831 10.2 76.2 70.7 58.8–87.1 73.31992 55 1 6.1 0.9 0.813 9.9 43.2 308.1 136.1–∞ 75.81993 73 2 16.7 0.7 0.834 9.7 152.4 143.0 103.2–222.6 147.62000 94 4 34.8 1.1 0.831 9.8 74.6 43.0 39.1–47.5 54.62003 94 1 1.5 0.1 0.822 10.0 138.6 303.5 229.8–437.8 190.32007 218 5 40.9 0.4 0.831 9.9 105.0 93.7 82.5–106.9 99.0

Grande Ronde 2007 47 0 10.6 0.9 0.816 8.9 naa 34.9 31.0–39.7 naHatchery 2008 150 10 100.0 2.9 0.822 8.8 naa 15.7 14.3–17.2 naUpper Grand Ronde 1991 57 6 66.7 2.8 0.808 8.6 22.6 19.0 16.6–21.8 20.6

River 1992 38 0 6.1 0.7 0.795 9.3 44.2 193.8 77.9–∞ 72.01993 62 1 25.8 1.1 0.818 9.3 107.9 39.0 34.8–44.1 57.32003 103 9 89.4 3.1 0.807 8.2 27.1 16.6 15.0–18.3 20.62008 95 1 21.2 2.9 0.844 9.3 53.6 83.1 69.6–101.5 65.2

Imnaha Hatchery 1988 39 1 15.2 1.1 0.838 9.7 68.8 50.9 43.3–61.0 58.51992 48 1 1.5 0.2 0.852 10.0 208.8 354.5 172.2–37,720.7 262.81997 95 3 24.2 0.9 0.833 9.8 44.6 56.8 51.1–63.5 50.02001 48 0 1.5 0.3 0.831 9.6 64.6 421.7 155.4–∞ 112.02002 47 0 6.1 0.5 0.840 9.7 58.0 169.5 123.9–260.7 86.42006 48 1 16.7 1.0 0.842 9.9 97.0 76.8 64.7–93.3 85.72007 48 2 19.7 1.9 0.827 9.5 64.8 35.9 32.2–40.1 46.22008 46 1 12.1 0.7 0.827 9.6 87.5 73.2 59.9–92.5 79.7

Imnaha River 1988 31 0 3.0 0.2 0.842 10.0 155.0 ∞ 692.0–∞ 310.01992 50 0 6.1 0.7 0.803 9.8 73.6 ∞ ∞–∞ 147.21995 40 0 19.7 1.5 0.826 9.1 26.1 28.6 23.5–35.5 27.31997 72 2 12.1 0.2 0.850 10.0 175.2 247.4 158.2–521.3 205.12001 47 0 1.5 0.1 0.841 9.9 451.8 375.0 214.8–1,294.4 409.82003 47 1 0.0 0.0 0.855 10.2 333.7 540.0 221.0–∞ 412.52005 47 0 0.0 0.2 0.843 10.0 174.9 713.6 251.1–∞ 280.92007 48 2 12.1 0.8 0.845 10.1 144.6 95.8 73.5–133.8 115.22008 47 0 1.5 0.0 0.845 10.0 258.9 2024.6 432.0–∞ 459.1

Lookingglass Creek 1959–1961 62 0 10.6 0.4 0.795 8.1 nab 168.8 95.3–553.0 na1967–1969 87 0 1.5 0.1 0.812 8.2 nab 426.3 227.9–2,154.9 na

1992 57 4 42.4 1.8 0.795 9.3 1,495.2 31.2 27.4–35.8 61.11993 62 0 15.2 0.8 0.831 9.7 51.3 53.8 43.4–68.5 52.51994 58 0 18.2 1.1 0.813 9.4 144.2 32.8 29.2–37.0 53.41997 60 3 40.9 1.9 0.818 8.5 29.9 25.3 22.4–28.8 27.4

2003–2005 31 0 0.0 0.4 0.832 9.7 nab 120.7 81.6–219.2 na2004–2006 60 0 7.6 0.5 0.841 10.1 nab 77.0 63.1–96.8 na

Lookingglass 1992 100 2 48.5 1.3 0.812 9.0 117.5 70.9 61.9 – 82.0 88.4Hatchery 1993 59 0 13.6 0.8 0.812 9.2 146.1 173.1 110.8–357.6 158.5

1997 94 1 40.9 0.5 0.817 9.1 106.5 127.7 99.5–172.8 116.12007 48 1 9.1 0.5 0.822 9.4 24.4 54.8 44.0–70.7 33.8

Lostine Hatchery 2005 47 1 4.5 0.5 0.807 8.7 naa 60.0 47.7–78.5 785Lostine River 1988 31 5 1.5 1.1 0.707 8.3 75.0 ∞ ∞–∞ 150.0

1991 60 2 10.6 0.6 0.798 8.7 38.7 39.7 34.9–45.6 39.21993 61 0 7.6 0.7 0.810 9.3 76.6 86.0 63.1–128.7 81.02001 127 5 37.9 0.7 0.816 8.9 51.1 58.1 51.2–66.4 54.42005 158 4 15.2 0.3 0.814 9.2 80.8 130.4 111.6–154.8 99.82006 80 0 12.1 0.5 0.821 9.0 119.2 115.2 89.2–157.9 117.2

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


TABLE 2. Continued.

Population BYNo. fishtyped

No. locinot in HWE

Sig. GD(%)

FS(%) HE AR

SALMON-Nb Nb

estimateLDNE Nb

estimateLDNE 95%

CIHM Nb

estimate

Minam River 1989 43 1 15.2 1.0 0.830 9.9 42.3 97.5 61.4–208.1 59.01995 40 1 13.6 1.3 0.826 9.0 29.9 25.5 22.1–29.6 27.51997 39 1 0.0 0.5 0.852 10.2 104.5 95.8 73.2–135.1 100.02001 49 1 0.0 0.3 0.825 10.4 ∞ ∞ ∞–∞ ∞2005 46 0 0.0 0.3 0.821 10.4 211.7 1,649.8 261.2–∞ 375.22006 47 1 3.0 0.2 0.836 10.2 123.2 328.1 178.1–1,588.5 179.1

Rapid River 1988 48 0 0.0 0.1 0.807 9.2 276.1 281.2 170.5–728.0 278.6Hatchery 1992 67 1 1.5 0.1 0.791 9.1 229.9 ∞ ∞–∞ 459.8

2000 48 0 3.0 0.0 0.803 9.1 465.1 417.1 173.4–∞ 439.82007 50 0 0.0 0.0 0.802 9.5 ∞ 559.4 235.1–∞ 1,118.8

Wenaha River 1991 47 2 25.8 1.8 0.810 9.1 47.4 31.0 26.8–36.2 37.51997 40 0 6.1 0.5 0.831 10.1 155.1 102.4 82.0–134.2 123.42003 51 1 9.1 0.6 0.830 9.8 81.2 92.8 73.9–122.2 86.62006 48 0 0.0 0.1 0.830 10.2 375.5 270.7 156.9–851.3 314.62007 48 0 4.5 0.4 0.850 10.4 73.7 107.0 80.6–154.1 87.3

aWe did not have three or more temporally spaced samples for this population, as needed by SALMONNb to estimate Nb.b SALMONNb requires that each sample consists of individuals from a single brood year.

FIGURE 3. Estimates of the effective number of breeders per year (Nb) and the total population size (N) for natural Chinook Salmon populations in the GrandeRonde and Imnaha River subbasins. No estimates of N were available for Lookingglass Creek.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


FIGURE 4. A neighbor-joining, consensus tree of Cavalli-Sforza and Edwards (1967) chord distances. Samples are labeled according to population name andthe last two digits of the brood year. Major groupings are identified by brackets, and bootstrap values greater than 50% are shown.

population groups from which they were derived. The samplesfrom Lookingglass Creek, however, suggest that a change hasoccurred in that population over the last 50 years. The earliest(presupplementation) samples from that population formed adistinct cluster together, and as a group appeared most similarto the Minam River. However, the Lookingglass Creek samplesfrom the 1990s, which were taken after Lookingglass Hatch-ery was built in 1982, formed a heterogeneous group alongwith samples from upper Grande Ronde River. This geneticcluster also included the Rapid River Hatchery samples, sup-porting the idea that introgression from that stock has occurredvia Lookingglass Hatchery. In contrast, the most recent samplesfrom Catherine Creek and Lookingglass Creek (BYs 2000–2007) were not part of that cluster, instead forming their owngroup. The Catherine Creek samples from BYs 1991–1993 donot cluster discretely, and are scattered among other samples.The two Carson National Fish Hatchery samples appeared to be

fairly distinct, most closely clustering with the BY 1991 samplefrom the upper Grande Ronde River.

DISCUSSIONIntrogression of a nonnative population into several native

populations has the potential to decrease the native populations’level of among-population diversity (Reisenbichler 2004). How-ever, we did not find evidence that the genetic population struc-ture has changed significantly for most of the populations inthis study. The overall amount of introgression from nonnativehatchery stocks appears to have been minimal for the Imnaha,Lostine, Minam, and Wenaha rivers, as temporal samples fromwithin each of those populations maintain genetic similarity toeach other over time. The Minam and Wenaha River m estimatesdo show evidence of some introgression from the Rapid RiverHatchery stock (Figure 2), but the dendrogram and pairwise FST

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


values indicate that these natural populations have maintainedtheir genetic distinctiveness from the hatchery stock. The lackof change in population structure for these stocks is notableconsidering just how great a proportion of the spawning popu-lation was composed of nonnative hatchery-origin fish in someyears. For example, Crateau (1997) reported that the percent-age of carcasses, averaged over the years 1990–1993, that wereof hatchery origin and found on the spawning grounds of theMinam and Wenaha rivers was 63.8% and 75.0%, respectively.The minimal gene flow that occurred from those hatchery-originfish into the natural populations could be because the hatcherystocks suffered from lower reproductive success in the riverthan the native stocks due to some factor that affected reproduc-tion. Previous studies comparing the reproductive success ofnatural-origin versus hatchery-origin salmonids have found re-duced fitness in hatchery-origin fish (Araki et al. 2007; Berntsonet al. 2011; Theriault et al. 2011; Hess et al. 2012), particu-larly when they were derived from a nonnative stock (Leideret al. 1990; Reisenbichler and Rubin 1999; Kostow et al. 2003).Lower reproductive success of the hatchery-origin fish could bedue to local adaptation in the native populations that would givethem a competitive advantage over the out-of-basin hatcherystocks. Local adaptation has been shown to occur frequently insalmonids (Fraser et al. 2011) and could explain why the CarsonNational Fish Hatchery and Rapid River Hatchery fish did notmore fully introgress into the Grande Ronde and Imnaha Riverbasin populations, despite there being abundant hatchery-originfish on the spawning grounds. As noted earlier, the Carson Na-tional Fish Hatchery stock and the Rapid River Hatchery stockare both derived from a mixture of stocks and thus are notnecessarily adapted to any single geographical location like thepresumed fine-scale local adaptation of native stocks. A studyby Matala et al. (2011) suggests that these hatchery stocks arebetter adapted to other Columbia River basins. These authorssuggest that introductions of the Carson National Fish Hatch-ery stock into the upper Columbia River basin and the RapidRiver Hatchery stock into the Clearwater River basin have beensuccessful due to better adaptation of these stocks to those ar-eas. In addition, other factors that have been shown to resultin the lower reproductive success of hatchery-reared salmonidsmay have been operating, such as wild males outcompetinghatchery males for mates, mate choice preference by spawningfemales for wild males, or better survival of eggs from wildfemales (Fleming and Gross 1993; Berejikian et al. 1997, 2001;Schroder et al. 2008).

Exceptions to the conclusion that the genetic populationstructure has not changed significantly for the populations inthis study are Lookingglass Creek and the upper Grande RondeRiver. It is believed that the native Lookingglass Creek stockwas extirpated soon after the construction of the LookingglassHatchery in 1982 (Boe et al. 2011) and replaced by the stocksbeing released from the hatchery. Our results cannot explicitlyprove or disprove this belief, but they do indicate that a ma-jor change in the genetic makeup of the Lookingglass Creek

population has occurred since the hatchery was established.We estimated substantial levels of introgression from the RapidRiver Hatchery stock into this population in the early 1990s(Figure 2), and the Lookingglass Creek samples from that eracluster more closely with the hatchery samples than they dowith the prehatchery Lookingglass Creek samples (Figure 4).This could be the result of introgression of the Rapid RiverHatchery stock into the native Lookingglass Creek populationor simply a complete replacement of it with the hatchery stock.It appears, though, that the current Lookingglass Creek Hatch-ery stock has quickly lost most, but not necessarily all, of thetraces of the Rapid River stock introgression/replacement it ex-perienced in the 1990s and by BY 2003 more closely resembledthe Catherine Creek stock. While the estimates of m show aninfluence of the Rapid River Hatchery stock on this popula-tion, the most recent samples we analyzed from LookingglassCreek and Hatchery and Catherine Creek cluster together on thedendrogram, most likely due to the fact that starting in 2004fish from Catherine Creek have been used as broodstock forthe Lookingglass Creek Hatchery program in the LookingglassCreek Hatchery (Monzyk et al. 2007). Presumably, the Cather-ine Creek stock represents a native Grande Ronde River basinstock, given that our most recent samples from there do not clus-ter with the Rapid River Hatchery stocks on the dendrogram andthe fact that our estimate of the amount of introgression fromthe Rapid River Hatchery stock into Catherine Creek has es-sentially decreased to zero in the most recent sample. Thus, thecurrent population in Lookingglass Creek more closely resem-bles other Grande Ronde basin populations than it did 20 yearsago. However, the true native Lookingglass Creek stock, as rep-resented by the distinctly clustering samples collected in 1964and 1972, has likely been extirpated or at least greatly changedby supplementation.

The upper Grande Ronde population shows a continual signalof introgression from the Rapid River Hatchery stock. Althoughwe do not have a sample from the upper Grande Ronde Riverbefore supplementation efforts began to compare with morerecent samples, as we do for Lookingglass Creek, the clustersthey form on the dendrogram and the estimates of m indicatethat the upper Grande Ronde River samples are not distinct fromthe Rapid River Hatchery samples. It is likely that Rapid RiverHatchery stock genes have become a part of this population.Why that only happened in this population is uncertain. It maybe that its persistently low effective population size has made itmore susceptible to introgression from hatchery-origin fish onits spawning grounds or its environmental conditions are morefavored by the Rapid River Hatchery stock (Matala et al. 2011).Habitat modifications in the basin that have reduced selectionto retain historical adaptations of the Grande River populationcould also have been a factor.

The populations in Catherine Creek and the upper GrandeRonde River are the only ones that showed any sign of intro-gression from the Carson National Fish Hatchery stock. The BY1991 samples from these populations had an estimate of m from

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


the Carson National Fish Hatchery stock that was significantlygreater than zero, and the upper Grande Ronde River populationwas the one most closely clustering with the Carson NationalFish Hatchery samples on the dendrogram. However, that is theonly brood year for which this is true, so whatever introgressionhas occurred appears to have been temporary.

Our results indicate that the supplementation programs im-plemented for Grande Ronde and Imnaha River basin ChinookSalmon populations have not caused any detectable, widespreadchanges in genetic diversity within populations or among sup-plementation history groups for the populations analyzed. Thisis similar to results found for nearby Snake River basin Chi-nook Salmon populations in the Salmon River, Idaho, by VanDoornik et al. (2011), who found little evidence of significantchanges in diversity among hatchery, supplemented, or non-supplemented populations in that system. Retention of geneticdiversity is an important goal for the supplementation programsimplemented in these locations (Hesse et al. 2006). However,within-population diversity loss and decreasing effective popu-lation sizes can be masked by migration between populations.Immigrants can introduce new alleles into a population, whichwould increase measures of genetic diversity in that population.A possible example of this can be seen in the upper GrandeRonde River samples from BY 1992, which showed a substan-tial increase in the Nb estimate for that year (Figure 3) thatcoincides with very high estimates of m from the Rapid RiverHatchery stock for that same year (Figure 2). However, thatincrease may have been temporary. Even in years when ourestimates of m are low, these populations have continued tomaintain their levels of diversity and effective population sizes.The Lookingglass Creek Nb estimates have decreased, but thismay be due to a significant change in the genetic makeup ofthis population after the Lookingglass Hatchery was built, asnoted earlier. And while these populations do appear to havepreserved their levels of genetic diversity and effective popula-tion size despite fluctuating census population sizes, they havenot achieved the desired census population sizes (HSRG 2009).Whether this is due to ineffective supplementation efforts usingmostly nonnative hatchery stocks or some other factor, such ashabitat loss or the presence of hydroelectric dams in the mi-gratory pathway of these populations, is debatable. Whateverthe reason, continued genetic monitoring of these populationswill be important to assure that the genetic diversity of thesepopulations is maintained.

ACKNOWLEDGMENTSThe authors thank the Confederated Tribes of the Umatilla

Indian Reservation, the Columbia River Inter-Tribal Fish Com-mission, and the Oregon Department of Fish and Wildlife forassistance with sample collecting and the Bonneville Power Ad-ministration for providing funding for this research. Jeff Hard,Linda Park, and David Teel provided valuable reviews of previ-ous versions of this manuscript.

REFERENCESAraki, H., B. A. Berejikian, M. J. Ford, and M. S. Blouin. 2008. Fitness of

hatchery-reared salmonids in the wild. Evolutionary Applications 1:342–355.Araki, H., B. Cooper, and M. S. Blouin. 2007. Genetic effects of captive breed-

ing cause a rapid, cumulative fitness decline in the wild. Science 318:100–103.

Belkhir, K., P. Borsa, L. Chikhi, N. Raufaste, and F. Bonhomme. 2003.GENETIX: programs for Windows for the genetic analysis of popula-tions. Laboratoire Genome, Populations, Interactions, CNRS UMR 5000,Universite de Montpellier II, Montpellier, France. Available: kimura.univ-montp2.fr/genetix. (February 2013).

Benjamini, Y., and D. Yekutieli. 2001. The control of the false discovery rate inmultiple testing under dependency. Annals of Statistics 29:1165–1188.

Berejikian, B. A., E. P. Tezak, L. Park, E. LaHood, S. L. Schroder, and E. Beall.2001. Male competition and breeding success in captively reared and wildCoho Salmon (Oncorhynchus kisutch). Canadian Journal of Fisheries andAquatic Sciences 58:804–810.

Berejikian, B. A., E. P. Tezak, S. L. Schroder, C. M. Knudsen, and J. J.Hard. 1997. Reproductive behavioral interactions between wild and cap-tively reared Coho Salmon (Oncorhynchus kisutch). ICES Journal of MarineScience 54:1040–1050.

Berntson, E. A., R. W. Carmichael, M. W. Flesher, E. J. Ward, and P. Moran.2011. Diminished reproductive success of steelhead from a hatchery supple-mentation program (Little Sheep Creek, Imnaha Basin, Oregon). Transactionsof the American Fisheries Society 140:685–698.

Boe, S. J., C. A. Crump, R. L. Weldert, and J. Wolf. 2011. LowerSnake River compensation plan, confederated tribes of the Umatilla In-dian Reservation: evaluation studies for 1 January 2008 to 31 Decem-ber 2008. U.S. Fish and Wildlife Service, Confederated Tribes of theUmatilla Indian Reservation, Project 421, La Grande, Oregon. Available:www.fws.gov/lsnakecomplan/Reports/CTUIRreports.html. (October 2012).

Carmichael, R. W., and R. T. Messmer. 1995. Status of supplementing ChinookSalmon natural production in the Imnaha River basin. Pages 284–291 inH. L. Schramm Jr. and R. G. Piper, editors. Uses and effects of culturedfishes in aquatic ecosystems. American Fisheries Society, Symposium 15,Bethesda, Maryland.

Cavalli-Sforza, L. L., and A. W. F. Edwards. 1967. Phylogenetic analysis: mod-els and estimation procedures. American Journal of Human Genetics 19:233–257.

CBFWA (Columbia Basin Fish and Wildlife Authority). 2008. 2008 status offish and wildlife resources in the Columbia River basin. CBFWA, Portland,Oregon.

Chapman, D. W. 1986. Salmon and steelhead abundance in the Columbia Riverin the nineteenth century. Transactions of the American Fisheries Society115:662–670.

Choi, J. H., H. Y. Jung, H. S. Kim, and H. G. Cho. 2000. PhyloDraw: a phylo-genetic tree drawing system. Bioinformatics 16:1056–1058.

Christie, M. R., M. L. Marine, R. A. French, R. S. Waples, and M. S. Blouin.2012. Effective size of a wild salmonid population is greatly reduced byhatchery supplementation. Heredity 109:254–260.

Crateau, E. 1997. Straying of hatchery origin spring/summer-run ChinookSalmon in the Grande Ronde basin. NOAA Technical Memorandum NMFS-NWFSC-30.

Eldridge, W. H., and K. Killebrew. 2008. Genetic diversity over multiple gen-erations of supplementation: an example from Chinook Salmon using mi-crosatellite and demographic data. Conservation Genetics 9:13–28.

Excoffier, L., G. Laval, and S. Schneider. 2005. Arlequin (version 3.0): an inte-grated software package for population genetics data analysis. EvolutionaryBioinformatics Online 1:47–50.

Faubet, P., and O. E. Gaggiotti. 2008. A new Bayesian method to identify theenvironmental factors that influence recent migration. Genetics 178:1491–1504.

Faubet, P., R. S. Waples, and O. E. Gaggiotti. 2007. Evaluating the perfor-mance of a multilocus Bayesian method for the estimation of migration rates.Molecular Ecology 16:1149–1166.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


Felsenstein, J. 2005. PHYLIP (phylogeny inference package) version 3.65.University of Washington, Seattle. Available: http://evolution.genetics.washington.edu/phylip.html. (January 2010).

Fleming, I. A., and M. R. Gross. 1993. Breeding success of hatchery and wildCoho Salmon (Oncorhynchus kisutch) in competition. Ecological Applica-tions 3:230–245.

Ford, M., P. Budy, C. Busack, D. Chapman, T. Cooney, T. Fisher, J. Geiselman,T. Hillman, J. Lukas, C. Peven, C. Toole, E. Weber, and P. Wilson. 2001.Upper Columbia River steelhead and spring Chinook Salmon populationstructure and biological requirements: final report, March 2001. NationalMarine Fisheries Service, Northwest Fisheries Science Center, Seattle.

Fraser, D. J., L. K. Weir, L. Bernatchez, M. M. Hansen, and E. B. Taylor.2011. Extent and scale of local adaptation in salmonid fishes: review andmeta-analysis. Heredity 106:404–420.

Goodman, D. 2004. Salmon supplementation: demography, evolution, and riskassessment. Pages 217–232 in M. J. Nickum, P. M. Mazik, J. G. Nickum, andD. D. MacKinlay, editors. Propagated fish in resource management. AmericanFisheries Society, Symposium 44, Bethesda, Maryland.

Guo, S. W., and E. A. Thompson. 1992. Performing the exact test of Hardy–Weinberg proportion for multiple alleles. Biometrics 48:361–372.

Hansen, M. M., D. J. Fraser, K. Meier, and K. L. D. Mensberg. 2009. Sixty yearsof anthropogenic pressure: a spatio-temporal genetic analysis of Brown Troutpopulations subject to stocking and population declines. Molecular Ecology18:2549–2562.

Hansen, M. M., K. Meier, and K. L. D. Mensberg. 2010. Identifying footprintsof selection in stocked Brown Trout populations: a spatio-temporal approach.Molecular Ecology 19:1787–1800.

Hedrick, P. W., D. Hedgecock, S. Hamelberg, and S. J. Croci. 2000. The impactof supplementation in winter-run Chinook Salmon on effective populationsize. Journal of Heredity 91:112–116.

Heggenes, J., M. Beere, P. Tamkee, and E. B. Taylor. 2006. Genetic diversityin steelhead before and after conservation hatchery operation in a coastal,boreal river. Transactions of the American Fisheries Society 135:251–267.

Hess, M. A., C. D. Rabe, J. L. Vogel, J. J. Stephenson, D. D. Nelson, and S. R.Narum. 2012. Supportive breeding boosts natural population abundance withminimal negative impacts on fitness of a wild population of Chinook Salmon.Molecular Ecology 21:5236–5250.

Hesse, J. A., J. R. Harbeck, and R. W. Carmichael. 2006. Monitoring andevaluation plan for northeast Oregon hatchery Imnaha and Grande Rondesubbasin spring Chinook Salmon. 2006 Technical Report to Bonneville PowerAdministration, Project 198805301, Portland, Oregon.

HSRG (Hatchery Scientific Review Group). 2009. Columbia River hatchery re-form system-wide report. Pacific Northwest Hatchery Reform Project, Seattle.Available: www.hatcheryreform.us. (May 2011).

Hulett, P. L., C. S. Sharpe, and C. W. Wagemann. 2004. Critical need for rig-orous evaluation of salmonid propagation programs using local wild brood-stock. Pages 253–262 in M. J. Nickum, P. M. Mazik, J. G. Nickum, andD. D. MacKinlay, editors. Propagated fish in resource management. Ameri-can Fisheries Society, Symposium 44, Bethesda, Maryland.

Kalinowski, S. T. 2005. HP-RARE 1.0: a computer program for performingrarefaction on measures of allelic richness. Molecular Ecology Notes 5:187–189.

Kalinowski, S. T., A. P. Wagner, and M. L. Taper. 2006. ML-RELATE: a computerprogram for maximum likelihood estimation of relatedness and relationship.Molecular Ecology Notes 6:576–579.

Kostow, K. E., A. R. Marshall, and S. R. Phelps. 2003. Naturally spawning hatch-ery steelhead contribute to smolt production but experience low reproductivesuccess. Transactions of the American Fisheries Society 132:780–790.

Laythe, L. L. 1950. The fishery development program in the lower ColumbiaRiver. Transactions of the American Fisheries Society 78:42–55.

Leider, S. A., P. L. Hulett, J. J. Loch, and M. W. Chilcote. 1990. Electrophoreticcomparison of the reproductive success of naturally spawning transplantedand wild steelhead trout through the returning adult stage. Aquaculture88:239–252.

Lichatowich, J. A. 1999. Salmon without rivers: a history of the Pacific salmoncrisis. Island Press, Washington, D.C.

Lowry, R. 2012. VassarStats: website for statistical computation. Vassar College,Poughkeepsie, New York. Available: VassarStats.net. (May 2012).

Marshall, S. L. 2010. A brief history of the lower Snake River compensationplan hatchery program for spring and summer Chinook Salmon. U.S. Fishand Wildlife Service, Lower Snake River Compensation Plan, Boise, Idaho.Available: www.fws.gov/lsnakecomplan. (January 2013).

Matala, A. P., J. E. Hess, and S. R. Narum. 2011. Resolving adaptive and de-mographic divergence among Chinook Salmon populations in the ColumbiaRiver basin. Transactions of the American Fisheries Society 140:783–807.

Matthews, G. M., and R. S. Waples. 1991. Status review for Snake Riverspring and summer Chinook Salmon. NOAA Technical Memorandum NMFSF/NWC-200.

McClure, M. M., F. M. Utter, C. Baldwin, R. W. Carmichael, P. F. Hassemer,P. J. Howell, P. Spruell, T. D. Cooney, H. A. Schaller, and C. E. Petrosky.2008. Evolutionary effects of alternative artificial propagation programs: im-plications for viability of endangered anadromous salmonids. EvolutionaryApplications 1:356–375.

Monzyk, F. R., T. L. Hoffnagle, R. W. Carmichael, D. L. Eddy, and P. J. Keniry.2007. Lower Snake River compensation plan: Oregon spring Chinook Salmonevaluation studies, 2004 annual progress report. Oregon Department of Fishand Wildlife, Salem.

Myers, J. M., R. G. Kope, G. J. Bryant, D. Teel, L. J. Lierheimer, T. C.Wainwright, W. S. Grant, F. W. Waknitz, K. Neely, S. T. Lindley, and R.S. Waples. 1998. Status review of Chinook Salmon from Washington, Idaho,Oregon, and California. NOAA Technical Memorandum NMFS-NWFSC-35.

Narum, S. R. 2006. Beyond Bonferroni: less conservative analyses for conser-vation genetics. Conservation Genetics 7:783–787.

NMFS (National Marine Fisheries Service). 1992. Endangered and threatenedspecies; threatened status for Snake River spring/summer Chinook Salmon,threatened status for Snake River fall Chinook Salmon. Federal Register57:78(22 April 1992):14653–14662.

NMFS (National Marine Fisheries Service). 2012. Northwest fisheries sciencecenter salmonid database: abundance. National Oceanic and Atmospheric Ad-ministration, NMFS, Northwest Fisheries Science Center, Seattle. Available:webapps.nwfsc.noaa.gov. (January 2012).

Olsen, E., P. Pierce, M. McLean, and K. Hatch. 1992. Stock summary reportsfor Columbia River anadromous salmonids. Annual Report to the BonnevillePower Administration, Document 94402-2, Project 88-108, Portland, Oregon.Available: efw.bpa.gov/searchpublications./(January 2012).

Pastor, S. M. 2004. An evaluation of fresh water recoveries of fish releasedfrom national fish hatcheries in the Columbia River basin, and observationsof straying. Pages 87–98 in M. J. Nickum, P. M. Mazik, J. G. Nickum, and D.D. MacKinlay, editors. Propagated fish in resource management. AmericanFisheries Society, Symposium 44, Bethesda, Maryland.

Peakall, R., and P. E. Smouse. 2006. GENALEX 6: genetic analysis in Excel—population genetic software for teaching and research. Molecular EcologyNotes 6:288–295.

Piggott, M. P., E. Bellemain, P. Taberlet, and A. C. Taylor. 2004. A multiplex pre-amplification method that significantly improves microsatellite amplificationand error rates for faecal DNA in limiting conditions. Conservation Genetics5:417–420.

RASP (Regional Assessment of Supplementation Project). 1992. Supplemen-tation in the Columbia basin summary report series. Report to the Bon-neville Power Administration, Project 85-62, Portland, Oregon. Available:efw.bpa.gov/searchpublications./(January 2012).

Reisenbichler, R. R. 2004. Uncertainty and research needs for supplementingwild populations of anadromous Pacific salmon. Pages 263–275 in M. J.Nickum, P. M. Mazik, J. G. Nickum, and D. D. MacKinlay, editors. Propa-gated fish in resource management. American Fisheries Society, Symposium44, Bethesda, Maryland.

Reisenbichler, R. R., and S. P. Rubin. 1999. Genetic changes from arti-ficial propagation of Pacific salmon affect the productivity and viability

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3


of supplemented populations. ICES Journal of Marine Science 56:459–466.

Rousset, F. 2008. GENEPOP’007: a complete re-implementation of the GENEPOP

software for Windows and Linux. Molecular Ecology Resources 8:103–106.Ryman, N., and L. Laikre. 1991. Effects of supportive breeding on the geneti-

cally effective population size. Conservation Biology 5:325–329.Schroder, S. L., C. M. Knudsen, T. N. Pearsons, T. W. Kassler, S. F. Young, C. A.

Busack, and D. E. Fast. 2008. Breeding success of wild and first-generationhatchery female spring Chinook Salmon spawning in an artificial stream.Transactions of the American Fisheries Society 137:1475–1489.

Schwartz, M. K., G. Luikart, and R. S. Waples. 2007. Genetic monitoring asa promising tool for conservation and management. Trends in Ecology andEvolution 22:25–33.

Small, M. P., K. Currens, T. H. Johnson, A. E. Frye, and J. F. Von Bargen.2009. Impacts of supplementation: genetic diversity in supplemented and un-supplemented populations of summer Chum Salmon (Oncorhynchus keta) inPuget Sound (Washington, USA). Canadian Journal of Fisheries and AquaticSciences 66:1216–1229.

Tessier, N., L. Bernatchez, and J. M. Wright. 1997. Population structure andimpact of supportive breeding inferred from mitochondrial and microsatelliteDNA analyses in land-locked Atlantic Salmon Salmo salar L. MolecularEcology 6:735–750.

Theriault, V., G. R. Moyer, L. S. Jackson, M. S. Blouin, and M. A. Banks. 2011.Reduced reproductive success of hatchery Coho Salmon in the wild: insightsinto most likely mechanisms. Molecular Ecology 20:1860–1869.

Van Doornik, D. M., R. S. Waples, M. C. Baird, P. Moran, and E. A. Berntson.2011. Genetic monitoring reveals genetic stability within and among threat-

ened Chinook Salmon populations in the Salmon River, Idaho. North Amer-ican Journal of Fisheries Management 31:96–105.

Waples, R. S. 1991. Genetic interactions between hatchery and wild salmonids:lessons from the Pacific Northwest. Canadian Journal of Fisheries and AquaticSciences 48(Supplement 1):124–133.

Waples, R. S. 2005. Genetic estimates of contemporary effective population size:to what time periods do the estimates apply? Molecular Ecology 14:3335–3352.

Waples, R. S., and C. Do. 2008. LDNE: a program for estimating effectivepopulation size from data on linkage disequilibrium. Molecular EcologyResources 8:753–756.

Waples, R. S., and C. Do. 2010. Linkage disequilibrium estimates of contempo-rary Ne using highly variable genetic markers: a largely untapped resource forapplied conservation and evolution. Evolutionary Applications 3:244–262.

Waples, R. S., M. Masuda, and J. Pella. 2007. SALMONNb: a program for com-puting cohort-specific effective population sizes (Nb) in Pacific salmon andother semelparous species using the temporal method. Molecular EcologyNotes 7:21–24.

Waples, R. S., and D. J. Teel. 1990. Conservation genetics of Pacific salmon: I.temporal changes in allele frequency. Conservation Biology 4:144–156.

Waples, R. S., D. J. Teel, and P. B. Aebersold. 1991. A genetic monitor-ing and evaluation program for supplemented populations of salmon andsteelhead in the Snake River basin. Annual Report of Research to the Bon-neville Power Administration, Project 89-096, Portland, Oregon. Available:efw.bpa.gov/searchpublications./ (October 2012).

Zar, J. H. 1984. Biostatistical analysis. Prentice-Hall, Englewood Cliffs, NewJersey.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:20

15

July

201

3



Evaluating Salmon Spawning Habitat Capacity UsingRedd Survey DataPhillip A. Groves a , James A. Chandler a , Brad Alcorn a , Tracy J. Richter a , William P.Connor b , Aaron P. Garcia b & Steven M. Bradbury ba Idaho Power Company , 1221 West Idaho Street, Boise , Idaho , 83702 , USAb U.S. Fish and Wildlife Service , Idaho Fishery Resource Office , Post Office Box 18,Ahsahka , Idaho , 83520 , USAPublished online: 03 Jul 2013.

To cite this article: Phillip A. Groves , James A. Chandler , Brad Alcorn , Tracy J. Richter , William P. Connor , Aaron P. Garcia& Steven M. Bradbury (2013) Evaluating Salmon Spawning Habitat Capacity Using Redd Survey Data, North American Journal ofFisheries Management, 33:4, 707-716, DOI: 10.1080/02755947.2013.793628







http://dx.doi.org/10.1080/02755947.2013.793628




MANAGEMENT BRIEF

Evaluating Salmon Spawning Habitat Capacity UsingRedd Survey Data

Phillip A. Groves,* James A. Chandler, Brad Alcorn, and Tracy J. RichterIdaho Power Company, 1221 West Idaho Street, Boise, Idaho 83702, USA

William P. Connor, Aaron P. Garcia, and Steven M. BradburyU.S. Fish and Wildlife Service, Idaho Fishery Resource Office, Post Office Box 18, Ahsahka,Idaho 83520, USA

AbstractManagement and recovery goals for fish populations often rely

on estimating the number of fish that can be supported by finitehabitats. In the absence of direct measures of carrying capacity,management decisions are commonly informed by results of habi-tat models. However, the shortcomings and spatially explicit natureof most habitat models result in making assumptions, often pre-clude inclusion of important variables, and are rarely validatedat the reach level. We analyze long-term, redd-count data for apopulation of Chinook Salmon Oncorhynchus tshawytscha to eval-uate spawning habitat capacity of a major river. Adult escapementgenerally increased during the years 1994–2012; consequently siteuse and total redd counts also increased. Together, the annual useof spawning sites and the redd counts (as functions of adult es-capement) provided evidence for density-dependent changes in theavailability and capacity of spawning sites. Redd counts exceededone recovery criterion specific to one spawning aggregate of thepopulation during 11 years of our survey data, supporting theconclusion that adequate spawning habitat is available to attain amanagement goal set for this aggregate.

In ecological systems, there are finite limits to the amountof habitat available, and those limits constrain realistic manage-ment goals. Knowledge of the carrying capacity of habitats iscritically important for establishing recovery goals and conser-vation measures for imperiled fish populations. Because directmeasures of carrying capacity may be lacking, researchers oftendevelop models to inform management decisions (e.g., see re-views by Rabeni 1992; Guisan and Zimmermann 2000). Modelscan be useful tools for determining the spawning habitat capac-ity of streams and rivers for various salmon species (e.g., Connoret al. 2001; Hatten et al. 2009).

*Corresponding author: [email protected] May 31, 2012; accepted March 27, 2013

For salmon, spawning habitat models can be data-intensive.They are usually spatially explicit and require that the physicalcharacteristics important to redd construction (e.g., depth, sub-strate, and velocity) be estimated for the reach of interest. Thisis often difficult for other factors, such as hyporheic upwellingand intergravel flow (Burner 1951; Geist and Dauble 1998; Geistet al. 2000), which are known to be important to fall ChinookSalmon Oncorhynchus tshawytscha but cannot easily be esti-mated over a broad area. Also, because the requisite spatial dataare often unavailable, estimates are commonly made for smallareas and then expanded to larger stream or river segments (e.g.,Connor et al. 2001; Hatten et al. 2009). This approach assumesthat all areas of similar habitat type have the same probability ofuse, which may or may not be accurate. Furthermore, spawninghabitat models, and estimates of potential redds that can be sup-ported, do not typically account for fish behaviors, such as pres-pawning movements, which can affect final spawning locations(Garcia et al. 2004; Connor and Garcia 2006); group spawn-ing, which can influence spawning site selection and use (Geistand Dauble 1998; Geist et al. 2000); and redd superimposition(Hayes 1987; Fukushima et al. 1998; Essington et al. 2000). Fi-nally, estimates of the number of redds that a river can supportare usually not validated for reasons that include populationsbeing at low abundance or lack of long-term redd survey data.

Connor et al. (2001) developed a spawning habitat model forlisted (i.e., threatened under the U.S. Endangered Species Act)fall Chinook salmon (NMFS 1992) that predicted that the unim-pounded Snake River downstream of Hells Canyon Dam couldsupport a recovery goal of 1,250 redds (based on a 1:1 femaleto male ratio; ICTRT 2007). However, they acknowledged that

707

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:57

15

July

201

3

708 GROVES ET AL.

they made many of the assumptions described earlier, and theywere not able to validate their estimate due to critically low pop-ulation sizes. They recommended collecting long-term annualredd survey data to determine if spawning site selection andredd construction were affected by density-dependent factors asthe population rebuilds. The Snake River fall Chinook Salmonpopulation has increased substantially since the early 1990s,and long-term redd count data are now available to validatethe spawning habitat capacity estimate of Connor et al. (2001).We reasoned that long-term redd count data would provide aconservative (assuming incomplete redd detection) measure ofabundance of semelparous females that typically construct onlyone redd prior to death (Barlaup et al. 1994; Esteve 2005).Further, the number of sites where redds are counted can beused as an estimate of spawning habitat availability. Thus, sim-ilar to populations that exhibit logistic growth (e.g., Hutchinson1957 and Caughley 1970), the number of spawning sites used(hereafter, sites) and redd counts might also exhibit logisticgrowth, with initial exponential increases followed by stabiliza-tion at the capacity of available habitat. Our objectives were to(1) describe the annual trends in total numbers of sites used andredds counted from 1994 to 2012, (2) evaluate the functionalgrowth relations of sites used and redds counted, in relation toan estimate of adult abundance, as indicators of spawning habi-tat capacity, and (3) evaluate the attainability of a managementgoal that is linked to redd capacity.

METHODSStudy population, location, and background.—The redd sur-

veys were conducted in the free-flowing lower Snake River,which supports one spawning aggregate of the threatened SnakeRiver basin fall Chinook Salmon evolutionarily significant unit(ESU; Waples 1991). While Snake River fall Chinook spawnin several tributaries (including the Clearwater, Grande Ronde,Imnaha, and Salmon rivers), our analyses for this paper focusonly on the portion of the population that uses the lower SnakeRiver. The free-flowing Snake River extends for 161 river kilo-meters (rkm) downstream from Hells Canyon Dam, to a reser-voir formed by Lower Granite Dam (Figure 1). The southernportion of the river forms the border between Idaho and Ore-gon, while the northern portion forms the border between Idahoand Washington. Lower Granite Dam is located on the lowerSnake River in Washington State, 173 rkm upriver from theconfluence with the Columbia River, and 748 rkm from the Pa-cific Ocean. The Hells Canyon Dam, at the southern extreme ofour study reach, forms an upstream migration barrier. One of thepopulation-level criteria for removing this ESU from the fed-eral list of threatened species requires that no fewer than 2,500natural-origin adults spawn within the lower Snake River (IC-TRT 2007). If only natural-origin fish were present upstream ofLower Granite Dam, this would equate to 1,000 redds, assuminga gender ratio of 1 female per 1.5 males (Milks et al. 2005), andeach female constructs one redd. We refer to the 1,000 redds

as the partial recovery goal. This is slightly different from therecovery goal discussed by Connor et al. (2001), who used anarbitrary female to male ratio of 1:1.

Data collection.—We obtained the counts of adult fallChinook Salmon (excluding jacks) that arrived at Lower Gran-ite Dam for 1994 through 2012 from the Columbia RiverData Access in Real Time website (DART at http://www.cbr.washington.edu/dart). To estimate total escapement upstreamof Lower Granite Dam, we subtracted the numbers of adultsthat were removed at the dam for hatchery broodstock, en-tered a hatchery located upstream of the dam, or were har-vested upstream of Lower Granite Dam (data supplied byWashington Department of Fish and Wildlife, Nez Perce TribeDepartment of Fisheries Resources, and Idaho Department ofFish and Game). Only a portion of the escapement past LowerGranite Dam is destined to spawn in the lower Snake River;many fish will eventually spawn within major tributaries suchas the lower Clearwater, Grande Ronde, and Imnaha rivers (Fig-ure 1). However, throughout our period of record the proportionof redds counted only within the lower Snake River, in rela-tion to the total count of redds within all major spawning areasupstream of Lower Granite Reservoir, has remained relativelystable at approximately 60% (Connor et al. 2011); thus, wereasoned that total escapement provides a consistent index ofspawner abundance in the lower Snake River.

Standard methods were used to count redds within thelower Snake River from 1994 to 2012 (Connor et al. 2011).Aerial surveys were scheduled at 7-d intervals throughout thefall Chinook Salmon spawning season (late October to earlyDecember), except after 2009 when flights were scheduled at14-d intervals to reduce flight risk. This change in procedurewas precipitated after a fatal helicopter crash occurred during aredd survey on the Clearwater River in the early fall of 2010. Itwas hoped this change would reduce the possibility of accidentsin the future. It was also recognized that a 14-d period betweensurveys would not diminish the accuracy of final redd countsbecause redds are generally visible for longer than 14 d. Theactual number of flights made each season was dependent onlogistics and weather (Table 1). Redds were counted from ahelicopter flown about 200 m above the river, which allowedobserving 100% of the river bottom at depths approximately<3 m. Each survey flight lasted about 3 h and covered theentire 161 km of the free-flowing portion of the lower SnakeRiver where fall Chinook Salmon spawn (i.e., between Asotin,Washington, [rkm 237] upstream to the Hells Canyon Dam [rkm398]; Figure 1). Redds were counted by the same experiencedobserver during all years with a secondary observer assisting bymaintaining data records and validating counts in areas wherequestionable redds occurred. Because only fall Chinook Salmonspawn in the main-stem Snake River during the fall, there wasno doubt as to the origin of redds. During each flight shallowredd locations (e.g., those approximately <3 m deep) werereferenced to a fifth of a river kilometer and general location(left or right bank, or mid-channel) on navigation charts, or with

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:57

15

July

201

3

MANAGEMENT BRIEF 709

FIGURE 1. Study area map illustrating the section of the Snake River that was monitored from 1994 to 2012, where one aggregate of the Snake River fallChinook salmon population spawns. [Figure available in color online.]

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:57

15

July

201

3

710 GROVES ET AL.

TABLE 1. The estimated adult escapement of fall Chinook Salmon from aerial surveys of shallow sites (<3 m deep) and from deep water (>3 m) searches inthe lower Snake River, 1994–2012.

Estimated Aerial Shallow Shallow Deep Deep DeepYear escapement surveys sites redds searches sites redds

1994 605 8 20 51 73 4 161995 636 7 23 41 42 3 301996 985 7 22 71 32 4 421997 1,004 8 15 49 63 2 91998 960 8 33 135 48 5 501999 1,931 9 54 273 73 17 1002000 2,316 9 50 255 60 8 912001 6,621 10 68 535 67 19 1752002 10,423 7 87 878 60 25 2352003 11,143 7 104 1130 40 25 3942004 12,470 8 103 1,218 67 41 4912005 9,666 9 99 1,042 41 27 4002006 6,892 6 80 696 56 34 3292007 8,930 8 73 714 63 39 4032008 12,740 8 104 1,233 71 48 5862009 11,584 8 126 1,511 64 36 5842010 37,202 4 124 1,950 87 57 9942011 21,184 4 108 1,949 67 46 8652012 29,103 4 107 1,412 64 33 416

GPS and mapping software. Specific spawning locations werecatalogued, maintained, and updated in a spreadsheet and withina geographic information system. The number of new reddsobserved at each location, during each flight was recorded.Questionable redds were often authenticated as the seasonprogressed by ground-checks. Evaluations on the ground werealso performed inconsistently at specific locations to obtain amore accurate total redd count when superimposition (i.e., oneredd overlapped with another) was noted during aerial surveys.

Each year, potential deep-water spawning locations (>3 m)were also searched for redds. Deep-water searches began justafter peak spawning in mid-November and were concluded bymid-December. Priority was given to 89 locations identified byGroves and Chandler (1999). At the onset of each spawningseason, the list of locations was divided and assigned to twocrews. Priority for searches was given to locations known to beused by spawning Chinook Salmon, but the crews attempted tosearch as many locations as possible that had the potential foruse (Table 1). Deep redds were counted from a boat by pass-ing a remote submersible camera over the river bottom alongtransects spaced 4–10 m apart at a given site, following the meth-ods described by Groves and Garcia (1998). Distance betweentransects was visually estimated (1994), determined by distancemarkers (1995–1998), or maintained by real-time GPS linkedto Geographic Information System (GPS/GIS) overlays (1999–2012). Staff attempted to view the entire river bottom at eachlocation by completely covering the area along and between

transects, and by continually adjusting camera depth and angleto maximize viewable area. Redds were counted and mappedwith shore-based survey equipment or boat-based GPS/GIS toprevent double counting of redds observed during aerial surveys.During deep water searches, if superimposition was encoun-tered, the river bottom was video-recorded, and redd boundarycoordinates were obtained with survey equipment or GPS/GIS,when possible. After reviewing the recorded video, while refer-encing plots of redd boundary coordinates, the number of reddswas estimated by dividing the total area (m2) of disturbed bot-tom substrates by 45.8 m2; this conservative estimate was basedon the range of 2.1–44.8 m2 reported by Chapman et al. (1986),allowing a small buffer for redd spacing. These deeper searchesdid not provide a 100% census of spawning in deep water be-cause most locations could only be visited once during eachspawning season, and we did not search every potential locationthat could have been used by spawners. Thus, we analyzed thedata collected during aerial surveys and deep-water searchesseparately to accomplish our objectives.

Data analysis.—In all of the analyses, we assumed thatchanges in counting accuracy were not the primary source of ob-served trends, that males outnumbered females by 1.5:1 (Milkset al. 2005), and that each female constructed only one redd. Wealso assumed that spawning habitat availability remained rela-tively static throughout our period of study. We expected thatgravels can and do move (being dynamic), but that the import ofgravels from tributaries, and the export of gravels downstream

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:57

15

July

201

3


out of the study area were similar in magnitude, resulting in nonet loss or gain of available spawning gravel. As far as we areaware, there is no evidence to support either a net gain or loss ofspawning gravel throughout the lower Snake River. We definedspawning sites in both shallow and deep water as distinct, con-tiguous patches of gravel where redds were encountered (seeConnor et al. 2001 for an example). We plotted the redd countsand the number of sites used against year to describe the annualtrends during 1994–2012.

The growth function fitted to the data was selected basedon examinations of scatter plots. It seemed appropriate to fit aBeverton–Holt stock–recruit function with 95% prediction in-tervals to the data (Ricker 1975). Each relationship used theestimated annual adult escapement as the independent variable.In doing this, we assumed that each group of data would be-have similar to a classic stock–recruit function, resulting in thefollowing equation:

R = 1/(a + β/P) (1)

in which R is the recruits (i.e., number of sites used, or numberof redds constructed), and P is the “stock” (i.e., number of adultsin the escapement). The Beverton–Holt function also providesan estimate of maximum recruits, 1/a, which we calculated foreach group of data, assuming that this would provide a goodestimate of capacity of both the total number of sites availableand number of redds that could be constructed. Analyses ofthese relationships tested for a density dependent change in siteavailability as well as the redd capacity throughout the lowerSnake River (in both shallow and deep water) as adult abundanceincreased.

RESULTSThe total estimated escapement of adult fall Chinook Salmon

increased substantially during our period of study, from a lowof 605 in 1994 to just over 37,000 in 2010, and it has remainedstrong through 2012 (Table 1; Figure 2). The use of both shal-low and deep sites for spawning within the lower Snake Rivergenerally increased over time (Figure 3A). The total number ofshallow sites used ranged from a low of 15 in 1997 to a high of126 in 2009, and the total number of deep sites used ranged froma low of 2 in 1997 to a high of 57 in 2010 (Table 1). Similarly,both shallow and deep redd counts generally increased since1994 (Figure 3B). The total shallow redd count ranged from alow of 41 in 1995 to a high of 1,950 in 2010, and the total deepredd count ranged from a low of 9 in 1997 to a high of 994 in2010 (Table 1). From 1994 to 2012, we documented spawningat 224 distinct sites. However, while many of the sites were usedconsistently throughout our period of study, redds were nevercounted at all 224 sites in a single year.

For both shallow and deep sites, we observed a rapid re-cruitment of sites being used as the estimated adult escapementinitially increased, but that rate leveled off as the annual escape-ment increased to more than 15,000 adults (Figure 4A, B). Inboth cases, the annual number of sites used showed a negativeexponential growth function in relation to annual escapement.The model of shallow sites available for use, based on adult es-capement, predicts that the maximum number of sites availableto be used should be 132 (confidence limit, ±29). The modelfor deep water sites predicts that a maximum of 62 (±26) sitesshould be available for use.

The annual number of redds counted was also a negativeexponential growth function of the number of adults present in

FIGURE 2. Annual trend in the estimated escapement of adult fall Chinook Salmon that passed Lower Granite Dam to spawn in the lower Snake River basin,1994–2012.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:57

15

July

201

3

712 GROVES ET AL.

FIGURE 3. Annual trends in (A) the number of sites used in shallow (depth, <3 m) and deep (depth, >3 m) water, and (B) in the number of shallow and deepredds counted, based on aerial surveys and deep-water searches conducted to count fall Chinook Salmon redds in the lower Snake River, 1994–2012.

the population; this was true for both shallow and deep redds(Figure 5A, B). Our model predicts that the maximum shallow-water redd capacity should be 2,976 (confidence limit, ±1,077).For deep-water habitat, our model predicts that the redd capacityis 1,466 (±976).

The combined models of shallow and deep redd capacitiespredict that the lower Snake River should be able to supportapproximately 4,442 redds. The actual survey counts have ex-ceeded 1,000 redds during 11 years, and 2,000 redds during3 years of our surveys, one of those years coming very close to

3,000 redds. Both the predicted capacity and the empirical reddcounts indicate that the partial recovery goal of 1,000 redds canbe achieved for this ESA-listed population (Figure 6).

DISCUSSIONWe did not observe use of all 224 documented sites during any

single year. This was true even in years with the highest numberof returning adults and corresponding high counts of redds. Inmany cases, models of habitat availability will not reflect actual

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:57

15

July

201

3


FIGURE 4. The functional growth relation between the annual number ofspawning sites used and the annual adult escapement of fall Chinook Salmon in(A) shallow water (<3 m) and (B) deep water (>3 m) within the lower SnakeRiver, 1994–2012. The predicted line ± 95% prediction intervals are shown.[Figure available in color online.]

use or preference (defined for our purposes as those sites thatwere both suitable, based on annual environmental conditions,and used, given nonrandom fish behavior) and the predictionsof available habitat; in addition, the expectations of managerswill be more than what the fish actually use. We believe thatthere are two reasons that the total number of documented sitespresent throughout a river or stream is an unreliable indicatorof habitat availability for most systems. First, the suitability ofa site for spawning is not fixed; it is dependent on depth, veloc-ity, and geomorphology (e.g., Orcutt et al. 1968; Shirvell andDungey 1983; Chapman et al. 1986; Geist and Dauble 1998;Groves and Chandler 1999; Baxter and Hauer 2000; Hanrahan2007). Depth and velocity are dependent on flow, which varieswithin and across years to some extent in all rivers. For exam-ple, even though the flows in the lower Snake River are heldstable throughout the spawning period by regulating releasesfrom Hells Canyon dam, those stable flows have ranged from alow of 241 m3/s to a high of 354 m3/s (Idaho Power Company,unpublished data). These differences in seasonal flows will not

FIGURE 5. The functional growth relation between the annual redd countsand the annual adult escapement of fall Chinook Salmon in (A) shallow (<3 m)and (B) deep (>3 m) water within the lower Snake River, 1994–2012. Thepredicted line and ± 95% prediction intervals are shown. [Figure available incolor online.]

necessarily alter substrate but will ultimately affect depth andwater velocity. While classic instream flow habitat models canreliably predict depth and water velocity (Bovee and Milhous1978), there are other factors they cannot predict. Geomorphol-ogy can be affected by extremely high annual flows (observedduring three years of our study), which can possibly shift gravels,alter horizontal hyporheic water movement, and affect the ver-tical interchange of water into and out of the shallow hyporheiczone. Thus, it is unlikely that every potentially available site ina river will be suitable for spawning in any single year, or evenover a series of years.

The second and perhaps most compelling reason that the to-tal quantity of known sites present throughout a river or streamis an unreliable indicator of habitat availability is that there arenonrandom social aspects of spawning. Chinook Salmon alterthe stream substrate when spawning (Field-Dodgson 1987) andoften construct clusters of redds in proximity to each other,even at low adult abundances, leaving apparent suitable habitatunused nearby. Geist et al. (2000) attributed such nonrandom

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:57

15

July

201

3

714 GROVES ET AL.

FIGURE 6. The total number of fall Chinook Salmon redds counted during aerial surveys and deep water searches (combined) in the lower Snake River during1994–2012, compared with the partial population level criterion for removing the ESU from the federal list of threatened species at hatchery (i.e., 1,000 redds)and the predicted total capacity of the lower Snake River.

distribution of spawning by Chinook Salmon to the influence ofunique habitat formed by the redd clusters. Essington et al.(1998) attributed redd clustering during spawning of BrookTrout Salvelinus fontinalis and Brown Trout Salmo trutta toan attraction to disturbed gravels from earlier spawning. Thus,initial use of any given site for spawning might ultimately affectthe final level of its saturation, while equally suitable sites areused to a lesser degree, or not at all.

Identifying the number of core, preferred sites used over time,across a wide range of returning adults and during periods ofenvironmental conditions that are normal for a given area mighthelp to provide a more useful indicator of spawning habitatavailability than predicting site availability via classic instreamflow models. The functional growth relations that we modeled,between the sites used per adult (Figure 4) provide evidencethat further increases in adult escapement will not result in asignificant increase of sites used. This implies that core siteshave likely been identified, and we should not expect to seemany more new sites used in the future. We hypothesize thatthere are limited core sites and that they are a subset of allavailable sites whose number varies based on environmentalstochasticity and nonrandom fish behavior. However, there willalways be the potential for slight increases in the total numberof sites used after the subset of preferred sites is attained and asterritoriality and opportunism promote spawning in less suitablehabitat. This pattern and predictions made for the number of sitesused would be expected if less competitive females were forcedto use less suitable habitat due to overlapping spawn timingand territoriality (e.g., Healey 1991; McPhee and Quinn 1998;Essington et al. 2000; Blanchfield and Ridgway 1997).

Given that we have probably identified the total number ofavailable spawning sites in the Snake River, the question be-comes how many redds can those sites support? The functionalgrowth relations of redds counted per adult is beginning to leveloff, but it does not appear to have attained clear maximums (Fig-ure 5). This suggests that although the number of sites availablehas been identified, the space available within those sites has notreached capacity. One reason for the apparent leveling off of theredd counts, especially for shallow water, is that we could haveviolated one of our critical assumptions and the redd count ac-curacy has declined as the number of adults and redds increased.We attempted to minimize bias in redd counts, associated withvariation in observer ability (e.g., Dunham et al. 2001) by hav-ing the same experienced observer count redds during all years.During periods of low adult escapement, individual redds wereclearly separated, and we are confident that this assumptionwas not violated. However, as the adult escapement and theirresulting redds increased, we began to note an increase in su-perimposition of redds, and we are no longer confident that ourredd counts are error free. Superimposition occurs when oneor more redds are constructed on top of previously constructedredds and can make it difficult to obtain an accurate redd countas multiple redds are erroneously identified as single redds. Su-perimposition usually occurs after initial redd clustering takesplace and has been observed in many salmon populations, whereearlier spawned females no longer defend their spawning terri-tories (Essington et al. 2000). During our period of study we didnot have the resources to consistently evaluate redd superimpo-sition. However, this phenomenon recently began to be noticedand has become more prevalent as the adult escapement has

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:57

15

July

201

3


increased. Redd superimposition has not been observed at allsites, but it is becoming common at sites that have been moreconsistently used throughout the period of our surveys, thosethat we refer to as core sites. For example, during the earlyyears of our surveys superimposition was rarely noted at anysite, while during 2012 some amount of superimposition wasobserved at 39% of all sites used, shallow and deep water com-bined. We hypothesize that as the adult abundance continues toincrease and because the fish tend to gregariously congregatefor spawning, redd superimposition will become more commonand pronounced, resulting in a stabilization in redd counts athigher levels of adult escapement (e.g., Hayes 1987); we definethis as a functional redd capacity. Further, we know that thespawning sites differ in size (e.g., Connor et al. 2001) and thatfocusing on the aggregate of sites ignored the possibility thatsome individual sites might have actually attained redd capacityat some time during 1994–2012.

Our results build upon the redd capacity estimates of Connoret al. (2001), and continue to support their conclusions. Usinga spatially explicit modeling approach, they predicted the lowerSnake River could support about 2,500 (CL, ±1,430) redds, de-pending on flow level, while our long-term surveys and resultantmodels predict that functional redd capacity is roughly 4,442(±2053). We use the term “functional redd capacity” becausethe effect of increased superimposition at higher adult numberswill certainly increase total redd numbers, limiting our abilityto make accurate counts, and we do not fully understand how itmight affect ultimate production. We recognize that the estimatemade by Connor et al. (2001) only applies to 106 spawning sitesknown at the time of their study. Their redd capacity estimatewould be expected to increase if applied to the larger numberof spawning sites we know to exist presently. It is important tostress that new sites have not been created throughout the lowerSnake River since Connor et al. (2001) conducted their work butthat the total escapement of adults has increased substantially,and the fish presently use a larger amount of available habitatthan previously documented. Analyzing a long-term data set ofredd counts enabled us to identify the trends in the number ofsites used as the number of adults increased, which reflected fishbehaviors that habitat models could not incorporate. Also, thislong-term data set revealed the dependency of the redd capacityof sites ultimately being controlled by redd superimposition. Fi-nally, this long-term data set indicates that the partial recoverygoal of 1,000 redds can be supported, and has actually beensurpassed, but it must be watched and evaluated carefully withrespect to the ultimate production of juveniles and hatchery pro-duced fish included in the spawning population. Two compellingquestions that arise from our knowledge of spawning habitat ca-pacity of the lower Snake River that require further study andanalysis are (1) is the productive potential of the overall capacityof redds enough to support a viable, self-sustaining populationof fall Chinook Salmon (especially in light of increased su-perimposition), and (2) can an overabundance of hatchery fish,used as supplementation for this population, adversely result in

competition and ultimate reduction in fitness of the natural pop-ulation? We maintain that continued collection of redd-countdata will aid in monitoring and managing the recovery of theSnake River fall Chinook Salmon population.

ACKNOWLEDGMENTSFor almost two decades, Snake River basin fall Chinook

Salmon research has been a highly cooperative venture involv-ing an interagency and tribal team of scientists and managers.We thank our team members, especially those who have ac-cumulated flight time. Bill Arnsberg, Jay Hesse, Yetta Jaeger,and several other anonymous reviewers improved this paper.Costs were covered by the Idaho Power Company and by theBonneville Power Administration through projects 199102900and 199801003 administered by D. Docherty. The use of tradenames does not imply endorsement by the U.S. Government.The findings and conclusions in this article are those of the au-thor(s) and do not necessarily represent the views of the U.S.Fish and Wildlife Service.

REFERENCESBarlaup, B. T., H. Lura, H. Saegrov, and R. C. Sundt. 1994. Inter- and intra-

specific variability in female salmonid spawning behaviour. Canadian Journalof Zoology 72:636–642.

Baxter, C. V., and F. R. Hauer. 2000. Geomorphology, hyporheic exchange,and selection of spawning habitat by Bull Trout (Salvelinus confluentus).Canadian Journal of Fisheries and Aquatic Sciences 57:1470–1481.

Blanchfield, P. J., and M. S. Ridgway. 1997. Reproductive timing and use ofredd sites by lake-spawning Brook Trout (Salvelinus fontinalis). CanadianJournal of Fisheries and Aquatic Sciences 54:747–756.

Bovee, K. D., and R. Milhous. 1978. Hydraulic simulation in instream flowstudies: theory and techniques. U.S. Fish and Wildlife Service, InstreamFlow Information Paper 5, Fort Collins, Colorado.

Burner, C. J. 1951. Characteristics of spawning nests of Columbia River salmon.U.S. Fish and Wildlife Service Fishery Bulletin 52:97–110.

Caughley, G. 1970. Eruption of ungulate populations, with emphasis onHimalayan thar in New Zealand. Ecology 51:53–72.

Chapman, D. W., D. E. Weitkamp, T. L. Welsh, M. B. Dell, and T. H. Schadt.1986. Effects of river flow on the distribution of Chinook Salmon redds.Transactions of the American Fisheries Society 115:537–547.

Connor, W. P., and A. P. Garcia. 2006. Prespawning movement of wild andhatchery fall Chinook Salmon adults in the Snake River. Transactions of theAmerican Fisheries Society 135:297–305.

Connor, W. P., A. P. Garcia, S. Bradbury, B. D. Arnsberg, and P. A. Groves.2011. Fall Chinook Salmon spawning ground surveys in the Snake Riverbasin upriver of lower Granite Dam, 2010. Report to the Bonneville PowerAdministration, Portland, Oregon.

Connor, W. P., A. P. Garcia, A. H. Connor, E. O. Garton, P. A. Groves, andJ. A. Chandler. 2001. Estimating the carrying capacity of the Snake River forfall Chinook Salmon redds. Northwest Science 75:363–371.

Dunham, J., B. Rieman, and K. Davis. 2001. Sources and magnitude of samplingerror in redd counts for Bull Trout. North American Journal of FisheriesManagement 21:343–352.

Essington, T. E., T. P. Quinn, and V. E. Ewert. 2000. Intra- and inter-specificcompetition and the reproductive success of sympatric Pacific salmon. Cana-dian Journal of Fisheries and Aquatic Sciences 57:205–213.

Essington, T. E., P. W. Sorensen, and D. G. Paron. 1998. High rate of redd super-imposition by Brook Trout (Salvelinus fontinalis) and Brown Trout (Salmo

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:57

15

July

201

3

716 GROVES ET AL.

trutta) in a Minnesota stream cannot be explained by habitat availabilityalone. Canadian Journal of Fisheries and Aquatic Sciences 55:2310–2316.

Esteve, M. 2005. Observations of spawning behaviour in Salmoninae: Salmo,Oncorhynchus and Salvelinus. Reviews in Fish Biology and Fisheries 15:1–21.

Field-Dodgson, M. S. 1987. The effect of salmon redd excavation on stream sub-strate and benthic community of two salmon spawning streams in Canterbury,New Zealand. Hydrobiologia 154:3–11.

Fukushima, M., T. J. Quinn, and W. W. Smoker. 1998. Estimation of eggs lostfrom superimposed Pink Salmon (Oncorhynchus gorbuscha) redds. CanadianJournal of Fisheries and Aquatic Sciences 55:618–625.

Garcia, A. P., W. P. Connor, D. J. Milks, S. J. Rocklage, and R. K. Steinhorst.2004. Movement and spawner distribution of hatchery fall Chinook Salmonadults acclimated and released as yearlings at three locations in the SnakeRiver basin. North American Journal of Fisheries Management 24:1134–1144.

Geist, D. R., and D. D. Dauble. 1998. Redd site selection and spawning habitatuse by fall Chinook Salmon: the importance of geomorphic features in largerivers. Environmental Management 22:655–669.

Geist, D. R., J. Jones, C. J. Murray, and D. D. Dauble. 2000. Suitability crite-ria analyzed at the spatial scale of redd clusters improved estimates of fallChinook Salmon (Oncorhynchus tshawytscha) spawning habitat use in theHanford Reach, Columbia River. Canadian Journal of Fisheries and AquaticSciences 57:1636–1646.

Groves, P. A., and J. A. Chandler. 1999. Spawning habitat used by fall ChinookSalmon in the Snake River. North American Journal of Fisheries Management19:912–922.

Groves, P. A., and A. P. Garcia. 1998. Two carriers used to suspend an underwatervideo camera from a boat. North American Journal of Fisheries Management18:1004–1007.

Guisan, A., and N. E. Zimmermann. 2000. Predictive habitat distribution modelsin ecology. Ecological Modelling 135:147–186.

Hanrahan, T. P. 2007. Bedform morphology of salmon spawning areas in a largegravel-bed river. Geomorphology 86:529–536.

Hatten, J. R., K. F. Tiffan, D. R. Anglin, S. L. Haeseker, J. J. Skalicky, and H.Schaller. 2009. A spatial model to assess the effects of hydropower operationson Columbia River fall Chinook Salmon spawning habitat. North AmericanJournal of Fisheries Management 29:1379–1405.

Hayes, J. W. 1987. Competition for spawning space between Brown(Salmo trutta) and Rainbow trout (S. gairdneri) in a lake inlet tributary,New Zealand. Canadian Journal of Fisheries and Aquatic Sciences 44:40–47.

Healey, M. C. 1991. Life history of Chinook Salmon (Oncorhynchustshawytscha). Pages 311–393 in C. Groot and L. Margolis, editors. Pacificsalmon life histories. University of British Columbia Press, Vancouver.

Hutchinson, G. E. 1957. An introduction to population ecology. Yale UniversityPress, New Haven, Connecticut.

ICTRT (Interior Columbia Technical Recovery Team). 2007. Appendix B:population size and complexity—interior Columbia Chinook and steel-head ESUs. Available: www.nwfsc.noaa.gov/trt/trt documents/appendixb viability 3 15 2007.pdf. (December 2011).

McPhee, M. V., and T. P. Quinn. 1998. Factors affecting the duration of nestdefense and reproductive lifespan of female Sockeye Salmon, Oncorhynchusnerka. Environmental Biology of Fishes 51:369–375.

Milks, D., M. Varney, M. Schuck, and N. J. Sands. 2005. Lyons Ferry hatcheryevaluation fall Chinook Salmon annual report: 2001 and 2002. Report to theU.S. Fish and Wildlife Service, Washington Department of Fish and Wildlife,Report FPA 05-02, Olympia.

NMFS (National Marine Fisheries Service). 1992. Endangered and threat-ened species: threatened status for Snake River spring/summer ChinookSalmon, threatened status for fall Chinook Salmon, final rule. Federal Register57:78(22 April 1992):14653–14663.

Orcutt, D. R., B. R. Pulliam, and A. Arp. 1968. Characteristics of steelheadtrout redds in Idaho streams. Transactions of the American Fisheries Society97:42–45.

Rabeni, C. F. 1992. Habitat evaluation in a watershed context. Pages 57–67in R. H. Stroud, editor. Fisheries management and watershed development.American Fisheries Society, Symposium 13, Bethesda, Maryland.

Ricker, W. E. 1975. Computation and interpretation of biological statistics offish populations. Fisheries Research Board of Canada Bulletin 191.

Shirvell, C. S., and R. G. Dungey. 1983. Microhabitats chosen by Brown Troutfor feeding and spawning in rivers. Transactions of the American FisheriesSociety 112:355–367.

Waples, R. S. 1991. Pacific salmon, Oncorhynchus spp., and the definitionof “species” under the Endangered Species Act. Marine Fisheries Review53:11–22.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:57

15

July

201

3



Post-Release Survival and Behavior of Adult Shoal Bassin the Flint River, GeorgiaTravis R. Ingram a , Josh E. Tannehill a & Shawn P. Young ba Georgia Department of Natural Resources , 2024 Newton Road, Albany , Georgia , 31701 ,USAb Department of Forestry and Natural Resources , Clemson University , 311 NaturalResources Drive, Clemson , South Carolina , 29634-0317 , USAPublished online: 03 Jul 2013.

To cite this article: Travis R. Ingram , Josh E. Tannehill & Shawn P. Young (2013) Post-Release Survival and Behaviorof Adult Shoal Bass in the Flint River, Georgia, North American Journal of Fisheries Management, 33:4, 717-722, DOI:10.1080/02755947.2013.806378







http://dx.doi.org/10.1080/02755947.2013.806378




MANAGEMENT BRIEF

Post-Release Survival and Behavior of Adult Shoal Bassin the Flint River, Georgia

Travis R. Ingram* and Josh E. TannehillGeorgia Department of Natural Resources, 2024 Newton Road, Albany, Georgia 31701, USA

Shawn P. YoungDepartment of Forestry and Natural Resources, Clemson University, 311 Natural Resources Drive,Clemson, South Carolina 29634-0317, USA

AbstractWe conducted a telemetry study of Shoal Bass Micropterus

cataractae in the lower Flint River, Georgia, during 2010. Our ob-jectives were to (1) characterize Shoal Bass migration from commu-nal spawning habitats and (2) evaluate the effects of translocationon survival of adult Shoal Bass and determine its value as a viablerestoration and management tool. Twenty-seven adult Shoal Bass(TL ≥305 mm) were fitted with hydroacoustic tags. Thirteen fish(control group) were released at the site of capture, and the re-maining 14 individuals were transported 75 river kilometers (rkm)downstream prior to release (translocated group). Telemetry datasuggested that spawning congregations are composed of fish fromlocal (<3 rkm) and distant (>3 rkm) home ranges. Initially, within14 d, translocated fish remained near their release site significantlylonger than control fish remained at the original capture site. After90 d at large, the distance that control Shoal Bass dispersed fromthe release site was similar to distances dispersed by the translo-cated group, 1.6–23.0 rkm and 2.6–28.6 rkm, respectively. After 90d, translocated Shoal Bass had not returned to the same river reachthat was occupied by control fish. No sex-specific differences werefound between groups. Postrelease survival after 90 d was 92% forboth translocated and control groups. Telemetry results indicatedthat Shoal Bass in the Flint River undergo substantial migrationsto spawning habitats. Stockpiling of Shoal Bass from live-releasetournaments may occur in the short term, and relocation awayfrom home ranges may occur in the long term, thus affecting adultShoal Bass distribution. The high survival rates and the eventualdispersal of most translocated fish observed in this study suggestShoal Bass can cope with translocation.

Shoal Bass Micropterus cataractae are endemic to theApalachicola–Chattahoochee–Flint River basin (ACF) and haveonly been recently described as a distinct species of black bassMicropterus spp. (Williams and Burgess 1999). The Flint River,

*Corresponding author: [email protected] February 24, 2012; accepted May 15, 2013

Georgia, supports the largest remaining Shoal Bass popula-tion in their native range. Shoal Bass generally prefer riverineshoal habitats and pools consisting of coarse substrate (cobble,boulder, and bedrock; Wheeler and Allen 2003; Johnston andKennon 2007; Stormer and Maceina 2009). This preferred habi-tat is typically a small proportion of the total available habitat;thus, Shoal Bass have been classified as a habitat specialist(Wheeler and Allen 2003; Johnston and Kennon 2007; Stormerand Maceina 2009). Habitat partitioning has been suggested asthe mechanism by which Shoal Bass have successfully evolvedin sympatry with other black basses in the ACF basin (Wheelerand Allen 2003).

Previous research studies were conducted on populationsin tributaries to the Chattahoochee River within Alabama andon the population within the Chipola River, Florida; however,study of the ecology and population dynamics of the Flint Riverpopulation have only been initiated within the last few years(Georgia Department of Natural Resources, unpublished data).These research efforts will be important as the different popu-lations may require unique management, and these populationsare managed under separate jurisdictions. For example, declin-ing abundance was identified within the Chattahoochee Riversubbasin (Williams and Burgess 1999; Stormer and Maceina2008, 2009), whereas a popular sport fishery has developedin the Flint River, prompting different management strategies.Overall decline has been attributed to dam construction and landuse changes in the ACF basin affecting habitat availability, butto a different extent among the several subbasins.

There is evidence that suggests Shoal Bass may ex-hibit spawning site fidelity and migrate from distant (>3 riverkilometer [rkm]) home ranges to congregate at a limited number

717

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3

718 INGRAM ET AL.

of spawning shoals (Georgia Department of Natural Resources,unpublished data). Spawning congregations of Shoal Bass in alimited number of locations may result in overexploitation. Inaddition, angler catches of spawning Shoal Bass during live-release tournaments may require transport of fish to weigh-insites, potentially reducing survival and spatial distribution ofthe population. Translocation of Shoal Bass may be needed tooffset the deleterious effects of habitat destruction, over harvest,and to ensure the genetic integrity of isolated populations.

Given the potential vulnerability of excess harvest andhabitat destruction at specific communal spawning sites andthe variable abundance of Shoal Bass populations within theACF basin, this study addressed objectives necessary for thelong-term management of Shoal Bass. Our objectives wereto (1) characterize Shoal Bass migration from communalspawning habitats and (2) evaluate the effects of translocationon survival of adult Shoal Bass and determine its value as aviable restoration and management tool.

METHODSStudy Area.—The study area encompassed 150 rkm of

the lower Flint River between Jim Woodruff Lock and Dam(JWLD), Florida, and the Georgia Power Dam above Albany,Georgia (Figure 1). This section is influenced by flow regula-tion at the Georgia Power Dam, which provides hydroelectricproduction and water level control. For this study, capture effortwas concentrated on a known spawning shoal approximately25 rkm downriver of the Georgia Power Dam (Figure 1). Thetranslocation release site, Bainbridge Boat Basin, is located atthe interface of the riverine and reservoir habitats at the up-permost portion of Lake Seminole, approximately 50 rkm fromJWLD. The Bainbridge Boat Basin was selected because numer-ous black bass fishing tournaments use this facility as the launchand weigh-in site. Live-released fish weighed during these tour-naments may come from as far as the Upper Apalachicola River,Lower Chattahoochee River, Lake Seminole, and the LowerFlint River. All fish are typically released into the BainbridgeBoat Basin.

Telemetry.—Twenty-seven adult Shoal Bass were capturedvia electrofishing at the aforementioned spawning site on theFlint River during April 2010 (Figure 1). Shoal Bass were mea-sured for TL (mm), wet weight (g), and fitted with hydroacous-tic transmitters. Hydroacoustic transmitters (V7-2 L, Vemco,Halifax, Nova Scotia) were surgically implanted into the ab-dominal cavity through a 2-cm ventral incision posterior to theleft pelvic fin. Incisions were closed with two interrupted ab-sorbable sutures (3-0 Coated Vicryl; Ethicon, Cincinnati, Ohio).Transmitters measured 7.0 mm in diameter and 20 mm in length,weighed 0.75 g in water, and possessed a minimum battery lifeof 95 d. Each transmitter was uniquely coded, programmed totransmit every 40–80 s, and operated at a frequency of 69.0 kHz.Shoal Bass were also tagged using a Floy tag (Floy Tag FM-84,Seattle, Washington) anchored in the abdominal wall within theincision created for the transmitter. Each tag was imprinted with

a unique identification number and contact information for tagreturn and reward. During transmitter implantation, sex was de-termined by visual examination of the gonads. Surgery lasted anaverage of 2 min. Following tagging, fish were held in a holdingtank for 15 min to recover and verify transmitter signal. Onlyfish that were able to swim freely were released.

Tagged fish were either released at the original capture site(control group) or translocated (translocated group). ThirteenShoal Bass (seven males and six females, mean ± SD = 422 ±65 mm in length and 1,093 ± 521 g in weight) comprised thecontrol group that was released immediately upon transmitterimplantation and recovery. Fourteen Shoal Bass (eight males andsix females, mean ± SD = 386 ± 59 mm in length and 810 ±403 g in weight) comprised the translocated treatment group.The translocated group was transported 75 rkm downriver in a1,100-L fiberglass tank with constant aeration to the BainbridgeBoat Basin and released (Figure 1).

Telemetry was conducted by boat within 24 h of the taggingevent and then every 2 weeks from the Georgia Power Dam toLake Seminole during April through August 2010. Shoal Basswere located using a Vemco VR 100 scanning receiver and VH110 directional hydrophone. The location of each Shoal Basswas recorded using a data-logging global positioning system(GPS) unit with a position accuracy of 3 m (GPSmap 76, GarminInternational, Olathe, Kansas).

Data Analyses.—We characterized the Shoal Bass spawningmigration by measuring the dispersal distance and dispersaldirection (upstream or downstream) of control fish from thespawning site. To examine the effects of translocation onShoal Bass, we calculated survival, time to dispersal, dispersaldistance, maximum displacement, and dispersal direction met-rics for comparison between control and translocated groups.Groups were stratified by sex (male and female) to examine sex-specific behavioral differences. Fish were considered to havedispersed once they moved further than 1 rkm from the spawningshoal. Time to dispersal was defined as the number of days untila fish moved greater than 1 rkm from the release site. Dispersaldistance was defined as the distance (rkm) moved by a fishfollowing dispersal, and maximum displacement was definedas the greatest distance each fish was observed from the releasesite. Dispersal direction was defined as the ultimate direction ofdisplacement relative to the release site for each tagged fish. Allmeasurements were calculated using GPS coordinates of indi-vidual fish relocations downloaded into Google Earth software(version 6.0, Google, Mountain View, California). All statisticalcomparisons of means (dispersal time, dispersal distance, andfarthest distance) between control and translocated groups wereconducted by using an unpaired Student’s two-tailed t-test withn-2 degrees of freedom and a significance level of 0.05.

RESULTSPrerelease survival for all telemetered Shoal Bass was 100%.

Two translocated fish were removed from the study; one wasnever located after release, and the other was harvested by an

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3


FIGURE 1. Map of the study area, Lower Flint River and Lake Seminole, within the mid-Apalachicola–Chattahoochee–Flint River basin and the southern UnitedStates of America. Prominent landmarks for the study are denoted by black stars.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3

720 INGRAM ET AL.

TABLE 1. Summary information for Shoal Bass with internal hydroacoustic transmitters monitored in the Flint River, Georgia. The C and T labels in the Groupcolumn represent control and translocated fish, respectively. The M and F labels in the Sex column designate male and female fish, respectively. Dispersal directionwas defined as the ultimate direction of displacement relative to the release site for each tagged fish and was classified as upstream (UPS) or downstream (DNS).

Times Farthest distance DispersalGroup Tagging date Sex TL (mm) located moved (rkm) direction

C April 19, 2010 M 512 3 4.3 UPSC April 19, 2010 M 346 2 7.7 DNSC April 19, 2010 F 532 2 17.3 DNSC April 19, 2010 F 361 3 7.6 UPSC April 19, 2010 M 490 3 23.4 DNSC April 19, 2010 M 361 4 19.4 UPSC April 19, 2010 F 409 2 1.4 DNSC April 19, 2010 F 490 5 4.9 UPSC April 19, 2010 M 441 2 96.1 DNSC April 19, 2010 M 367 3 2.1 UPSC April 19, 2010 M 424 3 0 DNSC April 19, 2010 F 391 5 3.6 UPSC April 19, 2010 F 369 4 1.9 UPST April 19, 2010 F 368 6 2.5 DNST April 19, 2010 F 347 7 3.2 UPST April 19, 2010 M 407 7 3.2 DNST April 19, 2010 M 372 7 3.2 UPST April 19, 2010 M 448 6 8.7 UPST April 19, 2010 F 530 0 0T April 19, 2010 M 397 6 3.9 UPST April 19, 2010 F 386 4 14.5 UPST April 19, 2010 F 420 2 1.2 UPST April 19, 2010 M 327 3 23.0 UPST April 19, 2010 M 373 2 1.5 UPST April 19, 2010 M 305 4 17.8 UPST April 19, 2010 F 369 4 2.5 DNST April 19, 2010 M 335 0

angler one day after release. The one mortality of a control fishwas assumed after telemetry indicated no movement throughoutthe study (Table 1). One Shoal Bass from the control group wasremoved from behavioral comparisons after being relocated bydata-logging receivers used for another study below the navi-gation lock at JWLD only 8 d after telemetering and release.Remaining Shoal Bass were relocated an average of four timesduring the 90-d period following release (Table 1).

No sex-based differences (females compared to males)within groups were observed to have influenced dispersaltime (translocated: t = 1.00, df = 10, P = 0.34; control:

t = 0.73, df = 10, P = 0.32), movement after initial dispersal(translocated: t = 0.73, df = 10, P = 0.48; control: t = 1.70,df = 10, P = 0.12), or farthest distance after 90 d at large(translocated: t = 0.89, df = 9, P = 0.39; control: t = 0.72,df = 10, P = 0.48). Therefore, comparisons between treatmentgroups were made with both sexes combined.

Shoal Bass from both groups eventually dispersed from therelease sites. However, the average time to dispersal was signif-icantly greater for translocated fish when compared to controlfish (t = 3.14, df = 21, P = 0.005; Table 2). Mean distancemoved after initial dispersal was not significantly different

TABLE 2. Mean dispersal time, distance moved after dispersal, and farthest distance moved from release sites for control and translocated groups of Shoal Bassin the Flint River.

Mean time Mean distance moved Mean farthest distance movedGroup to dispersal (days) after dispersal (rkm) from release site (rkm)

Control 12.3 7.1 9.1Translocated 27.8 34.6 10.2

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3


between groups (t = 1.78, df = 21, P = 0.09; Table 2). After 90d at large, the farthest distance moved from the release sites wasnot significantly different between the two groups (t = 0.28,df = 21, P = 0.78; Table 2). The final locations of the controlgroup ranged 1.6–23.0 rkm in both upstream and downstreamdirections from the spawning site (Table1). The translocatedfish dispersed 2.6–28.6 rkm and moved predominantly uprivertoward the original capture site (Table 1). Postrelease survivalrates at 90 d for both the translocated and the control groupswere 92% (95% CI: 77–100%).

DISCUSSIONOur results suggest that adult Shoal Bass may migrate long

distances between seasonal home ranges and spawning areas.The control group dispersed over a 40-rkm surrounding reachwithin 90 d postrelease. Tag returns from anglers have also indi-cated Shoal Bass volitionally dispersed from spawning habitatto a surrounding 60-rkm reach of the Flint River (Georgia De-partment of Natural Resources, unpublished data). These datasuggest that a portion of Flint River adult Shoal Bass under-take substantial migrations to spawning habitats. It is unknownexactly why some individuals travel greater distances to sim-ilar habitat farther up or down stream, but plausible explana-tions include spawning site fidelity and river flows. However,our findings that a portion of Shoal Bass migrate long dis-tances contradicted the only other reported telemetry resultsacquired from adult Shoal Bass. Stormer and Maceina (2009)telemetered adult Shoal Bass during April in a tributary to theChattahoochee River and reported that those Shoal Bass weresedentary and remained within a 1.5-rkm area the entirety ofthe 140-d transmitter life. River discharge and geomorphologyalong with the length of unimpeded river reach available to FlintRiver Shoal Bass compared to the populations studied in Chat-tahoochee River tributaries may lend to the different behavioralpatterns. These different spawning migrations may be cause fornew management issues in the Flint River. Shoal Bass popula-tions may need to be protected from exploitation at spawningsites and both critical spawning and postspawning habitat mayneed protection.

Translocated Shoal Bass initially remained within 1 rkmof the release site, but within 40 d most had dispersed. Themajority of Shoal Bass remained sedentary through the summerafter the initial dispersal event. Black bass behavior is knownto be seasonally influenced by river discharge, with movementdecreasing during summer. Stormer and Maceina (2009) foundthat Shoal Bass showed high site fidelity to meso-habitats evenas water levels decreased during summer. These results are verysimilar to postrelease stockpiling and dispersal behavior ex-hibited by other black bass species. Stockpiling of LargemouthBass Micropterus salmoides has been estimated at 49–64%,with released fish dispersing less than 1.6 rkm after 7–43 d(Richardson-Heft et al. 2000; Wilde 2003; Wilde and Paulson2003). The proportion of Smallmouth Bass Micropterus

dolomieu exhibiting sedentary posttournament behavior hasbeen estimated at 21–26% (Wilde 2003). Average long-termdispersal between translocated and control groups was similarand comparable to other black bass dispersal distances. Re-ported average long-term dispersal distances were 3.5–9.6 rkm(Richardson-Heft et al. 2000; Wilde 2003) for LargemouthBass and 7.3 rkm (Wilde 2003) for Smallmouth Bass. Notranslocated Shoal Bass returned to the original capture locationwithin 90 d and one was captured the following spring by anangler 5 rkm from the release location at the Bainbridge BoatBasin. It is possible that these translocated fish were removedpermanently from the original spawning aggregation.

Translocation of adult Shoal Bass to a distant location provedsuccessful. Adult Shoal Bass survived transport, dispersed froma distant release site, and generally moved upstream to pre-ferred habitat. Intentional relocation of individuals from a moreabundant population within a given species’ range in order torestore populations elsewhere has become an important fisherymanagement tool (Hallacher 1984; Minckley 1995; Harig et al.2000). However, these results are from transporting Shoal Bassunder ideal conditions and may not accurately represent inci-dental transport and release that occurs from tournament fish-ing. Shoal Bass are often stressed from being caught and againstressed through transport possibly in less than optimum condi-tions. Couple these stresses with the added stress of spawningor increased summer water temperature and an increase in mor-tality may occur. These results indicate translocation may be avaluable tool in future restoration efforts, but much is still un-known about the effects of tournament transportation on ShoalBass.

Although these results are encouraging and indicate highsurvival of captured and transported Shoal Bass, implicationsof negative effects on long-term behavior in terms of spawningand seasonal habitat use of live-released Shoal Bass are numer-ous and remained unanswered. Our findings indicate that ShoalBass in the Flint River demonstrate different migration behav-ior than populations of Shoal Bass in different systems. Thismigration behavior may be cause for unique management ofthis population to protect spawning Shoal Bass and the differenthabitats utilized by these fish throughout the year. Results of thisstudy may also indicate that transportation of Shoal Bass maybe a viable management tool to aid in restoration of fragmentedpopulations and to maintain genetic diversity in these popula-tions. Additional studies incorporating long-term telemetry andfishing tournaments across seasons should be initiated to ensureminimal impact on the endemic Shoal Bass populations.

ACKNOWLEDGMENTSThis research was funded through the Federal Aid in Sport-

fish Restoration Act. We gratefully acknowledge the person-nel of the Georgia Department of Natural Resources, FisheriesManagement Section for support and assistance. Thanks to JohnKilpatrick, Dean “Road Buzzard” Barber, and Rob Weller for

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3

722 INGRAM ET AL.

the countless hours spent shuttling trucks up and down the river.Special thanks to Todd Braswell for the volunteered time spentin the field.

REFERENCESHallacher, L. E. 1984. Relocation of original territories by displaced Black-

and-yellow Rockfish, Sebastes chrysomelas, from Carmel Bay, California.California Fish and Game 70:158–162.

Harig, A. L., K. D. Fausch, and M. K. Young. 2000. Factors influencing successof Greenback Cutthroat Trout translocations. North American Journal ofFisheries Management 20:994–1004.

Johnston, C. E., and R. A. Kennon. 2007. Habitat use of the Shoal Bass, Mi-cropterus cataractae, in an Alabama stream. Journal of Freshwater Ecology22:493–498.

Minckley, W. L. 1995. Translocation as a tool for conserving imperiled fishes:experiences in western United States. Biological Conservation 72:297–309.

Richardson-Heft, C. A., A. A. Heft, L. Fewlass, and S. B. Brandt. 2000. Move-ment of Largemouth Bass in northern Chesapeake Bay: relevance to sportfish-

ing tournaments. North American Journal of Fisheries Management 20:493–501.

Stormer, D. G., and M. J. Maceina. 2008. Relative abundance, distribution, andpopulation metrics of Shoal Bass in Alabama. Journal of Freshwater Ecology23:651–661.

Stormer, D. G., and M. J. Maceina. 2009. Habitat use, home range, and move-ment of Shoal Bass in Alabama. North American Journal of Fisheries Man-agement 29:604–613.

Wheeler, A. P., and M. S. Allen. 2003. Habitat and diet partitioning betweenShoal Bass and Largemouth Bass in the Chipola River, Florida. Transactionsof the American Fisheries Society 132:438–449.

Wilde, G. R. 2003. Dispersal of tournament-caught Black Bass. Fisheries28(7):10–17.

Wilde, G. R., and L. J. Paulson. 2003. Movement and dispersal of tournament-caught Largemouth Bass in Lake Mead, Arizona–Nevada. Journal of Fresh-water Ecology 18:339–342.

Williams, J. D., and G. H. Burgess. 1999. A new species of bass, Micropteruscataractae (Teleostei: Centrarchidae), from the Apalachicola River basin inAlabama, Florida, and Georgia. Bulletin of the Florida Museum of NaturalHistory 42:80–114.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3



Self-Reporting Bias in Chinook Salmon Sport Fisheriesin Idaho: Implications for Roving Creel SurveysJoshua L. McCormick a d , Michael C. Quist b & Daniel J. Schill ca Idaho Cooperative Fish and Wildlife Research Unit, Department of Fish and WildlifeSciences , University of Idaho , Post Office Box 441141, Moscow , Idaho , 83844 , USAb U.S. Geological Survey, Idaho Cooperative Fish and Wildlife Research Unit, Department ofFish and Wildlife Sciences , University of Idaho , Post Office Box 441141, Moscow , Idaho ,83844 , USAc Idaho Department of Fish and Game , 1414 East Locust Lane, Nampa , Idaho , 83686 , USAd Oregon Department of Fish and Wildlife , 3406 Cherry Avenue Northeast, Salem , Oregon ,97303 , USAPublished online: 15 Jul 2013.

To cite this article: Joshua L. McCormick , Michael C. Quist & Daniel J. Schill (2013) Self-Reporting Bias in Chinook SalmonSport Fisheries in Idaho: Implications for Roving Creel Surveys, North American Journal of Fisheries Management, 33:4,723-731, DOI: 10.1080/02755947.2013.808293







http://dx.doi.org/10.1080/02755947.2013.808293




ARTICLE

Self-Reporting Bias in Chinook Salmon Sport Fisheriesin Idaho: Implications for Roving Creel Surveys

Joshua L. McCormick*1

Idaho Cooperative Fish and Wildlife Research Unit, Department of Fish and Wildlife Sciences,University of Idaho, Post Office Box 441141, Moscow, Idaho 83844, USA

Michael C. QuistU.S. Geological Survey, Idaho Cooperative Fish and Wildlife Research Unit,Department of Fish and Wildlife Sciences, University of Idaho, Post Office Box 441141, Moscow,Idaho 83844, USA

Daniel J. SchillIdaho Department of Fish and Game, 1414 East Locust Lane, Nampa, Idaho, 83686, USA

AbstractSelf-reporting bias in sport fisheries of Chinook Salmon Oncorhynchus tshawytscha in Idaho was quantified by

comparing observed and angler-reported data. A total of 164 observed anglers fished for 541 h and caught 74 ChinookSalmon. Fifty-eight fish were harvested and 16 were released. Anglers reported fishing for 604 h, an overestimate of63 h. Anglers reported catching 66 fish; four less harvested and four less released fish were reported than observed.A Monte Carlo simulation revealed that when angler-reported data were used, total catch was underestimated by14–15 fish (19–20%) using the ratio-of-means estimator to calculate mean catch rate. Negative bias was reduced tosix fish (8%) when the means-of-ratio estimator was used. Multiple linear regression models to predict reportingbias in time fished had poor predictive value. However, actual time fished and a categorical covariate indicatingwhether the angler fished continuously during their fishing trip were two variables that were present in all of the topa priori models evaluated. Underreporting of catch and overreporting of time fished by anglers present challengeswhen managing Chinook Salmon sport fisheries. However, confidence intervals were near target levels and using moreliberal definitions of angling when estimating effort in creel surveys may decrease sensitivity to bias in angler-reporteddata.

Creel surveys are the primary means to collect informationon angler effort and catch rates and to estimate the numberof fish harvested and released in sport fisheries (Malvestuto1983; Pollock et al. 1994). Most creel surveys rely on angler-reported data to estimate metrics used in managing fisheries.Total catch is estimated by multiplying estimates of effort andmean catch rate (Robson 1961; Pollock et al. 1994). Mean catchrate is based on the angler-reported amount of time fished andnumber of fish caught. Bias with either of these metrics can

*Corresponding author: [email protected] address: Oregon Department of Fish and Wildlife, 3406 Cherry Avenue Northeast, Salem, Oregon 97303, USA.Received January 3, 2013; accepted May 20, 2013

result in inaccurate estimates of total catch. As with many hu-man behaviors, self-reporting of angling activity is subject tobias. Self-reporting bias in angler surveys is a form of nonsam-pling bias that includes recall, prestige, digit, and misreportingbias (Sudman and Bradburn 1974; Essig and Holliday 1991;Connelly et al. 2000). Recall bias refers to a respondent’s inabil-ity to recollect and accurately report events. Prestige bias refersto a respondent’s inaccurate response that they believe is so-cially desirable (Applegate 1984; Brown et al. 1986). Digit bias

723

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3

724 MCCORMICK ET AL.

refers to a respondent’s tendency to round numbers to those thatend in zeros or fives, and misreporting bias refers to a respon-dent’s attempt to deliberately mislead the survey administrator.In roving and access creel surveys in sport fisheries of ChinookSalmon Oncorhynchus tshawytscha in Idaho, estimates of effort(i.e., angling hours) and catch rely on angler self-reported dataand may be susceptible to any or all of these forms of bias.

Currently, several populations of spring- and summer-runChinook Salmon in the Snake River evolutionary significantunit are listed as threatened under the Endangered Species Act(ESA; WCCSBRT 1997). Because of their status, consumptivesport fisheries are limited to fish of hatchery origin, which arenot listed under the ESA. When targeting hatchery fish, anglersmay catch wild fish that must be released. The ESA limits theallowable incidental mortality of listed species that are caughtand released in fisheries that target hatchery fish. Fisheries mustbe closed when this condition is met (Apperson and Wilson1998). Fisheries may also be closed once a “harvest share” ofhatchery fish is reached. In Idaho, the total allowable harvest isshared equally between tribal and sport fisheries. Additionally,fisheries are often closed to allow fish to escape into upriverfisheries or hatcheries for use as broodstock (McCormick et al.2012). On-site creel surveys (i.e., roving and access surveys) areused to estimate the total number of fish harvested and releasedthroughout the season and depend on accurate self-reported datafrom anglers to properly manage the fisheries.

Much of the research on self-reporting bias in fisheries hasfocused on off-site survey techniques (Carline 1972; Fisher et al.1991; Roach et al. 1999). On-site survey techniques are gener-ally accepted as having less reporting bias because recall timeis short and harvest can be verified by creel clerks (Pollocket al. 1994; Newman et al. 1997; Roach et al. 1999). The resultsof research focused on on-site techniques have been inconsis-tent. Edwards (1971) found that a group of anglers fishing forRainbow Trout O. mykiss on the Colorado River substantiallyoverestimated their fishing time. The author also found thatunsuccessful anglers underestimated their fishing time, whileanglers who were successful were more accurate at estimatingfishing times. In contrast, no differences were found betweenactual and reported times for fisheries in Alberta lakes, and an-glers who were more successful generally underestimated theirfishing times (Radford 1973). Anglers who were less success-ful overestimated fishing times. No difference between actualand reported fishing times, and no correlation between lengthof trip and accuracy of reporting, was found in marine fisheriesin Texas (McEachron et al. 1986). Similarly, Steffe and Murphy(2010) found that angler-reported times were empirically unbi-ased in marine fisheries in New South Wales, Australia. Johnsonand Wroblewski (1962) found high variability in the accuracyof reported fishing times for individual anglers but no bias intotals hours fished for all anglers in a Walleye Sander vitreusfishery in Minnesota.

In addition to angler-reported fishing times, estimates ofcatch rate also rely on unbiased angler reports of number of

fish caught, harvested, and released. Since harvested fish canusually be observed by creel clerks in on-site surveys, lit-tle bias should arise from anglers self-reporting their harvest(Newman et al. 1997). However, self-reporting bias of harvestedfish was observed in fisheries with large bag limits in Floridalakes (Mallison and Cichra 2004). While harvested fish can beobserved, released fish cannot (Huntsman et al. 1978; Sullivan2003). Sullivan (2003) found that anglers overreported catch andrelease of protected-length Walleye by more than two-fold in Al-berta lakes. The author also found that overreporting of catchand release increased exponentially with decreasing catch.

Unlike fisheries where indices of catch are sufficient to makemanagement decisions, absolute values of catch and harvest areneeded to manage Chinook Salmon fisheries in Idaho. Knowl-edge of the extent and direction of possible self-reporting bias isdesired for proper management of sport fisheries in Idaho. Biasin self-reported catch or amount of time fished by anglers couldresult in reduced returns of wild fish or in insufficient return ofbroodstock fish to hatcheries, both of which can affect futurefisheries. Therefore, the objective of this research was to eval-uate self-reporting bias in Chinook Salmon creel surveys anddetermine its effect on mean catch rate and total catch estimates.

METHODSDiscreet observations of angling activity were conducted to

quantify self-reporting bias. Observed angling activity was thencompared with the angler-reported data given to Idaho Depart-ment of Fish and Game (IDFG) creel clerks conducting rovingand access surveys. Direct visual observations were conductedover a 26-d period during the 2011 and 2012 seasons at one siteon the Little Salmon and Middle Fork Clearwater rivers, twosites on the Salmon River, and three sites on the South ForkSalmon River. By regulation, legal fishing time on all fisherieswas limited to daylight hours only. Observers arrived beforelegal fishing time and remained until the end of legal fishingtime. Discreet observations were conducted so as not to influ-ence angler decisions on fishing locations, fishing times, fishharvest, or reporting to IDFG creel clerks. Angling activity wasobserved from afar (200–400 m) using spotting scopes on theLittle Salmon River and one site on the South Fork SalmonRiver. Two observers were used on the Middle Fork Clearwa-ter River and two sites on the South Fork Salmon River. Oneobserver fished and relayed information to the other observerwho discreetly recorded data. For both observation approaches,data were recorded on the time each angler entered and exitedthe fishery, total catch, time of catch, number and origin of fishreleased, and the number and origin of fish harvested. Hatcheryfish were identifiable to anglers and observers by an excised adi-pose fin. Sites were selected nonrandomly to increase samplesizes and allow for discreet observations. Survey days were se-lected to coincide with maximum angling activity on each river.

During the observation process, each angler was assigned aunique angler identification number at the start of their fishing

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3

SELF-REPORTING BIAS IN ROVING CREEL SURVEYS 725

episode, which was defined as the angler’s first cast. The end ofthe fishing event was defined as the time the angler exited thefishing area and was no longer available for a roving interviewor count. Anglers would frequently take a “break” from fishingor exit the fishery for short periods of time and were unavail-able for counting in a roving creel survey. If an angler took abreak for 5 minutes or more, they were assumed to be unavail-able for counting in a roving effort count and were recorded asnot fishing. An angler was assumed to be fishing and availablefor counting if they were actively fishing or changing tackle.Anglers that took a break or exited and reentered the fisherywere reassigned their initial identification number when theyreentered the fishery or resumed angling.

Anglers were interviewed by IDFG creel clerks during theirnormal creel survey schedule and asked the same questions asin a normal creel survey (e.g., number of fish harvested andreleased, amount of time fished). Roving interviews were con-ducted on the Little Salmon and Middle Fork Clearwater rivers,whereas a combination of access and roving interviews wereconducted on the Salmon and South Fork Salmon rivers. Im-mediately following the interview process, IDFG creel clerksprovided interview identification numbers to the observers wholinked interview numbers to angler identification numbers andrespective data.

The number of observed anglers interviewed was highly vari-able depending on the fishery and day. As a result, data werepooled from all fisheries for summary analysis. Mean catch ratefor all anglers was calculated using the ratio-of-means (ROM)estimator (Jones et al. 1995; Hoenig et al. 1997; McCormicket al. 2012). Reporting bias was calculated as the differencebetween the reported number of fish harvested, fish released,and observed values. A value of zero indicated the data werereported accurately, a negative value indicated that the anglerunderreported catch or amount of time fished, and a positivevalue indicated that the angler overreported catch or amount oftime fished.

An exploratory analysis was conducted using multiple lin-ear regression to examine relationships among reporting biasand amount of time fished, number of fish caught, and the timethe individual started fishing minus the legal start of fishingtime (Fox 2008). Additionally, two categorical covariates wereexamined that included the river where observations were con-ducted and whether the angler fished continuously during theirtrip. Because the sample size to evaluate reporting bias in thenumber of fish caught was small, regression models were onlycreated for reporting bias in the amount of time fished. Theassumptions of the linear models were evaluated by examininga suite of diagnostic plots including observed versus predictedvalues, externally studentized residuals versus predicted val-ues, and normal quantile–quantile plots (Fox 2008). Akaike’sinformation criterion corrected for small sample bias (AICc)was used to evaluate a priori candidate models (Akaike 1973;Burnham and Anderson 2002). Akaike weights (wi) were used toassess the relative plausibility of each candidate model. Model-

averaged regression coefficients were calculated using wi forall candidate models examined (Burnham and Anderson 2002).Model-averaged coefficients were only calculated for the predic-tor variables that were present in the most parsimonious model.The coefficient of determination (R2) was calculated to evalu-ate the goodness of fit of each candidate model. Additionally,the predictive performance of the most parsimonious model se-lected using AICc with and without model-averaged parameterestimates was evaluated using “leave-one-out” cross validation(Efron and Gong 1983; Efron and Tibshirani 1993). Duringthis procedure, one observation (i.e., angler-reported bias) wasomitted from the data and the model was fit with the remain-ing observations and the predicted error was calculated. This isknown as one fold. When model-averaged coefficient estimateswere evaluated, all candidate models were fit with the omittedobservation, model-averaged coefficients were estimated, andthe predicted error was calculated. This process was repeatedfor all observations creating a distribution of predicted errors.The mean square error (MSE) of the cross-validation procedurefor each model was estimated as follows:

MSE = 1

kn

k∑j=1

n∑j=1

(Yi j − Yi j )2, (1)

where Yi j is the observed ith value in the jth fold, Yi j is theestimated ith value in the jth fold, n is the number of observationsin one fold, and k is the number of folds.

Monte Carlo simulations were used to determine the influ-ence of potential self-reporting bias on estimates of mean catchrate and total catch. This evaluation was conducted to determinethe sensitivity of the mean catch rate estimators to self-reportingbias. Five thousand iterations were conducted where samples ofangler interviews were selected using simple random samplingwithout replacement (Cochran 1977). McCormick et al. (2012)showed that the ROM estimator provided the most accurate esti-mates of mean catch rate in these fisheries. However, the authorsused data that did not account for errors in angler-reported data.As a result, for each iteration mean catch rate was calculatedusing the ROM for the observed data and using the ROM andmeans-of-ratios (MOR) estimators for the angler-reported data(Jones et al. 1995; Hoenig et al. 1997; McCormick et al. 2012).Total catch was estimated for each iteration as the product ofmean catch rate and the observed hours of angling effort for allanglers at the time of the interviews. To determine the influenceof sample size on estimates of mean catch rate, total catch, andtheir confidence intervals, the number of simulated interviewsconducted varied from 25 to 150 by intervals of 25. Ninety-fivepercent nonparametric bootstrap confidence intervals were cal-culated for each estimate using the percentile method (Efronand Tibshirani 1993). Bias for each population parameter wasestimated as the difference between the mean of the empiri-cal sampling distribution from the true population parameter(i.e., mean catch rate, total catch). Coverage of 95% confidence

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3


TABLE 1. Observed and reported number of hatchery Chinook Salmon harvested and wild Chinook Salmon released by 164 anglers in three Idaho sport fisheriesduring the 2011 and 2012 seasons.

Observed catchObserved harvest (hatchery) (wild)

Reported harvestor catch 0 1 2 3 4 5 0 1

0 119 6 0 0 0 0 146 91 2 25 1 0 0 0 1 72 0 1 6 0 0 0 0 03 0 1 0 2 0 1 0 04 0 0 0 0 0 0 1 05 0 0 0 0 0 0 0 0

intervals (CIs) was also evaluated. In theory, 95% of all CIsshould encompass the true population parameter, and for itera-tions whose CIs did not encompass the true parameter values,50% should be below the true value and 50% should be abovethe true value. The percentage of CIs that encompassed the truepopulation parameter was determined. Simulations and statisti-cal analysis were conducted using the R statistical computinglanguage (R Development Core Team 2009).

RESULTSDuring the 2011 and 2012 Chinook Salmon fishing sea-

sons, 164 anglers were interviewed by IDFG creel clerks whiletheir trips were discreetly observed. Observed anglers caught 74Chinook Salmon; 58 were hatchery fish that were harvested and16 were wild fish that were released. No hatchery fish werereleased and no wild fish were illegally harvested. Of the 164anglers observed, 49 (30%) were successful in catching a fishand 10% of anglers caught 55% of the fish. One angler caughtfive fish, one angler caught four fish, four anglers caught threefish, 10 anglers caught two fish, and the remaining 33 anglerscaught one fish (Table 1). Observed anglers reported catching 66fish; 54 hatchery fish harvested and 12 wild fish released. Fourless hatchery fish and four less wild fish were reported than ob-served (Table 1). Forty-four percent of observed wild fish caughtand released were reported to creel clerks. No anglers falselyreported wild fish released as hatchery fish released.

Overall, anglers overreported their fishing time by 63 h. Ob-served anglers fished for a total of 541 h (mean, 3.44 h; range,0.17–13.4 h) and reported fishing for 604 h. A majority of an-glers reported their fishing time within 1 h of their actual amountof time fished (Figure 1). The observed angler catch rate was0.137 fish/h. Using self-reported data, the mean daily catch ratewas estimated at 0.110 fish/h. If angling effort was estimatedaccurately in a roving creel survey (i.e., 541 h) using a roving oraerial survey, multiplying the reported catch rate by the estimateof angler hours would result in an estimate of 60 fish. This isan underestimate of total catch by 14 fish (19%). The observedharvest rate of hatchery fish was 0.107 fish/h and the observed

catch rate of wild fish was 0.030 fish/h. The reported catch ratewas 0.090 fish/h and 0.020 fish/h for hatchery fish and wildfish, respectively. If angler effort was determined with 100%accuracy in a roving creel survey, the estimate of hatchery fishharvested would be 49 fish and the estimate of catch of wild fishwould be 11 fish. This is an underestimate of harvest by ninefish and an underestimate of wild fish released by five fish.

Examination of diagnostic plots indicated that the data metthe assumptions for linear regression analysis with the exceptionof the quantile–quantile plot, which indicated a slight departurefrom normality (Figure 2). Five of the 15 a priori regressionmodels to predict reporting bias accounted for all of the wi

(Table 2). All of the top five models contained the actual amountof time fished and the categorical variable indicating that theangler fished continuously during their trip. The top model in-cluded only these two variables and received 55% of wi. Thesecond best model, according to AICc, included these two vari-ables and the difference in time each angler started fishing from

Reported hours fished - observed hours fished

-4 -3 -2 -1 0 1 2 3 4 5

Num

ber

of a

ngle

rs

0

10

20

30

40

50

n =164

FIGURE 1. Distribution of the difference of observed and actual angler-reported fishing times of 164 anglers in three Chinook Salmon fisheries inIdaho during the 2011 and 2012 seasons.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3


Observed value

-4 -2 0 2 4 6

Pre

dict

ed v

alue

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Externally-studentized residual

-4 -3 -2 -1 0 1 2 3 4

Pre

dict

ed v

alue

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Quantiles of standard normal

-4 -2 0 2 4

Ext

erna

lly

stud

enti

zed

resi

dual

-4

-3

-2

-1

0

1

2

3

4

A

B

C

FIGURE 2. Diagnostic plots to evaluate the assumptions of a multiple linearregression model to predict angler-reported bias in fishing times in three ChinookSalmon fisheries in Idaho during the 2011 and 2012 seasons. Panel (A) showsthe observed versus predicted bias, panel (B) shows externally studentizedresiduals versus predicted bias, and panel (C) is a normal quantile–quantileplot.

the legal start time, and had an AICc value that was 1.68 greaterthan the top model. The number of fish caught and the differ-ence in time each angler started fishing from the start of legalfishing time also appeared in four of the top five models. Themodel containing the covariate for fishery location (i.e., river)only accounted for 1% of wi suggesting that reporting bias wassimilar between fisheries. The difference between reported andobserved fishing times was inversely related to actual amountof time fished, suggesting that as actual fishing time increasedanglers were more likely to underreport the time they fished(Table 2). The top model and model-averaged coefficients forthe categorical variable indicating whether the angler fished theentire duration of their trip were also negative. This suggeststhat anglers who fished continuously underreported their fish-ing times compared with anglers who took a break during theirfishing trip. All of the models examined exhibited low R2 (i.e.,R2 varied from 0.20 to 0.21 for the top five models; Table 2).Cross-validation MSE for the top model was 1.08 with and with-out model-averaged parameter estimates. With the exception ofthe intercept, which was 1.63 using model averaging comparedwith 1.60 without, model-averaged parameter estimates werenearly identical to the top model parameter estimates (Table 2).

Results from the Monte Carlo simulation showed that esti-mates of mean catch rate and total catch were unbiased usingobserved data (Table 3). Nonparametric CIs encompassed thetrue mean greater than 95% of the time for all samples sizes usingobserved data. While CIs were more than 95%, they were bi-ased low with sample sizes of 25 and 50 interviews. Confidenceintervals were relatively unbiased when 75 or more interviewswere conducted. When reported data were used, the MOR es-timator resulted in less-biased estimates of catch rate and totalcatch compared with the ROM estimator. The ROM estimatorand angler-reported data resulted in negatively biased estimatesthat varied from 14 to 15 fish (19–20%). Using the MOR estima-tor improved bias and resulted in estimates that were negativelybiased by six fish (8%). Bias in estimates of catch rate and totalcatch increased slightly with increasing sample size. Confidenceinterval coverage varied from 92% to 100% using the MOR es-timator and 93% to 97% using the ROM estimator. Confidenceinterval coverage increased for both estimators with increasingsample size. However, all CIs were biased low regardless of thesample size.

DISCUSSIONA major assumption when estimating mean catch rate, total

catch, and total harvest in creel surveys is that anglers accu-rately report data. While a majority of anglers were accurate inreporting their data, our results suggest that the minority thatviolated this assumption introduced bias into total catch andharvest estimates in various Chinook Salmon fisheries in Idaho.Estimates were negatively biased from 8% to 20% depending onthe estimator used to calculate mean catch rate. Sullivan (2003),

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3


TABLE 2. Multiple regression models and derived parameter estimates predicting angler-reported bias (observed-reported) in angling times in three ChinookSalmon fisheries in Idaho during the 2011 and 2012 seasons. Variables include the following: the actual amount of time fished (AT), the number of fish caught (C),the difference between the time the angler started fishing and the start of legal fishing time (�L), the categorical covariates for if the angler fished continuouslythroughout their fishing trip (FC), and the river where the fishery took place with the Little Salmon River serving as the reference for the Middle Fork Clearwater(MF), South Fork Salmon (SF), and Salmon (SR) rivers. Akaike’s information criteria corrected for small sample size (AICc), number of parameters (K), changein AICc value (�AICc), and AICc weights (wi) were used to select the top models from a set of a priori candidate models. The coefficient of determination (R2) isprovided as a measure of model fit.

Model K AICc �AICc wi R2

1.60 − 1.01(FC) − 0.15(AT) 4 479.29 0 0.55 0.211.63 − 1.00(FC) − 0.15(AT) − 0.01(�L) 5 480.97 1.68 0.24 0.211.80 − 1.18(FC) − 0.15(AT) − 0.01(C) − 0.10(DL)

+ 0.10(FC × �L)7 482.22 2.93 0.13 0.22

1.63 − 1.00(FC) − 0.16(AT) − 0.01(�L) − 0.01(C) 6 483.12 3.83 0.08 0.211.86 + 0.05(MF) − 0.07(SF) − 0.14(SR) − 1.17(FC)

− 0.01(C) − 0.17(AT) − 0.10(�L) + 0.10(FC × �L)10 488.38 9.1 0.01 0.20

1.05 − 0.95(FC) 3 495.27 15.98 0 0.131.08 − 0.91(FC) − 0.12(C) 4 496 16.71 0 0.131.04 − 0.95(FC) + 0.01(�L) 4 497.27 17.98 0 0.121.20 − 1.06(FC) − 0.11(C) + 0.08(�L) 6 498.38 19.09 0 0.130.81 − 0.13(AT) 3 506.38 27.09 0 0.070.83 − 0.13(AT) − 0.11(C) 4 507.46 28.17 0 0.070.87 − 0.14(AT) − 0.01(�L) 4 507.72 28.43 0 0.070.45 − 0.20(C) 3 515.4 36.11 0 0.010.47 − 0.01(�L) − 0.20(C) 4 517.41 38.12 0 0.010.37 − 0.01(�L) 3 518.88 39.59 0 0.01

who found that anglers overreported catch rates with decliningcatch, suggested that the failure of anglers to accurately reportdata could result in perceived hyperstability of the fishery andlead to collapse. While Chinook Salmon fisheries are managedbased on total catch estimates instead of mean catch rate es-timates, negative impacts of reporting bias are possible. Thesystematic underreporting of fish harvested and released andoverreporting of amount of time fished could lead to the over-harvest of hatchery fish and increase catch-and-release mortalityof wild fish beyond desired objectives. This could result in in-sufficient returns of fish to hatcheries, low escapement to upriverfisheries, or depletion of sensitive wild stocks.

With the exception of two anglers, one who overreportedcatch and release of wild fish by four and one angler who over-rerported by one, prestige bias (Applegate 1984; Brown et al.1986) did not appear to influence the estimates of total catch.More often anglers underreported the number of wild fish caughtand released. While the reasons for this are uncertain, anecdo-tal evidence suggests that anglers are doing so in an attemptto extend season lengths. This form of bias (i.e., misreportingbias) had a greater influence on the estimates of mean catch ratein this study than prestige bias. The low reporting rates of fishthat were caught and released suggests that managers shouldconsider using other methods to estimate catch rates of wild

TABLE 3. Estimated bias and percent confidence interval (CI) coverage of total catch estimates in three Chinook Salmon fisheries in Idaho during the 2011and 2012 seasons using Monte Carlo simulations. Estimates were calculated using observed and angler-reported data. Mean catch rate was estimated using themeans-of-ratios (MOR) estimator and the ratio-of-means (ROM) estimator and n represents the number of interviews conducted.

Observed MOR Reported ROM Reported MOR

n Catch bias % CI coverage Catch bias % CI coverage Catch bias % CI coverage

25 0.73 98 −14.12 93 −6.07 9250 0.55 99 −14.64 94 −5.91 9575 0.17 99 −14.76 94 −5.99 97100 0.12 100 −14.85 93 −6.25 98125 0.03 100 −14.97 94 −6.16 99150 −0.06 100 −15.05 97 −6.35 100

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3


Chinook Salmon. For instance, Sullivan (2003) used a test-fishery with a known catch rate to compare reported catch-and-release rates of Walleyes. It may be possible to create a ratio ofhatchery fish to wild fish caught based on a sample of trustwor-thy anglers in Chinook Salmon fisheries throughout Idaho. Thisratio could then be applied to correct the estimate of the num-ber of wild fish that were caught and released. However, cautionshould be taken when using this approach due to possible spatialand temporal heterogeneity in such a ratio.

Because harvested fish can typically be visually observed(Newman et al. 1997), prestige bias was low in the reportingof harvested fish. When anglers overreported the number offish harvested, it was likely a result of one angler taking creditfor the catch of another angler in their party. This occurredwhen one angler in the party was more successful than theother(s) and was an attempt to avoid reaching the bag limit ofthe successful angler. Because catch rate for the party is likelysimilar regardless of which angler actually harvests the fish,false reporting of catch by other members of a party may havelittle influence on estimates of harvest in creel surveys.

More commonly, anglers underreported harvest. This usuallyoccurred when an angler harvested a fish and transported itto their vehicle or campsite before being surveyed by a creelclerk. In Idaho, anglers are required to immediately record everyharvested fish on a harvest card. Estimates of harvest usingangler-reported data may be improved if creel clerks examineangler harvest cards to verify harvest or with increased lawenforcement presence. Similar to the underreporting of wildfish that are released, anglers may be underreporting harvest inattempt to extend season lengths. Due to relatively small samplesizes of fish caught, data were pooled across fisheries and theextent to which seasons may have been shortened based on self-reporting bias was not assessed. Additionally, a majority of thewild fish that were caught and released in this study were onthe South Fork Salmon River and relatively few wild fish werecaught in the other fisheries.

Although there was individual heterogeneity in the accu-racy of reported time spent angling, Chinook Salmon anglersin Idaho overreported the time they spent angling. In combina-tion with underreporting of catch, overreporting of time spentangling contributed to the underestimate of mean catch rateand total catch. Edwards (1971) found similar results; however,many other researchers found no significant difference betweenactual and reported times spent angling or its effect on totalcatch estimates (Radford 1973; McEachron et al. 1986; Phippenand Bergersen 1987; Phippen and Bergersen 1991; Steffe andMurphy 2010). No hypothesis tests were conducted in our studybecause sampling units were not selected randomly. Addition-ally, results of the Monte Carlo simulation are more meaning-ful than assigning an arbitrary significance level to evaluatedifferences.

Previous research has suggested that it is necessary to applya correction factor to accurately estimate mean catch rate whenobserved angling times are significantly different than angler-

reported data (Radford 1973). Various authors have found a re-lationship between other angling metrics (e.g., catch, time spentfishing) and the accuracy of angler-reported data that could beused to create a model-based correction factor (Edwards 1971;Radford 1973; Sullivan 2003). All of the models that we ex-amined to describe the relationship between various anglingmetrics and the accuracy of angler-reported time spent fishingexhibited low predictive value. This suggests that a correctionfactor is of little value in Idaho’s Chinook Salmon fisheriesbased on the explanatory variables evaluated. We attempted toincrease the predictive value and improve bias and precisionof our models by incorporating uncertainty of model selectionby averaging model coefficients (Burnham and Anderson 2002).However, model-averaging coefficients did not improve the pre-dictive value of the models.

While a correction factor based on the models we examinedmay be of low value, the modeling exercise helped elucidatefactors that influence reporting accuracy of time spent fishing.Perhaps the most meaningful of these factors was whether ornot anglers fished continuously during their trip. This variablewas included in 9 of the 15 candidate models, all of which wereranked higher than the 6 models in which it was not included.A large number of anglers who overestimated their time spentfishing did not fish continuously throughout their trip (i.e., tooka “break”). When these anglers were interviewed, they includedall or a portion of the time they did not fish. For some anglers thiswas an hour or more. Such errors may be accounted for in thequestionnaire process by asking anglers if they fished continu-ously. If not, the angler should be asked to estimate the amountof time they did not fish. The bias in overreporting the amountof time fished may also be corrected by adjusting the definitionof angling during effort counts (Phippen and Bergersen 1987;Phippen and Bergersen 1991). The definition of angling shouldbe the same when conducting angler counts as when anglersreport their amount of time fished. Adopting a more liberal def-inition of angling when conducting angler counts can increaseestimates of effort and correct for the negative bias observed inangler-reported catch rates.

Previous research has shown that using the ROM estima-tor based on completed trip data provided unbiased estimatesof total catch when multiplied by an unbiased estimate of an-gling effort (Jones et al. 1995; Hoenig et al. 1997; McCormicket al. 2012). Results of our simulation show that the MOR es-timator provided estimates that were less biased regardless ofsample size. Overall, anglers tended to overreport the time thatthey fished, which resulted in negatively biased estimates ofmean catch rate. Angler-reported fishing time has no effect onmean catch rate when using the MOR estimator for anglers whowere unsuccessful in catching a fish, which was the case witha majority of observed anglers. However, angling time of un-successful anglers is incorporated in the ROM estimator and alarger negative bias was observed. Additionally, the MOR esti-mator provided more accurate CIs at smaller sample sizes thanthe ROM estimator. In the absence of reporting bias, the MOR

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3


estimator will not provide an unbiased estimate of total catchwhen multiplied by an estimate of effort (Jones et al. 1995;McCormick et al. 2012). Using the ROM estimator in combina-tion with a more liberal definition of angling will likely providethe most unbiased estimate of total catch.

Although sample size may have been a limitation, the resultsof this study are based on the largest sample size of observedanglers in any published study. For instance, Edwards (1971)was based on a sample of 44 anglers, Radford (1973) was basedon a sample of 33 angling parties, and McEachron et al. (1986)was based on a sample of 94 interviews. The catch rates ofChinook Salmon in Idaho are relatively low compared withother fisheries where self-reporting bias studies have been con-ducted, which further limits the sample size of observations ofsuccessful anglers. Observation locations and times were se-lected to maximize observations of angling activity and wereassumed to be representative of the general angling population.This assumption was not explicitly evaluated. However, on manydays, angling activity was sparse and the same general trendsin self-reporting bias were observed on these days as on dayswhen more angling activity was taking place. Sample sizes inself-reporting studies are generally low due to the large amountof effort required to observe angling activity in fisheries sim-ilar to Chinook Salmon sport fisheries where angling activityis diffuse. Due to the conservation status of Chinook Salmonand their importance as a sport fishery, the implications of self-reporting bias in Chinook Salmon fisheries are arguably muchgreater than those in previous studies on this topic.

While previous research on self-reporting bias in other fish-eries has been mixed (Edwards 1971; Radford 1973; McEachronet al. 1986; Sullivan 2003), results of our study are somewhatdiscouraging. The underreporting of catch and overreporting ofthe amount of time fished produced negatively biased estimatesof total catch. Unlike other fisheries that may be managed basedon indices (e.g., mean catch rate or catch estimates over time),accurate estimates of total catch are necessary to properly man-age Chinook Salmon fisheries in Idaho. While overestimatingtotal catch can limit angling opportunity, underestimating totalcatch can potentially impact future fisheries. High variabilityin angler-reported data resulted in poor predictive performanceof regression models that could be used to account for self-reporting bias and suggests that correction factors may also beof little use. Nonetheless, we are encouraged that CIs in oursimulations encompassed the true population parameters neartargeted levels. This suggests that estimates of total catch may berobust to self-reporting bias if variability and CIs are consideredand fisheries are conservatively managed.

ACKNOWLEDGMENTSWe thank N. Porter, W. Field, J. Hansen, C. Smith, J. Walrath,

C. Watkins, S. Whitlock, J. Yates, and IDFG creel clerks for as-sistance with data collection; IDFG management biologists withexpertise on sampling locations; and F. Wilhelm, C. Williams, K.

Pope, and three anonymous reviewers for providing commentson earlier versions of this manuscript. Funding for this projectwas provided by the IDFG through Federal Aid in Sport FishRestoration. The Idaho Cooperative Fish and Wildlife ResearchUnit is jointly sponsored by the University of Idaho, U.S. Geo-logical Survey, IDFG, and Wildlife Management Institute. Theuse of trade, firm, or product names is for descriptive purposesonly and does not imply endorsement by the U.S. Government.

REFERENCESAkaike, H. 1973. Information theory and an extenstion of the maximum likeli-

hood principle. Pages 267–281 in B. N. Petrov and F. Csaki, editors. Secondinternational symposium on information theory. Akademiai Kiado, Budapest.

Apperson, K. A., and G. R. Wilson. 1998. 1997 Chinook Salmon fishery onthe South Fork Salmon River, Idaho. Idaho Department of Fish and Game,Report 98-32, Boise.

Applegate, J. E. 1984. Nongame tax check-off programs: a survey of New Jerseyresidents following the first year of contributions. Wildlife Society Bulletin12:122–128.

Brown, T. L., N. A. Connelly, and D. J. Decker. 1986. First-year results ofNew York’s “return a gift to wildlife”; tax checkoff. Wildlife Society Bulletin14:115–120.

Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodelinference: a practical information-theoretic approach, 2nd edition. Springer-Verlag, New York.

Carline, R. F. 1972. Biased harvest estimates from a postal survey of a sportfishery. Transactions of the American Fisheries Society 101:262–266.

Cochran, W. G. 1977. Sampling techniques, 3rd edition. Wiley, New York.Connelly, N. A., T. L. Brown, and B. A. Knuth. 2000. Assessing the relative

importance of recall bias and nonresponse bias and adjusting for those biasesin statewide angler surveys. Human Dimensions of Wildlife 5:19–29.

Edwards, G. B. 1971. Actual versus reported fishing times as a factor in creelcensus data. Progressive Fish-Culturist 33:54.

Efron, B., and G. Gong. 1983. A leisurely look at the bootstrap, the jackknife,and cross-validation. American Statistician 37:36–48.

Efron, B., and R. J. Tibshirani. 1993. An introduction to the bootstrap. Chapmanand Hall, New York.

Essig, R. J., and M. C. Holliday. 1991. Develpoment of a recreational fishingsurvey: the marine recreational fishery statistics survey case study. Pages245–254 in D. Guthrie, J. M. Hoenig, M. Holliday, C. M. Jones, M. J. Mills,S. A. Moberly, K. H. Pollock, and D. R. Talhelm, editors. Creel and anglersurveys in fisheries management. American Fisheries Society, Symposium12, Bethesda, Maryland.

Fisher, W. L., A. E. Grambsch, D. L. Eisenhower, and D. R. Morganstein.1991. Length of recall period and accurcy of estimates from the nationalsurvey of fishing, hunting, and wildlife-associated recreation. Pages 367–374 in D. Guthrie, J. M. Hoenig, M. Holliday, C. M. Jones, M. J. Mills,S. A. Moberly, K. H. Pollock, and D. R. Talhelm, editors. Creel and anglersurveys in fisheries management. American Fisheries Society, Symposium12, Bethesda, Maryland.

Fox, J. 2008. Applied regression analysis and generalized linear models, 2ndedition. Sage Publications, Thousand Oaks, California.

Hoenig, J. M., C. M. Jones, K. H. Pollock, D. S. Robson, and D. L. Wade.1997. Calculation of catch rate and total catch in roving surveys of anglers.Biometrics 53:306–317.

Huntsman, G. R., D. R. Colby, and R. L. Dixon. 1978. Measuring catches in theCarolina headboat fishery. Transactions of the American Fisheries Society107:241–245.

Johnson, M. W., and L. Wroblewski. 1962. Errors associated with a system-atic sampling creel census. Transactions of the American Fisheries Society91:201–207.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3


Jones, C. M., D. S. Robson, H. D. Lakkis, and J. Kressel. 1995. Properties ofcatch rates used in analysis of angler surveys. Transactions of the AmericanFisheries Society 124:911–928.

Mallison, C. T., and C. E. Cichra. 2004. Accuracy of angler-reported harvestin roving creel surveys. North American Journal of Fisheries Management24:880–889.

Malvestuto, S. P. 1983. Sampling the recreational fishery. Pages 397–419 inL. A. Nielsen and D. L. Johnson, editors. Fisheries techniques. AmericanFisheries Society, Bethesda, Maryland.

McCormick, J. L., M. C. Quist, and D. J. Schill. 2012. Effect of survey designand catch rate estimation on total catch estimates in Chinook Salmon fisheries.North American Journal of Fisheries Management 32:1090–1101.

McEachron, L. W., G. E. Saul, and S. J. Gutreuter. 1986. Accuracy of tripdurations reported by marine sport-boat fishermen. North American Journalof Fisheries Management 6:59–62.

Newman, S. P., P. W. Rasmussen, and L. M. Andrews. 1997. Comparison of astratified, instantaneous count creel survey with a complete mandatory creelcensus on Escanaba Lake, Wisconsin. North American Journal of FisheriesManagement 17:321–330.

Phippen, K. W., and E. P. Bergersen. 1987. Angling definitions and their effectson the accuracy of count–interview creel survey harvest estimates. NorthAmerican Journal of Fisheries Management 7:488–492.

Phippen, K. W., and E. P. Bergersen. 1991. Accuracy of a roving creel survey’sharvest estimate and evaluation of possible sources of bias. Pages 51–60in D. Guthrie, J. M. Hoenig, M. Holliday, C. M. Jones, M. J. Mills, S. A.Moberly, K. H. Pollock, and D. R. Talhelm, editors. Creel and angler surveys

in fisheries management. American Fisheries Society, Symposium 12,Bethesda, Maryland.

Pollock, K. H., C. M. Jones, and T. L. Brown. 1994. Angler survey methodsand their applications in fisheries management. American Fisheries Society,Special Publication 25, Bethesda, Maryland.

R Development Core Team. 2009. R: a language and environment for statisticalcomputing. R Foundation for Statistical Computing, Vienna.

Radford, D. S. 1973. A comparison between actual and reported angling times.Transactions of the American Fisheries Society 102:641–643.

Roach, B., J. Trial, and K. Boyle. 1999. Comparing 1994 angler catch andharvest rates from on-site and mail surveys on selected Maine lakes. NorthAmerican Journal of Fisheries Management 19:203–208.

Robson, D. S. 1961. On the statistical theory of a roving creel census of fisher-men. Biometrics 17:415–437.

Steffe, A. S., and J. J. Murphy. 2010. The importance of validating the dura-tion of fishing trips: are harvest rates from on-site recreational fishing sur-veys unbiased? North American Journal of Fisheries Management 30:506–510.

Sudman, S., and N. M. Bradburn. 1974. Response effects in surveys: a reviewand synthesis. Aldine Publishing, Chicago.

Sullivan, M. G. 2003. Exaggeration of Walleye catches by Alberta anglers.North American Journal of Fisheries Management 23:573–580.

WCCSBRT (West Coast Chinook Salmon Biological Review Team). 1997. Re-view of the status of Chinook Salmon (Oncorhynchus tshawytscha) fromWashington, Oregon, California, and Idaho under the U.S. EndangeredSpecies Act. NOAA Technical Memorandum NMFS-NWFSC-35.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:58

15

July

201

3



Precision Analysis of Three Aging Structures forAmphidromous Dolly Varden from Alaskan Arctic RiversJason T. Stolarski a & Trent M. Sutton ba Alaska Cooperative Fish and Wildlife Research Unit , University of Alaska–Fairbanks , PostOffice Box 757020, Fairbanks , Alaska , 99775 , USAb School of Fisheries and Ocean Science , University of Alaska–Fairbanks , 905 North KoyukukDrive, Fairbanks , Alaska , 99775 , USAPublished online: 22 Jul 2013.

To cite this article: Jason T. Stolarski & Trent M. Sutton (2013) Precision Analysis of Three Aging Structures for AmphidromousDolly Varden from Alaskan Arctic Rivers, North American Journal of Fisheries Management, 33:4, 732-740, DOI:10.1080/02755947.2013.806379







http://dx.doi.org/10.1080/02755947.2013.806379




MANAGEMENT BRIEF

Precision Analysis of Three Aging Structuresfor Amphidromous Dolly Varden from Alaskan Arctic Rivers

Jason T. Stolarski*Alaska Cooperative Fish and Wildlife Research Unit, University of Alaska–Fairbanks,Post Office Box 757020, Fairbanks, Alaska 99775, USA

Trent M. SuttonSchool of Fisheries and Ocean Science, University of Alaska–Fairbanks, 905 North Koyukuk Drive,Fairbanks, Alaska 99775, USA

AbstractThe accuracy of population statistics and the validity of manage-

ment actions they motivate are in part dependent on the acquisitionof quality age determinations. Such data for northern-form DollyVarden Salvelinus malma have been traditionally garnered usingotoliths, despite little research investigating the consistency of thisor alternative nonlethal techniques. To address these gaps, the pre-cision of age determinations generated from scales, otoliths, andfin rays was examined for 126 amphidromous Dolly Varden col-lected from two Arctic rivers. Three independent readers, age-biasplots, coefficients of variation (CVs), and percent agreement (PA)were used to estimate bias and precision for among-reader, within-structure comparisons and within-reader, among-structure com-parisons. Among-reader, within-structure tests of CVs suggestedthat otoliths produced more precise age determinations than finrays. Furthermore, the CV for scales was intermediate to and notsignificantly different from those for otoliths and fin rays. Age-bias plots suggested that, scales consistently underestimated agerelative to otoliths beginning at age 6. Underestimation was alsoapparent, but less distinct, within fin ray–otolith and scale–fin raycomparisons. Potential sources of error and management implica-tions are discussed. Because scale and otolith ages exhibited littlebias within cohorts younger than age 6, age may be determinednonlethally in these cohorts using scales; otoliths should be usedotherwise.

Northern-form Dolly Varden Salvelinus malma, hereafterreferred to as Dolly Varden, are distributed along the Arcticcoast of North America from the Mackenzie River in Canada,west and south through Alaska to the Seward Peninsula (Reistet al. 1997). Throughout their range, populations are largelyorganized by major river basin, which may contain both resi-

*Corresponding author: [email protected] April 27, 2012; accepted May 13, 2013

dent and sea-run individuals (McCart 1980; Everett et al. 1997).Amphidromous fish are generally larger and more abundant thanresident fish and support one of the largest and most importanttraditional subsistence fisheries in the Arctic coastal communi-ties of Alaska (McCart 1980; Pedersen and Linn 2005). Con-cerns regarding the potential ecological impacts of oil and gasexploration and climate change in the Arctic have strength-ened the need for sound management and monitoring practices(Hachmeister et al. 1991; Reist et al. 2006). The validity of thesepractices depends in part upon the acquisition of quality agedeterminations, as they are often developed using age-specificbiological data.

Northern fish species, such as Dolly Varden, are typicallyaged using calcified structures due to their longevity and slowrates of growth (McCart 1980; Howland et al. 2004). For a struc-ture to be useful for age determination, it must produce agesthat are both accurate (not addressed herein) and precise. DollyVarden age is almost exclusively estimated by using otoliths, ei-ther viewed whole or broken through the nucleus (Heiser 1966;Yoshihara 1973; Armstrong 1974; McCart 1980; Underwoodet al. 1995). Scale techniques have thus far been largely disre-garded because research within Arctic Char S. alpinus suggeststhat scale circuli patterns are unreliable predictors of fish age(Barber and McFarlane 1987; Baker and Timmons 1991); finray techniques have rarely been used (Heiser 1966; Barber andMcFarlane 1987). Nonlethal techniques using scales and fin raysconserve fish and allow age data to be collected from a greaterproportion of individuals within a population. This may be par-ticularly advantageous when aging Dolly Varden as the length

732

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:14

24

July

201

3


ranges of successive cohorts typically display considerable over-lap (Underwood et al. 1995). Such overlap can contribute toerror in age-specific biological statistics when extrapolating agedata from a subsample to a larger population, as is often thecase when using age–length keys. However, before nonlethaltechniques can be employed, the precision of age determinationgenerated from lethal (otoliths) and nonlethal (scales and finrays) techniques must be compared.

In the only study we could find that investigated the repro-ducibility of age determinations for Dolly Varden, Barber andMcFarlane (1987) concluded that otoliths generally producedolder age determinations relative to those based on pectoraland anal fin rays. However, this research did not use scales todetermine age; used only a single reader, which limited anal-yses to comparisons among structures; and was conducted onpooled samples containing both Dolly Varden and Arctic Char(Reist et al. 1997). As a result, the precision of Dolly Var-den aging techniques both within and among structures remainspoorly defined. The objective of the present study was to ad-dress this data gap by estimating the precision of scale, otolith,and fin ray age determinations for within- and among-structurecomparisons.

STUDY SITEThis study was conducted at spawning and overwintering

habitats of amphidromous Dolly Varden on the Ivishak andHulahula rivers, located on the coastal plain of the AlaskanArctic (Figure 1; Daum et al. 1984; Viavant 2005). The IvishakRiver is a north-flowing tributary of the Sagavanirktok River,the second largest basin on the North Slope of Alaska. Both

FIGURE 1. Map of the Eastern North Slope of Alaska with arrows indicatingthe general locations where postsmolt (A) and presmolt (B) Dolly Varden weresampled from the Ivishak and Hulahula rivers, respectively.

rivers originate in the Brooks Mountain range and drain into theBeaufort Sea: the Sagavanirktok River at Prudhoe Bay and theHulahula River near the coastal community of Kaktovik. Bothrivers contain resident and amphidromous populations of DollyVarden.

METHODSFish sampling.—Postsmolt Dolly Varden were captured via

angling from the Ivishak River during sampling events in earlySeptember of 2009, 2010, and 2011. Presmolt fish were col-lected using minnow traps from the Hulahula River duringAugust 2011. Sampling exclusively within habitats known tobe frequented by large numbers of amphidromous fish mini-mized the likelihood of capture and inclusion of resident fishinto the study. Upon capture, individuals were killed, weighedto the nearest 1 g, and measured for fork length to the nearest1 mm. Each fish was individually labeled, wrapped in plas-tic, and transported to University of Alaska Fairbanks, wherethey were frozen. In the laboratory, scales were sampled witha scalpel from an area posterior to the dorsal fin and above thelateral line and stored on waterproof paper (DeVries and Frie1996). The right pectoral fin was removed from each fish, rinsedin water, and stored on waterproof paper in a well-ventilatedarea to facilitate drying. Sagittal otoliths were removed us-ing the “open hatch” method of Secor et al. (1992), rinsedin water, dried, and stored dry in individually labeled plasticvials.

Structure preparation.—Fifteen to 20 scales from each fishwere wet-mounted on a glass slide and viewed with a compoundmicroscope under transmitted light at 40× magnification. Af-ter the subsample was screened for the presence of regeneratedscales, an image of a representative scale was captured usinga microscope-mounted 3.3 megapixel digital camera (Quantita-tive Imaging Co., Burnaby, British Columbia).

Fin rays were embedded in Epoxycure epoxy resin (Buehler,Lake Bluff, Illinois) following methods outlined in Koch andQuist (2007). Multiple transverse sections, each 0.5–0.75 mmthick, were cut using an Isomet low-speed saw (Buehler),equipped with a 102-mm-diameter diamond wafering blade ro-tating at 240 revolutions per minute. Care was taken to assurethat the first thin section encompassed or was slightly posteriorto the inflection point of the ray (Beamish 1981). Sections wereaffixed to a glass slide using Crystalbond thermoplastic cement(Structure Probe, Inc., West Chester, Pennsylvania) and viewedwith a compound microscope. A digital image was captured ofa representative fin ray at 20× and 40× magnifications undertransmitted light.

The right sagittal otolith of each fish was affixed to a glassslide using Crystalbond thermoplastic cement perpendicular tothe long axis of the otolith. Each otolith was ground to the corein the transverse plane using a thin-section machine (Hillquist,Inc., Denver, Colarado) and remounted to the slide flat side

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:14

24

July

201

3

734 STOLARSKI AND SUTTON

down before being ground to a final thickness of approximately0.3 mm. The otolith was hand-polished with a 1-µm diamondabrasive and viewed with a compound microscope under trans-mitted light. Digital images were captured at 20× and 40×magnifications. If the mounted otolith section was deemed in-adequate for age determination, the left sagitta was processedin the same fashion.

Age determination.—Age determinations were produced bythree independent readers trained in annulus identification. Eachreader estimated fish age from scales, otoliths, and fin rayswith knowledge of the capture date of the fish, but not thefish length. Scale annuli were identified as areas of greater cir-culi density or when successive circuli cut over each other.Annuli in fin ray and otolith sections were identified as alter-nating opaque and hyaline zones (DeVries and Frie 1996). Finray–based age estimates were derived from the first or secondray. Images were organized into separate libraries by structure,and no reader was allowed to determine age from multiple li-braries within a single day. To our knowledge, scale, otolith,and fin ray age determinations for Dolly Varden have yet to bevalidated.

Statistical analysis.—Age-bias plots were used to as-sess among-reader, within-structure and within-reader, among-structure bias (Campana et al. 1995). Age-bias plots depict themean age of fish determined by one reader that have been as-signed a given age by a second reader. Cohorts displaying com-plete agreement among ages assigned by each reader will fallon the 1:1 line of equivalence. Thus, bias is detected visually aspersistent (lasting more than two successive years) deviationsof 95% confidence intervals surrounding each mean from theline of equivalence. Detection of among-reader, within-structurebias is important as it indicates whether readers are using unifiedcriteria to identify and count annuli. If one or multiple readersconsistently over- or underestimate age relative to other readers,precision within that structure will reflect variability in agingmethods rather than the reproducibility of age determinations.If bias is detected, the criteria by which annuli are identifiedand counted must be revisited and agreed upon by readers, andages must be redetermined. Alternatively, within-reader, among-structure bias can be used to evaluate the relative strength ofany under- or overestimation of ages between the techniques.Precision of among-reader, within-structure and within-reader,among structure comparisons was estimated with percent agree-ment (PA), percent agreement to within 1 year (PA1), and co-efficient of variation (CV). Percent agreement statistics werecalculated as the number of pairwise comparisons in which agewas in total agreement (in the case of PA) or the number ofpairwise comparisons in which age was in agreement to within1 year (in the case of PA1), divided by the total number ofcomparisons made. Percent agreement statistics, once the pre-dominant means of accessing the precision of aging structuresare slowly being replaced with statistics such as CV, as the lat-ter measures do not account for age structure variation betweenspecies (Beamish and Fournier 1981). As such, these statistics

are mentioned only briefly and included primarily as a meansof comparison with past research. Coefficient of variation wascalculated as

CV j = 100 ·√∑R

i=1(Xi j −X j )2

R−1

X j,

where Xi j is the ith age determination for the jth fish, Xj isthe mean age of the jth fish, and R is the number of timesthe age of the fish is estimated (Chang 1982). Coefficient ofvariation was averaged across all fish for each structure inthe case of among-reader, within-structure comparisons andacross specific comparisons (i.e., scales versus fin rays) forwithin-reader, among-structure comparisons. Potential differ-ences among structures and comparisons were tested using anal-ysis of variance (ANOVA) with a post hoc Tukey’s honestlysignificant difference test when significant differences were de-tected. All statistical analyses were conducted using the statis-tical software package R (R development Core Team 2012) andevaluated at an α = 0.05.

RESULTSOf the total 143 pre- and postsmolt Dolly Varden collected

over the 3 years of sampling, 126 were included into the finalanalyses. Individuals ranged in fork length from 63 to 680 mmand encompassed ages 0–14 years (Figure 2). Annuli were iden-tified from digital images for all three structures (Figure 3). Scalecirculi patterns varied substantially between the freshwater andmarine periods of growth, with the marine phase exhibiting fargreater spacing between successive circuli. Visual examinationof among-reader, within-structure age-bias plots showed littlepersistent (i.e., did not last more than two consecutive years)deviation from the 1:1 equivalence line, indicating that readersused similar standards in identifying and counting annuli (Fig-ure 4). Mean PA and PA1 of among-reader, within-structurecomparisons were similar among structures (Table 1). How-ever, mean PA did not exceed 55% for any structure, whilemean PA1 exceeded 90% for all structures (Table 1). Age-bias plots of within-reader, among-structure comparisons in-dicated that scales began to underestimate fish age relative tootoliths beginning at age 6 (Figure 5); moreover, errors in-creased with age. Age-bias plots also indicated that fin raystended to underestimate age relative to otoliths and that scalestended to underestimate age relative to fin rays. For both ofthese comparisons, underestimation generally began at age 6and remained somewhat constant as age increased (Figure 5).However, trends within scale–fin ray and otolith–fin ray com-parisons were less distinct relative to scale–otolith comparisons.Mean PA and PA1 of within-reader, among-structure compar-isons were similar among comparisons but were generally lowerthan among-reader, within-structure estimates (Table 1). No dif-ferences in CV were detected for within-reader, among-structure

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:14

24

July

201

3


TABLE 1. Coefficient of variation (CV), mean percent agreement (PA), and mean percent agreement to within 1 year (PA1) of among-reader, within-structureand within-reader, among-structure comparisons of age determinations based on scales, otoliths, and fin rays for Dolly Varden sampled from the Ivishak andHulahula rivers. Within each comparison type, CV estimates with different lowercase letters are significantly different (P < 0.05) following ANOVA with a posthoc Tukey’s test.

Comparison type Structures CV Mean PA Mean PA1

Among-reader, within-structure Scales 9.08 zy 55.91 94.35Otoliths 7.91 y 55.02 94.18Fin rays 11.91 z 52.38 94.18

Within-reader, among-structure Scale–otolith 14.28 z 33.87 81.18Scale–fin ray 14.11 z 40.05 81.74Otolith–fin ray 13.59 z 33.07 83.33

FIGURE 2. Composite length (A) and age (B) data plotted as a function of sample proportion for Dolly Varden collected from the Ivishak and Hulahula riversbetween 2009 and 2011. Age data in B were derived from otoliths.

FIGURE 3. Digital image of an amphidromous, northern-form Dolly Varden scale (A), otolith (B), and fin ray (C) collected from the Ivishak River, Alaska. Eachstructure depicts four annuli. Note the “cutting over” of scale circuli at labeled annuli and the proximity of the first and second annuli on the fin ray.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:14

24

July

201

3


FIGURE 4. Age-bias plots with pairwise estimates of coefficient of variation (CV), percent agreement (PA), and percent agreement to within 1 year (PA1) foramong-reader, within-structure comparisons of scale, otolith, and fin ray age determinations for Dolly Varden collected from the Ivishak and Hulahula rivers. Errorbars represent 95% confidence intervals (for points with multiple observations) around the mean age assigned by one reader relative to all fish assigned a givenage by a second reader (Campana et al. 1995).

comparisons (Table 1; F2, 373 = 0.347, P = 0.707). Coefficientof variation of among-reader, within-structure comparisons dif-fered among structures (Table 1; F2, 373 = 3.143, P = 0.044).The post hoc Tukey test indicated that otoliths were more precisepredictors of Dolly Varden age than fin rays were.

DISCUSSIONThis research contributes to a growing body of literature in-

dicating that scales typically underestimate fish age relative tootoliths (Sikstrom 1983; Hubert et al. 1987; Sharp and Bernard1988; Graynoth 1996; Kruse et al. 1997; DeCicco and Brown2006; Stolarski and Hartman 2008). The onset of scale under-estimation corresponded well with estimates of the age at first

reproduction for Dolly Varden, suggesting that underestimationwas a result of ontogenetic reductions in growth and the for-mation of a “dense edge” on the scale margins (Nordeng 1961;Yoshihara 1973; Craig and Halderson 1981). A similar artifactwas often present in the interior of the scale, probably a resultof slow presmolt growth while in freshwater (McCart 1980).These features highlight the importance of training readers inidentification of both freshwater and marine annulus becausescale morphology and annuli appearance will change depend-ing upon the growth rate of the fish (Carlander 1974).

Within-reader comparisons of fin ray and otolith age deter-minations suggested that fin rays underestimated age relative tootoliths, beginning at age 6. Barber and McFarlane (1987) notedsimilar results studying age determinations from a mixed sample

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:14

24

July

201

3


FIGURE 5. Age-bias plots with pairwise estimates of coefficient of variation (CV), percent agreement (PA), and percent agreement to within 1 year (PA1) forwithin-reader, among-structure comparisons of scale, otolith, and fin ray age determinations for Dolly Varden collected from the Ivishak and Hulahula rivers. Errorbars represent 95% confidence intervals (for points with multiple observations) around the mean age assigned by one reader relative to all fish assigned a givenage by a second reader (Campana et al. 1995).

of Dolly Varden and Arctic Char. Fin ray underestimation hasalso been reported in Arctic Grayling Thymallus arcticus (Sik-strom 1983), Rainbow Trout Oncorhynchus mykiss and BrownTrout Salmo trutta (Graynoth 1996), Cutthroat Trout O. clarkii(Hubert et al. 1987), and Brook Trout Salvelinus fontinalis (Sto-larski and Hartman 2008). However, Zymonas and McMahon(2009) reported no bias between comparisons of ages derivedfrom pelvic fin rays and those derived from otoliths of BullTrout Salvelinus confluentus. Chronic misidentification of thefirst few annuli in fin ray sections is commonly cited as a po-tential cause of underestimation (Sikstrom 1983; Hubert et al.1987). Working with a population of slow-growing White Suck-ers Catostomus commersonii, Beamish (1973) noted that thefirst fin ray annulus was often too closely associated with the

ray center to be consistently identified, particularly within olderfish. For Dolly Varden, interior annuli clarity was often dimin-ished within older fish and declined as the location where the finray section was cut moved further from the inflection point ofthe ray. The effect of the latter phenomenon was minimized byderiving ages from one of the first three fin ray sections of theseries. However, evidence of more constant errors within olderfish suggests that misidentification of freshwater annuli couldbe occurring within these cohorts.

Within-reader comparisons of scale and fin ray age deter-minations suggested that scales often underestimated fish age,again beginning at age 6. However, this relationship was lesspronounced relative to scale–otolith and fin ray–otolith compar-isons. Previous research involving similar comparisons has been

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:14

24

July

201

3


generally inconsistent, with some studies confirming (Sikstrom1983; Stolarski and Hartman 2008) and others refuting (Hubertet al. 1987; Copeland et al. 2007) our results. Given suspectedsources of error within each of the two structures (see above),the lack of a consistent trend could be a function of the pro-portion of instances in which reader errors are isolated withina single structure versus when errors occur simultaneously inboth.

Percent agreement of among-reader, within-structure com-parisons was generally low compared with previous researchreports (Graynoth 1996; Zymonas and McMahon 2009). How-ever, such interspecies comparisons are made difficult by the factthat PA varies as a function of the age structure of the species inquestion (Beamish and Fournier 1981). The low PA seen herecould be a result of the age structure of the sample (Beamishand Fournier 1981; Zymonas and McMahon 2009), the use ofmultiple readers instead of multiple reads by the same reader(Ihde and Chittenden 2002), the relatively slow growth rates ofhigh-latitude fishes (Sikstrom 1983), or some combination ofthe three. Percent agreement of among-reader, within-structurecomparisons of scales, otoliths, and fin rays for species with sim-ilar age structures such as Arctic Grayling and Bull Trout rangebetween 50% and 67% and are comparable with PA observed inDolly Varden (Sikstrom 1983; Zymonas and McMahon 2009).Despite these contentions, PA1 was greater than 90% for allstructures, suggesting that gross disagreements in Dolly Vardenage were infrequent.

Tests of among-reader, within-structure CV suggested thatotoliths were more precise estimators of Dolly Varden age thanfin rays. Similar results have been reported in other speciesand are likely a result of misidentification of interior annuli, aspreviously discussed (Graynoth 1996; Stolarski and Hartman2008; Zymonas and McMahon 2009). However, the mean CVof age determinations garnered from scales and otoliths werefound to be similar, which is contrary to the findings of numer-ous studies reporting that, among all samples, scale-based agedeterminations are less precise indicators of fish age than areotoliths (Sikstrom 1983; Kruse et al. 1997; DeCicco and Brown2006; Zymonas and McMahon 2009; Schill et al. 2010). Thisfinding may be a direct result of the intensity and duration inwhich Dolly Varden grow throughout the year. Dolly Vardenacquire nearly 100% of their annual energy budget during theshort Arctic summers, spending the remainder of the year over-wintering in freshwater, where little to no food is consumed(Craig 1984; Boivin and Power 1990). Before reaching repro-ductive age, the intensity of growth within these periods and theconsistency with which they occur probably contribute to thedistinctiveness of scale annuli. The annulus formed followingthe first migration to the sea is particularly distinguishable dueto the contrast between it and the adjacent freshwater annuli.The consistency of annulus formation in scales stemming fromthis seasonal pattern probably rivals that of otoliths over thesame time interval and contributes to the similarity in precisionobserved between those two structures. However, if our sample

had contained a larger proportion of older fish, our results mighthave differed.

This research suggests that Dolly Varden age may be deter-mined nonlethally using scales within individuals age 5 andyounger. Our assertion is a result of our findings that biasof within-reader, among-structure comparisons of scales andotoliths is minimal within cohorts younger than age 6. Further-more, no statistical differences in CV calculated from among-reader, within-structure comparisons of scale and otolith agedeterminations were detected. However, beyond age 5, otolithsshould be used to generate age determinations for Dolly Var-den. The majority of Dolly Varden research and monitoringprojects have been conducted within nearshore coastal areasusing fyke nets. These catch data suggest that Dolly Vardendistribute themselves along the shore according to size, withthe smaller individuals occupying shallower habitats closer toshore (Craig and Halderson 1981; Hachmeister et al. 1991; Un-derwood et al. 1995; Fechhelm et al. 1997; Brown 2008). Agedata collected from random subsets of this catch indicate thatup to 70% of the individuals are younger than age 6 (Under-wood et al. 1995). While age composition probably varies overtime and space, it is reasonable to assume that many Dolly Var-den captured in nearshore fyke nets can be aged nonlethallyusing scales. Scale-based age determination may be particularlyvaluable for identifying first year smolts. This demographic isoften pooled for analysis purposes and can be easily and quicklyidentified using scales because of the contrast between freshwa-ter and marine circuli patterns (Fechhelm et al. 1997). Smoltidentification has been previously accomplished using graph-ical methods; however, these techniques are not as successfulwhen sampling locations are distant from river mouths (Fech-helm et al. 1997; Brown 2008). Nonlethal age determinationwill also allow age data to be collected from a greater pro-portion of the population, which may increase the precision ofage-specific statistics. However, it is always important to in-dependently verify the consistency of scale- and otolith-basedage determinations in the field within a subset of fish prior toimplementation of a particular technique.

ACKNOWLEDGMENTSWe thank William Carter, David Daum, and Dave Sowards

for field and logistical help and assisting with the age determi-nations. Scientific sampling was conducted under the authorityof Alaska Department of Fish and Game Fishery Resource Per-mit SF2011-208 and SF2011-046. Funding for this study wasprovided by the U.S. Fish and Wildlife Service, Arctic NationalWildlife Refuge, and the Fairbanks Fisheries Resource Office.The use of trade names of commercial products in this reportdoes not constitute endorsement or recommendation for use.Care and handling of all fish included in this study was in ac-cordance with approved protocols of the University of AlaskaFairbanks Institutional Animal Care and Use Committee Assur-ance 175440-3.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:14

24

July

201

3


REFERENCESArmstrong, R. H. 1974. Migration of anadromous Dolly Varden (Salvelinus

malma) in southeastern Alaska. Journal of the Fisheries Research Board ofCanada 31:435–444.

Baker, T. T., and L. S. Timmons. 1991. Precision of ages estimated from fivebony structures of Arctic Char (Salvelinus alpinus) from the Wood Riversystem, Alaska. Canadian Journal of Fisheries and Aquatic Sciences 48:1007–1014.

Barber, W. E., and G. A. McFarlane. 1987. Evaluation of three techniques to ageArctic Char from Alaskan and Canadian waters. Transactions of the AmericanFisheries Society 116:874–881.

Beamish, R. J. 1973. Determination of age and growth of populations of theWhite Sucker (Catostomus commersoni) exhibiting a wide range in size atmaturity. Journal of the Fisheries Research Board of Canada 30:607–616.

Beamish, R. J. 1981. Use of fin-ray sections to age Walleye Pollock, PacificCod, and Albacore, and the importance of this method. Transactions of theAmerican Fisheries Society 110:287–299.

Beamish, R. J., and D. A. Fournier. 1981. A method for comparing the precisionof a set of age determinations. Canadian Journal of Fisheries and AquaticSciences 38:982–983.

Boivin, T. G., and G. Power. 1990. Winter condition and proximate compositionof anadromous Arctic Charr (Salvelinus alpinus) in eastern Ungava Bay,Quebec. Canadian Journal of Zoology 68:2284–2289.

Brown, R. J. 2008. Life history and demographic characteristics of Arctic Cisco,Dolly Varden, and other fish species in the Barter Island region of northernAlaska. U.S. Fish and Wildlife Service, Alaska Fisheries Technical Report101, Fairbanks.

Campana, S. E., M. C. Annand, and J. I. McMillan. 1995. Graphical and statis-tical methods for determining the consistency of age determinations. Trans-actions of the American Fisheries Society 124:131–138.

Carlander, K. D. 1974. Difficulties in ageing fish in relation to inland fisherymanagement. Pages 200–205 in T. B. Bagenal, editor. The proceedings ofan international symposium on the ageing of fish. Unwin Brothers, Surrey,UK.

Chang, W. Y. B. 1982. A statistical method for evaluating the reproducibilityof age determination. Canadian Journal of Fisheries and Aquatic Sciences39:1208–1210.

Copeland, T., M. W. Hyatt, and J. Johnson. 2007. Comparison of methods usedto age spring–summer Chinook Salmon in Idaho: validation and simulatedeffects on estimated age composition. North American Journal of FisheriesManagement 27:1393–1401.

Craig, P. C. 1984. Fish use of coastal waters of the Alaskan Beaufort Sea: areview. Transactions of the American Fisheries Society 113:265–282.

Craig, P. C., and L. Halderson. 1981. Beaufort Sea barrier island-lagoon ecolog-ical process studies: final report, Simpson Lagoon, part 4, fish. Pages 384–678in S. R. Johnson and J. W. Richardson, editors. Environmental assessmentof the Alaskan Continental Shelf, final reports of principal investigators, vol-ume 8: biological studies. National Oceanic and Atmospheric Administra-tion, Outer Continental Shelf Environmental Assessment Program, Boulder,Colorado.

Daum, D., P. Rost, and M. W. Smith. 1984. Fisheries studies on the north slope ofthe Arctic National Wildlife Refuge, 1983. Pages 464–522 in G. W. Garnerand P. E. Reynolds, editors. Arctic National Wildlife Refuge coastal plainresource assessment, 1983 update report, baseline study of the fish, wildlife,and their habitats. U.S. Fish and Wildlife Service, Anchorage, Alaska.

DeCicco, A. L., and R. J. Brown. 2006. Direct validation of annual growthincrements on sectioned otoliths from adult Arctic Grayling and a comparisonof otolith and scale ages. North American Journal of Fisheries Management26:580–586.

DeVries, D. R., and R. V. Frie. 1996. Determination of age and growth. Pages483–512 in B. R. Murphy and D. W. Willis, editors. Fisheries techniques, 2ndedition. American Fisheries Society, Bethesda, Maryland.

Everett, R. J., R. L. Wilmot, and C. C. Krueger. 1997. Population geneticstructure of Dolly Varden from Beaufort Sea drainages of northern Alaska and

Canada. Pages 240–249 in J. B. Reynolds, editor. Fish ecology in Arctic NorthAmerica. American Fisheries Society, Symposium 19, Bethesda, Maryland.

Fechhelm, R. G., J. D. Bryan, W. B. Griffiths, and L. R. Martin. 1997. Summergrowth patterns of northern Dolly Varden (Salvelinus malma) smolts fromthe Prudhoe Bay region of Alaska. Canadian Journal of Fisheries and AquaticSciences 54:1103–1110.

Graynoth, E. 1996. Determination of the age of Brown and Rainbow trout in arange of New Zealand lakes. Marine and Freshwater Research 47:749–756.

Hachmeister, L. E., D. R. Glass, and T. C. Cannon. 1991. Effects of solid-fillgravel causeways on the coastal central Beaufort Sea environment. Pages81–96 in C. S. Benner and R. W. Middleton, editors. Fisheries and oil devel-opment on the continental shelf. American Fisheries Society, Symposium 11,Bethesda, Maryland.

Heiser, D. W. 1966. Age and growth of anadromous Dolly Varden Char, Salveli-nus malma (Walbaum) in Eva Creek, Baranof Island, southeastern Alaska.Alaska Department of Fish and Game, Federal Aid in Fish Restoration, ProjectF-5-R-4(5), Juneau.

Howland, K. L., M. Gendron, W. M. Tonn, and R. F. Tallman. 2004. Agedetermination of a long-lived coregonid from the Canadian north: comparisonof otoliths, fin rays and scales in Inconnu (Stenodus leucichthys). AnnalesZoologici Fennici 41:205–214.

Hubert, W. A., G. T. Baxter, and M. Harrington. 1987. Comparison of agedeterminations based on scales, otoliths and fin rays for Cutthroat Trout fromYellowstone Lake. Northwest Science 61:32–36.

Ihde, T. F., and M. E. Chittenden Jr. 2002. Comparison of calcified structuresfor aging Spotted Seatrout. Transactions of the American Fisheries Society131:634–642.

Koch, J. D., and M. C. Quist. 2007. A technique for preparing fin rays andspines for age and growth analysis. North American Journal of FisheriesManagement 27:782–784.

Kruse, C. G., W. A. Hubert, and F. J. Rahel. 1997. Using otoliths and scales todescribe age and growth of Yellowstone Cutthroat Trout in a high-elevationstream system, Wyoming. Northwest Science 71:30–38.

McCart, P. J. 1980. A review of the systematics and ecology of Arctic Char,Salvelinus alpinus, in the western Arctic. Canadian Technical Report of Fish-eries and Aquatic Sciences 935.

Nordeng, H. 1961. On the biology of char (Salmo alpinus L.) in Salangen, northNorway: I. age and spawning frequency determined from scales and otoliths.Nytt Magasin for Zoologi (Oslo) 10:67–123.

Pedersen, S., and A. Linn Jr. 2005. Kaktovik 2000–2002 subsistence fisheryharvest assessment. U.S. Fish and Wildlife Service, Office of SubsistenceManagement, Fishery Information Services Division, Federal SubsistenceFishery Monitoring Program, Final Project Report FIS 01-101, Anchorage,Alaska.

R Development Core Team. 2012. R: a language and environment for statisti-cal computing. R Foundation for Statistical Computing, Vienna. Available:www.R-project.org. (April 2012).

Reist, J. D., J. D. Johnson, and T. J. Carmichael. 1997. Variation and specificidentity of char from northwestern Arctic Canada and Alaska. Pages 250–261in J. B. Reynolds, editor. Fish ecology in Arctic North America. AmericanFisheries Society, Symposium 19, Bethesda, Maryland.

Reist, J. D., F. J. Wrona, T. D. Prowse, M. Power, J. B. Dempson, R. J. Beamish,J. R. King, T. J. Carmichael, and C. D. Sawatzky. 2006. General effectsof climate change on Arctic fishes and fish populations. Ambio 35:370–380.

Schill, D. J., E. R. J. M. Mamer, and G. W. LaBar. 2010. Validation of scalesand otoliths for estimating age of Redband Trout in high desert streams ofIdaho. Environmental Biology of Fishes 89:319–332.

Secor, D. H., J. M. Dean, and E. H. Laban. 1992. Otolith removal and preparationfor microstructural examination. Canadian Special Publication of Fisheriesand Aquatic Sciences 117:19–57.

Sharp, D., and D. R. Bernard. 1988. Precision of estimated ages of Lake Troutfrom five calcified structures. North American Journal of Fisheries Manage-ment 8:367–372.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:14

24

July

201

3


Sikstrom, C. B. 1983. Otolith, pectoral fin ray, and scale age determinations forArctic Grayling. Progressive Fish-Culturist 45:220–223.

Stolarski, J. T., and K. J. Hartman. 2008. An evaluation of the precision of finray, otolith, and scale age determinations for Brook Trout. North AmericanJournal of Fisheries Management 28:1790–1795.

Underwood, T. J., J. A. Gordon, M. J. Millard, L. A. Thorpe, and B. M.Osborne. 1995. Characteristics of selected fish populations of the ArcticNational Wildlife Refuge coastal waters, final report, 1988–1991. U.S. Fishand Wildlife Service, Alaska Fisheries Technical Report 28, Fairbanks.

Viavant, T. 2005. Eastern North Slope Dolly Varden stock assessment. AlaskaDepartment of Fish and Game, Fishery Data Series 05-07, Anchorage.

Yoshihara, H. T. 1973. Monitoring and evaluation of Arctic waters with emphasison the North Slope drainages. Alaska Department of Fish and Game, Divisionof Sport Fish, Federal Aid in Fish Restoration, Project F-9-5, Annual ProgressReport 1972–1973, Juneau.

Zymonas, N. D., and T. E. McMahon. 2009. Comparison of pelvic fin rays,scales and otoliths for estimating age and growth of Bull Trout, Salvelinusconfluentus. Fisheries Management and Ecology 16:155–164.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:14

24

July

201

3



Use of Stable Isotopes to Identify Redds of PutativeHatchery and Wild Atlantic Salmon and Evaluate TheirSpawning Habitat and Egg Thiamine Status in a LakeOntario TributaryJohn D. Fitzsimons a , Alex Dalton a , Brydon MacVeigh a , Mark Heaton b , Chris Wilson b &Dale C. Honeyfield ca Fisheries and Oceans Canada , 867 Lakeshore Road, Burlington , Ontario , L7R 4A6 , Canadab Ontario Ministry of Natural Resources, Aquatic Ecosystems Science Section , TrentUniversity , 300 Water Street, Peterborough , Ontario , K9J 8N8 , Canadac U.S. Geological Survey , 176 Straight Run Road, Wellsboro , Pennsylvania , 16901 , USAPublished online: 24 Jul 2013.

To cite this article: John D. Fitzsimons , Alex Dalton , Brydon MacVeigh , Mark Heaton , Chris Wilson & Dale C. Honeyfield(2013) Use of Stable Isotopes to Identify Redds of Putative Hatchery and Wild Atlantic Salmon and Evaluate Their SpawningHabitat and Egg Thiamine Status in a Lake Ontario Tributary, North American Journal of Fisheries Management, 33:4, 741-753,DOI: 10.1080/02755947.2013.806380







http://dx.doi.org/10.1080/02755947.2013.806380




ARTICLE

Use of Stable Isotopes to Identify Redds of PutativeHatchery and Wild Atlantic Salmon and Evaluate TheirSpawning Habitat and Egg Thiamine Status in a LakeOntario Tributary

John D. Fitzsimons,* Alex Dalton, and Brydon MacVeighFisheries and Oceans Canada, 867 Lakeshore Road, Burlington, Ontario L7R 4A6, Canada

Mark Heaton and Chris WilsonOntario Ministry of Natural Resources, Aquatic Ecosystems Science Section, Trent University,300 Water Street, Peterborough, Ontario K9J 8N8, Canada

Dale C. HoneyfieldU.S. Geological Survey, 176 Straight Run Road, Wellsboro, Pennsylvania 16901, USA

AbstractBoth wild and hatchery Atlantic Salmon Salmo salar can contribute to restoration but can exhibit differences

in spawning habitat selection (e.g., water depth, current speed) and egg nutritional quality (e.g., thiamine), whichaffect reproductive success. Hence, there is a need to be able to differentiate the spawning contribution of the twogroups in the wild. As diets of wild and hatchery-reared spawners are markedly dissimilar and diet is known toinfluence stable isotope signature, egg stable isotope signatures offer the potential to discriminate redds of each.Using stable isotope analysis of carbon and nitrogen (δ13C and δ 15N) of naturally spawned Atlantic Salmon eggs,we were able to discriminate the redds of putative wild (i.e., previously stocked life stage feeding in Lake Ontario;EWSR) from putative hatchery-reared Atlantic Salmon (EHSR). Eggs of EWSR were significantly more enriched innitrogen (δ 15N: 15.0 ± 0.5‰ [mean ± SE]) but more depleted in carbon (δ13C: −26.6 ± 0.3‰) than eggs of EHSR(δ 15N = 9.8 ± 0.6‰; δ13C = −17.5 ± 0.0‰). Eggs of EHSR were indistinguishable from eggs of known hatcheryAtlantic Salmon for both δ13C and δ 15N. Using stable isotopes to discriminate redd type, few differences were foundbetween the spawning habitat of putative wild and hatchery spawners. Similarly using the same criteria, thiaminelevels in eggs of EWSR (8,474 ± 840 pmol/g) were not significantly different from eggs of known wild Atlantic Salmon(3,691 ± 782 pmol/g) or of eggs of EHSR (14,865 ± 1,050 pmol/g), whose thiamine levels were indistinguishablefrom eggs of known hatchery Atlantic Salmon (14,200 ± 1,167 pmol/g). Egg thiamine levels for all groups wereabove established mortality thresholds. Our results indicate that both hatchery and wild Atlantic Salmon can makereproductive contributions, which can be differentiated and quantified using stable isotope signatures.

Historically Lake Ontario may have supported the world’slargest freshwater population of Atlantic Salmon Salmo salar(Webster 1982). By the late 1800s however, Atlantic Salmonwere virtually extinct in the lake due to damming of tribu-taries that blocked spawning migration, overharvest, deforesta-tion, and pollution (Netboy 1968; Parsons 1973; MacCrimmon

*Corresponding author: [email protected] August 24, 2012; accepted May 15, 2013

1977; Smith 1995). Attempts to reintroduce Atlantic Salmonto the Lake Ontario drainage through the 1900s had negligi-ble success (Crawford 2001), possibly due in part to ongoingeffects of an Alewife Alosa pseudoharengus diet-induced thi-amine deficiency (Ketola et al. 2000) to which Atlantic Salmonreproduction may be very sensitive (Fisher et al. 1996).

741

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3

742 FITZSIMONS ET AL.

Alewives contain high levels of thiaminase activity, which ispresumed to result in thiamine destruction in the gut of a preda-tor as a postmortem change (Tillitt et al. 2005). As a result,a diet dominated by Alewives can lead to thiamine deficiency.In Great Lakes salmonines other than Atlantic Salmon, dietsdominated by Alewives are associated with low egg thiamine(Fitzsimons and Brown 1998; Honeyfield et al. 2005), whilethe thiamine content of Atlantic Salmon eggs from the FingerLakes (Fisher et al. 1998a, 1998b) and Baltic Sea (Amcoff et al.1999; Borjeson et al. 1999) has been related to the thiaminenutrition of the maternal parent. The salmonine egg receivesall of its thiamine, an essential nutrient, from the maternal par-ent (Koski et al. 2005), such that the thiamine content of asalmonine egg will reflect the diet of the maternal parent. Thethiamine content of the diet of hatchery-reared salmonids tendsto be much higher than that of diets of wild salmonids, resultingin much higher egg thiamine concentrations (Fitzsimons et al.1998; Fisher et al. 1998b). Because of these differences, Fisheret al. (1998a) found that thiamine levels in Atlantic Salmonfed a hatchery diet were five-fold higher than those feeding onRainbow Smelt Osmerus mordax and from approximately 8- to80-fold higher than those feeding on Alewife. The problems ofmaternal thiamine nutrition associated with Alewife consump-tion may be further exacerbated in Atlantic Salmon because theybecome anorexic before entering the spawning river (Kadri et al.1995), which may occur several months prior to spawning (J.Kendell, Credit River Anglers Association, personal communi-cation), and do not feed during the several months they spendin the river (Grey and Tosh 1894 cited in Kadri et al. 1995). Theprolonged period of reduced thiamine intake during this timewould worsen a preexisting thiamine deficiency causing lowerthan expected thiamine levels in eggs leading to higher thanexpected embryonic mortality (Karlsson et al. 1999).

As a result of changes in the Lake Ontario food chain begin-ning in the late 1990s (Mills et al. 2003), Alewife underwent adramatic decline in abundance, thus reducing the potential forthiamine deficiency (Honeyfield et al. 2005; Fitzsimons et al.2009, 2010). In the early 2000s, there was evidence of natu-ral reproduction by salmonines other than Atlantic Salmon intributaries throughout Lake Ontario suggesting improvement inhabitat quality (McKenna and Johnson 2005; Connerton et al.2009) and that thiamine levels, for at least some species, wereadequate for reproduction. As a result of these changes, manage-ment agencies took actions involving the stocking of multiplelife stages of Atlantic Salmon, including adults, and the assess-ment of bottlenecks to determine if conditions for successfulsalmonid reproduction had improved from those existing ear-lier. Despite these efforts there remains considerable uncertaintyregarding the basic biology of Atlantic Salmon, including thediet of adults while in Lake Ontario.

The spawning habitat used by Atlantic Salmon has beenwidely studied, but there remain concerns about spawning habi-tat selection by hatchery-reared spawners relative to that of wildspawners. Based on laboratory experiments, hatchery-reared

Atlantic Salmon may select inferior habitat for spawning rel-ative to their wild counterparts (Fleming et al. 1996), and thiscan have negative implications for offspring survival and growth(Magee et al. 1996; Bernier-Borgault and Magnan 2002). Thespawning habitat used by wild Atlantic Salmon has been de-scribed for several rivers in Europe (Moir et al. 1998; Heggbergetet al. 1988) and eastern North America (Gibson 1993) but notfor tributaries of Lake Ontario. Scott et al. (2005) describedthe spawning habitat characteristics used by hatchery-rearedAtlantic Salmon in one Lake Ontario tributary that were foundto differ from the characteristics of random locations within thetributary. Initially, because there is no wild remnant stock, it isexpected most spawning by Atlantic Salmon in Lake Ontariotributaries will be the result of hatchery-reared fish with increas-ing contributions from fish resulting from natural reproductionover time but initially dependent on the reproductive success ofhatchery-reared fish.

Differentiating redds of wild and hatchery-reared AtlanticSalmon and evaluation of their respective spawning habitat andegg thiamine content is essential for understanding linkages withembryo survival. As wild and hatchery-reared Atlantic Salmonspawn at the same time, separation of redds on a temporal basisis not possible (Heggberget et al. 1988). Separation of juvenileand adult wild and hatchery-reared fish has been carried out us-ing scale pattern analysis (Hansen et al. 1993), morphologicalcharacteristics (Crozier 1998), or a combination of these meth-ods (Carr et al. 1997). Occasionally use of carotenoid pigmentanalyses have been applied (Youngson et al. 1997). In generalhowever, no methods are totally discriminatory and a combina-tion of methods is often recommended, but few are appropriatefor eggs (Lund and Hansen 1991). In addition, many of thesemethods require sampling of adults, which for Atlantic Salmonis difficult for a number of reasons. Prior to spawning, adultAtlantic Salmon are often cryptic, seeking cover in deep poolsor overhanging banks (Armstrong et al. 2003; J. D. Fitzsimons,unpublished data), and most likely at low abundance (Carr et al.2004). In addition, sampling of adults prior to spawning mayalter their spawning behavior and so affect redd characteristics(Thorstad et al. 2003). Moreover, once Atlantic Salmon havespawned, adults generally leave the spawning area (Bagliniereet al. 1990) making it difficult to associate adults with theircorresponding redds.

An alternative technique that could identify spawning byhatchery fish from that of wild fish postspawning is the anal-ysis of spawned eggs for stable isotope signatures (δ13C andδ15N). Stable isotopes have been used to infer movement pat-terns of fishes (Hesslein et al. 1991), distinguish anadromousfrom nonanadromous components of a stock (Doucett et al.1999a, 1999b), identify different spawning stocks or popula-tions of fishes (Gao and Beamish 1999), and provide informationabout feeding relationships in aquatic environments (Guigueret al. 2002). Stable isotopes were used to differentiate farmedAtlantic Salmon fed on a homogenous commercial aquaculturediet, from wild Atlantic Salmon, presumed to be feeding on a

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3

USE OF STABLE ISOTOPES TO DISCRIMINATE WILD FROM HATCHERY REDDS 743

range of wild prey (Dempson and Power 2004). Stable isotopesyield an assimilated history of feeding relationships (Fry andSherr 1984; Hobson and Welch 1992), which for freshwaterAtlantic Salmon probably reflects lacustrine feeding becausethey become anorexic prior to the spawning migration (Kadriet al. 1995).

We hypothesize that eggs spawned by hatchery-rearedAtlantic Salmon, fed a commercial hatchery diet, would yieldcharacteristically different stable isotope signatures than eggs ofAtlantic Salmon that feed on wild prey in Lake Ontario and thathad been stocked previously, being defined here as wild AtlanticSalmon. Further we hypothesize that if eggs can be differenti-ated on the basis of stable isotope signature, this information canbe used to (1) determine whether captive breeding and rearingof Atlantic Salmon in a hatchery environment affects spawn-ing site selection relative to wild spawners and (2) determinewhether the thiamine content of spawners using a wild diet con-taining unknown amounts of Alewives, which cause thiaminedeficiency, differs from spawners fed a hatchery diet assumedto have adequate thiamine levels (Fisher et al. 1998b)

The aim of the present study was to examine the utility ofstable isotope analyses to discriminate redds of hatchery-rearedand wild Atlantic Salmon in the Credit River and their respec-tive habitat attributes, egg thiamine content, and adequacy ofthiamine levels for reproduction. Redds were constructed bya combination of wild Atlantic Salmon that had migrated up-stream to the study area from Lake Ontario and hatchery-rearedAtlantic Salmon that were stocked as spawning adults directlyinto the study area.

METHODSAtlantic Salmon stocking: hatchery-reared fish.—Fish used

for stocking in the stocking area located on the upper CreditRiver approximately 40 km upstream of Lake Ontario (Figure 1)were captive bred age-4 + sexually mature adult LeHavre strainAtlantic Salmon. They had been reared at the Ontario Min-istry of Natural Resources Fish Culture Station (OMNRFCS) inHarwood, Ontario. The lineage of the fish used in this study isone that has been subject to artificial breeding and captivity forthe past 18 years (about 3–5 generations; Normandale OMNR-FCS) and artificial breeding since 1969 (P. Amiro, Fisheries andOceans Canada, personal communication). Stocking occurredat three locations within a 1,500-m section of the upper CreditRiver (Ontario) approximately 40 km upstream of its mouthon Lake Ontario. Stocking occurred on October 14, 2009, justprior to the 2009 spawning season. Each of the three stockinglocations corresponded to road crossings (Figure 2), and at eachlocation 114 Atlantic Salmon were released. Based on detailedhatchery records on gender and maturity, which we assumedapplied to the Atlantic Salmon released, we assumed that ofthe 114 fish released at each location, 48% (54) were maturefemales, 48% (54) were mature males, and 4% (6) were of un-known gender and unlikely to spawn (D. Rosborough, Ontario

FIGURE 1. Map of the location of the Credit River (Ontario), a tributary ofLake Ontario, located within the Great Lakes basin. Measurement bar indicates15 kilometers.

Ministry of Natural Resources, personal communication). His-torically Atlantic Salmon used the upper Credit River for spawn-ing, and earlier work (Fitzsimons, unpublished data) involvingstocked hatchery-reared Atlantic Salmon indicated that of sev-eral areas where they were stocked in equal numbers and sexratios, the highest spawning per adult female occurred in thisarea. Although Brown Trout Salmo trutta are also present inthe upper Credit River and spawning habitat characteristics ofBrown Trout are similar to those of Atlantic Salmon (Heggbergetet al. 1988; Armstrong et al. 2003), the area where AtlanticSalmon were released was upstream of the area reported used byresident Brown Trout for spawning (Zimmer and Power 2006).At each stocking location, individual Atlantic Salmon were dip-netted from a 92,270-L tank that was used for their transportfrom the OMNRFCS. Atlantic Salmon were placed directly intoa pool in the river with most fish dispersing either upstream ordownstream within hours of stocking. No wild Atlantic Salmonwere stocked in the area where hatchery Atlantic Salmon werereleased and thus it was assumed that any wild Atlantic Salmonpresent had migrated there from Lake Ontario although at thetime, no runs of wild Atlantic Salmon were known to exist.

Atlantic Salmon stocking: wild fish.—To evaluate redd habi-tat selection and construction by wild Atlantic Salmon in the ab-sence of hatchery-reared Atlantic Salmon, wild Atlantic Salmonfrom Lake Ontario were stocked into Rodgers Creek (Figure 1),a small tributary of the Credit River. Wild Atlantic Salmon forstocking into Rogers Creek were collected from the Credit Riverat Streetsville on their upstream migration from Lake Ontario.Streetsville is located approximately 10 km upstream of LakeOntario, while Rodgers Creek is approximately 30 km upstream.Atlantic Salmon were collected by angling or electrofishing be-low the Streetsville fishway or from the fishway itself betweenAugust 28 and November 2, 2009. On the day of capture, fish

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3


FIGURE 2. Locations (black triangles) where Atlantic Salmon stocking took place on the Credit River in 2009, and the five reaches (reach 5 is upstream, reach9 is downstream) evaluated for evidence of spawning activity and used for the collection of eggs from the redds. Road crossings demarcate the upstream anddownstream extent of the reaches. The dotted line denotes the approximate river course.

were measured for length (mm) and sexed using external mor-phology before being moved by road in a splash tank mountedon a truck and stocked into a pool just below a small damon Rodgers Creek. No hatchery-reared Atlantic Salmon werestocked into Rodgers Creek and because of obstructions nearits confluence with the Credit River, it was considered unlikelythat Atlantic Salmon would enter Rodgers Creek of their ownvolition from the Credit River.

Redd surveys and redd excavation.—To identify and enumer-ate Atlantic Salmon redds on the upper Credit River, a total offour relatively separate reaches (Figure 1) were traversed weeklystarting October 1, three weeks after Atlantic Salmon had beenstocked. Redd surveys were also conducted on Rodgers Creek.All redd locations were identified in comparison to surround-ing algal- or sediment-covered substratum as areas of clean

substratum with characteristic structure (Witzel and MacCrim-mon 1983) and further confirmed as putative redds throughthe identification of upstream pits and downstream tailspills(Schmetterling 2000). Global Positioning System (GPS) coordi-nates were taken for all redds found, and the downstream extentof the redd was marked by inserting a colored peg securely intothe substrate.

For each redd we measured flow rate (cm/s), water depth(cm), egg pocket depth (cm), redd area (m2), stream width (m),canopy (%), riparian vegetation type and cover, and stream habi-tat type (e.g., riffle, pool–riffle, run). Water depth was measuredat the center of the redd. Egg pocket depth was defined as thedownward vertical extent of loose gravel at the center of the redd.We did not make measures of substrate size as it has alreadybeen extensively reported that Atlantic Salmon use gravel-size

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3


substrate for spawning (Heggberget et al. 1988; Armstrong et al.2003; Scott et al. 2005) and hatchery and wild fish appear to usesimilar habitat for spawning (Økland et al. 1995; Fleming et al.1997).

As a basis for determining whether Atlantic Salmon spawn-ing habitat selection was deliberate or random in nature, weevaluated habitat at redd and nonredd sites on the Credit River.To identify nonredd random sites we used a table of randomnumbers to produce a set of coordinates for random sites withinthe study area. We compared the habitat characteristics for theserandom sites with that of redd sites. Occasionally coordinatesfor random sites would require sampling a location that wouldnot be possible for a fish to use for spawning, such as too closeto the shore or under a fallen tree, or that was not possible forus to sample, such as in a deep pool. In such cases, we did notcollect data from the location. Exclusion of such locations isreasonable because it is not representative to include unusablespawning sites in an assessment of site selection.

As spawners can construct test pits that superficially resembleredds but do not contain eggs, it was necessary to verify thatredds contained eggs; for verification we excavated each redd.As eggs immediately after spawning and for some time afterthat exhibit elevated mortality, possibly as a result of sensitivityto physical shock (Jensen and Alderdice 1989), we delayedexcavating redds until March, by which time eggs had reachedthe eyed stage.

Redds were excavated with a flat shovel (20 cm long) whiletwo overlapping collection nets (25-cm diameter; 800–900-µmmultifilament mesh) were held in the current immediately down-stream of the redd. While facing the stream bank, the shovelwas inserted vertically up to a depth of at least 20 cm into thesubstrate, and hence deeper than the average egg pocket depth(15 cm) reported for Atlantic Salmon (Heggberget et al. 1988)and other salmonines possibly involved in redd construction(Zimmer and Power 2006; Devries 1997). The shovel blade wasthen moved through a 90◦ arc while being lifted vertically untilhorizontal. The shovel blade and contents, while held in a hor-izontal orientation, were lifted to midway in the water columnwhere the shovel blade was gently rocked in the water currentto dislodge fine material and eggs into one of two downstreamcollection nets. This process was repeated until either eggs werecollected or 75% of the length of redd had been excavated andno eggs found. For Rodgers Creek, because of concerns aboutegg mortality resulting from the redd excavation procedure, only25% of the redd length was excavated. All eggs from an individ-ual redd were placed into an individual 250-mL plastic bottlefilled with river water and examined later the same day. Whenexamined, eggs were separated into translucent-live, or opaque-dead egg categories. Only live eggs were used for stable isotope,thiamine, and genetic analysis (see below).

Collection of eggs from migratory wild Atlantic Salmon.—Known wild Atlantic Salmon collected by angling or electrofish-ing below the Streetsville fishway that were initially intendedfor stocking into Rodgers Creek (see above) but which diedduring capture or shortly thereafter in transit in river water were

used to provide samples for stable isotope and thiamine anal-ysis. Because of death these samples were potentially biased,but stable isotopes are robust to a variety of preservation meth-ods (Barrow et al. 2008) and salmonine eggs have been heldfor periods of several months in a hatchery with no change inthiamine content (Brown et al. 1998a), suggesting that the max-imum 1-h holding would have had minimal effects. Moreover,thiamine levels in this group of eggs were all above those ofAtlantic Salmon that migrated successfully to a collection weiron the St. Mary’s River (Michigan) from Lake Huron, a dis-tance of 120 km (Fitzsimons, unpublished data). As there wereno barriers on the Credit River downstream of this point andthe location is only 10 km from Lake Ontario, Atlantic Salmoncollected at this location were considered representative of theAtlantic Salmon in the lake. Dead Atlantic Salmon were im-mediately placed on ice and transferred to a freezer (−20◦C)for storage prior to eggs being removed for stable isotope andthiamine analysis. Frozen fish were measured for total length(mm) and weight (g).

Stable isotope analysis.—Egg samples for stable isotopeanalysis of δ13C and δ 15N from either redds or females weredried to a constant mass at 65◦C and ground to a fine powder.Lipids were extracted from all tissues before analysis regardlessof isotope (see methods in Folch et al. 1957; Post and Parkinson2001) to account for the differential routing of stable isotopesin adipose versus dorsal muscle tissue (Kling et al. 1992). Al-though lipid extraction can significantly increase δ13C and δ 15Ncausing a positive shift in overall food web placement, the effectsdo not appear to be species-specific such that overall food webplacement in Lake Ontario should not have been altered (Murryet al. 2006). Stable isotope measurements were performed usinga Finnigan MAT Delta Plus continuous-flow elemental analyzerat Cornell University’s Boyce Thompson Stable Isotope Fa-cility (Ithaca, New York). Samples with high nitrogen contentwere compared with a methionine standard (δ13C = −25.40‰,δ15N = −0.81‰, 40.25% C, 9.39% N) and samples with lownitrogen content were compared with a burned cabbage stan-dard (δ13C = −27.22‰, δ15N = 0.77‰, 41.60% C, 3.35% N).The standard error from the mean of each isotopic run neverexceeded 0.15‰ with respect to the standards. Stable isotoperatios are expressed in δ notation as the deviation from standardsin ppt (‰) according to the following equation: δ 13C, δ 15N =(Rsample/Rstandard − 1) × 103, where R = 13C/12C or 15N/14N.The standard reference material was atmospheric nitrogen for15N and Pee Dee Belemnite carbonate for 13C.

Thiamine analysis.—Thiamine analysis of eggs from reddsor females was by the high pressure liquid chromatographymethod of Brown et al. (1998b). The individual thiamine vi-tamers thiamine pyrophosphate, thiamine monophosphate, andfree thiamine were quantified and thiamine concentration re-ported as the sum of all three vitamers.

Hatchery Atlantic Salmon.—To serve as a reference for thi-amine and stable isotopes for hatchery Atlantic Salmon eggscollected from redds, eggs of the same strain (LeHavre) held ata hatchery and fed the same diet as the Atlantic Salmon stocked

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3


(see above) were sampled for stable isotopes and thiamine anal-ysis (hereafter hatchery females).

Genetic analysis.—Species identification and confirmationthat redds were built by Atlantic Salmon and not Brown Trout,which are also known to spawn in the Credit River (Zimmerand Power 2006), was accomplished by genetic species iden-tification of eggs collected from each redd. The DNA fromindividual eggs was extracted, amplified, and sequenced for a500-nucleotide portion of the mitochondrial cytochrome b (cyt-b) gene, using methods described in Kyle and Wilson (2007).Species identification was confirmed using a local BLASTsearch (Altschul et al. 1997) against species-diagnostic ref-erence sequences in a wildlife forensic database (Kyle andWilson 2007). The DNA sequences from egg samples weretested for homology (identity) with reference sequences fromAtlantic Salmon, Brown Trout, and other salmonids (Kyle andWilson 2007); egg sequences with >99% homology withspecies-diagnostic reference sequences were recognized as be-longing to that species.

Statistics.— We used t-tests to compare the habitat criteriaof redds to those of random samples and to compare the habi-tat criteria of wild and hatchery Atlantic Salmon. Then G-testswere used to assess the independence between the type of habi-tat (e.g., run, riffle) where redds were constructed and eitherthe spawner type (e.g., wild, hatchery) or the habitat category(e.g., random location, redd). Comparison of stable isotopes andthiamine to assess the effect of putative parentage (e.g., hatch-ery, wild) and source (e.g., redd, female) of eggs was done withone-way analysis of variance using the general linear model pro-cedure in Systat. To conform to assumptions of normality andhomoscedasticity, all data were log10 transformed prior to anal-ysis and the appropriateness of this transformation confirmedwith normal probability plots. All tests were declared signifi-cant at P < 0.05.

RESULTS

Comparison of Hatchery and Wild Atlantic SalmonSpawning Activity and Stable Isotopes

A total of 79 potential Atlantic Salmon redds were identifiedin the reaches of the upper Credit River. These were associated

with an unknown number of wild spawners and the 342 hatcheryAtlantic Salmon stocked into the upper Credit River during 2009(Figure 2). Of the 79 suspected redds, 34 were excavated andfound to contain eggs, all of which were confirmed by geneticanalysis to be those of Atlantic Salmon. A much smaller numberof wild Atlantic Salmon (18) were stocked into Rodgers Creekconsisting of 10 males and 8 females having an average totallength of 531 mm. On Rodgers Creek, a total of 23 potentialredds were identified although eggs could only be extracted fromfive of these redds and all of these were confirmed as being thoseof Atlantic Salmon.

Due to limited numbers of eggs obtained from individualredds on the upper Credit River, eggs from only 20 redds fromthe Credit River were analyzed for stable isotopes, and thesewere compared to eggs obtained from the five known wildspawners collected from the fishway at Streetsville and fourbroodstock fish from the hatchery. Insufficient sample materialwas available from redds from Rodgers Creek for meaningfulstatistical analysis. Results for the Credit River for both δ13C(F = 860.1; df = 3, 24; P < 0.001) and δ15N (F = 266.1; df =3, 24; P < 0.001) indicated a significant effect of the type of eggsample (redds or females) and their putative parentage (hatch-ery or wild) (Table 1). The biplot of δ13C and δ15N for eggssampled from redds on the Credit River (Figure 3) showed clearseparation of one group of redds, presumed to be constructedby wild spawners, from another group of redds, presumed to beconstructed by hatchery spawners, and this was evident for bothδ13C and δ15N (Figure 3). This pattern was generally evidentacross the four reaches examined although insufficient numbersof eggs were collected to examine whether wild and hatcheryspawners preferred different reaches. The presumption that dis-similar isotopic signatures represented differences in parentage(wild versus hatchery) was further supported by the followingevidence. Stable isotope signatures differed significantly be-tween putative wild and hatchery types. For eggs collected fromknown wild spawners at Streetsville, δ15N (16.4‰) was sig-nificantly higher than that of putative wild redd eggs (15.0‰)collected from the upper Credit River, and much higher than thatof eggs from known hatchery fish (9.9‰) or eggs from reddsof putative hatchery fish (9.8‰). Similarly for putative wildredds, although δ13C (–26.6‰) was significantly less than thatof known wild spawners (−22.2‰), it was much less than that

TABLE 1. Summary of egg thiamine for wild and hatchery Atlantic Salmon eggs removed from females or redds on the Credit River and Rodgers Creek in2009. Means followed by the same lowercase letter in the same column are not significantly different (P > 0.05).

Source of eggs Type of female or redd Location N Thiamine mean (SE)

Females Wild, known Credit River 6 3,692 (783) yRedds Wild, putative Credit River 14 8,474 (840) zyRedds Hatchery, putative Credit River 6 14,865 (1,050) zRedds Wild, known Rodgers Creek 5 6,511 (1,931) yFemale Hatchery, known Hatchery 5 14,200 (1,167) z

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3


(a)

(b)

9

10

11

12

13

14

15

16

17

18

19

-30 -28 -26 -24 -22 -20 -18

13C

15N

9

10

11

12

13

14

15

16

17

18

-18.0 -17.8 -17.6 -17.4 -17.2 -17.0

13C

15N

FIGURE 3. Biplot of δ15N (‰) and δ13C (‰) for Atlantic Salmon eggs re-moved from redds from four reaches of the Credit River, Rodgers Creek, orwild Atlantic Salmon females for (a) all data and (b) a restricted set of δ13C(to highlight eggs of hatchery parents; shown as a box in the lower right ofFigure 3a). Symbols are as follows: filled diamond = reach 6, filled triangle =reach 7, open triangle = reach 8, filled square = reach 9, open diamond =adults.

of eggs of known hatchery fish (−17.7‰) or eggs from redds ofputative hatchery fish (−17.5‰). The differences among reddtypes are unlikely to be the result of different sized spawnersconstructing redds as neither δ13C nor δ15N in eggs of wildspawners was related (P > 0.05) to the size of parent whosetotal length ranged from 551 to 607 mm.

Comparison of Hatchery and Wild Atlantic SalmonSpawning Habitat Selection

Using the stable isotope data to assign putative redd parent-age (e.g., wild or hatchery), the spawning habitat used on theupper Credit River by hatchery spawners was largely indistin-guishable from that of wild spawners with two exceptions. Theseexceptions were the type of habitat Atlantic Salmon spawnedin and the proportion of riparian vegetation associated with thespawning area. The type of habitat used for spawning (e.g.,riffle, pool–riffle, run) differed (G = 15.3, df = 2, P < 0.05)between wild and hatchery spawners; wild Atlantic Salmon ap-peared to prefer pool–riffle (60%) over riffle (40%) or run (0%)

TABLE 2. Summary of habitat observations for hatchery and wild AtlanticSalmon spawners on the Credit River in 2009. An asterisk indicates P < 0.05.

Parameter

Hatcheryredd

mean (SE)Wild reddmean (SE) T , P-value

N 12 6Canopy (%) 5.9 (2.8) 9.0 (4.0) –0.6, 0.27Riparian vegetation

(%)58.6 (5.1) 86.0 (9.3) –2.6, 0.02*

Stream width (m) 14.8 (1.0) 14.9 (2.0) –0.1, 0.47Distance from

nearest shore (m)3.8 (0.7) 4.3 (0.9) –1.4, 0.34

Water flow (m/s) 0.6 (0.0) 0.7 (0.1) –1.4, 0.12Water depth (cm) 38.6 (4.4) 28.4 (8.0) –1.4, 0.10Pit depth (cm) 9.8 (0.4) 10.9 (0.9) –1.2, 0.14Redd width (m) 0.8 (0.1) 1.0 (0.2) –1.4, 0.12Redd length (m) 1.5 (0.2) 2.0 (0.3) –1.4, 0.10Redd area (m2) 1.1 (0.2) 1.9 (0.8) –1.10, 0.16Spawning days

after November 110.7 12.2

habitat, whereas hatchery-reared Atlantic Salmon preferred rif-fle (64%) over pool–riffle (27%) or run (9%) habitat. WildAtlantic Salmon spawned in areas with significantly greateramounts of riparian vegetation (86%) than hatchery-reared(59%) fish (Table 2). The limited data available for wild redds onRodgers Creek (Table 3) indicated that wild Atlantic Salmon se-lected similar habitat on this smaller system with the exceptionof stream width, which was as expected less (4.0 m) than for thelarger upper Credit River (14.8 m), and distance from the nearestshore (1.8 m), which was less than one half that of the CreditRiver (3.8 m).

TABLE 3. Summary of habitat observations for Atlantic Salmon redds onRodgers Creek in 2009.

ParameterWild reddmean (SE)

N 5Canopy (%) 16.0 (4.0)Riparian vegetation (%) 83.0 (7.7)Stream width (m) 4.0 (0.3)Distance from nearest shore (m) 1.3 (0.3)Water flow (m/s) 0.4 (0.0)Water depth (cm) 13.2 (1.5)Pit depth (cm) 8.6 (1.0)Redd width (m) 0.6 (0.1)Redd length (m) 1.3 (0.2)Redd area (m2) 0.9 (0.3)Spawning days after November 1 13.2

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3


TABLE 4. Summary of habitat observations for all Atlantic Salmon redds(wild, hatchery) and random locations on the Credit River in 2009. An asteriskindicates a significant difference (P < 0.05) between groups.

ParameterAll redds

mean (SE)

Randomlocations

mean (SE) T , P-value

N 18 41Canopy (%) 6.9 (2.2) 10.6 (2.5) −1.1, 0.13Riparian

vegetation (%)57.2 (5.4) 67.0 (3.4) 0.3, 0.50

Stream width(m)

14.8 (0.9) 11.6 (0.8) 2.7, 0.005*

Water flow (m/s) 0.6 (0.0) 0.4 (0.1) 3.8, 0.0002*Water depth (cm) 35.4 (4.0) 50.7 (4.4) −2.6, 0.007*

Spawning fish, whether putative wild or hatchery in origin,selected spatial habitat in proportion significantly different fromavailable habitat (Table 4). The width of stream used by spawn-ing Atlantic Salmon was wider (14.8 m) than was available(11.6 m), with spawning occurring in shallower water depths(35.4 cm) than were available (50.7 cm) but at higher flows(0.6 m/s) than were available (0.4 m/s). The proportion of typeof habitat used for spawning (e.g., riffle, pool–riffle, run, pool)did not differ from the proportions available (G = 7.5, df = 3,P > 0.05).

Comparison of Hatchery and Wild Atlantic SalmonThiamine Status

For the upper Credit River, when stable isotope data wereused to assign putative redd identity (e.g., wild or hatchery),egg thiamine status varied (F = 13.2; df = 4, 28; P < 0.001)depending on the source of eggs (redd versus female) but notwhether redds were of wild or hatchery origin (Table 1). Overallthe highest egg thiamine concentration occurred in eggs fromputative hatchery redds (14,865 pmol/g), but they were not dif-ferent from eggs of known hatchery fish (14,200 pmol/g). Simi-larly, there was no difference in thiamine levels among the eggsof known wild spawners (3,692 pmol/g) and eggs from reddsof known wild spawners on Rodgers Creek (6,511 pmol/g) andeggs from redds of putative wild spawners on the Credit River(8,474 pmol/g). Thiamine levels of eggs from putative hatch-ery redds were significantly higher (P < 0.05) than eggs ofknown wild spawners (3,692 pmol/g) but not eggs from reddsof putative wild spawners (8,474 pmol/g). The thiamine con-centration of eggs from known wild redds on Rodgers Creek(6,511 pmol/g) differed (P < 0.05) from that of eggs from pu-tative hatchery redds on the Credit River. In contrast, the eggsof putative wild redds (8,474 pmol/g) on the Credit River didnot differ from that of putative hatchery redds (P > 0.05). Thedifference among groups did not appear related to the size ofspawner since the thiamine concentration of eggs removed from

known wild spawners was unrelated (r = 0.37, df = 5, P > 0.05)to the size of spawner.

DISCUSSIONThere were two groups apparent in the stable isotope data

for redds possibly consistent with differing parentage, that be-ing hatchery and wild. The lack of differences between knownhatchery fish and redds of putative hatchery fish for both δ13C(−17.7‰ versus −17.5‰) and δ15N (9.9‰ versus 9.8‰) sup-ports the designation of the latter group as being of hatcheryorigin. Similarly, the data for redds of presumed wild originsupports a designation of wild origin. Levels of δ13C in eggsof putative wild redds (–26.6‰) were lower than that of eggsof known wild spawners (–22.2) and much lower than that ofeggs from either known hatchery spawners (–17.7‰) or puta-tive hatchery redds (–17.5‰), which we infer represent hatcheryfish. In addition, levels of δ15N in eggs from putative wild redds(16.4‰) were higher than that of eggs from known wild fish(15.0‰) and much higher than that of eggs from either knownhatchery spawners (9.9‰) or putative hatchery redds (9.8‰).

As the stable isotope signatures of the two groups of eggs(e.g., wild and hatchery) are related to the diet of their mater-nal parents, the higher nitrogen values in putative wild AtlanticSalmon eggs suggests a longer food chain for their maternalparents than was the case for the maternal parents of hatchery-reared Atlantic Salmon eggs (Vander Zanden and Rasmussen1999). It follows from this that the greater variation in theisotopic signatures of putative wild eggs compared to putativehatchery eggs probably represents greater heterogeneity in thediet of wild parents, possibly reflecting variation in the isotopicsignatures of prey items (Kiriluk et al. 1995). By comparison,the diets of hatchery parents appeared quite homogenous prob-ably due to the high level of regulation in feedstock materialsused to formulate hatchery diets based on minimal variation inthe isotopic signature of hatchery eggs. Dempson and Power(2004) found slight variation in isotopic signatures among var-ious salmon aquaculture feeds, and such variation would belargely eliminated by the adoption of a standardized feed, as isthe case for the OMNRFCS. More importantly however, hatch-ery diets use marine fish almost exclusively as the protein source(D. C. Honeyfield, personal communication) and it is known theδ13C value is distinctly different between marine and freshwater–terrestrial animals. Signatures of marine animals are typically7‰ more 13C enriched relative to those from freshwater equiv-alents, and this difference was similar to that between putativehatchery and wild salmon eggs (i.e., 9‰; Fry et al. 1983). Giventhe near universal use of marine protein in hatchery diets, theseparation observed in this study with Atlantic Salmon may beapplicable to a broad range of species allowing the differentia-tion of hatchery broodfish and their eggs from that of wild adultsand their eggs when diet consists of freshwater prey.

Although the use of stable isotope signatures of eggs offersconsiderable potential for differentiating the feeding origins ofparents, little is known about eggs relative to that of other fish

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3


tissues. Much of the literature pertaining to the use of stableisotopes for discerning feeding relationships among fish popu-lations is based on muscle (Vander Zanden et al. 1997, 1999;Perga and Gerdeaux 2005) and liver (Perga and Gerdeaux 2005),with relatively little reported for eggs (Grey 2001; Acolas et al.2008). Due to variation in tissue turnover rates, liver, becauseit is a regulatory tissue, appears to reflect annual changes indiet, whereas muscle, because isotope turnover depends pri-marily on growth rather than metabolic replacement, reflectsfood consumed during periods of growth (Perga and Gerdeaux2005). As gonadosomatic growth depends on current food in-take rather than tissues accumulated during past somatic growth(Roff 1983), it is reasonable to suppose that egg compositionreflects diet. This is supported by the observation that for At-lantic Salmon, the δ13C and δ15N concentration of the egg ishighly correlated with that of muscle (Fitzsimons, unpublisheddata).

Within the wild Atlantic Salmon eggs obtained for this study,there was evidence of spatial variation in isotopic signatures be-tween eggs from putative wild redds and known wild spawners,although the reasons for this variation are not clear. Significantdifferences in δ13C and δ15N were observed between eggs re-moved from known wild spawners collected 10 km from LakeOntario and eggs removed from putative wild redds located ap-proximately 40 km from the lake. The effect of diet on isotopicsignature is considered to be greater for δ15N than δ13C (Petersonand Fry 1987), but the difference in δ15N between the groupswas less than one trophic level (e.g., 3.4‰; Vander Zanden andRasmussen 1999). Although the diet of Atlantic Salmon whilein Lake Ontario is unknown, potential Lake Ontario prey fishshow variation for δ13C and δ15N among prey species (Kiriluket al. 1995), and depending on diet this may be reflected in eggstable isotope signatures.

The effect of maternal diet in influencing egg compositionas revealed by the isotopic signature was also evident for eggthiamine. Although levels in wild groups varied, possibly a re-flection of diet while in the lake (Fitzsimons and Brown 1998;Honeyfield et al. 2005), concentrations of thiamine in wildgroups, regardless of whether the eggs came from redds orfemales, appear to represent physiologically adequate levels ca-pable of preventing thiamine-dependent mortality. In all casesthiamine concentrations exceeded levels (1,600 pmol/g) consid-ered protective of thiamine-dependent larval mortality (Ketolaet al. 2000). Based on thiamine levels in Alewife-dependent At-lantic Salmon stocks reported by Fisher et al. (1998b) that aver-aged less than 1,000 pmol/g, the levels in the three groups of wildeggs examined here were three- to eight-fold higher. Althoughthis would suggest a relatively low level of Alewife consump-tion by Lake Ontario Atlantic Salmon, additional work wouldbe required to confirm this. Preliminary information based onthe use of stable isotope mixing models indicates that over a4-year period Alewives comprised no more than 10% of thediet of Lake Ontario Atlantic Salmon (Fitzsimons, unpublisheddata).

The relative abundance of wild to hatchery redds in ourstudy, when combined with the known number of hatchery fishstocked, can be used to estimate both the reproductive efficiency(i.e., redds built per female) of hatchery Atlantic Salmon aswell as the number of wild spawners present in the study area.Fleming et al. (1996) reported that Atlantic Salmon, whetherwild or farm reared, construct one redd per female. In our studythe proportion of hatchery to wild redds confirmed to have eggswas 74% based on 19 redds with assumed parental identity usingstable isotopes. Therefore of the 79 redds identified, 58 reddswould be attributable to hatchery females. The total number ofmature hatchery females out of the 342 hatchery fish stockedwas 162. Using this estimate of the number of female hatch-ery fish stocked and that spawned and built redds, it can becalculated there were 2.8 hatchery females per redd suggestinglow reproductive efficiency. Although we do not know if somehatchery spawners may have migrated either farther upstreamor downstream and not been censused in our surveys, whichcovered a restricted geographic area, this seems unlikely basedon the distribution of redds observed. Our surveys of spawn-ing activity were limited to 2,500 m of river, which included a500-m stretch of river upstream and downstream of each of theareas where fish were released. Most redds observed were in thearea where fish were released with far fewer redds found in thearea immediately upstream and downstream of the release areasuggesting that spawning occurred primarily in the area wherefish were released. Alternatively, the hatchery fish used in thisstudy, like their farmed counterparts (Fleming et al. 1996), mayconstruct fewer redds per female than wild fish supporting thenotion of low reproductive efficiency with over half the femalesnot spawning. In an experimental study, Fleming et al. (1996)observed that wild Atlantic Salmon constructed almost two-foldmore redds per female than hatchery-reared Atlantic Salmon. Ofthe 79 redds, we would estimate, based on proportions, that atotal of 21 redds would be attributable to putative wild spawn-ers, but since the number of wild adults was unknown we cannotestimate the number of wild spawners per redd. Based on oneredd per female, at least 21 wild spawners may have spawned inthe survey area, although if wild spawners performed similarlyto hatchery females (e.g., 2.8 females/redd), this number mayhave been closer to 59 wild spawners.

Selection of spawning habitat that is inappropriate for re-production can have profound fitness implications (Schjørring2002) and affect survival (Chapman 1988). In this study though,we detected few measurable differences between the habitatused by hatchery and wild spawners. Of those differences, theapparent preference for riffle habitat over pool–riffle habitat byhatchery spawners may be because of the reduced upwellingflow in riffles as opposed to pool–riffles. This has potential tonegatively affect embryonic survival, but more work would berequired to confirm this. The wild spawners responsible for thewild redds identified in this study were likely the result of earlierstocking activities of hatchery-reared Atlantic Salmon so are notrepresentative of a wild remnant stock. These Atlantic Salmon,

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3


based on numbers and life stages of fish stocked, were proba-bly of an earlier developmental stage than adults because of themuch higher stocking of yearling and younger stages that hasoccurred (M. Daniels, Ontario Ministry of Natural Resources,personal communication). Nevertheless, despite the period ofholding under hatchery conditions, the spawning habitat usedby wild Atlantic Salmon in this study was similar to that givenin reviews of the freshwater requirements of Atlantic Salmonbased on wild fish (Gibson 1993).

Due to wide-scale reductions in native stocks and conserva-tion needs (Carr et al. 2004), hatchery-reared Atlantic Salmonare often the only practical source of fish for restoring land-locked populations such as that in Lake Ontario, but they maybe associated with reduced reproductive efficiency. Wild andstocked fish can breed together, but detrimental effects of stock-ing on native stocks have been reported, notably a fitness declineof wild populations (Youngson and Verspoor 1998; Jonsson andJonsson 2006; Araki et al. 2007). In Lake Ontario, there hasbeen no wild stock to draw on as the original stock was extir-pated (COSEWIC 2006). As a result, restoration has been basedentirely on the use of hatchery-reared fish. However, because ofconcerns regarding the reproductive efficiency of captive bredAtlantic Salmon from laboratory experiments (Fleming et al.1996), assessing the proportion of wild fish during recovery andmaking corresponding adjustments to the stocking of hatcheryfish may be important for maintaining a positive recovery tra-jectory. In our study, while spawning habitat selection appearedsimilar between hatchery and wild Atlantic Salmon, there wasthe suggestion that a relatively low proportion of hatchery-rearedfish built redds. In addition, mortality of stocked fish is assumedto be elevated relative to their wild counterparts, given the rela-tively low percentages of released individuals returning to targetrivers (Eckert 2003; Ruzzante et al. 2004; COSEWIC 2006).Stocked fish are also more likely to stray than wild conspecificswhen returning to spawning grounds (Jonsson et al. 2003), andthis may affect spawning habitat use. For Lake Ontario, it is rec-ognized there is considerable heterogeneity in tributary habitatquality in the basin (Stanfield and Kilgour 2006; Randall 2010)and as a result the quality of spawning habitat probably variesas well (Stanfield and Jones 2003).

The complete loss of the historic Atlantic Salmon stock asoccurred with Lake Ontario will make it more difficult, butseemingly not impossible, to establish a self-sustaining stockthat initially, at least, relies almost entirely on hatchery-rearedfish. Restoration of Atlantic Salmon in Lake Ontario is beingpursued by the Ontario Ministry of Natural Resources throughan adaptive management process involving research into a vari-ety of rehabilitation measures, including habitat restoration andstocking at a variety of life stages. The timing, numbers, andappropriate life stages to use for this purpose are being evalu-ated on an experimental basis predicated in part on demographicanalysis that suggests that the most effective way of improvingpopulation fitness is by increasing survival at early life stages(age-0, age-1, and age-2 smolts), but this is an ongoing process

(M. Koops, Fisheries and Oceans Canada, personal communi-cation).

Although it is still uncertain as to the appropriate strategy touse for the restoration of Atlantic Salmon in Lake Ontario,the use of some life stages, like adults as used here, mayhave certain advantages over other life stages like juveniles.At low population size, being able to immediately access andpotentially saturate available high-quality spawning habitat iscritically important for achieving maximum reproductive rate(Myers et al. 1999). Restoration strategies employing stock-ing of mature hatchery-reared fish directly into spawning riverscan ensure that the best habitat in the best tributaries is usedfor spawning (Aprahamian et al. 2003). Juveniles and youngerstages can be released into streams believed to have superiorhabitat quality for spawning and rearing but can encounter mor-tality due to predation on their way downstream and in the lake(Klemetsen et al. 2003). Hence, the effectiveness of stocking ofjuveniles relative to the stocking of mature adults for rebuildingspawning stocks may be reduced. The effectiveness of juvenilestocking may be further reduced by straying prior to the initia-tion of the spawning migration (Jonsson et al. 2003). Those fishthat successfully migrate to a high-quality stream may, in ad-dition, encounter elevated stream temperatures before reachingthe colder temperatures of the upper reaches of a stream, alsoresulting in mortality (Eriksen et al. 2006; Rand et al. 2006).Exposure to elevated temperatures can lead to mortality depen-dent on run timing, availability of groundwater, and ambientconditions (Everitt 2006; Farrell et al. 2008). It is important tonote that Atlantic Salmon may be much more susceptible tothe effects of climate-change-mediated increases in river watertemperature (Eaton and Scheller 1996) than other salmonines.Atlantic Salmon, by entering spawning streams during sum-mer months when highest water temperatures occur (Webb andMcLay 1996), are likely to be exposed to much warmer tem-peratures than other salmonines, especially if coldwater poolhabitat is limiting (Ketola et al. 2009).

Although there is considerable uncertainty as to the contri-bution that stocking at a given life stage can make towards therestoration of Atlantic Salmon in Lake Ontario, the techniqueelaborated here appears to be able to provide an objective meansfor evaluating reproductive effort by putative hatchery and wildfish. These in turn can provide a means of objectively evaluatingrelative success in achieving conservation spawner requirements(O’Connell et al. 1997).

This work represents the first documentation using naturallyspawned eggs of natural reproduction by Atlantic Salmon inLake Ontario, and the second time in the history of the programthat natural reproduction has been observed. Natural reproduc-tion was first documented by Johnson et al. (2010) from age-0fish collected in the Salmon River (New York) in 2009. Taken to-gether these observations represent an important milestone in therestoration program and, although a marked improvement overthe previous decade (Crawford 2001), will require additionalwork towards linking the importance of individual restoration

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3


activities to the establishment and relative abundance of spawn-ing runs. Our findings follow a change in the Atlantic Salmonrestoration program in 2006 to increase stocking to 2.5 mil-lion fry. These stocking efforts focused on three high-qualityLake Ontario tributaries, including the Credit River, suggestingthe potential for an increased run size in the future (MarionDaniels, Ontario Ministry of Natural Resources, personal com-munication). Moreover, while it was evident from this study thatspawning Atlantic Salmon were able to ascend a partial barrierlocated at Norval and access high-quality spawning habitat lo-cated on the upper Credit River, in the future upstream move-ment will be facilitated by a fishway constructed at this barrierin 2012 (Mark Heaton, Ontario Ministry of Natural Resources,personal communication).

ACKNOWLEDGMENTSWe thank the staff of the Ontario Ministry of Natural Re-

sources for assistance during the study, and Marion Daniels,Bill Sloan, and the staff of the Ontario Ministry of Natural Re-sources Normandale Fish Culture station for making AtlanticSalmon available. Kimberlee Sparks facilitated the analysis ofstable isotope analysis at the Cornell Stable Isotope Facility.

REFERENCESAcolas, M. L., J. M. Roussel, and J. L. Bagliniere. 2008. Linking migratory

patterns and diet to reproductive traits in female Brown Trout (Salmo truttaL.) by means of stable isotope analysis on ova. Ecology of Freshwater Fish17:382–393.

Altschul, S. F., T. L. Madden, A. A. Schaffer, J. H. Zhang, Z. Zhang, W. Miller,and D. J. Lipman. 1997. Gapped BLAST and PSI-BLAST: a new generationof protein database search programs. Nucleic Acids Research 25:3389–3402.

Amcoff, P., H. Borjeson, P. Landergren, L. Vallin, and L. Norrgren. 1999.Thiamine (vitamin B1) concentrations in salmon (Salmo salar), Brown Trout(Salmo trutta) and cod (Gadus morhua) from the Baltic Sea. Ambio 28:48–54.

Aprahamian, M. W., K. Martin Smith, P. McGinnity, S. McKelvey, and J.Taylor. 2003. Restocking of salmonids—opportunities and limitations. Fish-eries Research 62:211–227.

Araki, H., B. Cooper, and M. S. Blouin. 2007. Genetic effects of captive breedingcause a rapid, cumulative fitness decline in the wild. Science 318:100–103.

Armstrong, J. D., P. S. Kemp, G. J. A. Kennedy, M. Ladle, and N. J. Milner.2003. Habitat requirements of Atlantic Salmon and Brown Trout in rivers andstreams. Fisheries Research 62:143–170.

Bagliniere, J. L., G. Maisse, and A. Nihouarn. 1990. Migratory and reproductivebehaviour of female adult Atlantic Salmon, Salmo salar L., in a spawningstream. Journal of Fish Biology 36:511–520.

Barrow, L. M., K. A. Bjorndal, and K. J. Reich. 2008. Effects of preserva-tion method on stable carbon and nitrogen isotope values. Physiological andBiochemical Zoology 81:688–693.

Bernier-Bourgault, I., and P. Magnan. 2002. Factors affecting redd site selec-tion, hatching, and emergence of Brook Charr, Salvelinus fontinalis, in anartificially enhanced site. Environmental Biology of Fishes 64:333–341.

Borjeson, H., P. Amcoff, B. Ragnarsson, and L. Norrgren. 1999. Reconditioningof sea-run Baltic salmon (Salmo salar) that have produced progeny with theM74 syndrome. Ambio 28:30–36.

Brown, S. B., J. D. Fitzsimons, V. P. Palace, and L. Vandenbyllaardt. 1998a.Thiamine and early mortality syndrome in Lake Trout. Pages 18–25 in G.McDonald, J. D. Fitzsimons, and D. C. Honeyfield, editors. Early life stagemortality syndrome in fishes of the Great Lakes and Baltic Sea. AmericanFisheries Society, Symposium 21, Bethesda, Maryland.

Brown, S. B., D. C. Honeyfield, and L. Vandenbyllaardt. 1998b. Thiamineanalysis in fish tissues. Pages 73–81 in G. McDonald, J. D. Fitzsimons, andD. C. Honeyfield, editors. Early life stage mortality syndrome in fishes ofthe Great Lakes and Baltic Sea. American Fisheries Society, Symposium 21,Bethesda, Maryland.

Carr, J. W., J. M. Anderson, F. G. Whoriskey, and T. Dilworth. 1997. Theoccurrence and spawning of cultured Atlantic Salmon (Salmo salar) in aCanadian river. ICES Journal of Marine Science 54:1064–1073.

Carr, J. W., F. Whoriskey, and P. O’Reilly. 2004. Efficacy of releasing captivereared broodstock into an imperilled wild Atlantic Salmon population as arecovery strategy. Journal of Fish Biology 65(Supplement 1):38–54.

Chapman, D. W. 1988. Critical review of variables used to define effects of finesin redds of large salmonids. Transactions of the American Fisheries Society117:1–21.

Connerton, M. J., B. A. Murry, N. H. Ringler, and D. J. Stewart. 2009. Majorityof age-3 Chinook Salmon (Oncorhynchus tshawytscha) in Lake Ontario werewild from 1992 to 2005, based on scale pattern analysis. Journal of GreatLakes Research 35:419–429.

COSEWIC (Committee on the Status of Endangered Wildlife in Canada). 2006.COSEWIC assessment and status report on the Atlantic Salmon Salmo salar(Lake Ontario population) in Canada. COSEWIC, Ottawa.

Crawford, S. S. 2001. Salmonine introductions to the Laurentian Great Lakes:an historical review and evaluation of ecological effects. Canadian SpecialPublication of Fisheries and Aquatic Sciences 132.

Crozier, W. W. 1998. Incidence of escaped farmed salmon, Salmo salar L., incommercial salmon catches and fresh water in Northern Ireland. FisheriesManagement and Ecology 5:23–29.

Dempson, J. B., and M. Power. 2004. Use of stable isotopes to distinguishfarmed from wild Atlantic Salmon, Salmo salar. Ecology of Freshwater Fish13:176–184.

DeVries, P. 1997. Riverine salmonid egg burial depths: review of published dataand implications for scour studies. Canadian Journal of Fisheries and AquaticSciences 54:1685–1698.

Doucett, R. R., W. Hooper, and G. Power. 1999a. Identification of anadromousand nonanadromous adult Brook Trout and their progeny in the TabusintacRiver, New Brunswick, by means of multiple-stable-isotope analysis. Trans-actions of the American Fisheries Society 128:278–288.

Doucett, R. R., M. Power, G. Power, F. Caron, and J. D. Reist. 1999b. Evidencefor anadromy in a southern relict population of Arctic Charr from NorthAmerica. Journal of Fish Biology 55:84–93.

Eaton, J. G., and R. M. Scheller. 1996. Effects of climate warming on fishthermal habitat in streams of the United States. Limnology and Oceanography41:1109–1115.

Eckert, T. H. 2003. Lake Ontario fishing boat census 2002. Page 59 in 2002New York State Department of Environmental Conservation Annual Report,Bureau of Fisheries Lake Ontario Unit and St. Lawrence River Unit to theGreat Lakes Fishery Commission’s Lake Ontario Committee, Cape Vincentand Watertown.

Eriksen, M. S., M. Bakken, Å. Espmark, B. O. Braastad, and R. Salte. 2006.Prespawning stress in farmed Atlantic Salmon Salmo salar: maternal cortisolexposure and hyperthermia during embryonic development affect offspringsurvival, growth and incidence of malformations. Journal of Fish Biology69:114–129.

Everitt, D. W. 2006. Natural reproduction and spawning site characteristicsof Chinook Salmon (Oncorhynchus tshawytscha) in the Salmon River, NewYork. Master’s thesis. State University of New York, College of Environmen-tal Science and Forestry, Syracuse.

Farrell, A. P., S. G. Hinch, S. J. Cooke, D. A. Patterson, G. T. Crossin, M.Lapointe, and M. T. Mathes. 2008. Pacific salmon in hot water: applyingaerobic scope models and biotelemetry to predict the success of spawningmigrations. Physiological and Biochemical Zoology 81:697–708.

Fisher, J. P., S. B. Brown, S. Connelly, T. Chiotti, and C. C. Krueger. 1998a.Interspecies comparisons of blood thiamine in salmonids from the FingerLakes, and effects of maternal size on blood and egg thiamine in Atlantic

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3


Salmon with and without Cayuga syndrome. Pages 112–123 in G. McDonald,J. D. Fitzsimons, and D. C. Honeyfield, editors. Early life stage mortalitysyndrome in fishes of the Great Lakes and Baltic Sea. American FisheriesSociety, Symposium 21, Bethesda, Maryland.

Fisher, J. P., S. B. Brown, G. W. Wooster, and P. R. Bowser. 1998b. Mater-nal blood, egg, and larval thiamin levels correlate with larval survival inlandlocked Atlantic Salmon (Salmo salar). Journal of Nutrition 128:2456–2466.

Fisher, J. P., J. D. Fitzsimons, G. F. Combs Jr., and J. M. Spitsbergen. 1996.Naturally occurring thiamine deficiency causing reproductive failure in FingerLakes Atlantic Salmon and Great Lakes Lake Trout. Transactions of theAmerican Fisheries Society 125:167–178.

Fitzsimons, J. D., and S. B. Brown. 1998. Reduced egg thiamine levels in inlandand Great Lakes Lake Trout and their relationship with diet. Pages 160–171 inG. McDonald, J. D. Fitzsimons, and D. C. Honeyfield, editors. Early life stagemortality syndrome in fishes of the Great Lakes and Baltic Sea. AmericanFisheries Society, Symposium 21, Bethesda, Maryland.

Fitzsimons, J. D., S. Brown, L. Brown, D. Honeyfield, J. He, and J. E. Johnson.2010. Increase in Lake Trout reproduction in Lake Huron following thecollapse of Alewife: relief from thiamine deficiency or larval predation?Aquatic Ecosystem Health and Management 13:73–84.

Fitzsimons, J. D., S. B. Brown, and L. Vandenbyllaardt. 1998. Thiamine levelsin food chains of the Great Lakes. Pages 90–98 in G. McDonald, J. D.Fitzsimons, and D. C. Honeyfield, editors. Early life stage mortality syndromein fishes of the Great Lakes and Baltic Sea. American Fisheries Society,Symposium 21, Bethesda, Maryland.

Fitzsimons, J. D., S. B. Brown, B. Williston, G. Williston, L. R. Brown, K.Moore, D. C. Honeyfield, and D. E. Tillitt. 2009. Influence of thiaminedeficiency on Lake Trout larval growth, foraging, and predator avoidance.Journal of Aquatic Animal Health 21:302–314.

Fleming, I. A., B. Jonsson, M. R. Gross, and A. Lamberg. 1996. An experimentalstudy of the reproductive behaviour and success of farmed and wild AtlanticSalmon (Salmo salar). Journal of Applied Ecology 33:893–905.

Fleming, I. A., A. Lamberg, and B. Jonsson. 1997. Effects of early experienceon the reproductive performance of Atlantic Salmon. Behavioral Ecology8:470–480.

Folch, J., M. Lees, and G. H. Sloane-Stanley. 1957. A simple method for theisolation and purification of total lipides from animal tissues. Journal ofBiological Chemistry 226:497–509.

Fry, B., R. S. Scalan, and P. L. Parker. 1983. 13C/12C ratios in marine food websof the Torres Strait, Queensland. Australian Journal of Marine and FreshwaterResearch 34:707–715.

Fry, B., and E. B. Sherr. 1984. δ13C measurements as indicators of carbonflow in marine and freshwater ecosystems. Contributions in Marine Science27:13–47.

Gao, Y. W., and R. J. Beamish. 1999. Isotopic composition of otoliths as a chem-ical tracer in population identification of Sockeye Salmon (Oncorhynchusnerka). Canadian Journal of Fisheries and Aquatic Sciences 56:2062–2068.

Gibson, R. J. 1993. The Atlantic Salmon in fresh water: spawning, rearing andproduction. Reviews in Fish Biology and Fisheries 3:39–73.

Grey, J. 2001. Ontogeny and dietary specialization in Brown Trout (Salmo truttaL.) from Loch Ness, Scotland, examined using stable isotopes of carbon andnitrogen. Ecology of Freshwater Fish 10:168–176.

Grey, R. H., and J. R. Tosh. 1894. 13th annual report, part II, note 2. FisheryBoard for Scotland, Glasgow.

Guiguer, K. R. R. A., J. D. Reist, M. Power, and J. A. Babaluk. 2002. Usingstable isotopes to confirm the trophic ecology of Arctic Charr morphotypesfrom Lake Hazen, Nunavut, Canada. Journal of Fish Biology 60:348–362.

Hansen, L. P., J. A. Jacobsen, and R. A. Lund. 1993. High numbers of farmedAtlantic Salmon, Salmo salar L., observed in oceanic waters north of theFaroe Islands. Aquaculture Research 24:777–781.

Heggberget, T. G., T. Haukebø, J. Mork, and G. Stahl. 1988. Temporal andspatial segregation of spawning in sympatric populations of Atlantic Salmon,

Salmo salar L., and Brown Trout, Salmo trutta L. Journal of Fish Biology33:347–356.

Hesslein, R. H., M. J. Capel, D. E. Fox, and K. A. Hallard. 1991. Stable isotopesof sulfur, carbon, and nitrogen as indicators of trophic level and fish migrationin the lower Mackenzie River basin, Canada. Canadian Journal of Fisheriesand Aquatic Sciences 48:2258–2265.

Hobson, K. A., and H. E. Welch. 1992. Determination of trophic relationshipswithin a high Arctic marine food web using δ13C and δ15N analysis. MarineEcology Progress Series 84:9–18.

Honeyfield, D. C., J. P. Hinterkopf, J. D. Fitzsimons, D. E. Tillitt, J. L. Zajicek,and S. B. Brown. 2005. Development of thiamine deficiencies and earlymortality syndrome in Lake Trout by feeding experimental and feral fishdiets containing thiaminase. Journal of Aquatic Animal Health 17:4–12.

Jensen, J. O. T., and D. F. Alderdice. 1989. Comparison of mechanical shocksensitivity of eggs of five Pacific salmon (Oncorhynchus) species and steel-head trout (Salmo gairdneri). Aquaculture 78:163–181.

Johnson, J. H., C. C. Nack, and J. E. McKenna Jr. 2010. Migratory salmonidredd habitat characteristics in the Salmon River, New York. Journal of GreatLakes Research 36:387–392.

Jonsson, B., and N. Jonsson. 2006. Cultured Atlantic Salmon in nature: a reviewof their ecology and interaction with wild fish. ICES Journal of MarineScience 63:1162–1181.

Jonsson, B., N. Jonsson, and L. P. Hansen. 2003. Atlantic Salmon straying fromthe River Imsa. Journal of Fish Biology 62:641–657.

Kadri, S., N. B. Metcalfe, F. A. Huntingford, and J. E. Thorpe. 1995. Whatcontrols the onset of anorexia in maturing adult female Atlantic Salmon?Functional Ecology 9:790–797.

Karlsson, L., E. Ikonen, A. Mitans, S. Hansson, and D. Uzars. 1999. Thiaminelevels in migrating salmon spawners (Salmo salar) in the Gulf of Riga and inthe Gulf of Bothnia. Pages 67–88 in B. E. Bengtsson, C. Hill, and S. Nellbring,editors. Nordic research cooperation on reproductive disturbances in fish:report from the Redfish project. TemaNord, Nordic Council of Ministers,Copenhagen.

Ketola, H. G., P. R. Bowser, G. A. Wooster, L. R. Wedge, and S. S. Hurst. 2000.Effects of thiamine on reproduction of Atlantic Salmon and a new hypothesisfor their extirpation in Lake Ontario. Transactions of the American FisheriesSociety 129:607–612.

Ketola, H. G., J. H. Johnson, J. Rinchard, F. J. Verdoliva, M. E. Penney, A. W.Greulich, and R. C. Lloyd. 2009. Effect of thiamine status on probability ofLake Ontario Chinook Salmon spawning in the upper or lower sections ofSalmon River, New York. North American Journal of Fisheries Management29:895–902.

Kiriluk, R. M., M. R. Servos, D. M. Whittle, G. Cabana, and J. B. Rasmussen.1995. Using ratios of stable nitrogen and carbon isotopes to characterizethe biomagnification of DDE, mirex, and PCB in a Lake Ontario pelagicfood web. Canadian Journal of Fisheries and Aquatic Sciences 52:2660–2674.

Klemetsen, A., P. A. Amundsen, J. B. Dempson, B. Jonsson, N. Jonsson, M. F.O’Connell, and E. Mortensen. 2003. Atlantic Salmon Salmo salar L., BrownTrout Salmo trutta L. and Arctic Charr Salvelinus alpinus (L.): a review ofaspects of their life histories. Ecology of Freshwater Fish 12:1–59.

Kling, G. W., B. Fry, and W. J. O’Brien. 1992. Stable isotopes and planktonictrophic structure in Arctic lakes. Ecology 73:561–566.

Koski, P., C. Backman, and O. Pelkonen. 2005. Pharmacokinetics of thiamine infemale Baltic salmon (Salmo salar L.) broodfish. Environmental Toxicologyand Pharmacology 19:139–152.

Kyle, C. J., and C. C. Wilson. 2007. Mitochondrial DNA identification of gameand harvested freshwater fish species. Forensic Science International 166:68–76.

Lund, R. A., and L. P. Hansen. 1991. Identification of wild and rearedAtlantic Salmon, Salmo salar L., using scale characters. Aquaculture Re-search 22:499–508.

MacCrimmon, H. R. 1977. Animals, man and change: alien and extinct wildlifeof Ontario. McClelland and Stewart, Toronto.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3


Magee, J. P., T. E. McMahon, and R. F. Thurow. 1996. Spatial variation inspawning habitat of Cutthroat Trout in a sediment-rich stream basin. Trans-actions of the American Fisheries Society 125:768–779.

McKenna, J. E., Jr., and J. H. Johnson. 2005. Juvenile Rainbow Trout productionin New York tributaries of Lake Ontario: implications for Atlantic Salmonrestoration. North American Journal of Fisheries Management 25:391–403.

Mills, E. L., J. M. Casselman, R. Dermott, J. D. Fitzsimons, G. Gal, K. T. Holeck,J. A. Hoyle, O. E. Johannsson, B. F. Lantry, J. C. Makarewicz, E. S. Millard,I. F. Munawar, M. Munawar, R. O’Gorman, R. W. Owens, L. G. Rudstam,T. Schaner, and T. J. Stewart. 2003. Lake Ontario: food web dynamics in achanging ecosystem (1970–2000). Canadian Journal of Fisheries and AquaticSciences 60:471–490.

Moir, H. J., C. Soulsby, and A. Youngson. 1998. Hydraulic and sedimentarycharacteristics of habitat utilized by Atlantic Salmon for spawning in theGirnock Burn, Scotland. Fisheries Management and Ecology 5:241–254.

Murry, B. A., J. M. Farrell, M. A. Teece, and P. M. Smyntek. 2006. Effectof lipid extraction on the interpretation of fish community trophic relation-ships determined by stable carbon and nitrogen isotopes. Canadian Journalof Fisheries and Aquatic Sciences 63:2167–2172.

Myers, R. A., K. G. Bowen, and N. J. Barrowman. 1999. Maximum reproductiverate of fish at low population sizes. Canadian Journal of Fisheries and AquaticSciences 56:2404–2419.

Netboy, A. 1968. The Atlantic Salmon: a vanishing species? Faber and Faber,London.

O’Connell, M. F., D. G. Reddin, P. G. Amiro, F. Caron, T. L. Marshall, G.Chaput, C. C. Mullins, A. Locke, S. F. O’Neil, and D. K. Cairns. 1997.Estimates of conservation spawner requirements for Atlantic Salmon (Salmosalar L.) for Canada. Department of Fisheries and Oceans, Canadian StockAssessment Secretariat, Research Document 97/100, Ottawa.

Økland, F., T. G. Heggberget, and B. Jonsson. 1995. Migratory behaviour ofwild and farmed Atlantic Salmon (Salmo salar) during spawning. Journal ofFish Biology 46:1–7.

Parsons, J. W. 1973. History of salmon in the Great Lakes, 1850–1970. U.S.Fish and Wildlife Service Bureau of Sport Fisheries and Wildlife TechnicalPaper 68.

Perga, M. E., and D. Gerdeaux. 2005. ‘Are fish what they eat’ all year round?Oecologia 144:598–606.

Peterson, B. J., and B. Fry. 1987. Stable isotopes in ecosystem studies. AnnualReview of Ecology and Systematics 18:293–320.

Post, J. R., and E. A. Parkinson. 2001. Energy allocation strategy in young fish:allometry and survival. Ecology 82:1040–1051.

Rand, P. S., S. G. Hinch, J. Morrison, M. G. G. Foreman, M. J. MacNutt, J. S.Macdonald, M. C. Healey, A. P. Farrell, and D. A. Higgs. 2006. Effects of riverdischarge, temperature, and future climates on energetics and mortality ofadult migrating Fraser River Sockeye Salmon. Transactions of the AmericanFisheries Society 135:655–667.

Randall, R. G. 2010. Quantifying the fluvial habitat needed for the re-introduction of Atlantic Salmon (Salmo salar) in Lake Ontario. CanadianScience Advisory Secretariat Research Document 2009/110.

Roff, D. A. 1983. An allocation model of growth and reproduction in fish.Canadian Journal of Fisheries and Aquatic Sciences 40:1395–1404.

Ruzzante, D. E., M. M. Hansen, D. Meldrup, and K. M. Ebert. 2004. Stockingimpact and migration pattern in an anadromous Brown Trout (Salmo trutta)

complex: where have all the stocked spawning sea trout gone? MolecularEcology 13:1433–1445.

Schjørring, S. 2002. The evolution of informed natal dispersal: inherent versusacquired information. Evolutionary Ecology Research 4:227–238.

Schmetterling, D. A. 2000. Redd characteristics of fluvial Westslope CutthroatTrout in four tributaries to the Blackfoot River, Montana. North AmericanJournal of Fisheries Management 20:776–783.

Scott, R. J., R. Kosick, M. Clement, D. L. G. Noakes, and F. W. H. Beamish.2005. Nest site selection and spawning by captive bred Atlantic Salmon,Salmo salar, in a natural stream. Environmental Biology of Fishes 74:309–321.

Smith, S. H. 1995. Early changes in the fish community of Lake Ontario. GreatLakes Fishery Commission Technical Report 60.

Stanfield, L. W., and M. L. Jones. 2003. Factors influencing rearing success ofAtlantic Salmon stocked as fry and parr in Lake Ontario tributaries. NorthAmerican Journal of Fisheries Management 23:1175–1183.

Stanfield, L. W., and B. W. Kilgour. 2006. Effects of percent imperviouscover on fish and benthos assemblages and instream habitats in Lake On-tario tributaries. Pages 577–599 in R. M. Hughes, L. Wang, and P. W.Seelbach, editors. Landscape influences on stream habitats and biologi-cal assemblages. American Fisheries Society, Symposium 48, Bethesda,Maryland.

Thorstad, E. B., T. F. Næsje, P. Fiske, and B. Finstad. 2003. Effects of hookand release on Atlantic Salmon in the River Alta, northern Norway. FisheriesResearch 60:293–307.

Tillitt, D. E., J. L. Zajicek, S. B. Brown, L. R. Brown, J. D. Fitzsimons, D. C.Honeyfield, M. E. Holey, and G. M. Wright. 2005. Thiamine and thiaminasestatus in forage fish of salmonines from Lake Michigan. Journal of AquaticAnimal Health 17:13–25.

Vander Zanden, M. J., G. Cabana, and J. B. Rasmussen. 1997. Comparingtrophic position of freshwater fish calculated using stable nitrogen isotoperatios (δ15N) and literature dietary data. Canadian Journal of Fisheries andAquatic Sciences 54:1142–1158.

Vander Zanden, M. J., and J. B. Rasmussen. 1999. Primary consumer δ13C andδ15N and the trophic position of aquatic consumers. Ecology 80:1395–1404.

Webb, J. H., and H. A. McLay. 1996. Variation in the time of spawning ofAtlantic Salmon (Salmo salar) and its relationship to temperature in theAberdeenshire Dee, Scotland. Canadian Journal of Fisheries and AquaticSciences 53:2739–2744.

Webster, D. A. 1982. Early history of Atlantic Salmon in New York. New YorkFish and Game Journal 29:26–44.

Witzel, L. D., and H. R. MacCrimmon. 1983. Redd-site selection by BrookTrout and Brown Trout in southwestern Ontario streams. Transactions of theAmerican Fisheries Society 112:760–771.

Youngson, A. F., and E. Verspoor. 1998. Interactions between wild and in-troduced Atlantic Salmon (Salmo salar). Canadian Journal of Fisheries andAquatic Sciences 55(Supplement 1):153–160.

Youngson, A. F., J. H. Webb, J. C. MacLean, and B. M. Whyte. 1997. Frequencyof occurrence of reared Atlantic Salmon in Scottish salmon fisheries. ICESJournal of Marine Science 54:1216–1220.

Zimmer, M. P., and M. Power. 2006. Brown Trout spawning habitat selectionpreferences and redd characteristics in the Credit River, Ontario. Journal ofFish Biology 68:1333–1346.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:15

24

July

201

3



A Comparison of Methods to Estimate ShovelnoseSturgeon Mortality in the Mississippi River Adjacent toMissouri and IllinoisQuinton E. Phelps a , Ivan Vining a , David P. Herzog a , Ross Dames a , Vince H. Travnichek a ,Sara J. Tripp a & Mark Boone aa Missouri Department of Conservation, 2901 West Truman Boulevard , Jefferson City ,Missouri , 65109 , USAPublished online: 24 Jul 2013.

To cite this article: Quinton E. Phelps , Ivan Vining , David P. Herzog , Ross Dames , Vince H. Travnichek , Sara J. Tripp & MarkBoone (2013) A Comparison of Methods to Estimate Shovelnose Sturgeon Mortality in the Mississippi River Adjacent to Missouriand Illinois, North American Journal of Fisheries Management, 33:4, 754-761, DOI: 10.1080/02755947.2013.808291







http://dx.doi.org/10.1080/02755947.2013.808291




ARTICLE

A Comparison of Methods to Estimate Shovelnose SturgeonMortality in the Mississippi River Adjacent to Missouriand Illinois

Quinton E. Phelps,* Ivan Vining, David P. Herzog, Ross Dames,Vince H. Travnichek, Sara J. Tripp, and Mark BooneMissouri Department of Conservation, 2901 West Truman Boulevard, Jefferson City,Missouri 65109, USA

AbstractMortality is a key parameter in understanding the dynamics of any fish population. We examined three methods

to evaluate Shovelnose Sturgeon Scaphirhynchus platorynchus mortality (e.g., ratio of first year recruits to all recruits[Heincke’s method], a linearized weighted catch curve, and an open system mark–recapture mortality approach). TheMississippi River was sampled in two distinct but connected geomorphic sections: upper Mississippi River (UMR;river kilometer 323–587) and the middle Mississippi River (MMR; river kilometer 0–322). All analyses were pooledacross these areas due to potential emigration or immigration throughout both study reaches. Heincke’s methodestimated annual mortality at 16.9% for the full range of ages (9–23 years) with generally increasing estimates withdecreasing age ranges considered (10–23, 11–23, 12–23, and so on). A linearized weighted catch curve that consideredincreasing estimates of mortality for the shorter age ranges generated an annual mortality estimate of 29.0%.d Fourmark–recapture models were considered using the program MARK. The model with the greatest support was themodel that provided estimates of annual mortality for each year and a single recapture probability. The annualmortality estimates from this model varied from 2.7% to 70.7% after correcting for tag loss. The best fit model witha single estimate of annual mortality was the one that estimated annual recapture probabilities for each year and hada mortality estimate of 34.6% (after correcting for tag loss, but not for immigration, emigration, or sampling effort).The three methods provided varying results, and our data indicated that a single method to estimate ShovlenoseSturgeon mortality rate may not be appropriate. As such, biologists must recognize that disparities in ShovelnoseSturgeon mortality rates may exist using various methods and should use caution when choosing which method willbe employed to estimate sturgeon mortality.

All 27 species of sturgeon (family Acipenseridae) worldwideare listed as threatened or endangered by the Convention on In-ternational Trade in Endangered Species of Wild Fauna andFlora (CITES) (Pikitch et al. 2005). Seven of the nine species inthe United States are listed under the U.S. Endangered SpeciesAct (Pikitch et al. 2005; Jelks et al. 2008). Three species of stur-geon co-occur in the Mississippi River basin (Pallid SturgeonScaphirhynchus albus, Lake Sturgeon Acipenser fulvescens, andShovelnose Sturgeon S. platorynchus). These three sturgeonstocks have been declining since the early 1900s (Coker 1929;

*Corresponding author: [email protected] August 1, 2012; accepted May 16, 2013

Birstein 1993; Boreman 1997; Keenlyne 1997; Saffron 2002;Colombo et al. 2007; Tripp et al. 2009) and Shovelnose Stur-geon have received minimal attention (Phelps and Tripp 2011).Shovelnose Sturgeon populations outside of Missouri have beena management concern during the last decade because of theirshrinking range, record harvest levels, declining catch rates,and increasing mortality rates (Keenlyne 1997; Morrow et al.1998; Quist et al. 2002; Colombo et al. 2007; Koch et al. 2009;Tripp et al. 2009). Several authors suggested that commercialexploitation of Shovelnose Sturgeon will increase dramatically

754

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:16

24

July

201

3

SHOVELNOSE STURGEON MORTALITY IN THE MISSISSIPPI 755

if other sturgeon populations from the Black Sea collapse ormarket pressures increase (Birstein 1993; Keenlyne 1997; Quistet al. 2002; Secor et al. 2002; Pala 2005).

Because of the Shovelnose Sturgeon population declines(Keenlyne 1997), the state of Missouri recently began enactingregulatory changes to protect Shovelnose Sturgeon populationsfrom overharvest. For example, Shovelnose Sturgeon commer-cial harvest regulations were implemented on the MississippiRiver in 2006. There was a change from a 762-mm maximumlength limit to a 610–813-mm harvest slot-length limit (whichis also the Illinois regulation). However, recent evidence of in-creasing mortality rates within the Mississippi River indicatesthat existing regulations may be inadequate (Colombo et al.2007) and that age-8 and older Shovelnose Sturgeon may ex-perience a 98% decline by 2016 (Tripp et al. 2009). This ev-idence supports a strong need for continued evaluation of theShovelnose Sturgeon population and harvest regulations. This isespecially true for some of the upper Mississippi River (UMR)pools because these pools represent one of only a few areas thatremains open to commercial Shovelnose Sturgeon harvest.

Because fish stocks are usually managed with a goal of main-taining sustainable harvests through regulation of fishing mor-tality (Hilborn and Walters 1992), mortality must be quantified.Of the several methods used to estimate mortality the mostcommon use mark–recapture (Pine et al. 2003), catch at age(Ricker 1975; Miranda and Bettoli 2007), maximum observedage (Hoenig 1983), life history theory (Roff 1984; Jensen 1996),and empirical relationships (Pauly 1980; Gunderson 1997). Thelatter three approaches provide estimates of natural mortality.Each method has its benefits and deficits. Analysis of mark–recapture data can be one of the most reliable methods to esti-mate total mortality, particularly natural mortality (Vetter 1988).Catch-at-age data are commonly used to estimate mortality forcommercially exploited species, but inaccurate age assignmentto individual fish in a sample can bias mortality estimates. Itis well documented that sturgeon species are difficult to age(Brennan and Cailliet 1989; Rossiter et al. 1995; Whitemanet al. 2004); however, Shovelnose Sturgeon pectoral fin raysprovide the most precise age estimates (Jackson et al. 2007)and are readily used (Colombo et al. 2007; Tripp et al. 2009).Mortality rates (e.g., rates above a predetermined threshold)of Shovelnose Sturgeon populations have been used to estab-lish harvest regulations (WDNR 2007; Koch et al. 2009; D. P.Herzog, unpublished; G. Scholten, Iowa Department of Natu-ral Resources, personal communication), but the method thatprovides the most reliable mortality estimate remains unclear.Our objective was to evaluate three mortality estimators forShovelnose Sturgeon captured by Missouri Department of Con-servation (MDC) employees in the upper Mississippi River andthe middle Mississippi River adjacent to Missouri and Illinoisby using the ratio of first-year recruits to all recruits (Heincke’smethod; Ricker 1975), a linearized weighted catch curve (Mi-randa and Bettoli 2007), and an open system mark–recapturemortality approach (Nichols 2005).

METHODSStudy area.—Sampling for all methods to estimate mortal-

ity (i.e., Heincke’s method, linearized weighted catch curve,and open systems mark recapture) was completed by MDCemployees. The Mississippi River was sampled in two geo-morphically distinct but connected sections: upper MississippiRiver (UMR; river kilometer [rkm] 323–587, from near Alton,Illinois, to Keokuk, Iowa) and the middle Mississippi River(MMR; rkm 0–322, from Cairo, Illinois, to the Missouri Riverconfluence). Specific sampling for each method to estimate mor-tality is described in detail below. Although river sections areseparated spatially, all analyses were pooled across these ar-eas due to potential emigration or immigration and subsequentmixing (Tripp and Garvey 2010; Phelps and Tripp 2011; Phelpset al. 2012; S. J. Tripp, R. C. Brooks, Kentucky Department ofFish and Wildlife Resources, D. P. Herzog, and J. E. Garvey,Southern Illinois University, unpublished results).

Heincke’s method and linerarized weighted catch curve.—Shovelnose Sturgeon used to generate mortality estimates usingHeincke’s method and the linearized weighted catch curve werecollected in the UMR below each of the following six lock anddams (LD) near tailwater habitats (stretch of river directly be-low dam): Winfield, Missouri (LD25, rkm 388); Clarksville,Missouri (LD24, rkm 439); Saverton, Missouri (LD22, rkm484); Quincy, Illinois (LD21, rkm 523); Canton, Missouri(LD20 rkm 552); and Keokuk, Iowa (LD19, rkm 586). TheMMR was sampled near Cape Girardeau, Missouri (rkm 72);near Chester, Illinois (rkm 161); and near Chain of Rocksweir (rkm 306). The specific habitats sampled were the chan-nel border (area outside the thalweg) and channel border dikes(area outside the thalweg with channel training structures). Weused stationary experimental gill nets (3.05-m-deep panels, each7.62 m in length with alternating mesh sizes of 3.81, 5.08, 7.62,and 10.16 cm, repeated twice, for a total net length of 60.96 m)that were set overnight during from November 2009 throughApril 2010. The gill nets used in these analyses were the samegear that were used in MDCs sturgeon monitoring project andprobably provide a representative sample of the ShovelnoseSturgeon population (MDC 2001). Shovelnose Sturgeon weremeasured (FL, mm) and weighed (g), and a pectoral fin ray wasremoved for age assignment.

Left pectoral fin rays were removed from approximately150 Shovelnose Sturgeon per site. They were removed distalto the articulating process using a knife following proceduresdescribed in Koch et al. (2008), then placed in coin envelopesand dried. The portion of the fin ray distal to the articulatingprocess was cross-sectioned for aging. Three sections were cutfrom the basal portion of each fin ray using a Buhler Isometlow speed saw. Each section (0.635 mm thick) was secured to amicroscope slide using cyanoacrylate. Cross sections were an-alyzed by two readers using a stereomicroscope under 7–45×magnification. Under transmitted light, the cross section wascomposed of alternating bands of translucent (fast growth) andopaque (slow growth) rings; each outer edge of an opaque band

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:16

24

July

201

3

756 PHELPS ET AL.

was considered an annulus (Everett et al. 2003). The annuli werecounted from the nucleus to the apex of each section. When read-ers disagreed, they examined the cross sections together to reachconsensus.

Using these age-frequency data, Heincke’s method was usedto quantify mortality rates for the population. Heincke’s methoduses the youngest fully-recruited age in proportion to the totalnumber of fish in the sample to generate a mortality estimate.The Heincke equation takes the form: A = n/N, where A istotal annual mortality, n is the number of fish in the youngestage-class considered fully recruited into the sampling gear, andN is the total number of recruited fish in the sample; SE was

calculated as SEA =√

A(1−A)N .

The linearized weighted catch curve estimates the total in-stantaneous mortality rate (Z) (Ricker 1975). The equation is alinear regression in which the natural log of the number of fishat age y is regressed against age x (Ricker 1975). The catch-curve analysis equation is loge(Nt) = loge(N0) – Z(t) where Nt

is the number alive at time t, N0 is the number alive initially(at time t0), Z is total instantaneous mortality rate, and t is thetime elapsed since t0. Additionally, frequency in each age-classwas weighted to reduce bias that might have occurred due to re-duced relative abundance of older individuals in the population(Slipke and Maceina 2000). The declining slope of this regres-sion equation represents Z, which is used to determine A usingthis equation: A = 1 − e−Z for the population. For the linearizedweighted catch curve, SE for annual mortality is the same asthe SE for the estimate of survival (S) and was calculated asSES = SEA = S (SEZ ), where SEZ is the SE for Z from theweighted regression (Quinn and Deriso 1999).

Using these mathematical methods, we altered the age inter-vals (by changing the first fully recruited age) of the descendinglimb of the age-frequency distribution (e.g., 9–23, 10–23, 11–23, and so on; see Table 1) to calculate eight distinct mortalityestimates using Heincke’s method and linearized weighted catchcurve. By doing this, we qualitatively tested the assumptions ofconstant recruitment, constant mortality, or aging discrepancies.If these assumptions were not violated, then the estimated an-nual mortality for each age range (9–23, 10–23, 11–23, and soon; see Table 1) should have been similar.

Open-system mark–recapture.—For the open-systemmark–recapture portion of this study, Shovelnose Sturgeonwere captured using a stratified random design incorporatingentanglement gear, trotlines, and bottom trawling using theLong Term Resource Monitoring Program (LTRMP) guidelines(Gutreuter et al. 1995). Missouri Department of Conservationemployees sampled Shovelnose Sturgeon using this approachin the MMR and UMR from January 2002 to September2010 to perform the open-system mark–recapture mortalityestimate (Nichols 2005). All Shovelnose Sturgeon collectedwere marked with individually numbered T-bar tags that wereinserted in the fleshy portion adjacent to the dorsal fin of eachfish. The size of fish caught during the project varied from 200to 850 mm FL, many of which were below the legal commercial

harvest slot-length limit of approximately 610–813 mm FL.Thus, only Shovelnose Sturgeon greater than 550 mm FLwere used in the mark–recapture analysis (see Figure 2). Weincluded fish smaller than the minimum length limit becauseof supposed incidental harvest of sublegal sturgeon. We alsochose this length because this is the start of the descendinglimb of the catch curve used for both Heincke’s method andlinearized weighted catch curve. As a result all three mortalityestimators were based on individuals of similar size and age.

Shovelnose Sturgeon recaptured during the tagging oper-ations were used to estimate recapture probabilities and sur-vival in an open population, which is sometimes referred to asconditional Cormack–Jolly–Seber modeling using the programMARK (Nichols 2005; White 2011). The four models testedwere the four general models as follows associated with anopen-population analysis for estimating survival (φ) and returnrate ( p) (Nichols 2005): (1) single estimates for all years ofsurvival and recapture probabilities, φ(.)p(.) (notation as used inWhite (2011) to represent specific models), (2) single estimatefor all years of survival with yearly (t) varying recapture prob-abilities (separate estimate of recapture probabilities for eachyear), φ(.)p(t), (3) yearly varying survival (separate estimate ofsurvival probabilities for each year) with a single estimate forall years of recapture probabilities, φ(t)p(.), and (4) both yearly-varying survival and recapture probabilities, φ(t)p(t). Commer-cial harvest data reported by commercial fishers from Illinoisand Missouri were available for the recapture years 2003–2010and were used as a covariate for yearly survival estimates, al-though the data did not improve any individual model. As such,these data are not provided in the results. These harvest data werethe only consistent harvest data available for the years being ana-lyzed; thus, no other harvest data were considered. Other covari-ates, such as environmental factors, were not used as covariatesdue to the variation in conditions of the Mississippi River overthe sampling period within a year and sparseness of data spa-tially and temporally throughout the upper and middle Missis-sippi River. Comparisons among the different mark–recapturemodels were performed using an information theoretic approach(Burnham and Anderson 2002). A goodness-of-fit test was per-formed using a median-c approach, a robust method available inMARK to test for overdispersion (White 2011). The model withthe highest Akaike’s information criterion (AIC) as adjusted forsmall sample sizes (AICc) or quasi-AIC (QAICc) weight (usedwhen adjustments due to overdispersion are necessary) was con-sidered the best model. The model with the best single survivalestimate for all years was also reported for comparison with theHeincke’s method and the linearized weighted catch curve.

RESULTS

Heincke’s Method and Linerarized Weighted Catch CurveDuring November 2009 through April 2010, 1,133 Shov-

elnose Sturgeon were captured using experimental gill netsby MDC employees in the UMR and the MMR to calculate

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:16

24

July

201

3


B.N = 1,133

Age (years)

5 10 15 20

Fre

qeu

ncy

0

20

40

60

80

100

120

140

160

A.N = 1,133Mean Length at age 9 = 571 (4.1)

Fork length (mm)

300 400 500 600 700 800

Fre

qeu

ncy

0

20

40

60

80

100

120

FIGURE 1. (A) Length-frequency distribution and (B) age-frequency distribu-tion for Shovelnose Sturgeon captured by MDC employees using experimentalgill nets from November 2009 through April 2010 in the upper Mississippi River(rkm 323–587, near Alton, Illinois, to Keokuk, Iowa) and the middle MississippiRiver (rkm 0–322, Cairo, Illinois, to the Missouri River confluence).

mortality using Heincke’s method and the linearized weightedcatch curve. In a completely seperate evaluation, 13,294 Shovel-nose Sturgeon from the UMR and MMR were collected by MDCemployees from January 2002 through September 2010 com-bined to complete an open-systems mark–recapture estimate ofmortality.

Fin rays were removed from the 1,133 Shovelnose Sturgeoncaught between November 2009 and April 2010 and all subse-quent aging analyses were based on those fish. Fork length var-ied from 270 to 770 mm, mean length at age 9 was 571 mm (SE =4.1), and age varied from 3 to 23 years (Figure 1). Heincke’sestimate of mortality was 16.9% for the full range of ages (9–23 years) and varied from 17.9% to 50.6% for the other ageranges indicating potential violation of assumptions associatedwith this model (Table 1). The linearized weighted catch curve

generated an annual mortality estimate of 29.0%, which variedfrom 31.1% to 41.0% for the other age ranges (Table 1).

Open-System Mark–RecaptureFrom January 2002 to September 2010, MDC employees

tagged 16,582 Shovelnose Sturgeon in the MMR, and an addi-tional 3,348 sturgeon were tagged in the UMR during this timeperiod. Due to the relatively recent tagging of some Shovel-nose Sturgeon in the UMR and the low number of fish taggedin the UMR (n = 3,348), the tagging data from the MMR andUMR were pooled for this analysis. To accommodate compar-isons with the other methods only Shovelnose Sturgeon at least550 mm FL were used, which provided 13,294 marked Shov-elnose Sturgeon for the mark–recapture analysis (Figure 2). Atotal of 266 Shovelnose Sturgeon were recaptured at least once

B.N = 266

Fork length (mm)

550 600 650 700 750

Fre

qu

ency

0

5

10

15

20

25

30

A. N = 13,294

Fork length (mm)

550 600 650 700 750 800

Fre

qu

ency

0

200

400

600

800

1000

1200

1400

FIGURE 2. Length-frequency distribution for Shovelnose Sturgeon capturedby MDC employees with entanglement gear, trotlines, and bottom trawling usingthe Long Term Resource Monitoring Program (LTRMP) guidelines (Gutreuteret al. 1995) from January 2002 through September 2010 in the upper MississippiRiver (rkm 323–587, near Alton, Illinois, to Keokuk, Iowa) and the middleMississippi River (rkm 0–322, Cairo, Illinois, to the Missouri River confluence).The top panel (A) represents Shovelnose Sturgeon marked and the bottom panel(B) are those Shovelnose Sturgeon that were recaptured during the evaluation.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:16

24

July

201

3

758 PHELPS ET AL.

TABLE 1. Annual mortality estimates for Shovelnose Sturgeon in the upper Mississippi River and middle Mississippi River combined, using three methods:linearized weighted catch curve (Catch curve), Heincke’s method (ratio of first year recruits to all recruits; Heincke’s), and open-systems mark–recapture(Mark–recapture).

Catch-curve Heincke’s Mark–recapture

Age (years)Point

estimate SEPoint

estimate SE YearPoint

estimate SE

Single estimate,all ages

0.290 0.0455 0.169 0.0133 Single estimate,all years

0.346 0.0647

9–23 0.290 0.0455 0.169 0.0133 2002–2003 0.365 0.253310–23 0.311 0.0453 0.179 0.0149 2003–2004 0.442 0.237811–23 0.335 0.0435 0.190 0.0169 2004–2005 0.585 0.186512–23 0.365 0.0400 0.230 0.0201 2005–2006 0.070 0.168713–23 0.394 0.0343 0.314 0.0252 2006–2007 0.027 0.164014–23 0.410 0.0371 0.401 0.0322 2007–2008 0.388 0.177215–23 0.408 0.0473 0.417 0.0418 2008–2009 0.473 0.212216–23 0.398 0.0622 0.506 0.0556 2009–2010 0.707 0..2976

Average 0.364 0.301 0.382Median 0.379 0.272 0.415

(Figure 2) (21 were recaptured twice and four were recapturedthree times), yielding an average recapture rate of 2.0%. Al-though the recapture rate was relatively low, the sample sizewas large enough to estimate survival and detection probabili-ties. Furthermore, these are similar to recapture rates provided inother sturgeon mark–recapture evaluations to evaluate survivaland mortality (Steffensen et al. 2010).

Four mark–recapture models were considered using the pro-gram MARK (Table 2). The median-c method estimated a cvalue of 1.62, indicating some overdispersion in the model fit.All model parameter SE-values were adjusted using this esti-mated c (Burnham and Anderson 2002; White 2011). The modelwith the greatest support (QAICc weight = 0.9350) was themodel that provided estimates of annual mortality for each yearand a single recapture probability (Table 2). The annual mortal-ity estimates from this model varied from 2.7% to 70.7% after

correcting for tag loss (Table 1). The best-fit model (QAICc

weight = 0.0538) with a single estimate of annual mortalitywas the one that estimated annual recapture probabilities foreach year and had a mortality estimate of 34.6% after correctingfor tag loss.

DISCUSSIONMortality rates are an integral component of developing har-

vest regulations for Shovelnose Sturgeon. However, there is alack of consensus in the literature on the preferred technique.Using common mortality estimation methods, we have demon-strated that different techniques can provide varying results.These differences are probably attributed to the assumptionsassociated with each method. Both Heincke’s method and thelinearized weighted catch-curve method assume accurate aging,

TABLE 2. Comparison of the four different upper Mississippi River and middle Mississippi River Shovelnose Sturgeon models used in the mark–recaptureanalysis using the number of parameters estimated, quasi-Akaike’s information criterion (QAICc), delta quasi-AIC (�QAICc), quasi-AIC weights (QAICc

weights).

ModelNumber ofparameters QAICc �QAICc QAICc weights

Different mortalities for each year, single recaptureprobability

9 1,783.15 0.0000 0.9350

Different mortalities for each year, differentrecapture probabilities for each year

15 1,792.13 8.9792 0.0105

Single mortality, different recapture probabilitiesby year

9 1,788.86 5.7101 0.0538

Single mortality, single recapture probability 2 1,797.49 14.3400 0.0072

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:16

24

July

201

3


constant recruitment, constant harvest, and constant mortality(Hilborn and Walters 1992; Miranda and Bettoli 2007).

First, Shovelnose Sturgeon aging has not been validatedand some authors report that counting annuli on sturgeon finrays is difficult (Rien and Beamesderfer 1994; Paragamian andBeamesderfer 2003; Hurley et al. 2004; Koch et al. 2011), butfin rays are the most precise aging method currently availablefor aging Shovelnose Sturgeon (Jackson et al. 2007). In ourstudy, FL frequency distributions, age estimates, and length-at-age data were within the range of other similar studies (Figure 1;Morrow et al. 1998; Kennedy et al. 2007), suggesting a poten-tial lack of aging error. As for other assumptions associated withcatch-at-age models, constant recruitment is rarely observed infish stocks (Maceina and Pereira 2007; Miranda and Bettoli2007; Phelps et al. 2010). We have no way to gauge the amountof variation in the exploitation rate of this stock of ShovelnoseSturgeon, because most fish stocks have varying harvest rates(Hilborn and Walters 1992; Quinn and Deriso 1999). Further-more, weather and river conditions may play a key role in theability of commercial fishers to successfully harvest riverinefishes (Scholten and Bettoli 2005), or harvest records are un-derreported (D. P. Herzog, unpublished data). Lastly, the knownage of full recruitment is also unclear because Shovelnose Stur-geon do not spawn every year, the female portion of the stockis not susceptible to the roe fishery every year, and males willrarely be harvested (though some nongravid females and malesare probably taken for their flesh). In many situations, includingthis study, the assumptions associated with catch-at-age mod-els may have been violated so biologists must recognize theseassumptions and minimize their influence.

Furthermore, the mark–recapture analysis also has uncer-tainty associated with it. Tag-retention rates may influence themortality rates generated by the MARK model. However, pre-liminary assessments suggest tag retention was high (>90%tag retention: V. H. Travnichek, unpublished data), and this in-formation was used to adjust mortality estimates in our study.These estimates are similar to those in another recently pub-lished Shovelnose Sturgeon tag-retention study (Hamel et al.2012). Immigration, emigration, and inconsistencies in sam-pling can also bias the results of the mark–recapture analysis(Nichols 2005; Koch et al. 2012). For example, more than 30Shovelnose Sturgeon tagged in the Missouri River have beenrecaptured in the MMR or UMR, and over 100 sturgeon taggedin the MMR and UMR have been recaptured in the MissouriRiver (MDC, unpublished data). However, due to low recapturerates we have no way of knowing whether the sturgeon returnto their original tagging area or stay in the location where theywere recaptured. As with the other methods of mortality esti-mation (above), minimizing the influence of model assumptionswill yield more precise estimates.

Management ImplicationsWe have amassed a comprehensive assessment of Shovel-

nose Sturgeon mortality in the upper and middle Mississippi

rivers using multiple methods. Although there are limitationsassociated with each method tested, these data provided insightinto the mortality rates of Shovelnose Sturgeon in the UMR andMMR. Many other studies have evaluated Shovelnose Sturgeonmortality to estimate either a reach-specific rate or a site-specificrate. Multiple studies have evaluated Shovelnose Sturgeon mor-tality in the Mississippi River and showed that mortality has var-ied between approximately 20% and 57%, and the consensus isthat mortality should be considered when forming managementstrategies (Morrow et al. 1998; Colombo et al. 2007; Casto-Yerty2009). Specifically, Goodyear (1993) and Bajer and Wildhaber(2007) modeled populations under their reach-specific mortalityrate and were able to predict future effects on the fishery undervarying levels of harvest. However, not accounting for uncer-tainty or variability in terms of mortality estimates may lead toerroneous management decisions.

We agree that mortality must be used in population evalu-ation; however, the three methods employed in this study pro-vided varying results and indicated a single mortality estimatefrom an individual method may not be appropriate. For both theHeincke’s method and linearized weighted catch curve, therewas a notable change in mortality from around age 12–14, atwhich point the linearized weighted catch-curve mortality es-timates increased and the associated SE declined (Table 1). Tominimize assumption violations with these catch-at-age models,managers should validate their aging technique and monitor re-cruitment and harvest to ensure their mortality models are valid(Hilborn and Walters 1992; Miranda and Bettoli 2007). For themark–recapture analysis, varying annual survival provided themost parsimonious fit to the data (qAICc weight = 0.9350; Ta-ble 2), indicating that the increase in uncertainty due to moreparameters is justified if it provides a better fit of the data to themodel (Burnham and Anderson 2002). Furthermore, to reducethis uncertainty and to acquire the best estimate of mortalityusing mark–recapture analyses, biologists must quantitativelyassess immigration and emigration, develop a standardized sam-pling protocol, and quantitatively evaluate tag retention.

To this end, all three methods to evaluate mortality provideddifferent results. Therefore, these discrepancies in mortalitymay lead to differing management decisions. For example, theShovelnose Sturgeon mortality estimate of 16.9% (Heincke’smethod) would probably be managed much differently (e.g.,more stringent regulations) than the 34.6% mortality rate es-timate generated with the open-system mark–recapture model,especially if fishing mortality accounts for the majority of to-tal mortality As such, fishery management biologists must usecaution to minimize associated assumption violations for eachmethod of estimating mortality. We suggest fishery manage-ment biologists use multiple mortality estimates (using severalmethods) in conjunction with ongoing monitoring of length-frequency and age-frequency distributions to ensure sustainabil-ity. Furthermore, developing a long-term standardized samplingprogram (i.e., continual sampling) would allow cohorts to be fol-lowed over time, thus minimizing the assumptions associated

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:16

24

July

201

3

760 PHELPS ET AL.

with mortality estimates. Therefore, more accurate estimates ofpopulation parameters could be determined.

ACKNOWLEDGMENTSWe would like to thank Fisheries and Resource Science di-

visions of the Missouri Department of Conservation for pro-viding support and funding to carry out this study. This studywas partially funded by the U.S. Army Corps of Engineers’Upper Mississippi River Restoration–Environmental Manage-ment Program’s Long Term Resource Monitoring componentimplemented by the U.S. Geological Survey, Upper MidwestEnvironmental Sciences Center and carried out by the MissouriDepartment of Conservation.

REFERENCESBajer, P. G., and M. L. Wildhaber. 2007. Population viability analysis of lower

Missouri River Shovelnose Sturgeon with initial application to the PallidSturgeon. Journal of Applied Ichthyology 23:457–464.

Birstein, V. J. 1993. Sturgeons and paddlefishes: threatened fishes in need ofconservation. Conservation Biology 7:773–787.

Boreman, J. 1997. Sensitivity of North American sturgeons and paddlefish tofishing mortality. Environmental Biology of Fishes 48:399–405.

Brennan, J. S., and G. M. Cailliet. 1989. Comparative age-determination tech-niques for White Sturgeon in California. Transactions of the American Fish-eries Society 118:296–310.

Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodelinference: a practical information-theoretic approach, 2nd edition. Springer-Verlag, New York.

Casto-Yerty, M. A. 2009. Population characteristics of exploited ShovelnoseSturgeon and assessment of a bridge-mounted hydroacoustic telemetry ar-ray on the lower Mississippi River, Tennessee. Master’s thesis. TennesseeTechnological University, Cookeville.

Coker, R. E. 1929. Studies of common fishes of the Mississippi River at Keokuk.U.S. Bureau of Fisheries Bulletin 45:141–225.

Colombo, R. E., J. E. Garvey, N. D. Jackson, R. Brooks, D. P. Herzog,R. A. Hrabik, and T. W. Spier. 2007. Harvest of Mississippi River sturgeondrives abundance and reproductive success: a harbinger of collapse? Journalof Applied Ichthyology 23:444–451.

Everett, S. R., D. L. Scarnecchia, G. J. Power, and C. J. Williams. 2003. Compar-ison of age and growth of Shovelnose Sturgeon in the Missouri and Yellow-stone rivers. North American Journal of Fisheries Management 23:230–240.

Goodyear, C. P. 1993. Spawning stock biomass per recruit in fisheries manage-ment: foundation and current use. Canadian Special Publication of Fisheriesand Aquatic Sciences 120:67–81.

Gunderson, D. R. 1997. Trade-off between reproductive effort and adult survivalin oviparous and viviparous fishes. Canadian Journal of Fisheries and AquaticSciences 54:990–998.

Gutreuter, S., R. Burkhardt, and K. Lubinski. 1995. Long term resource mon-itoring program procedures: fish monitoring. National Biological Service,Environmental Management Technical Center, Program Report 95-P002-1,Onalaska, Wisconsin.

Hamel, M. J., J. J. Hammen, and M. A. Pegg. 2012. Tag retention of T-baranchor tags and passive integrated transponder tags in Shovelnose Sturgeon.North American Journal of Fisheries Management 32:533–538.

Hilborn, R., and C. J. Walters. 1992. Quantitative fisheries stock assessment:choice, dynamics and uncertainty. Chapman and Hall, New York.

Hoenig, J. M. 1983. Empirical use of longevity data to estimate mortality rates.U.S. National Marine Fisheries Service Fishery Bulletin 81:898–903.

Hurley, K. L., R. J. Sheehan, and R. C. Heidinger. 2004. Accuracy and precisionof age estimates for Pallid Sturgeon from pectoral fin rays. North AmericanJournal of Fisheries Management 24:715–718.

Jackson, N. D., J. E. Garvey, and R. E. Colombo. 2007. Comparing agingprecision of calcified structures in Shovelnose Sturgeon. Journal of AppliedIchthyology 23:525–528.

Jelks, H. L., S. J. Walsh, N. M. Burkhead, S. Contreras-Balderas, E. Dıaz-Pardo,D. A. Hendrickson, J. Lyons, N. E. Mandrak, F. McCormick, J. S. Nelson,S. P. Platania, B. A. Porter, C. B. Renaud, J. J. Schmitter-Soto, E. B. Taylor,and M. L. Warren Jr. 2008. Conservation status of imperiled North Americanfreshwater and diadromous fishes. Fisheries 33:372–407.

Jensen, A. L. 1996. Beverton and Holt life history invariants result from optimaltrade-off of reproduction and survival. Canadian Journal of Fisheries andAquatic Sciences 53:820–822.

Keenlyne, K. D. 1997. Life history and status of the Shovelnose Sturgeon,Scaphirhynchus platorynchus. Pages 291–298 in V. J. Birstein, J. R. Waldman,and W. E. Bemis, editors. Sturgeon biodiversity and conservation. KluwerAcademic Publishers, New York.

Kennedy, A. J., D. J. Daugherty, T. M. Sutton, and B. E. Fisher. 2007. Populationcharacteristics of Shovelnose Sturgeon in the upper Wabash River, Indiana.North American Journal of Fisheries Management 27:52–62.

Koch, B., R. C. Brooks, A. Oliver, D. Herzog, J. E. Garvey, R. Hrabik, R.Colombo, Q. Phelps, and T. Spier. 2012. Habitat selection and movement ofnaturally occurring Pallid Sturgeon in the Mississippi River. Transactions ofthe American Fisheries Society 141:112–120.

Koch, J. D., M. C. Quist, C. L. Pierce, K. A. Hansen, and M. J. Steuck. 2009.Effects of commercial harvest on Shovelnose Sturgeon populations in theupper Mississippi River. North American Journal of Fisheries Management29:84–100.

Koch, J. D., W. J. Schreck, and M. C. Quist. 2008. Standardised removal andsectioning locations for Shovelnose Sturgeon fin rays. Fisheries Managementand Ecology 15:139–145.

Koch, J. D., K. D. Steffensen, and M. A. Pegg. 2011. Validation of age estimatesobtained from juvenile Pallid Sturgeon Scaphirhynchus albus pectoral finspines. Journal of Applied Ichthyology 27:209–212.

Maceina, M. J., and D. L. Pereira. 2007. Recruitment. Pages 121–185 inC. S. Guy and M. L. Brown, editors. Analysis and interpretation of freshwaterfisheries data. American Fisheries Society, Bethesda, Maryland.

MDC (Missouri Department of Conservation). 2001. Monitoring and evaluationproject proposal: monitoring of sturgeon populations in Missouri. MDC,Jefferson City.

Miranda, L. E., and P. W. Bettoli. 2007. Mortality. Pages 229–277 in C. S. Guyand M. L. Brown, editors. Analysis and interpretation of freshwater fisheriesdata. American Fisheries Society, Bethesda, Maryland.

Morrow, J. V., Jr., J. P. Kirk, K. J. Killgore, and S. G. George. 1998. Age, growthand mortality of Shovelnose Sturgeon in the lower Mississippi River. NorthAmerican Journal of Fisheries Management 18:725–730.

Nichols, J. D. 2005. Modern open-population capture–recapture models. Pages88–123 in S. C. Amstrup, T. L. McDonald, and B. F. J. Manly, editors. Hand-book of capture–recapture analysis. Princeton University Press, Princeton,New Jersey.

Pala, C. 2005. Ban on Beluga caviar points to sturgeon’s worldwide decline.Science 310:37.

Paragamian, V. L., and R. C. P. Beamesderfer. 2003. Growth estimates fromtagged White Sturgeon suggest that ages from fin rays underestimate trueage in the Kootenai River, USA and Canada. Transactions of the AmericanFisheries Society 132:895–903.

Pauly, D. 1980. On the interrelationships between natural mortality, growthparameters, and mean environmental temperature in 175 fish stocks. Journaldu Conseil, Conseil International pour l’Exploration de la Mer 39:175–192.

Phelps, Q. E., and S. J. Tripp. 2011. An investigation of age, growth, andmortality of Missouri Shovelnose Sturgeon populations from segments of thelower Missouri and Mississippi rivers. Missouri Department of Conservation,Jefferson City.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:16

24

July

201

3


Phelps, Q. E., S. J. Tripp, W. D. Hintz, J. E. Garvey, D. P. Herzog, D. E.Ostendorf, J. W. Ridings, J. W. Crites, and R. A. Hrabik. 2010. Water temper-ature and river stage influence mortality and abundance of naturally occurringMississippi River Scaphirhynchus sturgeon. North American Journal of Fish-eries Management 30:767–775.

Phelps, Q. E., G. W. Whitledge, S. J. Tripp, K. T. Smith, J. E. Garvey, D.P. Herzog, D. E. Ostendorf, J. W. Ridings, J. W. Crites, R. A. Hrabik,W. J. Doyle, and T. D. Hill. 2012. Identifying river of origin for age-0Scaphirhynchus sturgeons in the Missouri and Mississippi rivers using fin raymicrochemistry. Canadian Journal of Fisheries and Aquatic Sciences 69:930–941.

Pikitch, E. K., P. Doukakis, L. Lauck, P. Chakrabarty, and D. L. Erickson. 2005.Status, trends and management of sturgeon and paddlefish fisheries. Fish andFisheries 6:233–265.

Pine, W. E., K. H. Pollock, J. E. Hightower, T. J. Kwak, and J. A. Rice. 2003. Areview of tagging methods for estimating fish population size and componentsof mortality. Fisheries 28(10):10–23.

Quinn, T. J., II, and R. B. Deriso. 1999. Quantitative fish dynamics. OxfordUniversity Press, New York.

Quist, M. C., C. S. Guy, M. A. Pegg, P. J. Braaten, C. L. Pierce, and V. H.Travnichek. 2002. Potential influence of harvest on Shovelnose Sturgeonpopulations in the Missouri River system. North American Journal of Fish-eries Management 22:537–549.

Ricker, W. E. 1975. Computation and interpretation of biological statis-tics of fish populations. Fisheries Research Board of Canada Bulletin191.

Rien, T. A., and R. C. Beamesderfer. 1994. Accuracy and precision of WhiteSturgeon age estimates from pectoral fin rays. Transactions of the AmericanFisheries Society 123:255–265.

Roff, D. A. 1984. The evolution of life history parameters in teleosts. CanadianJournal of Fisheries and Aquatic Sciences 41:989–1000.

Rossiter, A., D. L. G. Noakes, and F. W. H. Beamish. 1995. Validation of ageestimation for the Lake Sturgeon. Transactions of the American FisheriesSociety 124:777–781.

Saffron, I. 2002. Caviar: the strange history and uncertain future of the world’smost coveted delicacy. Random House, New York.

Scholten, G. D., and P. W. Bettoli. 2005. Population characteristics and as-sessment of overfishing for an exploited Paddlefish population in the lowerTennessee River. Transactions of the American Fisheries Society 134:1285–1298.

Secor, D. H., P. J. Anders, W. Van Winkle, and D. A. Dixon. 2002. Can we studysturgeons to extinction? what we do and don’t know about the conservationof North American sturgeons. Pages 3–10 in W. Van Winkle, P. J. Anders,D. H. Secor, and D. A. Dixon, editors. Biology, management, and protectionof North American sturgeon. American Fisheries Society, Symposium 28,Bethesda, Maryland.

Slipke, J. W., and M. J. Maceina. 2000. Fishery analyses and simulation tools.Auburn University, Auburn, Alabama.

Steffensen, K. D., L. A. Powell, and J. D. Koch. 2010. Assessment of hatchery-reared Pallid Sturgeon survival in the lower Missouri River. North AmericanJournal of Fisheries Management 30:671–678.

Tripp, S. J., R. E. Colombo, and J. E. Garvey. 2009. Declining recruitment andgrowth of Shovelnose Sturgeon in the middle Mississippi River: implicationsfor conservation. Transactions of the American Fisheries Society 138:416–422.

Tripp, S. J., and J. Garvey. 2010. Fish passage in the upper Mississippi River.U.S. Army Corps of Engineers, Rock Island, Illinois.

Vetter, E. F. 1988. Estimation of natural mortality in fish stocks: a review. U.S.National Marine Fisheries Service Fishery Bulletin 86:25–43.

WDNR (Wisconsin Department of Natural Resources). 2007. Regulation to es-tablish length limit on Shovelnose Sturgeon in Iowa and Wisconsin boundarywater. WDNR, Madison.

White, G. C. 2011. Program MARK: mark and recapture parameter estimation,version 5.1. Colorado State University, Fort Collins.

Whiteman, K. W., V. H. Travnichek, M. L. Wildhaber, A. DeLonay, D. Papoulias,and D. Tillett. 2004. Age estimation for Shovelnose Sturgeon: a cautionarynote based on annulus formation in pectoral fin rays. North American Journalof Fisheries Management 24:731–734.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

0:16

24

July

201

3

This article was downloaded by: [Department Of Fisheries]On: 12 August 2013, At: 23:41Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK


Summer Habitat Use of Large Adult Striped Bass andHabitat Availability in Lake Martin, AlabamaSteven M. Sammons a & David C. Glover a ba Department of Fisheries and Allied Aquacultures , Auburn University , 203 Swingle Hall,Auburn , Alabama , 36849 , USAb Center for Fisheries, Aquaculture, and Aquatic Sciences , Southern Illinois University , 1125Lincoln Drive, Carbondale , Illinois , 62901 , USAPublished online: 26 Jul 2013.

To cite this article: Steven M. Sammons & David C. Glover (2013) Summer Habitat Use of Large Adult Striped Bass andHabitat Availability in Lake Martin, Alabama, North American Journal of Fisheries Management, 33:4, 762-772, DOI:10.1080/02755947.2013.806381







http://dx.doi.org/10.1080/02755947.2013.806381




ARTICLE

Summer Habitat Use of Large Adult Striped Bassand Habitat Availability in Lake Martin, Alabama

Steven M. Sammons* and David C. Glover1

Department of Fisheries and Allied Aquacultures, Auburn University, 203 Swingle Hall, Auburn,Alabama 36849, USA

AbstractTo identify the summer habitat use by Gulf strain Striped Bass Morone saxatilis in Lake Martin, Alabama, 36 fish

greater than 4 kg were tracked weekly from June through September in 2009 and 2010. Temperature and dissolvedoxygen profiles were collected biweekly from July to October each year and were incorporated into a hypsographiccurve to estimate summertime volumes of Striped Bass habitat. Striped Bass moved downstream in the reservoirand deeper in the water column as the summer progressed in both years. Across all temperatures and dissolvedoxygen concentrations (DOCs) measured, fish generally selected cooler temperatures when DOCs were greater than3.1 mg/L, but below that level, fish selected warmer temperatures. By early July each year, quality Striped Basshabitat (≤21.0◦C and ≥3.2 mg/L DOC) comprised 40–50% of the total habitat available in Lake Martin. However,in 2009 all quality habitat was gone from the reservoir by August 1, and total habitat (≤25.0◦C and ≥1.6 mg/LDOC) rapidly decreased until there was no suitable habitat found in Lake Martin by mid-September. The quantity ofquality and total habitat declined more rapidly in 2009 than in 2010; although quality habitat was eliminated from thereservoir by the end of August 2010, total habitat persisted throughout the summer. In both years, habitat availabilityincreased when reservoir destratification began. Analysis of historical data indicated that quality habitat was foundin the lower portion of the reservoir during the latter half of August in only 6 of 15 years; total Striped Bass habitatwas available in all but 1 year. The amount of water flowing through the system during the spring and summer was akey determinant of Striped Bass habitat availability by late summer.

Originally restricted to marine and estuarine systems, StripedBass Morone saxatilis have become important sport fish in manyreservoir systems across the southeastern USA (Coutant 1987;Jackson and Hightower 2001; Young and Isely 2002). StripedBass are active, pelagic piscivores that commonly reach a largesize (>15 kg) and prey on other pelagic fishes such as clu-peids (Axon and Whitehurst 1985). Waters in which the tem-perature exceeds 25◦C and the dissolved oxygen concentration(DOC) is below 2 mg/L are considered to be unusable habi-tat for adult Striped Bass (Coutant and Carroll 1980; Coutant1985), and field studies of telemetered Striped Bass have shownthat adult fish commonly use waters that are less than 20◦Cand have a DOC of at least 4 mg/L (Coutant 1985; Bettoli

*Corresponding author: [email protected] address: Center for Fisheries, Aquaculture, and Aquatic Sciences, Southern Illinois University, 1125 Lincoln Drive, Carbondale,

Illinois 62901, USA.Received September 26, 2012; accepted May 15, 2013

2005). Therefore, many populations of Striped Bass are limitedby the amount of cool, oxygenated water found in reservoirsystems during summer (Coutant 1985; Young and Isely 2002;Bettoli 2005). Summer mortality of adult Striped Bass (>5 kg)has been linked to poor environmental conditions that lead tolower body condition, increased disease, and ultimately death(Coutant 1985; Matthews 1985; Moss 1985; Braschler et al.1989). Given that preferred temperatures typically decrease asfish age (Coutant 1985; Matthews 1985), it follows that sum-mer mortality of Striped Bass can be more severe in largerfish (>5 kg). Thus, availability of suitable summertime watertemperatures and DOCs are probably the most important lim-iting factors governing the ability of a reservoir to develop a

762

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013

STRIPED BASS SUMMER HABITAT USE 763

trophy (>10 kg) Striped Bass fishery (Axon and Whitehurst1985; Coutant 1985).

Oxygen is a nonrenewable resource in the cool, deep watersof reservoirs during summer, and availability of this vital sum-mertime habitat for adult Striped Bass can be governed by sev-eral factors. Hydropower generation provides abnormally coolwater downstream from the dam (possibly providing a thermalrefuge for fishes in that system), but discharging from the coolerhypolimnion can deplete the lower strata of waters in the reser-voir contained upstream from the dam (Cole and Hannan 1990).Increases in nutrient inputs either from upstream sources or fromlakeshore development may increase the depletion of dissolvedoxygen from lower water strata through increased respirationor decay (Coutant 1987). Obviously, all of these factors canand often do occur simultaneously, thus affecting the volume ofavailable summertime habitat for Striped Bass in unpredictableways. Coutant (1987) recommended that water resource man-agers should specifically identify water temperatures and DOCsrequired by the species of interest in each water body. Zonesthat meet requirements of these species should be quantifiedand maintained, especially during critical periods when thesezones are likely to become constricted. To date, little attempthas been made to quantify available summer habitat of StripedBass in reservoir systems or to examine how reservoir operationor other factors may affect that quantity.

Lake Martin is a large (16,188 ha) oligomesotrophic trib-utary storage reservoir on the Tallapoosa River in east-centralAlabama (Figure 1). Reaching depths in excess of 45 m, LakeMartin typically stratifies in May and remains stratified untilOctober in most years (D. Bayne, Auburn University, personalcommunication). Since 1978, Striped Bass have been stockedinto Lake Martin on an annual basis by the Alabama Depart-ment of Conservation and Natural Resources (ADCNR). Soonafter these stockings began, a quality fishery for Striped Bassdeveloped, from which anglers caught numerous fish larger than10 kg annually (N. Nichols, ADCNR, personal communication).These Striped Bass were initially Atlantic strain fish, but Gulfstrain fish were stocked into the reservoir beginning in 1989.After the development of this fishery, periodic summer mortal-ities of adult Striped Bass have occurred on Lake Martin (J.Lochamy, Alabama Power Company [APC], personal commu-nication). These fish kills typically occurred in late August tomid-September in the lower section of the reservoir and pre-dominantly affected large (>5 kg) Striped Bass; the most recentof these mortalities occurred during 1991, 1994, and 2001. Thecauses of these mortalities in adult Striped Bass are not knownbut are generally thought to be related to the availability ofcool, oxygenated water during the summertime (N. Nichols,personal communication). Therefore, objectives of this studywere to: (1) quantify movement and habitat use of adult StripedBass in Lake Martin during summer, (2) estimate the approx-imate volume of suitable Striped Bass habitat present in LakeMartin during summer, and (3) examine possible factors affect-ing this volume.

FIGURE 1. Top: Lake Martin, Alabama, showing locations of temperatureand dissolved oxygen profiles taken biweekly in Lake Martin from June toOctober in 2009 and 2010. Bottom: Subbasins designated for determination ofStriped Bass habitat existing in Lake Martin during the summers of 2009 and2010. [Figure available in color online.]

METHODSStriped Bass telemetry.—In March and April 2009, 30

Striped Bass (≥4 kg) were collected using an electrofishing boatwith a boom-mounted electrofisher (Smith Root 7.5 GPP unit)and long lines baited with Goldfish Carassius auratus (Mosset al. 2005). An additional six Striped Bass were collected byelectrofishing in April 2010. All fish were implanted with 25-gradio tags (model F1850, Advanced Telemetry Systems) and 22-g ultrasonic tags (model CTT-83-3I, Sonotronics) following theprocedures of Maceina et al. (1999). Combined, these tags wereless than 2% of the fish’s body weight, following the recommen-dation of Winter (1996). The radio tags had a life expectancyof 1,086 d and were fitted with a mortality sensor. Specifically,if these tags were motionless for at least 24 h due to death or

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013

764 SAMMONS AND GLOVER

expulsion, the signal rate doubled from 50 to 100 pulses per sec-ond. The ultrasonic tags had a life expectancy of 36 months andwere temperature sensitive; transmitted pulses were dependenton the surrounding temperature (Bettoli and Osborne 1998). Therelationship between pulse rate and water temperature was mod-eled for each tag to ensure that the correct temperature occupiedby Striped Bass was determined.

Striped Bass were tracked by boat using either a directionalhydrophone (Sonotronics) or a four-element Yagi directionalantenna (Advanced Telemetry Systems). The location of eachfish was determined by moving along the reservoir until a signalwas detected. Tracking would continue until the radio signalwas strongest when the antenna was pointed at the water orthe sonic pulses were of similar strength as the hydrophone wasrotated 360◦. At this point, the location of each fish was recordedusing a GPS unit. All tracking was conducted by experiencedindividuals who had been trained using known-tag locations(Brenden et al. 2004). In addition, a temperature and dissolvedoxygen profile was taken (1-m intervals) at each fish locationto determine the depth of water where the fish were found todifferentiate these data from the sonic tag temperature data.

In 2009, all Striped Bass equipped with transmitters weretracked once per week from June until the end of September toidentify summer movement and habitat use. In 2010, due to lownumbers of tagged fish, Striped Bass were tracked over a 24-hperiod weekly beginning in late June and ending late Septemberfor a total of twelve 24-h tracking periods. During 2010 trackingevents, up to nine fish were selected randomly and found every4 h for 24 h; these data were used to assess temperature anddissolved oxygen concentration (DOC) selection only. Duringeach tracking period, the precise location (within 5 m) of eachfish was mapped using GPS, and the sonic tag’s pulse rate wasrecorded. Temperature and dissolved oxygen profiles were takenin the vicinity of fish locations to identify water depths andDOCs that fish were using when located.

Daily movements were calculated for each fish in 2009 bydividing the distance moved between locations by the amountof time elapsed (in days) between locations, which representedthe absolute minimum distance traveled (Wilkerson and Fisher1997; Sammons et al. 2003). Fish movements were comparedbetween months using a mixed-model ANOVA (PROC MIXED;SAS 2004) using individual fish as the sampling unit to avoidpseudoreplication (Sammons et al. 2003; Rogers and White2007). Weekly use of water temperatures, DOCs, and depths bytelemetered fish were compared among months in 2009 usingmixed-model ANOVA as described above. All values were loge

transformed to homogenize variances.Relationships among fish TL and temperature, DOC, depth

use, and movement were explored using 2009 data in Pearsoncorrelations (PROC CORR; SAS 2004). Movement and depth-,temperature-, and DOC-use data were divided into two timeperiods: June–July and August–September, and compared withStriped Bass length to explore whether fish size had any ef-fect on movement or habitat use as the summer progressed and

physicochemical conditions changed in Lake Martin. Also,Striped Bass were grouped into two size-classes (<10 kg and>10 kg) and movements, depth use, temperature use, and DOCuse were compared between these size-groups using each fishwithin each size-group and month combination as an individ-ual observation in an ANOVA (SAS 2004). Significance for allstatistical tests was judged at an α-level of 0.10.

Striped Bass habitat selection.—To determine how tempera-ture and DOC affected the probability of Striped Bass presencein the water column, a repeated-measures logistic regressionanalysis was conducted (PROC GENMOD; SAS 2004). EachStriped Bass relocated via telemetry in 2009 and 2010 repre-sented a set of observations taken at 1-mdepth intervals with thecorresponding temperature and DOC, of which only one depthobservation was set to presence (i.e., 1) and all others wereset to absence (i.e., 0). A first-order autoregressive correlationstructure was used to account for repeated observations on eachfish through time, which assumes that observations taken moreclosely in time are more correlated than those farther apart intime. Data were pooled across months (May–October) and years(2009–2010) to determine whether general patterns of selectionby Striped Bass were evident. The logistic regression model was

loge [p/(1 − p)] = α + (β1 × temp) + (β2 × DOC)

+ (β3 × temp × DOC),

where p is the probability of Striped Bass presence, α is theintercept or minimum log odds ratio of Striped Bass presenceexcluding the effects of temperature (temp), DOC, and theirinteraction (temp × DOC), and β1, β2, and β3 are coefficientsdetermining the effect size of temperature, dissolved oxygen,and the interaction between temperature and dissolved oxygen,respectively. Results were then back-transformed to odds ratiosor probabilities, or both, for purposes of interpretation.

Striped Bass habitat availability.—Striped Bass habitat char-acteristics were estimated by examination of temperatures anddepths of Striped Bass equipped with transmitters in 2009 and2010 to define quality and total habitat. Quality habitat wasconsidered to be water with temperatures and DOCs contain-ing 75% of the Striped Bass locations in June and July of bothyears, when the reservoir was stratified, but a wide range of tem-peratures and dissolved oxygen was available to the fish. TotalStriped Bass habitat was defined as water with temperatures andDOCs that contained 95% of the Striped Bass locations duringJune–August of both years, when the reservoir contained a widerange of temperatures but DOCs were more limited. To examinethe effect of fish size on habitat use, Striped Bass were groupedinto the two size-classes described above, and then quality andtotal habitat definitions were examined between size-groups us-ing box plots.

Temperature and dissolved oxygen profiles were collectedin 2009 and 2010 every 2 weeks from July to September inthe main river channels of the three major embayments of Lake

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013


TABLE 1. Definitions of environmental variables used in correlation analysis to determine the factors influencing the availability of Striped Bass habitat presentin Lake Martin, Alabama, in August.

Variable Definition

Cooling degree-days Calculated for Alexander City, Alabama (station 010160), cooling degree days = averagetemperature (◦F) ([High + Low]/2) − 65, must be ≥0, summed for each day over period ofinterest.

Number of days ≥ 32◦C Number of days where the high temperature was ≥32◦C in Alexander City, Alabama (station010160), summed over period of interest.

Precipitation Daily precipitation (cm) at Alexander City, Alabama (station 010160) summed over period ofinterest.

Turbine-hours Turbine-hours for Lake Martin Dam, summed over period of interest; 1 turbine-hour = one turbineoperating during any hour in a 24-h period.

Discharge Discharge from Lake Martin Dam (ha-m), summed over period of interest.Inflow Inflow (ha-m) from two main tributaries to Lake Martin, Hillabee Creek (gauge 0241500) and

Tallapoosa River (gauge 02414715), data from U.S. Geological Survey stream gauges. Inflowswere summed over period of interest.

Martin (Figure 1). Profiles were taken approximately every 2 kmfrom the dam to the point where conditions likely to supportStriped Bass were unable to be found. The digital topographicmap of Lake Martin was divided into 15 subbasins to accountfor longitudinal variations in temperature and dissolved oxy-gen caused by gradients in productivity and flow commonlyobserved in reservoirs (Cole and Hannon 1990; Figure 1). Datafrom this map were used to create a hypsographic curve for eachsubbasin, and volume was calculated for every 1-m stratum ofwater within each subbasin (assuming full pool elevation) usingthe equation in Cole (1983),

VZ x−Z x−1 = 1

3(AZ x−1 + AZ x +

√AZ x × AZ x−1)(Zx − Zx−1),

where V = volume, Z = depth, x = stratum, and A = area ofdepth strata. These volumes were used to estimate the quan-tity of Striped Bass habitat during the summer based on ourtemperature and dissolved oxygen profile stations. Temperatureand dissolved oxygen profiles within each subbasin were aver-aged, a mean profile was created, and Striped Bass habitat wasestimated for each subbasin as described above. Striped Basshabitat estimates from each subbasin were then summed to esti-mate total Striped Bass habitat in the reservoir on each samplingdate.

Striped Bass habitat was estimated in the lower lake sub-basin of Lake Martin from 1994 to 2008 using historical datacollected by the Alabama Department of Environmental Man-agement (ADEM). Two ADEM profile stations approximatelybounded the lower and upper ends of the lower lake subbasinarea, which was probably the most important area of the reser-voir for summertime Striped Bass habitat. Therefore, data fromthese profiles were averaged to create a mean profile and habi-tat was estimated for this area only. These stations were usu-ally sampled each year once per month from April to October;

however, sampling frequency was changed periodically due tologistical constraints (G. Curva, ADEM, personal communica-tion). Examination of these data revealed that data were availablefor mid to late August in every year from 1994 to 2008 exceptfor 1995 and 2001, when no profiles were taken by ADEM any-where in Lake Martin during any month. Thus, Striped Basshabitat availability was estimated for mid to late August eachyear in Lake Martin.

Striped Bass habitat estimates for the lower lake subbasin ob-tained during this study in 2009 and 2010 were added to ADEMdata to create a history of late-August Striped Bass habitat avail-ability in this basin of Lake Martin from 1994 to 2010. Theseestimates were compared with a suite of abiotic variables usingcorrelation analysis (PROC CORR; SAS 2004) in an attempt todescribe factors that may mediate availability of August StripedBass habitat in Lake Martin. Abiotic variables were obtainedfrom the National Weather Service, the U.S. Geological Survey,and APC (Table 1). Because most environmental variables werehighly correlated with one another, a multiple regression modelusing these variables to predict availability of quality or totalstriped bass habitat in Lake Martin was inappropriate. There-fore simple linear regression models were constructed using thestrongest correlated environmental variable for each habitat type(PROC REG; SAS 2004) to assess relationships among envi-ronmental conditions and habitat availability (Maceina 1992).

RESULTS

Striped Bass Tagging and TrackingStriped Bass tagged in 2009 (N = 30) ranged from 740 to

1,063 mm TL and from 4.78 to 17.0 kg. Of these fish, two werenever found after tagging, one was harvested by an angler, andtwo others died from unknown causes. Contact was lost withseven fish by the end of tracking activities in 2009; three werefound dead in 2010 and four were never located again. Also,

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013


TABLE 2. Monthly mean depth, temperature, dissolved oxygen concentration (DOC), and daily movement of tagged Striped Bass tracked weekly over a 4-monthperiod in Lake Martin, Alabama, during 2009. Means with the same letters were similar (Tukey’s HSD test; P > 0.10), SD-values in parentheses.

Month Depth (m) Temperature (◦C) DOC (mg/L) Movement (m/d)

June 14.8 (5.3) z 18.3 (2.6) x 4.6 (1.4) z 523 (549) yJuly 14.7 (6.1) z 19.6 (3.4) xy 4.0 (1.6) z 279 (282) yAugust 16.9 (5.9) z 19.8 (4.5) y 2.5 (1.1) y 379 (300) ySeptember 10.3 (5.0) y 24.7 (2.0) z 4.0 (2.4) z 2,039 (1,816) z

only seven fish tracked in 2009 were relocated alive in 2010;three of these fish died during tracking in 2010 and another wasnever relocated after the initial contact in 2010. Striped Basstagged in 2010 (N = 6) ranged from 766 to 821 mm and from4.42 to 7.42 kg. Of these fish, one was harvested by an anglerand one died from unknown causes. During tracking activitiesin 2010, dead Striped Bass (most < 10 kg) were commonlyobserved floating in the lower half of Lake Martin, particularlythroughout the month of August.

Daily movement and spatial distribution.—Striped Basswere highly mobile in Lake Martin, often exhibiting displace-ments in excess of 10 km between weekly locations. Meanmovement was variable among fish, but mean movement rateswere greater in September than in any other month in 2009(Table 2; F3, 45 = 7.67, P < 0.001).

Two tagged Striped Bass were found outside of Lake Martinduring the study period. One fish was located by plane on June2, 2009, approximately 62 km upstream from Lake Martin inthe Tallapoosa River near Wadley, Alabama. That fish was nextrelocated 3 weeks later approximately 85 km downstream inLake Martin. The fish remained in the lower area of the reservoirand was last located on August 26, 2009. This fish was notrelocated in 2010 and may have spent the entire year upstream inthe Tallapoosa River. Another fish presumably spent the summerof 2010 in the Tallapoosa River upstream from Lake Martin,as this fish was never located in the reservoir during trackingand was found in the Tallapoosa River approximately 10 kmupstream from Lake Martin via aerial tracking conducted onOctober 14, 2010.

Seasonal habitat use.—When considered together, patternsof depth, temperature, and DOC use by Striped Bass displayedmarkedly different monthly patterns throughout the tracking pe-riod in 2009. Modal depth of Striped Bass generally increasedfrom June to August, as reservoir water surface temperatures in-creased; this behavior continued until September, when taggedfish abruptly shifted to shallower depths (Figure 2). Accord-ingly, mean depth of these fish was greater in June, July, andAugust than in September (Table 2; F3, 53 = 11.08, P < 0.001).In June, when stratification was well established in Lake Martin,these fish were found in water where temperatures were below23◦C, but they were found in waters with a wide range of DOCs(Figure 2). As summer progressed, Striped Bass depth distri-butions began displaying bimodality, and one group was foundnear the thermocline and another was found in the hypolimnion.

The group associated with the hypolimnion was found in waterswhere temperatures were less than 20◦C and DOC was less than3 mg/L; whereas, the thermocline group was most often foundin water where temperatures centered on 25◦C and DOC rangedfrom 1 to 6 mg/L (Figure 2). In September, almost all fish werefound in waters where temperatures centered on 25◦C and DOCranged from near 0 to 8 mg/L. The mean temperature of waterused by Striped Bass was highest in September and lowest inJune, and July and August was intermediate (Table 2; F3, 53 =34.11, P < 0.001). Similarly, the mean DOC of water used byStriped Bass was higher in June, July, and September than inAugust (Table 2; F3, 53 = 12.95, P < 0.001).

Effect of fish size on movement and habitat use.—Distributions of movement, depth, DOC use, and temperatureuse of tagged Striped Bass were similar between size-classes ineach month of 2009 (Kolmogorov–Smirnov analysis ≤ 1.15, P≥ 0.145). Also, mean daily movement of Striped Bass (F1, 19 ≤0.19, P ≥ 0.677), as well as depth (F1, 19 ≤ 2.11, P ≥ 0.162)and temperature (F1, 19 ≤ 1.26, P ≥ 0.275) use, were all similarbetween size-classes in all months. Striped Bass > 10 kg usedgreater mean DOC than did smaller fish in September (F1, 19

= 5.45, P = 0.031), but mean DOC was similar between size-classes in the other months (F1, 19 ≤ 1.96, P ≥ 0.175).

Annual mean daily movement was not related to StripedBass length (P = 0.635), nor was mean daily movement in Juneand July (P = 0.120). Mean daily movement in August andSeptember was correlated to fish length (r = 0.39, P = 0.095).Mean temperature, mean dissolved oxygen use, and depth werenot related to Striped Bass TL in any month of 2009 (P ≥ 0.14).

Striped Bass Habitat SelectionThe logistic regression analysis of Striped Bass presence and

absence based on temperature and DOC resulted in the model,

loge[p/(1 − p)] = −6.32 + (0.14 × temp) + (1.10 × DOC)

− (0.05 × temp × DOC),

in which all terms have a significant effect on Striped Bass habi-tat selection (Table 3). The model indicated that Striped Basswere selecting for one of two options: low temperature and highDOC, or high temperature and low DOC. The other two ex-tremes, low temperature and low DOC or high temperature andhigh DOC, were more numerous than those preferred combina-tions, but were not selected by Striped Bass. The model further

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013


FIGURE 2. Depth distribution, water temperatures, and dissolved oxygen concentrations of tagged Striped Bass in Lake Martin, Alabama, over 4 months in2009. Dotted lines on the depth panels denote approximate location of the thermocline.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013


TABLE 3. Results of the logistic regression analysis to predict Striped Bass presence or absence based on the temperature and dissolved oxygen content of thewater.

Parameter Estimate SELower limitof 95% CI

Upper limit of95% CI Z P-value

Intercept −6.3208 0.8768 −8.0394 −4.6023 −7.2 <0.0001Temperature 0.1444 0.0482 0.0499 0.2389 2.99 0.003Dissolved oxygen 1.0091 0.1501 0.7149 1.3033 6.72 <0.0001Temperature × dissolved oxygen interaction −0.045 0.007 −0.0589 −0.0305 −6.2 <0.0001

predicted that Striped Bass tended to select cooler temperaturesif DOC was at least 3.2 mg/L, but once DOC declined below thatlevel, the fish selected for warmer temperatures. Specifically, atDOC at or above 3.2 mg/L, the probability of Striped Bass pres-ence was predicted to increase by 9.4% with each 1◦C increasein temperature (χ2

1 = 88.24, P < 0.001); the probability ofStriped Bass presence decreased only by 2.2% with each 1◦Cincrease in temperature when DOC was below 3.2 mg/L, thoughthis effect was not significant (χ2

1 = 0.32, P = 0.572). This in-dicates that habitat selection by Striped Bass was more stringentwhen DOC was at or above 3.2 mg/L in terms of temperature.Likewise, Striped Bass selected for high DOC when water tem-peratures were at or below 22.6◦C, but at higher temperaturesthey selected for lower DOC. Specifically, the probability ofStriped Bass presence decreased by 21.4% with each unit in-crease in DOC at temperatures exceeding 22.6◦C (χ2

1 = 25.99,P < 0.001), but increased 24.3% with each unit increase inDOC when temperatures were below that level (χ2

1 = 60.62,P < 0.001).

Summer Striped Bass Habitat AvailabilityOf all Striped Bass locations 75% had temperatures in June

and July that were at or below 21.0◦C and had DOC that wasat least 3.2 mg/L; thus, waters with these characteristics weredesignated as quality Striped Bass habitat. Furthermore, 95%of Striped Bass locations in June–August had temperatures ator below 25◦C and had a DOC of at least 1.6 mg/L, and wa-ter meeting these criteria was designated as total Striped Basshabitat. The amount of marginal habitat was thus defined asthe difference between these two volumes. Little difference wasobserved in 25% quartiles, 75% quartiles, or medians of tem-peratures or DOCs occupied by Striped Bass > 10 kg comparedwith smaller fish for either quality or total habitat designations.Thus, Striped Bass habitat definitions based on the entire rangeof sizes tracked in this study were considered valid.

By the beginning of July each year, quality Striped Basshabitat comprised 40–50% of the total habitat available in LakeMartin, but in 2009 all quality habitat was gone from the reser-voir by August 1 (Figure 3). Marginal habitat rapidly decreaseduntil there was no suitable Striped Bass habitat available by mid-September. In 2010, quality habitat did not decline as rapidly asin 2009, but was completely gone from the reservoir by the endof August (Figure 3). Similarly, marginal habitat did not decline

as rapidly in 2010 as in 2009, and was not completely eliminatedduring that summer, although habitat availability was very lowin late September into early October. In both years, the reser-voir began destratifying by the end of September and marginalhabitat availability rapidly increased (Figure 3).

In the second half of August, quality Striped Bass habitatwas found in the lower subbasin of Lake Martin in only 6 of

FIGURE 3. Volumes of Striped Bass habitat present in Lake Martin, Alabama,from July to October in 2009 and 2010. Quality habitat was defined as waterwith temperatures ≤21.0◦C and dissolved oxygen concentrations ≥3.2 mg/L.Marginal habitat was defined as water with temperatures > 21.0◦C but ≤25◦Cand dissolved oxygen concentrations < 3.2 but ≥1.6 mg/L.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013


FIGURE 4. Estimates of Striped Bass habitat availability in the lower lakesubbasin of Lake Martin, Alabama, during the second half of August from 1994to 2010. Zeroes represent years when no habitat was present; blank years (1995and 2001) were years in which no profile data were available to estimate habitat.

the 15 years of data that were available between 1994 and 2010(Figure 4). Conversely, total Striped Bass habitat was availablein the reservoir during this time in all but 1 year (2003). Volumeof quality Striped Bass habitat available was less than 10,000hectare-meters (ha-m) every year, with the exception of 2007(Figure 4). Total Striped Bass habitat present in the reservoirranged from 0 to 31,487 ha-m and averaged 17,486 ha-m acrossthe 15 years of data.

The amount of quality habitat available in the second halfof August was not correlated with cooling degree-days, num-ber of days with a high temperature of 32◦C or greater, ordischarge from Martin Dam (Table 4). The amount of qualityhabitat present in Lake Martin was correlated with precipitation,hours of turbine operation (turbine-hours) at Martin Dam, andinflow into Lake Martin. The amount of total Striped Bass habi-tat present in the second half of August was not correlated withcooling degree-days, but was correlated with all other variables(Table 4). The strongest correlation was with turbine-hours fromMartin Dam. Data plots showed that the quality habitat correla-tions were strongly influenced by one outlier (2007), but thoseof total habitat were less influenced by outliers. Consequently,correlations of habitat availability and environmental variableswere generally stronger for total habitat than for quality habitat(Table 4). For quality habitat, the best model was:

Quality habitat (ha-m) = 19,384 − 65.73

× [precipitation (cm)].

This model was significant (P = 0.013) and explained 39%of the variation in the availability of quality habitat. For totalhabitat, the best model was:

Total habitat (ha-m) = 33,026 − 4.66 × (turbine-hours).

This model was highly significant (P < 0.001) and ex-plained 79% of the variation in availability of total Striped Basshabitat.

DISCUSSIONStriped Bass in Lake Martin were found to undergo a “habitat

squeeze” during summer as they became trapped in limited ther-mal refugia between warm, oxygenated, epilimnetic waters andcool, anoxic or hypoxic, hypolimnetic waters, which is a com-mon phenomenon in other systems (Coutant 1985; Matthewset al. 1985). The general behavior pattern of Striped Bass inLake Martin was to select cooler temperatures if DOC was above3.2 mg/L, but below that level, fish would generally select forwarmer temperatures, similar to what Thompson et al. (2010)described for Striped Bass in a shallow, eutrophic reservoir in

TABLE 4. Results (Pearson correlation coefficient [r] and P-value) of correlation analyses between the amount of quality habitat and the total Striped Basshabitat available during the second half of August from 1994 to 2010 and selected environmental variables. Environmental variables are defined in Table 1.

Environmental variable Quality habitat Total habitat

Cooling degree-days r = 0.31, P = 0.257 r = 0.28, P = 0.306Number of days ≥ 32◦C r = 0.22, P = 0.430 r = 0.49, P = 0.062Precipitation r = −0.62, P = 0.013 r = −0.81, P = 0.001Turbine-hours r = −0.51, P = 0.050 r = −0.89, P < 0.001Discharge r = −0.40, P = 0.135 r = −0.77, P = 0.001Inflow r = −0.48, P = 0.071 r = −0.78, P = 0.001

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013


North Carolina. As summer progressed, Striped Bass in LakeMartin displayed dual strategies for persisting throughout thesummer, with fish found either near the thermocline or deep inthe hypolimnion. Use of either of these strategies did not appearto be size dependent, as there was no relation between length offish and depths used. Some Striped Bass moved above the ther-mocline in September, where they existed in water temperatureswarmer than 25◦C and DOC greater than 4 mg/L. This apparentindividualistic behavior has been described for other fish species(Sammons and Maceina 2005), and it appears that Striped Bassin Lake Martin likewise employ various strategies to cope withchanging water quality variables during the summer (Thompsonet al. 2010).

Striped Bass have been found to use springs or hypolim-netic discharges for summer refugia if ambient conditionsduring summer are unfavorable (Cheek et al. 1985; Moss1985; Braschler et al. 1989; Van Den Avyle and Evans 1990;Wilkerson and Fisher 1997). However, thermal refugia viasprings are not known to exist in Lake Martin (C. Greene,ADCNR, personal communication), and none were detectedduring tracking activities on Lake Martin; thus, this cannot ex-plain the individualistic behavior documented in this study. TwoStriped Bass were located upstream in the Tallapoosa River,where a hypolimnetic discharge exists below Harris Dam. Ingeneral, most Striped Bass in Lake Martin typically remained inthe reservoir and used hypolimnetic waters for summer refuge;this observation is similar to previous findings (Hampton et al.1988; Jackson and Hightower 2001; Moss et al. 2005).

Striped Bass in Lake Martin were sometimes observed usingwaters of surprisingly high temperatures and low DOC, evenduring periods when large volumes of habitat were still avail-able in the reservoir. Striped Bass were commonly observedupstream in middle reaches of the three main reservoir armsin July and August of 2009. These fish persisted in narrowbands of marginal Striped Bass habitat found near the reservoirbottom; this habitat was typically characterized by DOC below2.5 mg/L. The fish continued to use these areas until this suitablehabitat was no longer present, even though a sizeable volumeof water with higher DOCs and acceptable temperatures ex-isted downstream in the reservoir, well within the demonstratedmovement capabilities of these fish. Many studies have reportedthat Striped Bass avoid water with DOC less than 3 mg/L (Cheeket al. 1985; Coutant 1985; Matthews et al. 1985, 1989); how-ever, this was not observed in Lake Martin. Striped Bass canuse suboptimal habitat (both temperature and DOC) for peri-ods up to 4–6 weeks (Zale et al. 1990; Jackson and Hightower2001; Young and Isely 2002; Thompson et al. 2010), but inmost cases this only occurred after suitable Striped Bass habi-tat had disappeared from the system. This was not the case inLake Martin, where 23% of the Striped Bass locations in Julyoccurred in areas with DOC below 3.0 mg/L, and 20% of loca-tions were in water where temperatures were above 22◦C, eventhough higher quality habitat was still available in the reservoir.Thompson et al. (2010) described a similar behavior pattern for

Striped Bass in a North Carolina reservoir, which they linked toforage fish distribution. The diet of Striped Bass in Lake Martinwas unknown, but was presumed to be primarily Threadfin ShadDorosoma petenense and Gizzard Shad D. cepedianum, as thishas been found in other studies (Raborn et al. 2007; Shepherdand Maceina 2009). These species are both warmwater fishesand primarily found in epilimnetic waters, thus Striped Bassmay be selecting warmer water due to forage distribution, asfound by Thompson et al. (2010). Similarly, Bevelhimer (1997)observed Smallmouth Bass Micropterus dolomieu to remain inmuch warmer water than in those of their preferred tempera-ture in Melton Hill Reservoir, Tennessee, even though coolerwater was readily available. Certainly, there were a number ofunmeasured factors that could have affected the distribution ofStriped Bass in Lake Martin, such as access to forage fishes,which may have allowed Striped Bass to persist in physiolog-ically stressful conditions. Nevertheless, this study and thosepreviously mentioned demonstrate the continued importance oftelemetry studies to refine understanding of temperature andDOC preferenda and their effects on fish behavior in the wild.

Striped Bass found in Lake Martin are Gulf strain fish, whichoriginally occurred in coastal rivers of the Gulf of Mexico anddid not oversummer in the ocean like their Atlantic strain coun-terparts (Wooley and Crateau 1983). These fish may be betteradapted to higher temperatures and lower DOCs than Atlanticstrain fish (Wooley and Crateau 1983; Van Den Avyle and Evans1990). Most studies of Striped Bass habitat use have been con-ducted with Atlantic strain fish, and results from these studiesmay not be directly comparable with systems having Gulf strainfish, particularly in terms of temperature and dissolved oxygenrequirements. As mentioned previously, the fish used in the LakeMartin study were all large adults (>4.5 kg), which in Atlanticstrain fish can have significantly lower temperature preferendathan smaller fish (Coutant 1985). Few Striped Bass field studies,and no laboratory studies, on temperature and DOC selectionhave been conducted with fish of this size. However, Bettoli(2005) conducted a habitat use study on Atlantic strain fish ofsimilar sizes to the ones used in the Lake Martin study and foundthat Striped Bass selected for water temperatures of 17.5◦C dur-ing the summer and were never found in water above 24◦C. Incontrast, 25% of Striped Bass locations in Lake Martin in Julyand August of both years were in water temperatures > 22◦C.These considerations deserve future study to more clearly definehabitat requirements of Gulf strain Striped Bass.

Habitat availability is a major limitation of Striped Bass fish-eries across the southeastern USA and often determines qualityof these fisheries in terms of numbers and size (Coutant 1987).Compared with many reservoirs in this region, Lake Martin sup-ports a large volume of Striped Bass habitat throughout much ofthe year, probably owing to the size, depth, and relative infertil-ity of the reservoir. However, this study has demonstrated thatStriped Bass habitat availability is subject to drastic seasonaland annual changes even in reservoirs that appear to providegood opportunities for quality Striped Bass fisheries such as

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013


Lake Martin. Unusually high amounts of precipitation in sum-mer can result in large volumes of water generated through thedam to maintain reservoir levels near full-pool levels and avoidflooding. During these events, cool hypolimnetic waters are dis-charged through the dam and replaced with warm epilimneticwaters, resulting in large changes in habitat availability (Coleand Hannan 1990). Also, frequent rain events can flush nutri-ents downstream and from shoreline habitats such as lawns fromlakeside homes, which can temporarily increase nutrient levelsand primary production in the reservoir, causing more rapiddepletion of oxygen through decomposition (Coutant 1987).

Striped Bass habitat in Lake Martin declined rapidly through-out the summer of 2009, culminating in a 2-week period whereno suitable habitat was present in the reservoir. Conversely,although there was about 20% less total Striped Bass habitatpresent in Lake Martin at the beginning of July in 2010 thanin 2009, habitat declined more slowly and the reservoir neverreached the point where no suitable Striped Bass habitat existed.Weather patterns in 2009 and 2010 were drastically different. Al-though 2009 was characterized by high precipitation and coolertemperatures, especially during the summer, 2010 was one ofthe hottest years on record, but annual precipitation was about10% below average. Thus, high temperatures and low precip-itation may result in a greater amount of Striped Bass habitatto be present in Lake Martin in late August than during timeswhen temperatures are cooler and annual precipitation is high.A correlation of historical late-August Striped Bass habitat inthe lower forebay of Lake Martin with environmental variablessupport this interpretation, and suggests that flow through thesystem is more important than the relative summer climate indetermining Striped Bass habitat availability in late summer.

The most recent Striped Bass mortality events in LakeMartin before this study occurred in 1991, 1994, and 2001, all inthe period from late August to mid-September (J. Lochamy, per-sonal communication). Unfortunately, temperature or dissolvedoxygen profile data were not collected by ADEM in 1991 or2001, making the estimation of Striped Bass habitat availabil-ity in those years impossible. However, the estimate of StripedBass habitat availability in late August 1994 was the secondlowest estimate of August habitat availability over the periodfrom 1994 to 2010. Precipitation amounts from January 1 to thedate of habitat estimation in August in 1991, 1994, and 2001were the fourth, sixth, and second highest, respectively, in the20-year period from 1991 to 2010. Similarly, turbine-hours atMartin Dam were the third, fourth, and seventh highest in 1991,1994, and 2001, respectively, over that same 20-year period. Noquality habitat was found in the reservoir in late August 1994,and regression equations developed in this study likewise pre-dicted that no quality habitat existed in August of 1991 or 2001.Similarly, total Striped Bass habitat was predicted to be 9,843ha-m in 1991 and 7,072 ha-m in 2001, far lower than the esti-mated mean habitat volume from 1994 to 2010 (17,704 ha-m).It appears that low availability of Striped Bass habitat in Augustis a factor in the occurrence of Striped Bass mortality events.

However, Striped Bass habitat availability in late August waslow in other years when mortality events were not observed,most notably in 2003 and 2005. Although a lack of Striped Basshabitat caused by high flow-through conditions in the reservoirmay not always result in a large Striped Bass mortality event,it probably increases the likelihood of such events. The 3 yearsthat large Striped Bass mortality events were observed (i.e.,1991, 1994, and 2001) all fell within the top six wettest yearsover the 20-year period from 1991 to 2010. Likewise, all 3 yearsfell within the top 7 years of heaviest turbine use at Martin Damfrom April through August over that 20-year period.

Results from this study revealed that cool, wet summers af-fected Striped Bass habitat availability to a greater degree thandid hot, dry summers, which are the two most common summerweather patterns in Alabama. Understanding factors that affectavailability of Striped Bass habitat allows biologists to predictperiods of low habitat availability and the likely effects on thesefisheries. Also, it enables them to foster discussions with damoperators or other regulatory agencies to conserve Striped Basshabitat. However, in many cases, such as Lake Martin, dam oper-ators may have few options available to them to reduce impactsto Striped Bass habitat from operation procedures. Because LakeMartin has a high degree of residential development along itsshorelines and water levels are kept at full pool throughout thesummer, APC must generate large volumes of water through thedam during periods of high rainfall to reduce flooding effects inthe reservoir. Changes in water levels within Lake Martin alsohave economic impacts on property owners that must be takeninto account when trying to manage for Striped Bass habitat(Hanson et al. 2002). However, other options, such as forebayor hypolimnion aeration (Ruane et al. 1986), exist to increasethe amount of quality Striped Bass habitat within reservoirs. Incontrast, increases in reservoir productivity resulting from up-stream or in-basin habitat modifications will probably decreasethe quantity of Striped Bass habitat available at the beginningof each summer, as well as increase the rate of decline of thishabitat during the summer. Striped Bass fisheries have great eco-nomic value, especially those with trophy components (Schorret al. 1995; Lothrop 2012), and factors influencing the qualityof the Striped Bass fishery could have substantial effects on theregional economy. The major factors mediating availability ofsummertime Striped Bass habitat found in this study appear tobe controllable to some extent, and resource agencies should beable to use the results of this study to protect this fishery andmaximize its value to surrounding communities.

ACKNOWLEDGMENTSKeith Floyd, ADCNR, assisted in the use of long lines to

collect Striped Bass for tagging. Jim Lochamy, Jim Crew, An-gela Segars-Anderegg, and others at APC provided data andsupport throughout this project. Gina Curvin, ADEM, providedthe temperature and dissolved oxygen profile data used for thehistorical Striped Bass habitat analysis. Amanda Flemming andHenry Mealing at Kleinschmidt Associates coordinated several

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013


focus-group meetings that helped focus and foster this project.Aaron Kern and Jonathan Brown, Auburn University, conductedmuch of the sampling during this project. Laurie Earley, AuburnUniversity, assisted in the GIS analyses. This study was fundedby APC in 2009 and ADCNR in 2010 and 2011. Commentsfrom three reviewers, and especially the two editors, substan-tially improved this manuscript.

REFERENCESAxon, J. R., and D. K. Whitehurst. 1985. Striped Bass management in lakes with

emphasis on management problems. Transactions of the American FisheriesSociety 114:8–11.

Bettoli, P. W. 2005. The fundamental thermal niche of adult landlocked StripedBass. Transactions of the American Fisheries Society 134:305–314.

Bettoli, P. W., and R. S. Osborne. 1998. Hooking mortality and behavior ofStriped Bass following catch and release angling. North American Journal ofFisheries Management 18:609–615.

Bevelhimer, M. S. 1997. Smallmouth Bass habitat use and movement patternswith respect to reservoir thermal structure. Proceedings of the Annual Confer-ence Southeastern Association of Fish and Wildlife Agencies 49(1995):240–249.

Braschler, D. W., M. G. White, and J. W. Foltz. 1989. Movements and habitatselection of Striped Bass in the Santee-Cooper reservoirs. Proceedings of theAnnual Conference Southeastern Association of Fish and Wildlife Agencies42(1988):27–34.

Brenden, T. O., B. R. Murphy, and E. M. Hallerman. 2004. Estimating total errorin locating radiotelemetry transmitters by homing in a riverine environment.Journal of Freshwater Ecology 19:295–304.

Cheek, T. E., M. J. Van Den Avyle, and C. C. Coutant. 1985. Influences of waterquality on distribution of Striped Bass in a Tennessee River impoundment.Transactions of the American Fisheries Society 114:67–76.

Cole, G. A. 1983. Textbook of limnology. Waveland Press, Prospect Heights,Illinois.

Cole, T. M., and H. H. Hannan. 1990. Dissolved oxygen dynamics. Pages 71–107 in K. W. Thornton, F. E. Payne, and B. L. Kimmel, editors. Reservoirlimnology: ecological perspectives. Wiley, New York.

Coutant, C. C. 1985. Striped Bass, temperature, and dissolved oxygen: a specula-tive hypothesis for environmental risk. Transactions of the American FisheriesSociety 114:31–61.

Coutant, C. C. 1987. Thermal preference: when does an asset become a liability?Environmental Biology of Fishes 18:161–172.

Coutant, C. C., and D. S. Carroll. 1980. Temperatures occupied by ten ultrasonic-tagged Striped Bass in freshwater lakes. Transactions of the American Fish-eries Society 109:195–202.

Hampton, K. E., T. L. Wenke, and B. Z. Zamrzla. 1988. Movements of adultStriped Bass tracked in Wilson Reservoir, Kansas. Prairie Naturalist 20:113–125.

Hanson, T. R., L. U. Hatch, and H. C. Clonts. 2002. Reservoir water levelimpacts on recreation, property, and nonuser values. Journal of the AmericanWater Resources Association 38:1007–1018.

Jackson, J. R., and J. E. Hightower. 2001. Reservoir Striped Bass movements andsite fidelity in relation to seasonal patterns in habitat quality. North AmericanJournal of Fisheries Management 21:34–45.

Lothrop, R. L. 2012. Economic impact of Striped Bass angler visitation at LewisSmith Lake, Alabama. Master’s thesis. Auburn University, Auburn, Alabama.

Maceina, M. J. 1992. A simple regression model to assess environmental effectson fish growth. Journal of Fish Biology 41:557–565.

Maceina, M. J., J. W. Slipke, and J. M. Grizzle. 1999. Effectiveness of threebarrier types for confining Grass Carp in embayments of Lake Seminole,Georgia. North American Journal of Fisheries Management 19:968–976.

Matthews, W. J. 1985. Summer mortality of Striped Bass in reservoirs of theUnited States. Transactions of the American Fisheries Society 114:62–66.

Matthews, W. J., L. G. Hill, D. R. Edds, and F. P. Gelwick. 1989. Influenceof water quality and season on habitat use by Striped Bass in a large south-western reservoir. Transactions of the American Fisheries Society 118:243–250.

Matthews, W. J., L. G. Hill, and S. M. Schellhaass. 1985. Depth distribution ofStriped Bass and other fish in Lake Texoma (Oklahoma–Texas) during sum-mer stratification. Transactions of the American Fisheries Society 114:84–91.

Moss, J. L. 1985. Summer selection of thermal refuges by Striped Bass inAlabama reservoirs and tailwaters. Transactions of the American FisheriesSociety 114:77–83.

Moss, J. L., K. B. Floyd, J. C. Greene, J. M. Piper, T. D. Berry, and P. D. Ekema.2005. Seasonal distribution and movement of Striped Bass in Lewis SmithReservoir, Alabama. Proceedings of the Annual Conference SoutheasternAssociation of Fish and Wildlife Agencies 57(2003):141–149.

Raborn, S. W., L. E. Miranda, and M. T. Driscoll. 2007. Prey supply and predatordemand in a reservoir of the southeastern United States. Transactions of theAmerican Fisheries Society 136:12–23.

Rogers, K. B., and G. C. White. 2007. Analysis of movement and habitat usefrom telemetry data. Pages 625–676 in C. S. Guy and M. L. Brown, editors.Analysis and interpretation of freshwater fisheries data. American FisheriesSociety, Bethesda, Maryland.

Ruane, R. J., C. E. Bohac, W. M. Seawell, and R. M. Shane. 1986. Improvingthe downstream environment by reservoir release modifications. Pages 270–277 in G. E. Hall and M. J. Van Den Avyle, editors. Reservoir fisheriesmanagement: strategies for the 80s. American Fisheries Society, SouthernDivision, Reservoir Committee, Bethesda, Maryland.

Sammons, S. M., and M. J. Maceina. 2005. Activity patterns of LargemouthBass in a subtropical U.S. reservoir. Fisheries Management and Ecology12:331–339.

Sammons, S. M., M. J. Maceina, and D. G. Partridge. 2003. Changes in behav-ior, movement, and home ranges of Largemouth Bass following large-scalehydrilla removal in Lake Seminole, Georgia. Journal of Aquatic Plant Man-agement 41:31–38.

SAS (Statistical Analysis Systems). 2004. SAS system for linear models, release9.1. SAS Institute, Cary, North Carolina.

Schorr, M. S., J. Sah, D. F. Schreiner, M. R. Meador, and L. G. Hill. 1995.Regional economic impact of the Lake Texoma (Oklahoma–Texas) StripedBass fishery. Fisheries 20(5):14–18.

Shepherd, M. D., and M. J. Maceina. 2009. Effects of Striped Bass stocking onLargemouth Bass and Spotted Bass in Lewis Smith Lake, Alabama. NorthAmerican Journal of Fisheries Management 29:1232–1241.

Thompson, J. S., J. A. Rice, and D. S. Waters. 2010. Striped Bass habitatselection rules in reservoirs without suitable summer habitat offer insightinto consequences for growth. Transactions of the American Fisheries Society139:1450–1464.

Van Den Avyle, M. J., and J. W. Evans. 1990. Temperature selection by StripedBass in a Gulf of Mexico coastal river system. North American Journal ofFisheries Management 10:58–66.

Wilkerson, M. L., and W. L. Fisher. 1997. Striped Bass distribution, movements,and site fidelity in Robert S. Kerr Reservoir, Oklahoma. North AmericanJournal of Fisheries Management 17:677–686.

Winter, J. 1996. Advances in underwater biotelemetry. Pages 555–590 in B. R.Murphy and D. W. Willis, editors. Fisheries techniques, 2nd edition. AmericanFisheries Society, Bethesda, Maryland.

Wooley, C. M., and E. J. Crateau. 1983. Biology, population estimates, andmovement of native and introduced Striped Bass, Apalachicola River, Florida.North American Journal of Fisheries Management 3:383–394.

Young, S. P., and J. J. Isely. 2002. Striped Bass annual site fidelity and habitatutilization in J. Strom Thurmond Reservoir, South Carolina–Georgia. Trans-actions of the American Fisheries Society 131:828–837.

Zale, A. V., J. D. Wiechman, R. L. Lochmiller, and J. Burroughs. 1990. Lim-nological conditions associated with summer mortality of Striped Bass inKeystone Reservoir, Oklahoma. Transactions of the American Fisheries So-ciety 119:72–76.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:41

12

Aug

ust 2

013



Insight into Age and Growth of Northern Snakehead inthe Potomac RiverJohn Odenkirk a , Catherine Lim a , Steve Owens a & Mike Isel aa Virginia Department of Game and Inland Fisheries , 1320 Belman Road, Fredericksburg ,Virginia , 22401 , USAPublished online: 26 Jul 2013.

To cite this article: John Odenkirk , Catherine Lim , Steve Owens & Mike Isel (2013) Insight into Age and Growth ofNorthern Snakehead in the Potomac River, North American Journal of Fisheries Management, 33:4, 773-776, DOI:10.1080/02755947.2013.806382







http://dx.doi.org/10.1080/02755947.2013.806382




ARTICLE

Insight into Age and Growth of Northern Snakeheadin the Potomac River

John Odenkirk,* Catherine Lim, Steve Owens, and Mike IselVirginia Department of Game and Inland Fisheries, 1320 Belman Road, Fredericksburg,Virginia 22401, USA

AbstractA Northern Snakehead Channa argus population was documented in the Potomac River system in 2004. Since that

time, the population has expanded in range and number, yet relatively little is known about key population variablesincluding age and growth. Lack of known-age fish with which to compare otoliths from the sampled population hashindered verification, and Northern Snakehead otoliths can be difficult to interpret. We compared growth increments(mm/d) of recaptured fish marked with T-bar anchor tags (n = 51; mean time at large, 310 d [SD = 302]) to otolithannuli from fish sacrificed in 2011 and 2012 (n = 192). While immersed (solution of 80% water and 20% glycerin)otolith transverse perspectives were viewed “cracked” with transmitted light. Readings from fish aged 1–4 yearssuggested initial growth was much faster than previously reported, but length at age was highly variable. Annualgrowth increments for fish aged 1–4 (mean length at age of fish, respectively: 394, 563, 644, and 721 mm TL) wereconverted to estimated daily growth (mm/d), which was reasonably similar to the daily growth of recaptured taggedfish. Von Bertalanffy growth parameters were L∞ = 780, K = 0.48, and t0 = −0.56, where L∞ is the asymptotic length,K is a growth coefficient, and t0 is a time coefficient at which length would theoretically be zero. The implications ofrapid growth include the potential for the earlier onset of sexual maturity, which could represent enhanced chancesfor successful colonization.

Northern Snakeheads Channa argus were first documentedin the Potomac River system in 2004 (Odenkirk and Owens2005), and they have subsequently expanded in range and num-ber (Odenkirk and Owens 2007). Concern remains about thepotential of this introduction to cause biological and ecologicalimpacts to native and naturalized fish populations (Courtenayand Williams 2004; Herborg et al. 2007), and there is a paucityof information concerning basic life history traits of this po-tentially invasive fish. The ability to accurately determine agesof fishes without bias is critical to effective management andresearch for any species (Isely and Grabowski 2007), and thusfurther study of Northern Snakehead life history has been rec-ommended (Jiao et al. 2009).

Accurate and precise age and growth information is essentialto understand Northern Snakehead biology in the Potomac Riverand is a precursor for the ability to evaluate recruitment mech-anisms and determine mortality estimates. Lack of known-agefish, known growth increments, or both with which to compare

*Corresponding author: [email protected] November 20, 2012; accepted May 15, 2013

otolith annuli has hindered verification of ages derived fromotolith viewing. Additionally, Northern Snakehead otoliths canbe irregular, opaque, and may even show variability in mor-phology between individuals. The only known North Americanevaluations of Northern Snakehead age and growth (Odenkirkand Owens 2007; Cohen et al. 2012) were conducted withoutthe benefit of known-age fish or other “corroborating” evidence.Thus, this study’s objectives were to determine mean lengthat age for a sample of the Potomac River Northern Snakeheadpopulation and compare otolith-derived growth increments withgrowth increments of recaptured tagged fish as a means to verifythe efficacy of otoliths for accurate age determination.

METHODSNorthern Snakeheads were captured during surveys con-

ducted by the Virginia Department of Game and Inland Fish-eries (VDGIF) with DC electrofishing during daytime from a

773

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:42

12

Aug

ust 2

013

774 ODENKIRK ET AL.

TABLE 1. Length-at-age data for 192 Northern Snakeheads from otolith interpretations. Fish were sampled in 2011 and 2012; TL = mean total length (mm);n = sample size; AGI = annual growth increment in mm; mm/d = estimated daily growth (AGI/365). Asterisk (*) indicates uncalculated value due to low samplesize.

Age n TL (mm) SD AGI (mm) Daily growth (mm/d) Minimum TL Maximum TL

0 2 191 65 145 2371 17 394 86 203 0.56 292 5302 78 563 72 169 0.46 390 7703 47 644 77 81 0.22 504 7824 30 707 79 63 0.17 531 8645 12 712 69 5 0.01 592 8646 2 789 17 * * 777 8017 4 717 103 * * 575 818

5.2-m aluminum johnboat using twin anodes with six dropperseach. The pulse box was a Smith Root Type VI-A set at max-imum output (1,061 V DC) run at 6–7 A and powered by a5,000-W generator. Crews consisted of a boat driver and twonetters. Virginia tidal creeks (Potomac River tributaries) weresampled annually between March and September and included(from north to south) Little Hunting Creek, Dogue Creek, Ac-cotink Creek, Pohick Creek, Occoquan River, and Aquia Creek.Electrofishing was concentrated in shallow water (<2 m) alongchannel margins and in embayments along aquatic vegetationtransition lines (e.g., submersed to floating).

Under a protocol developed in 2009 among a cooperativefisheries management regional group (U.S. Fish and WildlifeService, Maryland Department of Natural Resources, and D.C.Department of Fisheries), three out of every four fish capturedwere measured for TL (nearest millimeter), tagged with a FloyT-bar anchor tag at the base of the dorsal fin, and released. Everyfourth fish was sacrificed for otolith extraction and other scien-tific information. Recaptured, tagged fish were not included inthe sacrifice rotation, as recaptured fish were released (afterTL was measured) to maximize potential recaptures for mark–recapture population estimates. The cooperative regional studyhad broader context and objectives (e.g., determining NorthernSnakehead movement and exploitation), but an unforeseen ben-efit was the recapture of 51 fish by VDGIF crews (through 2012)that allowed for calculations of growth increments over knowntime periods with which to compare growth increments derivedfrom otolith-aged fish. Validation of annual growth incrementscan be accomplished through the recapture of physically markedfish by comparing a reference sample with samples collectedfrom recaptured marked fish (Isely and Grabowski 2007). Al-though the ages of tagged fish were not known, future recap-tures of tagged fish could provide discrete growth incrementsfor known time periods.

Crews from VDGIF sacrificed a total of 201 Northern Snake-heads in 2011 and 2012 and removed both sagittal otoliths,which were viewed by one experienced reader. Otolith trans-verse perspectives were viewed “cracked” with transmitted lightwhile immersed (solution of 80% water and 20% glycerin;

Heidinger and Clodfelter 1987). Mean length at age was de-termined for each age-group, and the resulting annual growthincrements were converted to estimates of daily growth (mm/d),which were then compared with growth increments of recap-tured tagged fish. Growth increments of recaptured tagged fishwere stratified by 100-mm-TL groups based on fish size at tag-ging for comparison to increments derived from otolith evalua-tions for fish of similar length.

Von Bertalanffy growth parameters were derived using FASTversion 2.0 (Slipke and Maceina 2001) for fish aged 0–5 years.Fish aged 6 and 7 were not used in the model due to low samplesize. Trends of growth increments were evaluated with linearregression models.

RESULTSAges were assigned to 192 of the 201 sacrificed fish (nine sets

of otoliths were unreadable). Eight year-classes were encoun-tered and the oldest was age 7 (Table 1). However, growth washighly variable, as the longest fish encountered (two individualsat 864 mm) were designated age 4 and age 5. Total length in-creased rapidly each year before reaching an asymptote after age4 (Figure 1). Correspondingly, annual growth increments were

FIGURE 1. Von Bertalanffy growth curve for Northern Snakeheads capturedfrom the Potomac River system.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:42

12

Aug

ust 2

013

NORTHERN SNAKEHEAD IN THE POTOMAC RIVER 775

TABLE 2. Growth increments for 51 recaptured Northern Snakeheadsmarked with T-bar anchor tags and sorted by 100-mm-TL groups based ontagging size; n = sample size; mm/d = mean growth rate in of all fish cap-tured per 100-mm-TL group based on length gain from tagging and recapturemeasurements.

Size group(mm) n

Growth rate(mm/d) SD

200 1 0.89300 5 0.53 0.27400 4 0.63 0.39500 18 0.25 0.15600 13 0.24 0.22700 6 0.1 0.08800 3 0.01 0.01900 1 0.03

largest for young fish and declined from 0.56 (age 1) to 0.01(age 5) mm/d, (R2 = 0.97, P = 0.0002, n = 5). Von Bertalanffygrowth parameters were L∞ = 780, K = 0.48, and t0 = −0.56,where L∞ is the asymptotic length, K is a growth coefficient,and t0 is a time coefficient at which length would theoreticallybe zero.

Growth increments of recaptured tagged fish displayed asimilar trend of inverse growth rate to length (R2 = 0.88, P =0.001, n = 8). Recaptured fish had a mean “time at large” of310 d (SD = 302). Fish tagged in the 200-mm size-group hadthe largest growth increment (0.89 mm/d), and a decreasingtrend occurred thereafter (Table 2). In some cases growth in-crements of tagged fish were similar to those derived from fishsacrificed for otolith analysis. For instance, fish tagged in the400-mm size-group had a 0.63-mm/d growth increment, whichwas comparable with sacrificed fish assigned age 1 (mean TLof 394 mm) at 0.56 mm/d. However, fish tagged in the 500-mm size-group had a 0.25-mm/d growth increment, which wasless than that for sacrificed fish assigned age 2 (mean TL of563 mm) at 0.46 mm/d; but the next set (those groups approxi-mating 600 mm TL) were more uniform at 0.24 and 0.22 mm/d,respectively, for those tagged and sacrificed.

DISCUSSIONThese data contribute important insight to the understand-

ing of a basic life history tenet of Northern Snakehead biologyin the Potomac River system. Accurate age and growth infor-mation is needed as a prerequisite to obtain reliable estimatesof recruitment and mortality. This information can be used bymanagers to better understand Northern Snakehead populationdynamics. Although data presented here are not comprehensive,and sample sizes were low, they illustrate a population of in-dividuals that is growing much faster than previously reported.The discrete growth increments obtained for 51 tagged, recap-tured fish at large in the system for an average of nearly 1 yearfor comparison is an added benefit.

Current otolith-based length-at-age estimates are nearly dou-ble those previously described for Northern Snakeheads in thissystem by Odenkirk and Owens (2007). While it is possiblethat growth rates changed over time, it is more plausible thatthe earlier study suffered from inexperience in the interpreta-tion of Northern Snakehead otoliths. Initial reads by Odenkirkand Owens (2007) were predominately made by viewing wholeotoliths and interpreting bands of pigmentation as annuli. Sometransverse perspectives were also examined at that time, but fewotoliths had clear annuli, despite a wide range of sizes. Thelargest otoliths sometimes had one visible band, but it was notpossible to verify it as the first annulus; also, this band was notpresent until a fish was larger than seemed reasonable for first-year growth. Several readers with experience in age and growthstructures examined a number of otoliths and produced simi-lar age estimates using the whole view method. Assigned agesseemed reasonable for fish size, and without known-age fish forcomparison and a seeming lack of discernible annuli visible inthe transverse view, this method was continued by Odenkirkand Owens (2007). Otolith evaluations in the current study ben-efitted from exclusively using transverse perspectives (viewedcracked with transmitted light while immersed) combined withexperience from the first study and insights of several years ofrecaptured, tagged fish.

Growth increments for recaptured, tagged fish were not iden-tical to those of otolith-aged fish of similar size, but the sim-ilarities suggested growth was more rapid than initially per-ceived. This observation should provide managers with a meansto more reliably estimate Northern Snakehead growth potentialand mean length at age than what has been previously available.In most cases, growth increments from recaptured tagged fishwere slightly larger than those for otolith-aged fish of similarsize. This may be because the latter fish were exposed to win-ters thereby lowering the average annual growth increments,while not all recaptured tagged fish were at large over a win-ter period. Respiratory function almost ceases while NorthernSnakeheads hibernate (Uchida and Fujimoto 1933). It is alsolikely that growth increments would have been more similar ifsample sizes were higher.

Mean lengths at age of Northern Snakeheads in the currentstudy were remarkably similar to those recently reported by Co-hen et al. (2012) for a Northern Snakehead population in NewYork. For example, mean length at age based on interpretationsof scales for fish assigned age 1 in New York was 400 mm TL(versus 394 mm TL in the current study). However, these growthrates are substantially faster than those reported for NorthernSnakeheads in previous studies outside North America. Zhanget al. (1999) evaluated Northern Snakehead scales from WanghuLake, China, (n = 75 specimens from 1997 and 1998) and de-termined age-4 fish averaged 450 mm TL. Similarly, Courtenayand Williams (2004) summarized several Northern Snakeheadage-and-growth studies and observed sexual maturity at aboutage 3 at a length of 300–350 mm TL in the Amur River, China,and a size of 300 mm TL at age 2 in Japan. Gascho Landis

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:42

12

Aug

ust 2

013

776 ODENKIRK ET AL.

et al. (2011) suggested that Northern Snakeheads grew moreslowly in the Potomac River than populations in China, Rus-sia, and Uzbekistan, but the length-at-age data they used for thePotomac River were provided by Odenkirk and Owens (2007).Thus, potential biases previously discussed may have limited theutility of those comparisons. However, Gascho Landis (2011)did include growth increments for 11 recaptured tagged fish,and the resulting average annual growth increment was 99 mm.The average annual growth increment in the current study was104 mm. Indeed, growth rates in the current study surpassedall those described by Gascho Landis (2011), so it seems likelyNorthern Snakeheads are growing faster in newly colonizedNorth American waters than in waters where the fish is nativeor has been naturalized for an extensive period. Nonindigenousspecies may express different life history traits as they adapt tonew environments (Jiao et al. 2009), and some fish are knownto have elevated degrees of piscivory (relative to body size)when they become established outside their native range (Janget al. 2006). Of germane interest is that unverified interpreta-tion of calcified structures has been a struggle, and it is likelymany literature assertions based solely on such interpretationsare biased. A growing body of evidence provided by recapturedtagged fish suggests rapid growth.

Implications of rapid growth of young fish (elevated lengthat age) include the potential for earlier onset of sexual matu-rity representing enhanced chances for successful colonizationand increased fecundity. In Korea, Northern Snakeheads werereported to reach sexual maturity at about 300 mm TL or age2 (Uchida and Fujimoto 1933), which supports reviews madeby Courtenay and Williams (2004). Based on the size of gravidfish sampled in the Potomac River, many Northern Snakeheadsappear to be sexually mature by their first year. Response effortsto control or eradicate newly discovered populations would nec-essarily need to be rapidly implemented prior to reproductionfor greatest chances of success. Additionally, a great deal ofuncertainty remains surrounding spawning success and recruit-ment variables driving this population. Accurate assessmentsof age will be needed to correlate these with biotic and abioticfactors relating to recruitment, as efforts continue to documentmechanisms affecting year-class strength.

Continued emphasis and research should center on valida-tion of otolith-aging techniques using marked, known-age fish.Batch-marking fingerlings with oxytetracycline immersion orother indelible marking methods should be used to track indi-vidual cohorts through time. Plans are underway to implementthis strategy using a pond in a secure location within the cur-rent established range of Northern Snakehead. Added benefitsof this particular evaluation may include insights to ecologi-cal impacts on existing fish populations (primarily nonnative

centrarchids) within this lentic environment. Additionally,marginal increment analysis would be useful to determine whenannuli are formed and provide short-term verification that onlyone ring is formed per year.

ACKNOWLEDGMENTSThis study was funded, in part, through Federal Aid in Sport-

fish Restoration Grant F-111-R. We thank all the VDGIF biolo-gists, Conservation Police Officers, technicians, and volunteerswho assisted in field collections.

REFERENCESCohen, M. K., J. A. MacDonald, and A. M. Brinton. 2012. Northern Snakeheads

in two New York City lakes over four years. American Fisheries Society,Fisheries Reports (March 2012), Bethesda, Maryland. Available: fisheriesre-ports.org/2012/03/. (November 2012).

Courtenay, W. R., Jr., and J. D. Williams. 2004. Snakeheads (Pisces, Channidae):a biological synopsis and risk assessment. U.S. Geological Survey Circular1251.

Gascho Landis, A. M., N. W. R. Lapointe, and P. L. Angermeier. 2011. Indivi-udual growth and reproductive behavior in a newly established population ofNorthern Snakehead (Channa argus), Potomac River, USA. Hydrobiologia661:123–131.

Heidinger, R. C., and K. Clodfelter. 1987. Validity of the otolith for determiningage and growth of Walleye, Striped Bass, and Smallmouth Bass in power plantcooling ponds. Pages 241–251 in R. C. Summerfelt and G. E. Hall, editors.Age and growth of fish. Iowa State University Press, Ames.

Herborg, L. M., N. E. Mandrak, B. C. Cudmore, and H. J. MacIsaac. 2007. Com-parative distribution and invasion risk of snakehead (Channidae) and Asiancarp (Cyprinidae) species in North America. Canadian Journal of Fisheriesand Aquatic Sciences 64:1723–1735.

Isely, J. J., and T. B. Grabowski. 2007. Age and growth. Pages 187–228 inC. S. Guy and M. L. Brown, editors. Analysis and interpretation of freshwaterfisheries data. American Fisheries Society, Bethesda, Maryland.

Jang, M. H., G. J. Joo, and M. C. Lucas. 2006. Diet of introduced LargemouthBass in Korean rivers and potential interactions with native fishes. Ecologyof Freshwater Fish 15:315–320.

Jiao, Y., N. W. R. Lapointe, P. L. Angermeier, and B. R. Murphy. 2009. Hi-erarchical demographic approaches for assessing invasion dynamics of non-indigenous species: an example using Northern Snakehead (Channa argus).Ecological Modelling 220:1681–1689.

Odenkirk, J., and S. Owens. 2005. Northern Snakeheads in the tidal PotomacRiver system. Transactions of the American Fisheries Society 134:1605–1609.

Odenkirk, J., and S. Owens. 2007. Expansion of a Northern Snakehead popu-lation in the Potomac River system. Transactions of the American FisheriesSociety 136:1633–1639.

Slipke, J. W., and M. J. Maceina. 2001. Fisheries analysis and simulation tools(FAST), version 2.0. Auburn University, Auburn, Alabama.

Uchida, K., and M. Fujimoto. 1933. Life history and method of culture of theKorean snakehead fish, Ophicephalus argus. Bulletin of the Fishery Experi-ment Station of the Government-General of Chosen 3, Series C, Number 1,Busan, South Korea.

Zhang, X., S. Gong, J. Liu, X. He, and J. Gao. 1999. Studies on age and growth ofChanna argus in the Wanghu Lake. Acta Hydrobiologica Sinica 23:600–603.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:42

12

Aug

ust 2

013



Homing of Sockeye Salmon within Hidden Lake, Alaska,Can Be Used to Achieve Hatchery Management GoalsChristopher Habicht a , Terri M. Tobias b , Gary Fandrei c , Nathan Webber c , Bert Lewis a &W. Stewart Grant aa Alaska Department of Fish and Game , Commercial Fisheries Division , 333 Raspberry Road,Anchorage , Alaska , 99518 , USAb Alaska Department of Fish and Game , Commercial Fisheries Division , 43961 KalifornskyBeach Road, Suite B, Soldotna , Alaska , 99669 , USAc Cook Inlet Aquaculture Association , 40610 Kalifornsky Beach Road, Kenai , Alaska , 99611 ,USAPublished online: 29 Jul 2013.

To cite this article: Christopher Habicht , Terri M. Tobias , Gary Fandrei , Nathan Webber , Bert Lewis & W. Stewart Grant(2013) Homing of Sockeye Salmon within Hidden Lake, Alaska, Can Be Used to Achieve Hatchery Management Goals, NorthAmerican Journal of Fisheries Management, 33:4, 777-782, DOI: 10.1080/02755947.2013.808290



Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained inthe publications on our platform. Taylor & Francis, our agents, and our licensors make no representations orwarranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Versionsof published Taylor & Francis and Routledge Open articles and Taylor & Francis and Routledge Open Selectarticles posted to institutional or subject repositories or any other third-party website are without warrantyfrom Taylor & Francis of any kind, either expressed or implied, including, but not limited to, warranties ofmerchantability, fitness for a particular purpose, or non-infringement. Any opinions and views expressed in thisarticle are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. Theaccuracy of the Content should not be relied upon and should be independently verified with primary sourcesof information. Taylor & Francis shall not be liable for any losses, actions, claims, proceedings, demands,costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly inconnection with, in relation to or arising out of the use of the Content.




http://dx.doi.org/10.1080/02755947.2013.808290



North American Journal of Fisheries Management 33:777–782, 2013Published with license by American Fisheries SocietyISSN: 0275-5947 print / 1548-8675 onlineDOI: 10.1080/02755947.2013.808290

MANAGEMENT BRIEF

Homing of Sockeye Salmon within Hidden Lake, Alaska,Can Be Used to Achieve Hatchery Management Goals

Christopher HabichtAlaska Department of Fish and Game, Commercial Fisheries Division, 333 Raspberry Road, Anchorage,Alaska 99518, USA

Terri M. TobiasAlaska Department of Fish and Game, Commercial Fisheries Division, 43961 Kalifornsky Beach Road,Suite B, Soldotna, Alaska 99669, USA

Gary Fandrei and Nathan WebberCook Inlet Aquaculture Association, 40610 Kalifornsky Beach Road, Kenai, Alaska 99611, USA

Bert Lewis and W. Stewart Grant*Alaska Department of Fish and Game, Commercial Fisheries Division, 333 Raspberry Road, Anchorage,Alaska 99518, USA

AbstractThe supplementation of natural populations of Pacific salmon

Oncorhynchus spp. with hatchery fish poses unique managementchallenges. Two such challenges addressed in this study are limitingthe number of hatchery fish spawning with natural-origin fish andmaximizing the number of natural-origin fish in the supplemen-tation broodstock. In this study, we focus on stock enhancementof Sockeye Salmon O. nerka in Hidden Lake, Alaska, where theTrail Lakes Hatchery supplements the natural population withhatchery-raised fry. Production in Hidden Lake is limited by theavailability of spawning habitat and not by juvenile rearing capac-ity. The hatchery collects broodstock from the lake and releasesfry with thermally marked otoliths at one of two primary naturalspawning sites in Hidden Lake each year. During this study, an av-erage of 58% of the fish returning to the lake through a weir on theoutlet stream were of hatchery origin. However, an average of 88%of the fish at the release site were hatchery-origin fish, indicating anonrandom distribution of hatchery-origin spawners. This patternis consistent with homing to specific sites within the lake of either orboth hatchery- and wild-origin fish. However, this distribution re-sults in a larger-than-desirable proportion of hatchery-origin fishspawning with natural-origin fish at the release site. The proportionof hatchery-origin fish used for brood is also larger than desirablebecause the site is also the broodstock collection site. We proposethat releasing hatchery fish at a new location removed from the pri-mary spawning areas and the hatchery broodstock collection site

C© Christopher Habicht, Terri M. Tobias, Gary Fandrei, Nathan Webber, Bert Lewis, and W. Stewart Grant*Corresponding author: [email protected] May 12, 2012; accepted May 20, 2013

will reduce the proportion of hatchery-origin fish spawning withwild-origin fish and increase the proportion of wild-origin fish inthe broodstock, if our results are due, at least in part, to homing ofhatchery fish.

Pacific salmon Oncorhynchus spp. typically spawn in streamsor lakes, spend a variable amount of time as fry in freshwater,and move into marine waters for 2 to 7 years before returningas adults to spawn in freshwater (Quinn 2005). Sockeye SalmonO. nerka are finely adapted to local conditions that influencereproductive success and survival (Taylor 1991; Fraser et al.2011) and hence home to their natal sites on a small spatial scaleto spawn (Quinn et al. 2006). The tendency to home to natal sitesto spawn produces reproductive isolation between populationsand demographic independence among populations that mustbe taken into account in the management of wild and enhancedpopulations supporting commercial fisheries.

In the face of declining harvests and habitat changes, largesalmon hatchery programs were developed in Alaska, BritishColumbia, and the Pacific Northwest beginning in the late 19thcentury (Roppel 1982; Mahnken et al. 1998). These hatcheryprograms currently produce large numbers of fish that may poseecological and genetic risks to wild populations (Reisenbichler

777

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013

778 HABICHT ET AL.

FIGURE 1. Bathymetric map of Hidden Lake, Alaska, showing Sockeye Salmon spawning areas, broodstock collection sites, and fry release sites. Depth isshown in meters.

and McIntyre 1977; Campton 1995; Naish et al. 2007; Grant2012). Recent studies show that hatchery rearing can reducefitness in the wild (Kostow 2004; Araki et al. 2007, 2008) andthat hybridization between hatchery and wild fish can lower theoverall fitness of wild populations (Ford 2002).

In recognition of these risks, the Alaska Department of Fishand Game (ADFG) has implemented regulations to protectwild salmon stocks from the undesirable effects of hatcheries.These guidelines are incorporated into hatchery managementprotocols to address specific areas of concern. One particularunwanted effect is interbreeding between hatchery-reared andwild fish. The ADFG has attempted to reduce the incidence ofhatchery–wild fish interbreeding in the enhancement of smallpopulations of salmon at several sites. For example in an en-hancement program for Chinook Salmon O. tshawytscha onthe Ninilchik River, Alaska, returning hatchery fish are identi-fied by adipose fin clips and not allowed to pass a weir to thespawning grounds (Booz and Kerkvliet 2011). In other cases,the otoliths of hatchery-reared fish are thermally marked (Volket al. 1990, 1999) so that the origins of the broodstock can beidentified. The progeny of these fish can be allocated to specificprograms to meet release goals. The English Bay Lakes, Alaska,Sockeye Salmon enhancement program provides an example

of this type of program, in which progeny of wild-origin par-ents are released back into the lake of origin and progeny fromhatchery-origin parents are released into terminal areas withno wild stocks. These procedures reduce interactions betweenhatchery- and natural-origin fish. The use of otolith marking,however, is somewhat limited, because it permits only retro-spective analysis and cannot be used in real time to managespawning populations.

In this study, we address similar enhancement managementchallenges by describing the distributions of hatchery-rearedand natural Sockeye Salmon in Hidden Lake, Alaska (Figure 1),and by using this information to guide program management toreduce hatchery–wild interactions. The goal of this evaluationwas to determine whether returning hatchery- or wild-originSockeye Salmon home within Hidden Lake. To test this wecompared the proportions of hatchery- and natural-origin fishat the hatchery release site to the proportions of fish enteringthe lake. The null hypothesis is that the distribution of hatchery-and wild-origin fish would be the same at the entry site as therelease site. Based on the results of our study, the program canbe modified to take advantage of the homing behavior to reducerisks associated with hatchery- and wild-origin fish spawningand hatchery broodstock collection.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013


The homing of salmon to natal spawning areas, especiallySockeye Salmon, has been the focus of much research (Dittmanand Quinn 1996; Quinn et al. 1999). Mature Sockeye Salmongenerally spawn in streams associated with a lake, but someadults spawn in the lake itself. Fry spend 0–2 years in fresh-water before migrating to sea as smolts where they spend 1–4 years, before returning as adults to natal streams or lakes tospawn (Quinn et al. 1999). Sockeye Salmon tend to show higherlevels of homing fidelity than other Pacific salmon (Hendryet al. 2004). This homing behavior isolates spawning groupsso that random genetic drift and local selection produce differ-ences among groups, which can often be detected with molec-ular markers (Grant et al. 1980; Hendry et al. 1996, 2000;Seeb et al. 2000). Although lake and stream spawners oftendiffer genetically, less variation among spawning groups oc-curs within lakes (Varnavskaya et al. 1994; Burger et al. 1995;Seeb et al. 2000; Habicht et al. 2007; Gomez-Uchida et al.2011).

The Hidden Lake Sockeye Salmon enhancement programwas started by the State of Alaska in 1976 and has been operatedby the Cook Inlet Aquaculture Association since 1988 to takeadvantage of unused rearing capacity (Kyle et al. 1990). Pop-ulation size in Hidden Lake appears to be limited by availablespawning habitat and not by juvenile rearing capacity (Simpsonand Edmundson 1999). Hidden Lake does not have significantperennial streams flowing into it so that fish spawn almost exclu-sively on beaches. The lake has some of the largest zooplanktonbiomass in lakes on the Kenai Peninsula and produces some ofthe largest Sockeye Salmon smolts of any lake system in Alaska(Simpson and Edmundson 1999). The current goal of the en-hancement program is to produce an annual return of at least30,000 spawners to the lake, and this is achieved by stocking anaverage 830,000 unfed fry (2006–2011).

Eggs are stripped from mature beach-spawning fish, fertil-ized, and incubated at nearby Trail Lakes Hatchery. The otolithsof developing embryos are thermally marked (Volk et al. 1990,1999). Hatchery-origin spawners can be identified with 100%certainty, because a sample of fry is certified by ADFG beforefish are released into the lake. Developing larvae are maintainedat the hatchery for a few months until fry have absorbed theiryolk sac. Unfed fry are trucked from the hatchery, transferredto a boat, which takes them to a single release site (Site B,Figure 1), where they are released from the transport tank. Fryare typically released in the late morning or early afternoonand have been observed to swim down into the substrate in theimmediate area (G. Fandrei, personal observation). When theproject was initiated in 1976, naturally spawning fish were usedas broodstock, but after hatchery fish returned to the lake tospawn, broodstock were taken without regard to the origins ofthe spawners, because it was not possible to determine the ori-gins of fish in real time. Otolith temperature marking was firstapplied in 1996, so that returning hatchery-origin fish could beidentified in 1999.

METHODSStudy Site.—Hidden Lake (60◦29’ N 150◦16’ W) is an

oligomesotrophic system 86 m above sea level and drainsa watershed of 37.4 km2 (Kyle et al. 1990) with aver-age annual precipitation of 44 cm and estimated water res-idence time of 11.7 years (http://www.ciaanet.org/Projects/HIDDEN%20RPT%2007.pdf). The lake is 683 hectares in size,with a mean depth of 20 m and a maximum depth of 45 m. Ithas 22.5 km of shoreline and one seasonally inflowing streamat the southwest end of the lake. The lake is drained by an out-let stream flowing into Skilak Lake, which is part of the KenaiRiver system and is 96 km upstream from Cook Inlet. The fishassemblage is dominated by beach-spawning Sockeye Salmonwith an average return to the lake of 23,300 fish (2006–2011),counted at a weir across the outlet stream. The lake also supportskokanee or residual Sockeye Salmon, Coho Salmon O. kisutch,Dolly Varden Salvelinus malma, and Lake Trout S. namaycush.Two primary Sockeye Salmon spawning areas are located onbeaches at the northwest end of the lake about 1 kilometer apart(Figure 1, Sites A and B). There are other spawning locations inthe lake but the numbers of fish are generally low (<50) on thesesmall scattered areas. Specific spawner counts are unavailable,but similar numbers of fish generally spawn at Sites A and B (G.Fandrei, personal communication). Broodstock are collected atonly Site B, where unfed fry are released.

Sample collection.—Total escapement to Hidden Lake, ascounted at the weir, ranged from 11,002 (2009) to 40,503 (2010)fish during this study. Returning fish were sampled at the weirthree times (1–2 week apart) during the run and were repre-sentative of the run. Fish in spawning condition were collectedusing a beach seine at Sites A and B. Otoliths were collectedfrom these fish or from freshly spawned-out carcasses at leastthree times at Site B throughout the egg-take period from 2008to 2010. Otoliths were also collected from spawners at Site A atthe same times. No Sockeye Salmon spawn in the lake’s tribu-taries. Otoliths were mounted on a glass slide and ground beforeexamining them with a microscope for thermal banding (Volket al. 1990).

Statistical analysis.—In order to determine if year shouldbe controlled across these tests, we first tested the hypothesisthat the proportion of hatchery-origin fish at the weir was ho-mogenous across years. The goal of this project was to test thehypothesis that after entering the lake hatchery- and wild-originfish are randomly distributed among spawning areas. We testedthe hypothesis that the proportion of hatchery-origin fish foundat Site B (hatchery release site) was equal to the proportion ofhatchery-origin fish sampled at the weir. Secondarily, we testedthe hypothesis that the proportion of hatchery-origin fish foundat Site A was equal to the proportion of hatchery-origin fishsampled at the weir. These tests were performed through chi-square tests of contingency tables followed by Fisher’s methodto combine the results across annual tests. Critical value wasset at 0.05 and adjusted for multiple tests using the Bonferroni

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013

780 HABICHT ET AL.

TABLE 1. Number of hatchery-origin (otolith-marked) and natural-origin Sockeye Salmon at the outlet weir and at Site A and Site B (Figure 1), where hatcheryfish are not stocked and are stocked, respectively, in Hidden Lake, Alaska, by year. Percent of hatchery fish and total sample sizes (N) are provided.

Year Site Hatchery Natural Percent hatchery N

2008 Weir 51 75 40.5 126Site A 123 109 53.0 232Site B 197 37 84.2 234

2009 Weir 268 134 67.0 402Site A 124 126 49.4 250Site B 213 31 87.3 244

2010 Weir 314 250 55.7 564Site A 133 108 55.2 241Site B 228 16 93.4 244

All years Weir 633 459 58.0 1,092Site A 380 343 52.6 723Site B 638 84 88.4 722

correction when applicable. Sample sizes provided statisticalpower to detect an expected difference as low as 10%, 95% ofthe time.

RESULTSThermal banding patterns on otoliths in sampled fish from

three sites (2008–2010) indicated that, on average, 58% of re-turning adults sampled at the weir were of hatchery origin, andthat 53% sampled at Site A and 88% sampled at Site B were ofhatchery origin (Table 1). The chi-square test of the hypothesisthat the proportions of hatchery-origin fish at the weir acrossyears was highly significant (P < 0.001) indicating that theseproportions were not consistent across years. This result madeit necessary to control for year in the remaining tests. Chi-square statistics testing the null hypothesis that the proportionof hatchery-origin fish found at Site B (hatchery release site)was equal to the proportion of hatchery-origin fish sampled atthe weir was highly significant for every year (P < 0.001), withhigher proportions of hatchery-origin fish at Site B than at theweir in all three years. Fisher’s Method result across all yearswas highly significant for this hypothesis test (P < 0.001). Thechi-square test of the proportion of hatchery-origin fish foundat Site A was equal to the proportion of hatchery-origin fishsampled at the weir was highly significant only in 2009 (P <

0.001), with a lower proportion of hatchery-origin fish at Site Athan at the weir. In 2008 and 2010, no deviation was observedbetween the proportions of hatchery-origin fish at Site A and atthe weir, after adjustment for multiple tests (P = 0.02 and 0.89,respectively). Fisher’s method result across all years was alsowas highly significant (P < 0.001) for this hypothesis test.

DISCUSSIONThe results of this study show a complex pattern of migration

to spawning areas in Hidden Lake. While 58% of fish enter-

ing the lake were of hatchery origin, 88% of the fish collectedat release Site B were of hatchery origin. A larger proportionof hatchery-origin fish returning to Site B was detected overall 3 years, indicating that this behavior is a persistent fea-ture of Sockeye Salmon spawning in Hidden Lake. However,the origins of these differences are uncertain. Consistent withthese results, a significant chi-square test suggested a lowerproportion of hatchery-origin fish at Site A in 2009, relativeto the proportion of fish entering the lake but not in 2008 or2010.

The nonrandom distribution of spawning fish may be due tohoming within the lake by either wild-origin or hatchery-originfish. The higher proportion of hatchery-origin fish at Site B maybe due to the homing of hatchery-origin fish to Site B wherethey were released or to the homing of wild-origin fish. Becauseabundance estimates at the two sites are unavailable, it is notpossible to determine which explanation is correct. Importantly,both explanations indicate homing behavior within a smalllake system. Natural populations of stream and lake spawningSockeye Salmon are thought to home to natal sites with someprecision (Groot and Margolis 1991; Hendry et al. 1996; Quinnet al. 1999; Gomez-Uchida et al. 2011), so that genetic differ-ences between spawning populations might be expected. How-ever, genetic differences among spawning groups of SockeyeSalmon within a lake were not observed in three Bristol Baylakes (Varnavskaya et al. 1994; Habicht et al. 2007), in threelakes in Cook Inlet (Burger et al. 1995; Seeb et al. 2000), and inone of the two lakes in Russia (Varnavskaya et al. 1994). Only asingle instance of genetic differentiation between lake-spawningaggregations has been reported. In this case, Kuril Lake is ge-ographically much larger than Hidden Lake and supports thelargest lake-spawning Sockeye Salmon population in Russia(Varnavskaya et al. 1994). While the amount of genetic segre-gation between spawning groups in Hidden Lake is unknown,the identification of hatchery-origin and wild-origin fish using

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013


otolith marks indicates the nonrandom mixing among spawningareas.

The results of our study can help to guide the enhancementprogram in Hidden Lake, regardless of the behavioral processesproducing differences between spawning areas. A routine prac-tice in this program is the release of early-stage unfed fry intothe lake. This reduces the amount of time fry are cultured inthe hatchery, reducing opportunities for domestication, and theearly release facilitates exposure to the same environmental cuesat a time when these fish are imprinted. Coho Salmon, releasedas unfed fry, also exhibit life history traits that are more similarto wild salmon than to fish released as smolts (Theriault et al.2010). After release, unfed hatchery fry seek cover in the beachgravel in the immediate vicinity (G. Fandrei, personal obser-vation), increasing the amount of time spent at that location.Fed fry have already made the ontogenic shift from benthicto pelagic habitats selection. When older fry leave the bottomand move around the lake to forage, they abandon their placeof birth (Groot and Margolis 1991), reducing opportunities forimprinting on the physical cues of the release site.

Most hatcheries in Alaska attempt to segregate the hatch-ery population from wild populations by using only hatcherysalmon returning to the hatchery for broodstock. Even thoughpopulation sizes of segregated hatcheries are large so that changethrough random drift is small, genetic changes can still occurthrough the selection of traits that enhance survival during hatch-ery culture (domestication). In fact, selection is more efficientin large than in small populations because the countering effectsof random drift are less important in large populations. In manyhatcheries, the intensity of domestication selection has been re-duced by more closely integrating broodstock for hatchery pro-duction with wild populations (HSRG 2004; Naish et al. 2007).

Based on the results of this study, several steps can be takento reduce the impact of hatchery-origin fish on wild populationsin Hidden Lake. One step would be to collect broodstock atspawning sites with a high proportion of natural fish. In particu-lar, broodstock collection is being moved from Site B to Site A,where the proportion of hatchery-origin fish is significantly lessthan at Site B, where 88% of the fish were of hatchery origin.Our results show that 33–60% of spawners at site B would likelybe natural-origin adults.

Another step would be to release unfed hatchery fry at a newrelease site farther from the broodstock collection site (Site B)and closer to the outlet stream (Figure 1). If our results are dueto homing of hatchery-origin fish to their release site, this stepwould direct returning hatchery-origin fish away from beacheswith naturally spawning fish (Sites A and B). This procedurecould have three effects. First, it would decrease the numberof hatchery-origin fish breeding with natural-origin fish in thewild. Second, it would decrease the proportion of hatchery-origin fish used as hatchery broodstock. Third, it would decreasethe survival of hatchery-origin progeny because of suboptimalenvironmental conditions (coarse substrate) at the new release

site. These desired effects hinge on homing fidelity to the newrelease site for spawning.

Even though these practices would not entirely eliminate theinadvertent inclusion of hatchery-origin fish into the broodstock,they would nevertheless reduce the proportion of hatchery-origin fish and help to reduce genetic risks to natural popu-lations. The Cook Inlet Aquaculture Association, the ADFG,and the U.S. Fish and Wildlife Service have developed a workplan that details program monitoring. This includes ongoingbaseline data collection and sampling to examine how theseprogram modifications alter the proportions of hatchery-originfish in wild and hatchery-broodstock aggregates when the firstfish from the modified program return in 2017. The inclusion ofdetailed fish counts at all sites within Hidden Lake will facili-tate better understanding of hatchery- and wild-origin SockeyeSalmon homing behavior as this project continues in the future.

ACKNOWLEDGMENTSWe thank J. Jasper for providing statistical advice and C.

Cline at the Cook Inlet Aquaculture Association for prepar-ing and reading the otoliths used in the study. Sara Turner,Bill Templin, and Eric Volk provided insightful commentson the manuscript and the comments of four anonymous re-views helped to shape the manuscript considerably. Funding formanuscript preparation and publication were provided by Stateof Alaska general funds. This is Professional Publication Num-ber PP-273 of the Commercial Fisheries Division of the AlaskaDepartment of Fish and Game.

REFERENCESAraki, H., B. A. Berejikian, M. J. Ford, and M. S. Blouin. 2008. Fitness of

hatchery-reared salmonids in the wild. Evolutionary Applications 1:342–355.Araki, H., B. Cooper, and M. S. Blouin. 2007. Genetic effects of captive breeding

cause a rapid, cumulative fitness decline in the wild. Science 318:100–103.Booz, M. D., and C. M. Kerkvliet. 2011. Ninilchik River Chinook

Salmon stock assessment and supplementation, 2008. Alaska Departmentof Fish and Game, Fishery Data Series 11-54, Anchorage. Available:www.adfg.alaska.gov/FedAidpdfs/FDS11-54.pdf. (May 2012).

Burger, C. V., J. E. Finn, and L. Holland-Bartels. 1995. Pattern of shorelinespawning by Sockeye Salmon in a glacially turbid lake: evidence for sub-population differentiation. Transactions of the American Fisheries Society124:1–15.

Campton, D. E. 1995. Genetic effects of hatchery fish on wild populations ofPacific salmon and steelhead: what do we really know? Pages 337–353 inH. L. Schramm Jr. and R. G. Piper, editors. Uses and effects of culturedfishes in aquatic ecosystems. American Fisheries Society, Symposium 15,Bethesda, Maryland.

Dittman, A. H., and T. P. Quinn. 1996. Homing in Pacific salmon: mechanismsand ecological basis. Journal of Experimental Biology 199:83–91.

Ford, M. J. 2002. Selection in captivity during supportive breeding may reducefitness in the wild. Conservation Biology 16:815–825.

Fraser, D. J., L. K. Weir, L. Bernatchez, M. M. Hansen, and E. B. Taylor.2011. Extent and scale of local adaptation in salmonid fishes: review andmeta-analysis. Heredity 106:404–420.

Gomez-Uchida, D., J. E. Seeb, M. J. Smith, C. Habicht, T. P. Quinn, and L. W.Seeb. 2011. Single nucleotide polymorphisms unravel hierarchical divergence

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013

782 HABICHT ET AL.

and signatures of selection among Alaskan Sockeye Salmon (Oncorhynchusnerka) populations. BMC Evolutionary Biology [online serial] 11:article 48.

Grant, W. S. 2012. Understanding the adaptive consequences of hatchery–wildinteractions in Alaska salmon. Environmental Biology of Fishes 94:325–342.

Grant, W. S., G. B. Milner, P. Krasnowski, and F. M. Utter. 1980. Use of biochem-ical genetic variants for identification of Sockeye Salmon (Oncorhynchusnerka) stocks in Cook Inlet, Alaska. Canadian Journal of Fisheries andAquatic Sciences 37:1236–1247.

Groot, C., and L. Margolis, editors. 1991. Pacific salmon: life histories. Univer-sity of British Columbia Press, Vancouver.

Habicht, C., L. W. Seeb, and J. E. Seeb. 2007. Genetic and ecological divergencedefines population structure of Sockeye Salmon populations returning toBristol Bay, Alaska, and provides a tool for admixture analysis. Transactionsof the American Fisheries Society 136:82–94.

Hendry, A. P., V. Castric, M. T. Kinnison, and T. P. Quinn. 2004. The evolutionof philopatry and dispersal: homing versus straying in salmonids. Pages 52–91 in A. P. Hendry and S. C. Stearns, editors. Evolution illuminated: salmonand their relatives. Oxford University Press, New York.

Hendry, A. P., T. P. Quinn, and F. M. Utter. 1996. Genetic evidence for thepersistence and divergence of native and introduced Sockeye Salmon (On-corhynchus nerka) within Lake Washington, Washington. Canadian Journalof Fisheries and Aquatic Sciences 53:823–832.

Hendry, A. P., J. K. Wenburg, P. Bentzen, E. C. Volk, and T. P. Quinn. 2000. Rapidevolution of reproductive isolation in the wild: evidence from introducedsalmon. Science 290:516–518.

HSRG (Hatchery Scientific Review Group). 2004. Hatchery reform: principlesand recommendations of the HSRG. Long Live the Kings, Seattle. Available:www.hatcheryreform.org. (January 2012).

Kostow, K. E. 2004. Differences in juvenile phenotypes and survival betweenhatchery stocks and a natural population provide evidence for modified se-lection due to captive breeding. Canadian Journal of Fisheries and AquaticSciences 61:577–589.

Kyle, G. B., D. S. Litchfield, and G. L. Todd. 1990. Enhancement of Hidden LakeSockeye Salmon (Oncorhynchus nerka): summary of fisheries production(1976–1989). Alaska Department of Fish and Game, Division of FisheriesRehabilitation, Enhancement, and Development, Report 102, Juneau.

Mahnken, C., G. Ruggerone, W. Waknitz, and T. Flagg. 1998. A historical per-spective on salmonid production from Pacific Rim hatcheries. North PacificAnadromous Fish Commission Bulletin 1:38–53.

Naish, K. A., J. E. Taylor III, P. S. Levin, T. P. Quinn, J. R. Winton, D. Huppert,and R. Hilborn. 2007. An evaluation of the effects of conservation and fisheryenhancement hatcheries on wild populations of salmon. Advances in MarineBiology 53:61–194.

Quinn, T. P. 2005. The behavior and ecology of Pacific salmon and trout.American Fisheries Society, Bethesda, Maryland.

Quinn, T. P., I. J. Stewart, and C. P. Boatright. 2006. Experimental evidenceof homing to site of incubation by mature Sockeye Salmon, Oncorhynchusnerka. Animal Behaviour 72:941–949.

Quinn, T. P., E. C. Volk, and A. P. Hendry. 1999. Natural otolith microstruc-ture patterns reveal precise homing to natal incubation sites by Sock-eye Salmon (Oncorhynchus nerka). Canadian Journal of Zoology 77:766–775.

Reisenbichler, R. R., and J. D. McIntyre. 1977. Genetic differences in growthand survival of juvenile hatchery and wild steelhead trout, Salmo gairdneri.Journal of the Fisheries Research Board of Canada 34:123–128.

Roppel, P. 1982. Alaska’s salmon hatcheries, 1891–1959. Alaska Departmentof Fish and Game, Division of Fisheries Rehabilitation, Enhancement, andDevelopment, Juneau.

Seeb, L. W., C. Habicht, W. D. Templin, K. E. Tarbox, R. Z. Davis, L. K.Brannian, and J. E. Seeb. 2000. Genetic diversity of Sockeye Salmon of CookInlet, Alaska, and its application to management of populations affected bythe Exxon Valdez oil spill. Transactions of the American Fisheries Society129:1223–1249.

Simpson, E. M., and J. A. Edmundson. 1999. Hidden Lake Sockeye enhance-ment project: technical review. Alaska Department of Fish and Game, Di-vision of Commercial Fisheries, Regional Information Report 2A99-16,Anchorage.

Taylor, E. B. 1991. A review of local adaptation in Salmonidae, with par-ticular reference to Pacific and Atlantic salmon. Aquaculture 98:185–207.

Theiault, V., G. R. Moyer, and M. A. Banks. 2010. Survival and life history char-acteristics among wild and hatchery Coho Salmon (Oncorhynchus kisutch)returns: how do unfed fry differ from smolt releases? Canadian Journal ofFisheries and Aquatic Sciences 67:486–497.

Varnavskaya, N. V., C. C. Wood, R. J. Everett, R. L. Wilmot, V. S. Varnavsky,V. V. Midanaya, and T. P. Quinn. 1994. Genetic differentiation of subpop-ulations of Sockeye Salmon (Oncorhynchus nerka) within lakes of Alaska,British Columbia, and Kamchatka, Russia. Canadian Journal of Fisheries andAquatic Sciences 51(Supplement 1):147–157.

Volk, E. C., S. L. Schroder, and K. L. Fresh. 1990. Inducement of unique otolithbanding patterns as a practical means to mass-mark juvenile Pacific salmon.Pages 203–215 in N. C. Parker, A. E. Giorgi, R. C. Heidinger, D. B. Jester Jr.,E. D. Prince, and G. A. Winans, editors. Fish-marking techniques. AmericanFisheries Society, Symposium 7, Bethesda, Maryland.

Volk, E. C., S. L. Schroder, and J. J. Grimm. 1999. Otolith thermal marking.Fisheries Research 43:205–219.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013



Evaluation of Internal Tag Performance in Hatchery-Reared Juvenile Spotted SeatroutJonathan P. Wagner a , Reginald B. Blaylock a & Mark S. Peterson aa Department of Coastal Sciences , University of Southern Mississippi , 703 East Beach Drive,Ocean Springs , Mississippi , 39564 , USAPublished online: 29 Jul 2013.

To cite this article: Jonathan P. Wagner , Reginald B. Blaylock & Mark S. Peterson (2013) Evaluation of Internal TagPerformance in Hatchery-Reared Juvenile Spotted Seatrout, North American Journal of Fisheries Management, 33:4, 783-789,DOI: 10.1080/02755947.2013.808292







http://dx.doi.org/10.1080/02755947.2013.808292




ARTICLE

Evaluation of Internal Tag Performance in Hatchery-RearedJuvenile Spotted Seatrout

Jonathan P. Wagner, Reginald B. Blaylock,* and Mark S. PetersonDepartment of Coastal Sciences, University of Southern Mississippi, 703 East Beach Drive,Ocean Springs, Mississippi 39564, USA

AbstractStock enhancement programs rely on the ability to recapture and identify stocked fish to evaluate stocking

effectiveness. Since 2006, the Seatrout Population Enhancement Cooperative (SPEC) has released almost 600,000Spotted Seatrout Cynoscion nebulosus, about 100 mm TL, tagged with opercular coded wire tags (CWTs) into coastalMississippi waters. However, only about 50 fish have been recaptured and initial retention of the opercular CWThas rarely exceeded 75%. This study first evaluated the suitability of visible implant alpha (VIA) and visible implantelastomer (VIE) tags for use in juvenile Spotted Seatrout. The VIA tags performed poorly. Based on those results,the study evaluated the effects of tagging site and fish size on survival, growth, and retention of CWTs and VIE tagsand VIE tag fragmentation in juvenile Spotted Seatrout. Three separate growth experiments with juvenile SpottedSeatrout that had mean initial TLs of 93, 138, and 152 mm, respectively, were conducted to assess the effects oftagging. Each growth experiment had nine treatments consisting of a control, fish with either an opercular CWT, adorsal muscle CWT, a ventral caudal fin VIE tag, or a jaw VIE tag, and four false-tagged treatments correspondingto each tagged treatment. Dorsal CWTs were retained better than opercular CWTs; VIE tags were equally retainedregardless of body location. However, VIE tag quality was affected over the long term by pigmentation overlap andfragmentation. Growth rates and survival were not different within any size-class experiment or among treatments.This study has shown that CWTs and VIE tags are effective marking methods for juvenile Spotted Seatrout.

Almost 40% of known fisheries stocks are considered over-fished, depleted, or recovering from depletion (FAO 2010). Eventhough capture-fisheries production has been stable since thelate 1980s (Diana 2009; FAO 2010), a burgeoning human pop-ulation, which is predicted to grow from 7 to 9 billion by 2050(Cohen 2003), and shifts in social and economic factors havesubstantially increased the demand for fisheries products forboth food and recreation. Management responses to decliningand overfished stocks as required by the Magnuson–StevensAct have routinely included two approaches: the regulation offishing effort through restrictions on catch, season, or gear andhabitat restoration. Although potentially effective, these tech-niques often do not produce quick results.

Stock enhancement, or the release of cultured fish into thewild to supplement wild populations, constitutes a third ap-proach. Stock enhancement is a well-developed and accepted

*Corresponding author: [email protected] October 11, 2012; accepted May 16, 2013

practice in inland freshwater systems dating back to the 17thcentury (Stickney 2000). Marine stock enhancement (MSE)historically suffered from the difficulties in culturing marinelarvae and monitoring releases in an open system (Leber 2004).After 80 years of stocking billions of unmarked fry of a va-riety of species into marine waters, the expected improve-ments in fishery yields never materialized and the practice fellout of favor. Despite management efforts, the decline of wildfisheries continued throughout the latter part of the 20th cen-tury and renewed interest in alternative management strategies.Blankenship and Leber (1995) argued that technology had ad-vanced such that a comprehensive quantitative assessment ofMSE could be accomplished to explain the success or failureof the practice and, as such, this approach should be consid-ered a viable tool in a comprehensive fisheries managementstrategy.

783

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013

784 WAGNER ET AL.

Along with improvements in feeding and filtration technol-ogy, the success of MSE depends on the ability to identify andtrack hatchery-reared fish as they recruit to and supplementwild populations. Tagging has permitted the assessment of im-portant ecological questions such as dispersal (Hendon et al.2002; Miller and Able 2002; Able et al. 2012), straying (Candyand Beacham 2000), and the use of essential habitat (Able et al.2006, 2012). While oxytetracyline and genetic marking play animportant role, physical tags are the most widely used methodof identifying individuals or batches of fish. Different tag de-signs, however, elicit different host responses and have differentretention and recovery rates. The demand for a tag that is wellretained and inflicts little physical damage has led to the devel-opment of visible implant elastomer (VIE) and visible internalalpha (VIA) tags, which are embedded just beneath the epider-mis (Guy et al. 1996) and are externally visible, and the codedwire tag (CWT), which is injected below the epidermis and notexternally visible.

The CWT is a 1-mm-long piece of magnetized stainless steelimprinted with an identifying code. This small size makes itideal for use in small fish (Leblanc and Noakes 2012). The im-printed codes are typically in a sequential format that allows forhundreds of thousands of coding possibilities that facilitates theefficient, long-term marking of large numbers of fish in a rela-tively short period of time. However, CWTs require postmortemrecovery.

The Spotted Seatrout Cynoscion nebulosus is an obligate es-tuarine species with high natal fidelity and limited migration(Hendon et al. 2002; Comyns et al. 2008; Johnson et al. 2011)and is the most popular recreational catch in the Gulf of Mexico(Perret et al. 1980; Hettler 1989; Johnson et al. 2011). As withmany heavily fished populations, the spawning potential ratios(SPRs) often have been less than what is generally consideredideal. In Mississippi, Fulford and Hendon (2010) noted that theannual fishing mortality is close to that for maximum sustain-able yield and the population is highly dependent on annualrecruitment. The combination of the Spotted Seatrout’s biologi-cal characteristics and its popularity, therefore, make it a poten-tially suitable candidate for MSE. As a result, an evaluation ofthe feasibility of MSE as part of a comprehensive managementstrategy for Spotted Seatrout began in 2005, and since 2006almost 600,000 juvenile seatrout with opercular CWTs havebeen released in Mississippi waters. Few tagged fish, however,have been recovered, and 30-d laboratory tag retention evalu-ations have rarely exceeded 75% retention (authors’ personalobservations), which limits the ability to assess the potentialeffectiveness of the program.

To improve the ability to assess the success of a potentialMSE program, we examined the performance of internal tagsin juvenile hatchery-reared Spotted Seatrout. We first assessedthe efficacy of VIA and VIE tags for use in Spotted Seatroutin an attempt to narrow the choices for potential tagging siteson the fish. Based on those results, we then used three separatesize-classes (mean TL of 93, 138, and 152 mm) of fish to eval-

uate the effects of the tagging site (including the currently usedCWT location) and fish size on survival, growth, and retentionof CWTs and VIE tags in juvenile Spotted Seatrout. We alsoevaluated VIE tag fragmentation.

METHODSSource of fish.—Spotted Seatrout for this study were main-

tained at the University of Southern Mississippi (under Institu-tional Animal Care and Use Protocol 08050801). Larvae werespawned from captive adults held in closed recirculating sys-tems under controlled environmental conditions in a mannersimilar to Arnold et al. (1978). To standardize age and growth,all experimental fish, except for those used in the preliminaryevaluations, were derived from a single spawn. After about 50d of rearing in a series of recirculating tanks in which the lar-vae were weaned from rotifers (Brachionus sp.) to brine shrimp(Artemia sp.) and finally onto dry pelleted food, the fish weremaintained on commercial pelleted food (Skretting, Stavanger,Norway) at about 3% of their body weight per day until theywere needed for the growth experiments. Dissolved oxygen wasmaintained above 4.5 mg/L, water temperature was maintainedat about 27◦C, and salinity was maintained at 25‰. Ammonia(as ammonia nitrogen, NH3-N) and nitrite (as nitrite-nitrogen,NH2-N), maintained at <0.25 mg/L through the use of somecombination of Polygeyser bead filters, sand filters, foam frac-tionators, activated carbon, and ozone, were measured dailyusing Hach test strips (Hach, Loveland, Colorado).

Preliminary research.—Pilot research was conducted withonly VIA and VIE tags (both Northwest Marine Technology)to narrow the choices for suitable tagging locations in juvenileSpotted Seatrout (105–146 mm TL). Twenty-four fish were in-jected with red VIE tags either anteroposteriorly into the ventralsurface of the caudal peduncle (n = 8), the ventral jaw tissue (n =8), or the ventral surface just anterior to the pelvic girdle (n =8). These tags were all retained with no mortalities at 18 d posttagging (DPT); however, pelvic tag retention had declined to63%. Mortality associated with VIE tags was low, but cannibal-ism was greater than 50%, which thus compromised long-termretention data on all tags. Actual retention is unknown, but at 51DPT six of the initial 16 ventral jaw and caudal peduncle tagswere still present and of good quality. Thus, the caudal pedun-cle and ventral jaw localities were selected as the experimentaltagging sites.

Twenty fish were implanted with VIA tags anteroposteriorlyin either the caudal peduncle (n = 10) or ventral jaw tissue (n =10) and by 4 d after tagging, all VIA tags had been lost or thefish had died. Because all fish had been lost, a second trial wasconducted with 20 additional fish. In this trial tag retention at51 DPT was 30% (ventral jaw) and 0% (caudal peduncle). Thetags that were retained were unreadable due to pigmentationoverlap. Due to their poor performance, VIA tags were omittedfrom the remainder of the study.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013

TAG PERFORMANCE IN HATCHERY-REARED SPOTTED SEATROUT 785

Experimental design.—The experimental system for themain portion of the study consisted of thirty 60-L blackpolyethylene tanks, each supplied with its own aeration, but con-nected to a common water supply and filtration system. Threeseparate 30-d growth experiments using progressively largerfish were conducted in succession. Experiment 1 used 135 fish(mean length, 93 mm TL at the beginning of the experiment),experiment 2 used 135 fish (mean length, 138 mm TL), andexperiment 3 used 120 fish (mean length, 152 mm TL). At thebeginning of each of the three experiments, fish were movedindividually from the holding tank, anesthetized with tricainemethanesulfonate (MS-222) until they lost equilibrium whileheld in water treated with StressCoat (Aquarium Pharmaceuti-cals, Chalfront, Pennsylvania), weighed (wet weight [WW, g],measured [TL, mm], tagged before being allowed to recover,and moved into the experimental system. Fifteen fish (or 12 fishin the case of experiment 3) were tagged in each of the followingways before five (or four) of each were randomly assigned toeach of three replicate tanks for a total of 27 tanks: (1) a stan-dard sequential CWT (Northwest Marine Technology) injectedin the opercular musculature in a dorsoventral direction usinga Mark-IV automatic injector (Northwest Marine Technology);(2) a standard sequential CWT injected in the epaxial dorsalmusculature in a posteroanterior direction using a Mark-IV au-tomatic injector; (3) a red VIE tag manually injected into theventral caudal fin tissue in a posteroanterior direction using a29-gauge, hand-pressurized, hypodermic syringe; or (4) a redVIE tag injected into the ventral jaw tissue in an anteroposte-rior direction using a 29-gauge, hand-pressurized, hypodermicsyringe. Each of the four tag treatments had a corresponding“handling” control in which an equal number of fish in an equalnumber of tanks were false-tagged, meaning the tag-specificprotocol for each of the four groups including the insertion ofthe needle was performed on the appropriate body part withouttag injection. An additional overall experimental control groupconsisted of fish that were moved from the grow-out tank, anes-thetized, weighed, measured, and handled only when tagged fishwere assessed for growth at days 15 and 30. The three remain-ing tanks were stocked with five unmarked fish each to serve asreplacement fish for treatments that experienced mortality in or-der to maintain similar growth conditions and densities betweentreatments. Fish were fed 3.0- or 4.0-mm dry feed (Skretting) at3% of total body weight per day based on initial measurements.

Tag retention was evaluated on days 3, 6, 9, 12, and 15 dur-ing the first 15 d to assess immediate retention (Wagner 2009).On retention assessment days, only fish belonging to the false-tagged and tagged treatments were netted, removed from theirexperimental tank, anesthetized, checked for tag retention, elas-tomer quality (if applicable), and signs of infection before beingreturned to their respective tank. Elastomer quality was judgedas being a full tag (continuous elastomer) or a fragmented tag (nocontinuous elastomer). On days 15 and 30, the same retentionassessment was conducted; however, fish from all treatments,including the handling control group, were removed and eachfish was anesthetized, weighed, and measured. Based on these

measurements, feed rate and pellet size were adjusted to ac-commodate fish growth and maintain a 3% body weight per dayfeeding rate. Upon completion of each 30-d growth experiment,all tagged fish were grouped together and moved into a 600-L tank within a closed recirculating raceway for estimation oflong-term tag retention. Retention and tag quality were moni-tored once a month for a total of 4 months after tagging. Thesefish were also weighed and measured at 60, 90, and 120 d aftertransfer.

Statistical analyses.—A two-way ANOVA was used to com-pare 30-d growth rate (g/d) between each tagging treatment (n =9) and size-class (n = 3). Sidak’s post hoc tests were used toseparate mean growth rates among the main factors.

Analysis of percent tag fragmentation (expressed as the num-ber of fish with fragmented tags over the total number of VIE-tagged fish per location) of jaw (n = 3) and caudal VIE (n = 3)tag treatments only (between-subjects factors) across four timeperiods (30, 60, 90, and 120 d; within-subject factors) was con-ducted with a split-plot ANOVA for each growth experiment.All tag fragmentation percentage data were arcsine-transformedprior to analysis. Mauchly’s test was also conducted to assessthe sphericity assumption of the repeated measures analyses.When significant F-values were found for the between-subjectsfactors, mean values were separated using Sidak’s post hoc test.If sphericity was violated, the Greenhouse–Geisser correctionwas used to adjust the degrees of freedom in order to determinesignificance for the within-subjects tests.

All analyses were performed with SPSS (version 15 or 20); allvalues were considered significant when P ≤ 0.05. All data usedin ANOVA procedures were tested for normality (one-sampleKolgomorov–Smirnov [K–S] test) and heterogeneity (Levene’stest) assumptions prior to analysis (Field 2005).

RESULTS

SurvivalAt 30 DPT, Spotted Seatrout survival was high in all treat-

ments across all growth experiments and with both tag types.There was only a single tagging-associated mortality through-out the duration of the 30-d study that occurred at 6 DPT in afish with a VIE tag inserted in the jaw during experiment 2.

Fish in all treatments within experiments were present at the120 DPT assessment except for (1) the loss of a replicate of afish with a VIE tag inserted in the caudal fin in experiment 2 dueto an air line malfunction at the day-30 assessment and just priorto combining the fish for continued assessment, and (2) the lossof all fish with dorsal CWTs and fish with VIE tags in the jawat 93 DPT in experiment 2 as a result of a filter malfunction.

Tag RetentionCoded wire tags.—Tag retention after 30 d varied slightly

between experiments for fish tagged with a CWT in the oper-cular musculature (Table 1). The retention of opercular CWTsfor growth experiments 1, 2, and 3 was 87, 80, and 92%, re-spectively, indicating high retention across size-classes. These

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013

786 WAGNER ET AL.

TABLE 1. Tag retention after 30 d for coded wire tags (CWTs) and visible implant elastomer (VIE) tags for three successive growth experiments in juvenileSpotted Seatrout. A bacterial infection prior to growth experiment 3 reduced replicate totals to n = 4 instead of 5 fish per replicate and subsequently to n = 12 fishper treatment instead of 15 as in experiments 1 and 2. Initial TL is included for each experiment.

Experiment 1 Experiment 2 Experiment 3(fish TL, 93 mm) (fish TL, 138 mm) (fish TL, 152 mm)

Tag and location n Tags lost % retention n Tags lost % retention n Tags lost % retention

Dorsal CWT 15 0 100 15 0 100 12 0 100Opercular CWT 15 2 86.7 15 3 80 12 1 91.7Caudal VIE tag 15 0 100 15 0 100 12 0 100Jaw VIE tag 15 0 100 15 0 100 12 0 100

percentages remained constant to 120 DPT. No tag loss in fishwith the opercular CWT was observed after 9 DPT. Dorsal CWTretention was 100% at 120 DPT for all experiments.

Visible implant elastomer tags.—All VIE tags were 100% re-tained at 120 DPT regardless of growth experiment (size-class)or tag location. The use of ultraviolet light and amber-shadedsunglasses was often required to increase the visibility of bothtags, especially in the caudal location. However, tag fragmen-tation was present in fish in all growth experiments for bothtag locations. Fragmentation did not change significantly bytime (split-plot ANOVA: df = 3, P = 0.075) nor tag location(split-plot ANOVA: df = 1, P = 0.613; Figure 1) for eithergrowth experiment. However, there was a possible trend overtime for decreased fragmentation with increased fish size in thejaw-tagged fish (Figure 1A) and final 120-d fragmentation per-centages were 66.7, 50.0, and 41.7% for growth experiments 1,2, and 3, respectively. In contrast, final caudal tag fragmentationwas 20, 90, and 25% for experiments 1, 2, and 3, respectively(Figure 1B). Caudal VIE tags experienced the highest fragmen-tation rates in experiment 2 (fish TL, 138 mm) with percentagesranging from 26.7% at 30 DPT to 90% at 120 DPT. Additionally,darker pigments in the tail region of juvenile Spotted Seatroutresulted in pigmentation overlap, which affected tag visibilityin the caudal VIE treatment.

GrowthMean 30-d growth rates (g/d) were not significantly different

(Figure 2; Table 2) among treatments within an experiment(ANOVA: F1, 8 = 0.514, P = 0.844), but there was a significantsize-class effect (ANOVA: F1, 2 = 21.914, P < 0.001). Sidaktests indicated that growth rate was significantly greater in size-class 1 (experiment 1) compared with those in experiments 2 and3 (all P < 0.001) and that growth rate in size-class 2 (experiment2) equaled that in size-class 3 (experiment 3) (P = 0.619).However, while the data were normally distributed (K–S test:all P > 0.428) there was a slight deviation in the homogeneity ofvariances (Levene’s test: P = 0.036). We consider this differenceneglible, however, based on Underwood’s (1997) discussion ofthe robustness of ANOVA to these minor violations.

DISCUSSIONSurvival was excellent for Spotted Seatrout of all sizes tagged

with either a VIE tag or a CWT. Except for a single fish with aVIE tag inserted in the jaw that died at 6 DPT, no large-scale,unexplained mortalities occurred during the 120-d duration of

FIGURE 1. Percent VIE tag fragmentation over 120 d for (A) jaw and (B)caudal body locations in three successive experiments using progressively largerjuvenile Spotted Seatrout. Mean starting lengths were 93 mm TL (experiment 1),138 mm (experiment 2), and 152 mm TL (experiment 3). A tag was consideredfragmented when no continuous piece of elastomer was present in the tissue. Thepercent fragmentation is expressed as the number of fish with fragmented tagsover the total number of VIE-tagged fish for that body location. In experiment1, the 30-d caudal tags (shown in panel B) were unfragmented, thus there is nobar for that entry. The asterisk indicates the final fragmentation assessment forjaw-tagged fish in experiment 2 was at 93 d instead of 120 d.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013


TABLE 2. Summary of mean 30-d growth rate (g/d) of juvenile Spotted Seatrout for three growth experiments. Each experiment used control fish and fishwith dorsal CWTs (DCWT), opercular CWTs (OPCWT), caudal VIE tags (CFVIE), jaw VIE tags (JVIE), as well as four corresponding false-tag (F plus tagabbreviation, e.g., FDCWT) treatments. Initial TL is included for each experiment.

Growth rate (g/d)

Experiment 1 Experiment 2 Experiment 3Tag treatment (fish TL, 93 mm) (fish TL, 138 mm) (fish TL, 153 mm)

Control 0.41 0.23 0.07DCWT 0.32 0.23 0.11OPCWT 0.28 0.17 0.25CFVIE 0.27 0.20 0.18JVIE 0.28 0.23 0.24FDCWT 0.32 0.16 0.12FOPCWT 0.26 0.23 0.15FCVIE 0.30 0.23 0.21FJVIE 0.31 0.18 0.28

the studies. We therefore concluded that neither CWTs or VIEtags themselves nor the process of using them caused juvenileSpotted Seatrout increased mortality. Reeves and Buckmeier(2009) found that VIE tag-induced mortality was species- andsize-specific. We are unaware of any previous studies of CWTsin juvenile Spotted Seatrout. Although this study did not ad-dress the issue, it is possible that an externally visible tag like aVIE tag could contribute more to postrelease mortality than aninvisible CWT due to increased visibility to predators. Reevesand Buckmeier (2009) found no evidence for such an effectassociated with VIE tags, but noted that the literature is some-what equivocal on the issue. Additionally, retention for CWTsimplanted into epaxial dorsal and opercle muscle averaged 93%

FIGURE 2. Effects of tagging and false-tagging with CWTs and VIE tags on30-d growth rate (g/d; mean ± SE) of juvenile Spotted Seatrout across threesuccessive 30-d growth experiments. Mean starting lengths were 93 mm TL(growth experiment 1), 138 mm TL (growth experiment 2), and 152 mm TL(growth experiment 3). Treatment codes are as follows: DCWT, CWT placedin epaxial dorsal muscle; OPCWT, CWT placed in opercular cheek muscle;CFVIE, VIE tag placed in ventral caudal fin tissue; JVIE, VIE tag placed in lowerjaw tissue; FDCWT, false-tagged (CWT) in epaxial dorsal muscle; FOPCWT,false-tagged (CWT) in opercular cheek muscle; FCVIE, false-tagged (VIE tag)in ventral caudal fin tissue; and FJVIE, false-tagged (VIE tag) in lower jawtissue. False-tagged treatments were those where the appropriate needle for thetag was inserted but no tag was injected.

for both body locations through 120 DPT, and VIE tags in boththe lower jaw and ventral caudal fin were 100% retained at 120DPT. Thus, the CWT and VIE tag are both well suited for usein juvenile Spotted Seatrout.

Although CWT retention rates across both body locationswere high, the only tag that was lost during the study was theopercular CWT, with experiment 2 showing the greatest degreeof tag loss (20%). Additionally, there was a potential negativepattern between fish size and tag loss, and the highest retentionrate for the opercular CWT occurred in the largest size-group offish. Brennan et al. (2005) observed a similar pattern in juvenileCommon Snook Centropomus undecimalis. Perhaps the size ofthe area targeted for tagging plays a key role in initial retention.The opercle muscle of juvenile Spotted Seatrout is relativelythin, thus considerable precision is required when tags are in-serted. Therefore, opercular tags easily could be injected toodeeply or at an inappropriate angle in small fish, which may re-sult in tags being pushed through to the buccal cavity where theycould be initially detected in standard quality control checks butejected soon thereafter (Bumguardner et al. 1992).

One of the goals of this study was to evaluate a possiblealternative site for CWTs because of the lower than expected re-tention of opercular CWTs in the mass tagging program (∼75%,authors’ personal observations). The retention rates observed inthis study for opercular CWTs (mean = 86.1%) were, how-ever, higher than those observed during mass tagging events.Although we cannot explain the difference, one possible expla-nation is that the speed required in a mass tagging event reducesaccuracy and retention. Another possible explanation is that wedid not use the quality-control device (QCD) that is used in masstagging operations to verify the presence of a tag before release.The QCD acts in coordination with the Mark IV automatic taginjectors to sort successfully tagged fish from those tagged un-successfully. This is accomplished by passing the fish through atunnel where a detector separates tagged and untagged fish. In

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013

788 WAGNER ET AL.

the situation where high precision is required for proper inser-tion, as the case may be for Spotted Seatrout, water current inthe QCD or the process of dropping the fish into water as theyexit the QCD could facilitate loss of tags that were not ideallyplaced. Therefore, this study suggests that, with some modifica-tion of procedure, opercular CWT retention in the mass taggingprogram can be increased.

This study also demonstrated that the epaxial dorsal mus-culature was an excellent body location for the use of CWTsbecause no tags were lost at 120 DPT and there was no effect onsurvival or growth. In a short-term study with Red Snapper Lut-janus campechanus Brennan et al. (2007) found dorsal CWTsto be highly successful with a retention rate of 90.4% at 6 weeksafter tagging with no mortality. Dorsal CWTs also have beenused successfully in Rainbow Trout Oncorhynchus mykiss withretention percentages ranging from 92% to 100% (Hale andGray 1998). Despite its success in this study, the dorsal tag bringswith it the criticism of being in a potentially edible area of thefish, which could influence the success of fisheries-independentmonitoring due to the potential unwillingness of anglers to partwith an edible portion of the fish. Opercular tags are not in aportion of the body that is typically consumed and, therefore,are conducive to angler participation. Consumption of the tag,albeit unlikely due to the small size (length, 1.1 mm; diameter,0.25 mm) is also a possibility, but the risk to the consumer wouldbe low.

All VIE tags were visible at 120 DPT, but ultraviolet light andamber-shaded glasses were required throughout the duration ofthe study for effective tag detection. At the base of the caudal finthere was a proliferation of dark, white, and silvery pigments thatincreased with age and resulted in overlap with the VIE tag. Inmarine fishes, VIE tags can be retained well over the short term(Olsen et al. 2004; Bushon et al. 2007), but tag fragmentationis common with VIE tags as muscle tissue spreads with fishgrowth, which can lead to increased fragmentation over time.Brennan et al. (2005) observed pigmentation overlap of VIEtags in the caudal peduncle of older Common Snook and didnot recommend this body location for long-term use. The jawtag was not affected by pigmentation and was more visible thanthe caudal tag.

For both anteroposteriorly tagged VIE tag locations therewas a general increase in fragmentation over time. During ex-periment 2, however, there was a much higher degree of caudaltag fragmentation (90%) compared with experiments 1 and 3and the jaw tag location (50%) in experiment 2. This could not,however, be explained by differential growth over the 30-d timeperiod; fish in experiment 2 had intermediate growth rates. As-torga et al. (2005) demonstrated that anteroposteriorly injectedtags in juvenile Gilthead Seabream Sparus auratus fragmentedless than dorsoventrally injected tags because fish tend to growthin length more than depth. Thus, although we did not specif-ically examine it in this study, anteroposterior tag orientationmay be a better option than a dorsoventral orientation. Size attagging did have an effect on the amount of fragmentation in

the jaw tag. Fish tagged at larger sizes experienced less frag-mentation (41.7%) than those tagged at smaller sizes (61.7%).This finding may be attributable to size of the target area, whichis proportional to the size of the fish and, therefore, inverselyproportional to the difficulty in applying the tag. Transparent tis-sue is limited in juvenile Spotted Seatrout; thus, more precisionis required when injecting a VIE tag into smaller individuals.Therefore, we conclude that VIE tags, particularly those in thejaw, work well for juvenile Spotted Seatrout. However, becauseof fragmentation they may be most ideal for short-term studiesin larger juveniles in which visual, nonlethal identification of thefish is required. The combination of the lack of automation intheir application and the limited number of coding possibilitiescompared with CWTs limits the use of VIE tags in mass taggingoperations.

As has been determined in previous studies (Heidinger andCook 1988; Peterson and Key 1992; Astorga et al. 2005; Hoeyand McCormick 2006) there was no effect of tag type, location,or handling on 30-d growth rate of juvenile Spotted Seatrout inany of the experiments. In our study, small fish had a greatergrowth rate than large fish, which is not unusual during fishontogeny (Weatherley and Gill 1987; Wootton 1992).

The data from this study provide new information regardingthe applicability of current and potential tagging methods forjuvenile Spotted Seatrout. Overall, except for VIA tags, whichclearly require further research, internal tags appear to have littlenegative effect on the well-being of juvenile Spotted Seatroutand fish can be tagged effectively at small sizes without affectinggrowth. Target size and the selection of the tagging site shouldbe considered carefully when choosing tagging locations as bothmay play a role in the overall performance of the tag.

The study suggests that for the critical task of evaluat-ing large-scale enhancement programs (Blankenship and Leber1995), properly applied CWTs are ideal. When combined withthe tag’s cost effectiveness and its inherent capacity for complexcoding patterns, the excellent retention provides the capabilityfor robust experimental designs and long-term reliable identi-fication of hatchery-reared fish. The study also establishes thatVIE tags are practical and reliable, but it also confirms theconclusions of Leblanc and Noakes (2012) that their use forlong-term identification is inadvisable pending further research.Future research also should evaluate how to effectively increasethe retention of CWTs in mass tagging operations for Spot-ted Seatrout as well as improved methods for the use of theCWT in other body locations that are not vulnerable to humanconsumption.

ACKNOWLEDGMENTSWe thank the Science Consortium for Ocean Replenishment

(SCORE) and the Mississippi Tidelands Trust Fund for fundingthis project, and the Seatrout Population Enhancement Cooper-ative (SPEC) staff (J. Snawder, A. Apeitos, B. Schesny, and J.Lemus) for assistance. We thank Northwest Marine Technology,

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013


Inc. for assistance in the use of their products. We also thank J.Lotz for serving on the lead author’s graduate committee and anumber of graduate students who sacrificed their time to helpus with this research.

REFERENCESAble, K. W., S. M. Hagan, and S. A. Brown. 2006. Habitat use, movement,

and growth of young-of-the-year Fundulus spp. in southern New Jersey saltmarshes: comparisons based on tag/recapture. Journal of Experimental Ma-rine Biology and Ecology 335:177–187.

Able, K. W., D. N. Vivian, G. Petruzzelli, and S. M. Hagan. 2012. Connectivityamong salt marsh subhabitats: residency and movements of the Mummichog(Fundulus heteroclitus). Estuaries and Coasts 35:743–753.

Arnold, C. R., J. L. Lasswell, W. H. Bailey, T. D. Williams, and W. A. Fable Jr.1978. Methods and techniques for spawning and rearing Spotted Seatrout inthe laboratory. Proceedings of the Annual Conference Southeastern Associ-ation of Fish and Wildlife Agencies 30(1976):167–178.

Astorga, N., J. M. Afonso, M. J. Zamorano, D. Montero, V. Oliva, H. Fernandez,and M. S. Izquierdo. 2005. Evaluation of visible implant elastomer tagsfor tagging juvenile Gilthead Seabream (Sparus auratus L.): effects ongrowth, mortality, handling time and tag loss. Aquaculture Research 36:733–738.

Blankenship, H. L., and K. M. Leber. 1995. A responsible approach to marinestock enhancement. Pages 167–175 in H. L. Schramm Jr. and R. G. Piper,editors. Uses and effects of cultured fishes in aquatic ecosystems. AmericanFisheries Society, Symposium 15, Bethesda, Maryland.

Brennan, N. P., K. M. Leber, and B. R. Blackburn. 2007. Use of coded-wire andvisible implant elastomer tags for marine stock enhancement with juvenileRed Snapper Lutjanus campechanus. Fisheries Research 83:90–97.

Brennan, N. P., K. M. Leber, H. L. Blankenship, J. M. Ransier, and R. DeBrulerJr. 2005. An evaluation of coded wire and elastomer tag performance in juve-nile Common Snook under field and laboratory conditions. North AmericanJournal of Fisheries Management 25:437–445.

Bumguardner, B. W., R. L. Colura, and G. C. Matlock. 1992. Long-term codedwire tag retention in juvenile Sciaenops ocellatus. U.S. National MarineFisheries Service Fishery Bulletin 90:390–394.

Bushon, A. M., G. W. Stunz, and M. M. Reese. 2007. Evaluation of visibleimplant elastomer for marking juvenile Red Drum. North American Journalof Fisheries Management 27:460–464.

Candy, J. R., and T. D. Beacham. 2000. Patterns of homing and straying in south-ern British Columbia coded-wire tagged Chinook Salmon (Oncorhynchustshawytscha) populations. Fisheries Research 47:41–56.

Cohen, J. E. 2003. Human population: the next half century. Science 302:1172–1175.

Comyns, B. H., C. F. Rakocinski, M. S. Peterson, and A. M. Shiller. 2008.Otolith chemistry of juvenile Spotted Seatrout Cynoscion nebulosus reflectslocal natal regions of coastal Mississippi, USA. Marine Ecology ProgressSeries 371:243–252.

Diana, J. S. 2009. Aquaculture production and biodiversity conservation. Bio-Science 59:27–38.

FAO (Food and Agriculture Organization of the United Nations). 2010. The stateof world fisheries and aquaculture 2010. FAO, Fisheries and Aquaculture De-partment, Rome. Available: www.fao.org/docrep/013/i1820e/i1820e00.htm.(December 2010).

Field, A. 2005. Discovering statistics using SPSS, 2nd edition. Sage Publica-tions, London.

Fulford, R. S., and J. R. Hendon. 2010. Evaluating management ac-tions for Spotted Seatrout, Cynoscion nebulosus, in Mississippi with an

age-structured projection model. Gulf and Caribbean Research 22:51–61.

Guy, C. S., H. L. Blankenship, and L. A. Nielsen. 1996. Tagging and mark-ing. Pages 353–383 in B. R. Murphy and D. W. Willis, editors. Fisheriestechniques, 2nd edition. American Fisheries Society, Bethesda, Maryland.

Hale, R. S., and J. H. Gray. 1998. Retention and detection of coded wire tagsand elastomer tags in trout. North American Journal of Fisheries Management18:197–201.

Heidinger, R. C., and S. B. Cook. 1988. Use of coded wire tags for markingfingerling fishes. North American Journal of Fisheries Management 8:268–272.

Hendon, J. R., J. R. Warren, J. S. Franks, and M. V. Buchanan. 2002. Movementsof Spotted Seatrout (Cynoscion nebulosus) in Mississippi coastal waters basedon tag–recapture. Gulf of Mexico Science 20:91–97.

Hettler, W. F., Jr. 1989. Food habits of juveniles of Spotted Seatrout and GraySnapper in western Florida Bay. Bulletin of Marine Science 44:155–162.

Hoey, A. S., and M. I. McCormick. 2006. Effects of subcutaneous fluorescenttags on the growth and survival of a newly settled coral reef fish, Pomacentrusamboinensis (Pomacentridae). Pages 420–424 in Y. Suzuki, T. Nakamori,M. Hidaka, H. Kayanne, B. E. Casareto, K. Nadaoka, H. Yamano, and M.Tsuchiya, editors. Proceedings of the 10th international coral reef symposium.Japanese Coral Reef Society, Tokyo.

Johnson, M. W., S. A. Bortone, B. W. Klement, and R. L. Shipp. 2011. Populationchanges and location-specific differences for otolith-derived age and growthof recreationally harvested Spotted Seatrout (Cynoscion nebulosus) fromAlabama in 2007. Gulf of Mexico Science 29:13–24.

Leber, K. M. 2004. Marine stock enhancement in the USA: status, trends andneeds. Pages 11–24 in K. M. Leber, S. Kitada, H. L. Blankenship, and T.Svasand, editors. Stock enhancement and sea ranching: developments, pitfallsand opportunities. Blackwell Scientific Publications, Oxford, UK.

Leblanc, C. A., and D. L. Noakes. 2012. Visible implant elastomer (VIE) tagsfor marking small Rainbow Trout. North American Journal of Fisheries Man-agement 32:716–719.

Miller, M. J., and K. W. Able. 2002. Movements and growth of tagged young-of-the-year Atlantic Croaker (Micropogonias undulatus L.) in restored andreference marsh creeks in Delaware Bay, USA. Journal of ExperimentalMarine Biology and Ecology 267:15–33.

Olsen, E. M., J. Gjøsæter, and N. C. Stenseth. 2004. Evaluation of the useof visible implant tags in age-0 Atlantic Cod. North American Journal ofFisheries Management 24:282–286.

Perret, W. S., J. E. Weaver, R. O. Williams, P. L. Johansen, T. D. McIlwain,R. C. Raulerson, and W. M. Tatum. 1980. Fishery profiles of Red Drumand Spotted Seatrout. Gulf States Marine Fisheries Commission, InformationSeries 6, Ocean Springs, Mississippi.

Peterson, M. S., and J. P. Key. 1992. Evaluation of hand-tagging juvenileWalleyes with binary-coded wire microtags. North American Journal of Fish-eries Management 12:814–818.

Reeves, K. S., and D. L. Buckmeier. 2009. Mortality, predation, and tag visibilityof fish marked with visible implant elastomer tags. North American Journalof Fisheries Management 29:323–329.

Stickney, R. R. 2000. History of aquaculture. Pages 436–446 in R. R. Stickney,editor. Encyclopedia of aquaculture. Wiley, New York.

Underwood, A. J. 1997. Experiments in ecology. Cambridge University Press,Cambridge, UK.

Wagner, J. 2009. Evaluation of internal tag performance in hatchery-rearedjuvenile Spotted Seatrout, Cynoscion nebulosus. Master’s thesis. Universityof Southern Mississippi, Hattiesburg.

Weatherley, A. H., and H. S. Gill, editors. 1987. The biology of fish growth.Academic Press, London.

Wootton, R. J. 1992. Fish ecology. Chapman and Hall, New York.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:43

12

Aug

ust 2

013



Posthandling Survival and PIT Tag Retention byAlewives—A Comparison of Gastric and SurgicalImplantsTheodore Castro-Santos a & Volney Vono ba U.S. Geological Survey, Leetown Science Center , S.O. Conte Anadromous Fish ResearchCenter , Post Office Box 796, One Migratory Way, Turners Falls , Massachusetts , 01376 , USAb Fish Passage Center , Federal University of Minas Gerais , Avenida Antônio Carlos, 662731270-901, Belo Horizonte , Minas Gerais , BrazilPublished online: 02 Aug 2013.

To cite this article: Theodore Castro-Santos & Volney Vono (2013) Posthandling Survival and PIT Tag Retention by Alewives—AComparison of Gastric and Surgical Implants, North American Journal of Fisheries Management, 33:4, 790-794, DOI:10.1080/02755947.2013.811130







http://dx.doi.org/10.1080/02755947.2013.811130




MANAGEMENT BRIEF

Posthandling Survival and PIT Tag Retention by Alewives—AComparison of Gastric and Surgical Implants

Theodore Castro-Santos*U.S. Geological Survey, Leetown Science Center, S.O. Conte Anadromous Fish Research Center,Post Office Box 796, One Migratory Way, Turners Falls, Massachusetts 01376, USA

Volney Vono1

Fish Passage Center, Federal University of Minas Gerais, Avenida Antonio Carlos, 6627 31270-901,Belo Horizonte, Minas Gerais, Brazil

AbstractWe compared survival and tag retention of Alewives Alosa pseu-

doharengus tagged with PIT tags, using intraperitoneal (IP) sur-gical implants, gastric implants (GI), and untagged controls heldfor 38 d. Retention was 100% for IP-tagged Alewives and 98%for GI-tagged implants. No significant difference in survival wasobserved among any of these groups. These results lend support tothe use of PIT telemetry for studying fish passage and migrationof anadromous herring. Both methods hold promise for improv-ing estimates of freshwater survival of adult anadromous clupeids;further research should make it also possible to refine estimates ofadult marine survival.

Alewives Alosa pseudoharengus and other river herring wereonce plentiful keystone species on the Atlantic coast of NorthAmerica. They were widely used for food, fertilizer, and bait,and their importance to both riverine and coastal ecology werewidely recognized (Kulik 1985). With industrialization anddamming of coastal streams, however, herring populations be-gan to decline. By some estimates, current populations are lessthan 1% of their historic abundance (ASMFC 2009), and riverherring are currently under review by the National Oceano-graphic and Atmospheric Administration for protection underthe Endangered Species Act. For better understanding of the ef-fectiveness of restoration efforts, improved methods for markingand monitoring individuals are needed.

The causes of declines in river herring are diverse and notwell understood, but barriers to migration, particularly thosecaused by dams for hydropower and other industrial develop-ment, rank high among probable factors (ASMFC 2009). The

*Corresponding author: [email protected] February 4, 2013; accepted May 29, 2013

primary mitigation measure for overcoming these barriers isfishways, particularly the roughened chute type, and increas-ingly of naturelike design (Franklin et al. 2012).

Performance of these fishways is poorly documented, andwhere informative evaluations have been performed, the resultsvary widely (Bunt et al. 2012; Franklin et al. 2012). To betterunderstand fishway performance, it will be necessary to greatlyincrease the number of fishways that are evaluated (Bunt et al.2012; Castro-Santos et al. 2009). Meaningful evaluations ofperformance must quantify not only the number of individualsthat ascend fishways but also the rate and proportion of suc-cess. The best methods currently available for performing suchevaluations involve various forms of telemetry. Active radio-and acoustic telemetry can be informative and offer the abilityto reliably quantify both entry and passage rates (Bunt et al.1999; Castro-Santos and Haro 2010; Castro-Santos and Perry2012). The tags are expensive, however, and these costs precludewidespread evaluations of sufficient sample sizes to provide pre-cise performance estimates.

An alternative to active transmitters are PIT tags. These arerelatively inexpensive (about US$3 each) and have the addedadvantage of an unlimited lifespan (PIT tags have no batteriesand are instead activated by the antenna coils they pass throughor past). They have a disadvantage in that detection ranges areshort compared with those of active transmitters. They are, how-ever, particularly well-suited to quantifying movements of fish inconstrained environments such as fishways, culverts, and smallrivers (Castro-Santos et al. 1996; Franklin et al. 2012). Withinthese environments, the restricted detection range confers otherimportant advantages: locations of tagged fish can be determined

790

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013


with greater precision, and multiple tagged individuals can bepresent without interference among tags.

To reliably interpret fishway and other performance data fromPIT telemetry, one must evaluate whether implantation of tagsresults in added mortality and whether tag retention is longenough for studies to provide meaningful and accurate measuresof performance. Because stress and infection from surgery arewidely expected to affect both survival and behavior, we wereinterested in exploring whether gastric tagging might provide aviable alternative. Gastric implants (GI), where tags are insertedinto the stomach via the esophagus, are widely used in activeradiotelemetry studies. For adult clupeids that are not feedingduring their freshwater migrations, retention of GI active trans-mitters is high, and tag-induced mortality appears to be low(Harris and Hightower 2011). We therefore wanted to knowwhether gastrically implanted PIT tags are retained as well assurgically implanted tags, and whether the gastric method offersany advantage in terms of reduced mortality.

At present, few studies are available that document ratesof tagging-induced mortality or tag loss for PIT-tagged fish,and those studies that are available focus almost exclusively onsalmonids (Ombredane et al. 1998; Gries and Letcher 2002;Skov et al. 2005). We know of no published studies on PIT tagretention and survival for clupeids; such information is impor-tant, however, because clupeids are widely known to be muchmore sensitive to handling than salmonids are, further raisingconcerns of acute and latent tagging effects.

Assessment of tagging effects is further complicated becauseof the high freshwater mortality of anadromous clupeids, com-monly ranging from 40% to 70% during the spawning migration(Havey 1961; Durbin et al. 1979; Franklin et al. 2012). To ac-count for this there is a need for controlled tests of incrementaleffects of tagging on survival. This paper provides a detaileddescription of surgical and gastric implantation methods formarking Alewives with PIT tags and evaluates the effects ofboth methods on tag retention and survival.

METHODSCapture and transport.—Adult migrant Alewives were col-

lected from a fishway on the Monument River at Bournedale,Massachusetts. The fishway is approximately 200 m upstreamfrom the mouth of the Monument River at the Cape Cod Canal(seawater). Alewives were seined from a resting pool, trans-ferred by dip net to a tank truck filled with simulated brackish(7.2‰ salinity) seawater, and transported 13.5 km to the EastSandwich State Fish Hatchery (Massachusetts Division of Ma-rine Fisheries, Sandwich, Massachusetts). This facility had tworound holding tanks (6.1 m diameter) supplied with untreatedstream water at a rate of 15 L/min.

Immediately upon arrival, Alewives were individually dip-netted out of the truck and sequentially assigned to one of twotreatments or a control. No anesthesia was used because the

added handling needed for anesthesia was thought to impartgreater stress than the rapid surgical techniques described here.

Tagging methods.—Treatment fish were tagged withuniquely coded half-duplex PIT tags (TIRIS model RI-TRP-WRHP; Texas Instruments, Dallas, Texas, USA). These glass-encapsulated tags, 23 mm long by 3.8 mm in diameter, canreadily be detected with large, inexpensive antennas suitablefor monitoring full-scale fishways and up to third-order streams(Castro-Santos et al. 1996; Franklin et al. 2012). The half-duplexsystems have the added advantage—particularly important tocoastal anadromous species such as Alewives—of being insen-sitive to changes in water level and salinity (Castro-Santos et al.1996).

Intraperitoneal (IP) surgeries were performed with a no. 15scalpel blade (Figure 1). The width of this blade was 3.75 mm,which made it particularly well-suited for this size PIT tag. Itproduced an incision that the tag had to stretch in order to passthrough, but which closed after insertion without the need forsutures. The incision was cut on the right side, just posterior tothe pelvic fin, three rows of scales dorsal to the ventral midline.Typically this required removal of one or two scales, althoughwhere possible we placed the incision below the scales withoutremoving them. The incision location was selected because thebody wall is thin there and the wound is less likely to affectswimming ability than it would if it were placed on the midline.This method can be superior to syringe methods commonlyused for smaller tags, because the syringe method requires thatthe needle open a round hole larger than the tag, which doesnot close as neatly (Gries and Letcher 2002). Not only did thescalpel method offer greater control, but the combination ofincision with manual insertion reduced risk of injuring internalorgans. Blades and tags were sanitized with ethanol and allowed

FIGURE 1. Illustration of equipment used for inserting PIT tags. Above is ascalpel handle fitted with a no. 15 blade (note the size of the blade relative tothe size of the PIT tag). Below is the applicator device used for the GI implant.[Figure available in color online.]

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013

792 CASTRO-SANTOS AND VONO

to air-dry between each surgery. Total handling time using thismethod was less than 30 s per fish.

The second treatment was the gastric or esophageal implant(GI). This was performed with a specially designed applica-tor, a 15-cm length of stiff-walled polypropylene air hose tube(6.4 mm OD and 4.0 mm ID; Figure 1). We evaluated several dif-ferent types of tubing and found this material had characteristicsof flexibility/stiffness and resilience that made it ideal for thisapplication. The end of the tube was heated over a flame to bur-nish the end (smoothing out any burrs) and gently compressed,reducing the inside diameter at one end to 3.8 mm. Tags wereindividually inserted into the unheated end and pushed downthe tube to the constriction by using a 3.2-mm-diameter stain-less steel rod, cut and fitted with a cap to ensure that it couldreach the end of the applicator tube and no further (Figure 1).Loaded with a single tag, the tube was gently inserted down thethroat of the Alewife until mild resistance indicated that the tubehad reached the stomach. The tube was then slowly withdrawn,while the plunger was simultaneously depressed. This depositedthe tag in the stomach while minimizing the pressure applied tothe internal organs. Total handling time for the GI method wasabout 20 s per fish.

Control fish were individually dipnetted in sequence with thetreatment fish and conveyed directly from the transport tank totheir respective holding tanks without any additional handling.In this way we hoped to identify the total effect of either handlingmethod in comparison with the control group.

Alewives were introduced into the holding tanks individuallyand in sets of three (one from each treatment group and onecontrol), alternating tanks each time a single individual fromeach treatment group had been introduced. In this way bothtanks received an equal number of Alewives of each treatmentgroup, and any cumulative effect of handling and holding timewas equally distributed between the two tanks. Surplus fish wereincluded as additional controls.

Monitoring and statistics.—Alewives were held for 38 d tomonitor survival and tag loss. The floors of the tanks wereinspected twice daily for shed tags or mortalities. Any deadAlewives or found tags were removed, placed in individualsealed plastic bags, and frozen for later examination. Each bagwas labeled with the time of discovery, which was assumed tocorrespond to time of death or ejection. All mortalities wereautopsied by opening the abdominal cavity and evaluating tagplacement, looking for evidence of injury or any other complica-tions associated with tagging. At the end of the study, survivingAlewives were scanned for tags, released to the river, and al-lowed to return to the ocean.

The effect of tagging method on survival was analyzed us-ing a Cox proportional hazards regression, with treatment typeand holding tank included as categorical variables. This methodallowed for comparison of continuous rates of mortality amongthe three groups over the entire duration of the study. The methodcalculates effects of covariates on the “hazard” (mortality rate),which was individually calculated from the time elapsed since

the beginning of the study until the final extant observationor time of death. Mortalities were included as complete ob-servations, and surviving Alewives were included as censoredobservations (Cox 1972; Hosmer and Lemeshow 1999; Castro-Santos and Haro 2003). Individual effects were also tested usingthe Kaplan–Meier method (Wilcoxon and log-rank tests; Kaplanand Meier 1958).

RESULTS AND DISCUSSIONWe used a total of 156 herring in this study, 49 in each treat-

ment group plus 57 controls (Table 1). Average fork length forthis population is 237 ± 11 mm (Castro-Santos 2005), meaningthat tags were about 10% of the body length. All Alewives werecollected and tagged on May 4, 2009, and the study was termi-nated 38 d later, on June 11, 2009. By May 4, 65% of the run hadpassed through the fishway. By June 11 (the termination date),99% of the run had passed. Alewives present in the collectionpool may have included some downstream migrants and otherindividuals that had been present in the pool for some time.Because the Alewives were collected late in the run, we assumethey were more susceptible to handling stress than were thosethat arrived earlier in the run (Leonard and McCormick 1999;Castro-Santos and Letcher 2010). Thus any mortality effectsobserved in this study should be conservative, that is, tending tooverestimate the magnitude of handling effect.

A total of 78 herring (49.7%) died during this study (Table 1),more than half of these deaths occurring after 30 d; only 6 (3.8%)died within the first week (Figure 2). The only tag ejected wasfrom a GI-tagged fish; it was found on the tank floor 1.8 d afterthe start of the experiment. Thus overall tag retention rates were99.0% (100% for IP-tagged Alewives and 98% for GI-taggedAlewives). The increased mortality during the final 8 d promptedus to terminate the study. Mortalities were not equally distributedbetween the two tanks, with 48 (59%) deaths in Tank 1 and 30deaths (40%) in Tank 2 (Wilcoxon and log-rank tests P < 0.01).Although we did not test for disease, the fact that both tanksreceived identical treatments suggests that disease is the mostlikely cause of this difference. The observed mortality is typical

TABLE 1. Survival and tag retention by PIT-tagged Alewives. Data aregrouped by tank and tagging method.

N

Tank Method Alive Dead Ejected Total

1 Control 13 18 — 31a

GI 7 18 1 26IP 14 11 0 25

2 Control 18 10 — 28GI 13 10 0 23IP 14 10 0 24

aIncludes one Alewife that ejected its tag; it was not possible to determine whether thisfish survived.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013


Time (d)

0 10 20 30

Sur

vivo

rshi

p (%

)

0

20

40

60

80

100

FIGURE 2. Survivorship curves comparing effects of tagging with 23-mm PITtags on survival of Alewives. Data are for controls (solid lines), gastric implants(short dashes), and intraperitoneal implants (long dashes). Dotted lines indicatethe 95% confidence interval of the survival function for the control group.

of the freshwater residence period for adults of this species (40–70%: Havey 1961; Dalton et al. 2009; Franklin et al. 2012).

Neither tagging method increased mortality significantlycompared with controls (Table 2). Surprisingly, the best survivalwas among the IP-tagged herring, and differences between IP-and GI-tagged herring approached significance (P = 0.098).

Five (10.2%) of the IP-tagged Alewives showed evidence ofdamage to the gonads—in one case the testes were inflamed; in asecond they showed evidence of rupture, presumably caused bytagging. In the remaining three cases, the tag was embedded inthe gonads (two ovaries, one testis). It was not evident whetherany of these injuries had led to systemic infection, and noneof the other dead IP-tagged Alewife mortalities showed anyevidence of tag-induced trauma. Of the GI-tagged Alewives,one tag was found outside of the gastrointestinal tract in theperitoneum. We assume that this tag penetrated the stomachduring insertion, which could have contributed to the death ofthis individual (at 9.1 d after tagging). Only one other GI-taggedindividual showed signs of trauma, having inflammation of boththe stomach and the pyloric caeca. Combined with the control

TABLE 2. Results of the Cox proportional hazards regression of tank andtagging effects on the log-mortality rate. Tank effect was coded as binary vari-ables (0 for Tank 1, 1 for Tank 2). Tagging methods were also coded as binaryvariables. Estimates indicate the effect on log mortality rate [ln(hazard)] of eachtank and each treatment compared with controls. The hazard ratio given is theexponential value of the estimate and indicates the change in mortality rateassociated with each tank or treatment.

Effect DF Estimate ± SE Pr > ψ2 Hazard ratio

Tank 1 0.617 ± 0.233 0.008 1.853IP 1 −0.151 ± 0.289 0.600 0.859GI 1 0.324 ± 0.265 0.221 1.383

data, the necropsy data suggest that, while there was doubtlesssome additional stress and risk of trauma associated with theimplantation methods, neither method was a driving factor inthe deaths of the tagged individuals.

This study demonstrates that neither GI nor IP tagging meth-ods increased mortality of anadromous Alewives held in fresh-water for 38 d; moreover, retention rates for both methods ap-proached 100%. The duration of this study was nearly twice thetypical duration of freshwater migration for individual Alewives(Franklin et al. 2012), suggesting that PIT telemetry using eithermethod is potentially an effective tool for monitoring migratorybehavior, passage success, and postspawning survival.

These results are consistent with those documented previ-ously for other taxa. Retention was slightly better than has beenobserved in juvenile Atlantic Salmon Salmo salar and BrownTrout S. trutta tagged with 12-mm PIT tags (Ombredane et al.1998; Gries and Letcher 2002). The effect of tagging on sur-vival also compared favorably with that found for salmonids.Ostrand et al. (2012) found that tags similar in size to thoseused here caused a temporary reduction in growth rates amongtagged Coho Salmon Oncorhynchus kisuch and SteelheadO. mykiss, but not among Cutthroat Trout O. clarkii or Bull TroutSalvelinus confluentus, and had no significant effect on survival.

Despite these results, concern remains that the stress of han-dling could cause fish to lose migratory motivation, or to per-form less well than unhandled fish once they enter fishways. Inthis respect, the fact that PIT-tagging does not appear to posea significant survival risk is serendipitous: because herring areiteroparous, a proportion of tagged individuals are likely to re-turn to the ocean and survive to spawn in subsequent years.This means that studies performed over several years (and thisis recommended, given the known interannual variations in mi-gratory behavior and performance) may recruit data from repeatspawners. Repeat spawners are particularly valuable becausethey can be considered effectively unhandled after their initialyear of tagging. Even if only a small proportion of repeat spawn-ers survive and retain their tags, passage performance of theseindividuals can then be compared with that of untagged fish andbe used to quantify the magnitude and variability of the effectof tagging on behavior. This will improve estimates of passageperformance and survival.

Despite these encouraging results, further study is neededto better understand sublethal and latent effects of tagging andhandling on migratory fish and to determine which method isbest for long-term studies. Because our Alewives were held fora brief period and in a fasted state, we could not compare theeffects of the two tagging methods on feeding, growth, and long-term survival. Also, once the Alewives resumed feeding, the GItags probably would be passed and so not be retained by repeatspawners. A similar risk exists for IP-tagged Alewives: tags canbe ejected through transintestinal absorption and also duringspawning (Marty and Summerfelt 1986). More work is neededto better understand how retention and survival are affected overthe longer term, and how implantation methods can be improved

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013

794 CASTRO-SANTOS AND VONO

to maximize survival and minimize behavioral effects. Whenthese data become available, it will become possible to use PITtelemetry to evaluate not only passage performance but also thefreshwater and marine survival of this and related species.

ACKNOWLEDGMENTSAny use of trade, product, or firm names is for descriptive

purposes only and does not imply endorsement by the U.S.Government. Philips Brady, Edward Clark, and Luis Carmo ofthe Massachusetts Division of Marine Fisheries, and Amy Tefferof the University of Massachusetts Amherst provided essentialsupport throughout this study. Shortly after conducting the workdescribed in this paper, Volney Vono was diagnosed with braincancer, from which he died in 2011. Vono was a committedconservationist and scientist whose gentle integrity is missed byall who knew him. This paper is dedicated to his memory.

REFERENCESASMFC (Atlantic States Marine Fisheries Commission). 2009. Ammendment 2

to the Interstate Fishery Management Plan for Shad and River herring (RiverHerring management). ASMFC, Arlington, Virginia.

Bunt, C. M., T. Castro-Santos, and A. Haro. 2012. Performance of fish passagestructures at upstream barriers to migration. River Research and Applications28:457–478.

Bunt, C. M., C. Katopodis, and R. S. McKinley. 1999. Attraction and passageefficiency of White Suckers and Smallmouth Bass by two Denil fishways.North American Journal of Fisheries Management 19:793–803.

Castro-Santos, T. 2005. Optimal swim speeds for traversing velocity barriers:an analysis of volitional high-speed swimming behavior of migratory fishes.Journal of Experimental Biology 208:421–432.

Castro-Santos, T., A. Cotel, and P. W. Webb. 2009. Fishway evaluations for betterbioengineering – an integrative approach. Pages 557–575 in A. J. Haro, K.L. Smith, R. A. Rulifson, C. M. Moffit, R. J. Klauda, M. J. Dadswell, R.A. Cunjak, J. E. Cooper, K. L. Beal, and T. S. Avery, editors. Challengesfor diadromous fishes in a dynamic global environment. American FisheriesSociety, Symposium 69, Bethesda, Maryland.

Castro-Santos, T., and A. Haro. 2003. Quantifying migratory delay: a newapplication of survival analysis methods. Canadian Journal of Fisheries andAquatic Sciences 60:986–996.

Castro-Santos, T., and A. Haro. 2010. Fish guidance and passage at barriers.Pages 62–89 in P. Domenici and B. G. Kapoor, editors. Fish locomotion: aneco-ethological perspective. Science Publishers, Enfield, New Hampshire.

Castro-Santos, T., A. Haro, and S. Walk. 1996. A passive integrated transponder(PIT) tagging system for monitoring fishways. Fisheries Research 28:253–261.

Castro-Santos, T., and B. H. Letcher. 2010. Modeling migratory bioenergeticsof Connecticut River American Shad (Alosa sapidissima): implications forthe conservation of an iteroparous anadromous fish. Canadian Journal ofFisheries and Aquatic Sciences 67:806–830.

Castro-Santos, T., and R. W. Perry. 2012. Time-to-event analysis as a frameworkfor quantifying fish passage performance. Pages 427–452 in N. S. Adams,J. W. Beeman, and J. Eiler, editors. Telemetry techniques. American FisheriesSociety, Bethesda, Maryland.

Cox, D. R. 1972. Regression models and life tables. Journal of the RoyalStatistical Society 34:187–220.

Dalton, C. M., D. Ellis, and D. M. Post. 2009. The impact of double-crestedcormorant (Phalacrocorax auritus) predation on anadromous alewife (Alosapseudoharengus) in south-central Connecticut, USA. Canadian Journal ofFisheries and Aquatic Sciences 66:177–186.

Durbin, A. G., S. W. Nixon, and C. A. Oviatt. 1979. Effects of the spawningmigration of the Alewife, Alosa pseudoharengus, on freshwater ecosystems.Ecology 60:8–17.

Franklin, A. E., A. Haro, T. Castro-Santos, and J. Noreika. 2012. Evaluation ofnature-like and technical fishways for the passage of Alewife (Alosa pseu-doharengus) at two coastal streams in New England. Transactions of theAmerican Fisheries Society 141:624–637.

Gries, G., and B. H. Letcher. 2002. Tag retention and survival of age-0Atlantic Salmon following surgical implantation with passive integratedtransponder tags. North American Journal of Fisheries Management 22:219–222.

Harris, J. E., and J. E. Hightower. 2011. Movement patterns of American Shadtransported upstream of dams on the Roanoke River, North Carolina andVirginia. North American Journal of Fisheries Management 31:240–256.

Havey, K. A. 1961. Restoration of anadromous Alewives at Long Pond, Maine.Transactions of the American Fisheries Society 90:281–286.

Hosmer, D. W., and S. Lemeshow. 1999. Applied survival analysis. Wiley, NewYork.

Kaplan, E. L., and P. Meier. 1958. Nonparametric estimation from incompleteobservations. Journal of the American Statistical Association 53:457–481.

Kulik, G. 1985. Dams, fish, and farmers: defense of public rights in eighteenth-century Rhode Island. Pages 25–50 in S. Hahn and J. Prude, editors. Thecountryside in the age of capitalist transformation: essays in the social historyof rural America. University of North Carolina Press, Chapel Hill.

Leonard, J. B. K., and S. D. McCormick. 1999. Effects of migration distanceon whole-body and tissue-specific energy use in American Shad (Alosa sa-pidissima). Canadian Journal of Fisheries and Aquatic Sciences 56:1159–1171.

Marty, G. D., and R. C. Summerfelt. 1986. Pathways and mechanisms forexpulsion of surgically implanted dummy transmitters from Channel Catfish.Transactions of the American Fisheries Society 115:577–589.

Ombredane, D., J. L. Bagliniere, and F. Marchand. 1998. The effects of passiveintegrated transponder tags on survival and growth of juvenile Brown Trout(Salmo trutta L.) and their use for studying movement in a small river.Hydrobiologia 372:99–106.

Ostrand, K. G., G. B. Zydlewski, W. L. Gale, and J. D. Zydlewski. 2012. Longterm retention, survival, growth, and physiological indicators of juvenilesalmonids marked with passive integrated transponder tags. Pages 135–145in J. McKenzie, B. Parsons, A. Seitz, R. K. Kopf, M. G. Mesa, and Q.Phelps, editors. Advances in fish tagging and marking technology. AmericanFisheries Society, Symposium 76, Bethesda, Maryland.

Skov, C., J. Brodersen, C. Bronmark, L. A. Hansson, P. Hertonsson, and P.A. Nilsson. 2005. Evaluation of PIT-tagging in cyprinids. Journal of FishBiology 67:1195–1201.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013



Juvenile Movement among Different Populations ofCutthroat Trout Introduced as Embryos to VacantHabitatTessa M. Andrews a f , Bradley B. Shepard b , Andrea R. Litt c , Carter G. Kruse d , AlexanderV. Zale e & Steven T. Kalinowski ca Division of Biological Sciences , University of Georgia , 831 Biological Sciences, 120 CedarStreet, Athens , Georgia , 30602 , USAb Wildlife Conservation Society , Yellowstone Rockies Program , 301 North Willson Avenue,Bozeman , Montana , 59715 , USAc Department of Ecology , Montana State University , Post Office Box 173460, Bozeman ,Montana , 59717 , USAd Turner Enterprises, Incorporated , 1123 Research Drive, Bozeman , Montana , 59718 , USAe U.S. Geological Survey, Montana Cooperative Fishery Research Unit , Montana StateUniversity , Post Office Box 173460, Bozeman , Montana , 59717 , USAf Department of Genetics , University of Georgia , 405 Biological Sciences, 120 Cedar Street,Athens , Georgia , 30602 , USAPublished online: 06 Aug 2013.

To cite this article: Tessa M. Andrews , Bradley B. Shepard , Andrea R. Litt , Carter G. Kruse , Alexander V. Zale & StevenT. Kalinowski (2013) Juvenile Movement among Different Populations of Cutthroat Trout Introduced as Embryos to VacantHabitat, North American Journal of Fisheries Management, 33:4, 795-805, DOI: 10.1080/02755947.2013.812582







http://dx.doi.org/10.1080/02755947.2013.812582




MANAGEMENT BRIEF

Juvenile Movement among Different Populationsof Cutthroat Trout Introduced as Embryos to Vacant Habitat

Tessa M. Andrews*1

Division of Biological Sciences, University of Georgia, 831 Biological Sciences, 120 Cedar Street,Athens, Georgia 30602, USA

Bradley B. ShepardWildlife Conservation Society, Yellowstone Rockies Program, 301 North Willson Avenue, Bozeman,Montana 59715, USA

Andrea R. LittDepartment of Ecology, Montana State University, Post Office Box 173460, Bozeman,Montana 59717, USA

Carter G. KruseTurner Enterprises, Incorporated, 1123 Research Drive, Bozeman, Montana 59718, USA

Alexander V. ZaleU.S. Geological Survey, Montana Cooperative Fishery Research Unit, Montana State University,Post Office Box 173460, Bozeman, Montana 59717, USA

Steven T. KalinowskiDepartment of Ecology, Montana State University, Post Office Box 173460, Bozeman,Montana 59717, USA

AbstractTranslocations are frequently used to increase the abundance

and range of endangered fishes. One factor likely to affect the out-come of translocations is fish movement. We introduced embryosfrom five Westslope Cutthroat Trout Oncorhynchus clarkii lewisipopulations (both hatchery and wild) at five different locationswithin a fishless watershed. We then examined the movement ofage-1 and age-2 fish and looked for differences in movement dis-tance among source populations and among introduction sites; wealso examined the interactions among age, population, and intro-duction site. At age 1, most individuals (90.9%) remained within1,000 m their introduction sites. By age 2, the majority of individu-als (58.3%) still remained within 1,000 m of their introduction site,but considerably more individuals had moved downstream, somemore than 6,000 m from their introduction site. We observed a sig-nificant interaction between age and source population (F4, 1077 =15.45, P < 0.0001) as well as between age and introduction site

*Corresponding author: [email protected] address: Department of Genetics, University of Georgia, 405 Biological Sciences, 120 Cedar Street, Athens, Georgia 30602, USA.Received April 20, 2012; accepted June 3, 2013

(F41, 1077 = 11.39, P < 0.0008), so we presented results in the contextof these interactions. Within age-groups, we observed differences inmovement behavior among source populations and among donorpopulations of Westslope Cutthroat Trout. We discuss these find-ings in light of previous research on juvenile salmonid movement.

Translocating fish is an important conservation strategy formany imperiled fish species. Translocations can create new pop-ulations by reestablishing fish in habitats that were historicallyoccupied or by establishing new populations in historicallyfishless habitats (e.g., U.S. Fish and Wildlife Service 1998;Colorado Division of Wildlife 2004). Both of these types oftranslocations can increase the range and abundance of species atrisk, which should decrease the risk of extinction (Griffith et al.1989). In addition to creating new populations, fish managers

795

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013

796 ANDREWS ET AL.

use translocations to supplement existing populations. Introduc-ing individuals into existing populations can reduce the effects ofinbreeding and the demographic risk of extinction (e.g., Madsenet al. 1999; Pimm et al. 2006; Bouzat et al. 2009). Introducingindividuals can also speed the recovery of populations follow-ing other management interventions, such as nonnative speciesremoval or habitat restoration (e.g., Jones 2010).

Westslope Cutthroat Trout Oncorhynchus clarkii lewisi ex-emplify an imperiled taxon that would benefit from translo-cations. Westslope Cutthroat Trout have been extirpated frommuch of their historical habitat (Shepard et al. 2005), and manyof the remaining genetically pure populations are restricted toheadwater streams where they are isolated from other trout pop-ulations by barriers that prevent upstream movement of fish.These barriers protect populations from hybridization with non-native Rainbow Trout O. mykiss and from competition withnonnative Brook Trout Salvelinus fontinalis, but increase therisk of demographic stochasticity and inbreeding depression(Peterson et al. 2008; Fausch et al. 2009). Translocations ofWestslope Cutthroat Trout could facilitate conservation by ame-liorating the negative effects of inbreeding, simulating gene flowamong isolated populations, and establishing additional popula-tions (e.g., GCTRT 1998; Alves et al. 2004; CRCT CoordinationTeam 2006; Teuscher and Capurso 2007; MDFWP 2007; Gress-well 2011). However, before translocations are widely used as aWestslope Cutthroat Trout conservation tool, we need to knowmore about the factors affecting the success of these projects.

Data gathered from translocations of other Cutthroat Troutsubspecies (Greenback Cutthroat Trout O. clarkii stomias andRio Grande Cutthroat Trout O. clarkii virginalis) have helpedidentify factors that can influence the success of these conser-vation efforts. Habitat features like cold summer water temper-ature, narrow stream width, and a lack of deep pools has limitedthe success of previous Cutthroat Trout translocations (Harigand Fausch 2002), while translocation sites with at least 2 ha ofhabitat that previously supported reproducing trout populationshave had the highest rates of success (Harig et al. 2000).

Movement following translocation is another factor likely tobe important in Cutthroat Trout translocation projects. Stream-dwelling trout can exhibit extensive movement (e.g., Gowan andFausch 1996; Hilderbrand and Kershner 2000; Schmetterlingand Adams 2004; Gresswell and Hendricks 2007). Extensivedownstream movement over a protective barrier could seriouslycompromise a Cutthroat Trout translocation because individualsmoving past the barrier would be “lost” to the project, as wouldthe genetic diversity they could add to the population. Stream-dwelling trout can also exhibit restricted movement, remainingwithin ∼100-m home ranges (Clapp et al. 1990; Rodrıguez2002). If translocated fish remain in small home ranges neartheir introduction site, multiple introduction sites throughout asystem—rather than a single introduction site—may be neces-sary to meet restoration goals.

The Cherry Creek project provided an opportunity to studytrout movement in a translocation project. The goal of the CherryCreek project was to create a genetically diverse Westslope Cut-

throat Trout population in a secure refuge within the MadisonRiver basin by translocating almost 35,000 embryos from mul-tiple populations into habitat vacated after a series of pisci-cide treatments (Bramblett 1998). The translocated populationis thriving and should soon be the largest genetically pure West-slope Cutthroat Trout population east of the Continental Dividein Montana (Lee Nelson, Montana, Fish, Wildlife, and Parks,personal communication).

Our goal for this study was to describe how translocatedWestslope Cutthroat Trout in the Cherry Creek project moved.We looked for differences in movement between age-1 and age-2 individuals, among individuals from different source popula-tions, and among individuals introduced to different locationsin the study system.

METHODSStudy design.—The Cherry Creek project consisted of two

phases. During the first phase, about 90 km of Cherry Creekthat is isolated from downstream habitats by an 8-m-high wa-terfall was treated with the piscicides antimycin and rotenone toremove nonnative fish species. Historically, there were no fishabove the waterfall. However, nonnative Brook Trout, RainbowTrout, and Yellowstone Cutthroat Trout O. clarkii bouvieri oc-cupied this habitat in the 20th century, most probably as a resultof introductions to the watershed (P. Clancey, Montana Fish,Wildlife, and Parks, personal communication). During the sec-ond phase of the project, in which this study took place, weintroduced Westslope Cutthroat Trout embryos and then exam-ined the movement of juvenile fish.

Source populations.—Embryos introduced to the study sitecame from two hatchery and three wild-source populations. Oneof the hatchery populations was the state of Montana’s captiveWestslope Cutthroat Trout conservation population, which isreared at Washoe Park Hatchery in Anaconda, Montana. Thispopulation was founded in 1984 from populations of trout inthe upper Flathead and Clark Fork river drainages. The popula-tion was infused with additional gametes from the FlatheadRiver drainage about 20 years later (M. Sweeney, MontanaFish, Wildlife and Parks, personal communication). The otherhatchery population was from a collaborative public–privateWestslope Cutthroat Trout hatchery located on the Sun Ranch(44.965◦N, 111.605◦W) within the Madison River drainage.This population was founded in 2002 using individuals fromthe same three wild populations that donated embryos to thisproject (described below).

We introduced embryos from three wild-sourcepopulations—Ray, Muskrat, and White’s creeks. All three ofthese creeks supported genetically pure Westslope CutthroatTrout. The population from Ray Creek (46.411◦N, 111.267◦W)was estimated to have 2,000–3,000 age-1 and older individualsinhabiting over 8 km of stream when surveyed in 2007. Thispopulation was isolated from nonnative fish by a perchedculvert (L. Nelson, Montana Fish, Wildlife and Parks, per-sonal communication). The population from Muskrat Creek

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013


(46.302◦N, 112.032◦W) was estimated to have 3,500–4,000age-1 and older individuals inhabiting over 8 km of streamwhen surveyed in 2007 (L. Nelson, personal communication).This population may have been as small as 100 individualsbefore aggressive management began in 1997 (L. Nelson,personal communication). The population from White’s Creek(46.618◦N, 111.491◦W) was estimated to have about 1,000age-1 and older individuals inhabiting slightly more than3 km of stream when surveyed in 2006. This population alsodecreased to about 100 individuals in the 1990s (Shepard et al.2002). The populations in Muskrat and White’s creeks requiredintensive restoration management in recent years, including theconstruction of human-made barriers, Brook Trout removal,and habitat restoration (Shepard et al. 2002; L. Nelson, personalcommunication).

Since hybridization is a constant threat to Cutthroat Troutpopulations, we screened all source populations for poten-tial introgression with Rainbow Trout and Yellowstone Cut-throat Trout by genetically testing all source adults to ensurethat no evidence of genetic introgression was present. In addi-tion, all source populations were screened to ensure they werefree of the following fish pathogens: Renibacterium salmoni-narum, Aeromonas salmonicida, Yersinia ruckeri, Myxoboluscerebralis, infectious hematopoetic necrosis virus, infectiouspancreatic necrosis virus, and viral hemorrhagic septicemiavirus. We screened for these disease pathogens by samplingat least 60 fish from each source population or from a surrogatespecies in sympatry with the source population prior to transferof embryos (Ken Staigmiller, Montana Fish, Wildlife and Parks,personal communication). All individuals collected for screen-ing and for introduction to Cherry Creek were collected underpermits issued by the state of Montana and in compliance withstate policy regarding the transfer of fish among locations.

Embryo collection and introduction.—We collected embryosfrom source populations in 2007, 2008, and 2009. In the wild, wecaptured adult trout using a backpack electrofisher and confinedthem instream in flow-through containers until they were readyto spawn. After we spawned wild fish, we marked them byclipping their dorsal fins and then released them. This assuredthat wild fish were only spawned once for this project. We alsocollected a small pelvic fin clip from each spawning adult forgenetic analysis.

At the Washoe Park Hatchery, we collected gametes fromripe adults once a week. Our goal was to introduce embryosfrom Washoe Park Hatchery into Cherry Creek at the same timeas we introduced all wild embryos. Since we could not predictexactly when wild embryos would be at the eyed stage—andtherefore ready for introduction—we collected embryos at thehatchery over the course of a month to maximize our chances ofhaving embryos from Washoe Park Hatchery at the eyed stageat same time as the wild embryos.

At the Sun Ranch Hatchery, we attempted to spawn all ripeadults. We captured adults from the facility’s holding pond withseine nets three times a week for the entire spawning season

and spawned all ripe males and females that had not previouslycontributed gametes to the project. We held captured adultsin mesh containers within the pond and released them fromcontainers after completing embryo collection.

We followed the same spawning protocol for all fishcontributing gametes to the project. We stripped each female ofeggs and divided the eggs into groups, which were distributedamong 1-L insulated bottles. The eggs in each bottle werefertilized with milt from a different male to produce a uniquemale × female cross. Next we used water to rinse away remain-ing milt and left the fertilized eggs (embryos) undisturbed for atleast 30 min to water harden in a water–iodophor solution. Wildembryos were packed in coolers for transport to the hatcheryat Sun Ranch, where they were incubated alongside embryosfrom the Sun Ranch Hatchery population. Embryos from theWashoe Park Hatchery population were incubated on site atthat hatchery. We held all embryos in Heath tray incubatorsuntil the eyed stage and treated them with formalin every 3–7d to prevent fungus growth. After the embryos had reached theeyed stage, we removed dead embryos, counted the number ofsurvivors, and transported the surviving embryos to the studysite in 1-L insulated bottles.

At the study site, we introduced embryos at the headwatersof the Cherry Creek basin in 2007 and then moved introductionsites down the basin in subsequent years (Figure 1). In 2007,we introduced embryos to the headwaters of Cherry Creek andCherry Lake Creek. In 2008, we introduced embryos to twotributaries: an unnamed tributary of main-stem Cherry Creekand a tributary of Cherry Lake Creek called Pika Creek. In2009, we again introduced embryos to two tributaries: CarpenterCreek and South Fork Cherry Creek.

The number of embryos introduced from each populationvaried depending on their availability (Table 1). Because em-bryos from Sun Ranch Hatchery were not available in 2009,only four populations were introduced that year. In 2007 and2008, embryos from all source populations were introducedto all sites (Table 1). Embryos from a single male × femalecross were introduced to the same incubator, with the exceptionof male × female crosses from Washoe Park Hatchery, whichwere sometimes divided among several incubators.

We used instream remote-site incubators (RSIs) to plant eyedembryos at introduction sites. The RSIs are designed to con-sistently supply embryos with freshwater, while avoiding sedi-mentation problems associated with buried incubators, and havebeen used previously to successfully introduce other species ofsalmonid embryos (e.g., Donaghy and Verspoor 2000; Kaedingand Boltz 2004; Al-Chokhachy et al. 2009). Hatched fry ab-sorbed the yolk sac while in the RSI. After swim-up, fry exitedthe incubators via an outflow tube, through which the waterflowed into a 19-L (5 gal) bucket. After putting embryos in anRSI, we checked them every 2–3 d until the last fry was in thebucket. When we checked RSIs, we made sure they were stillsupplying embryos with freshwater and we counted and releasedany fry in the receiving bucket.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013

798 ANDREWS ET AL.

FIGURE 1. Upper Cherry Creek study area showing remote-site incubator (RSI) sites and Westslope Cutthroat Trout sampling sections (Sampling) by year. Thesymbols representing incubator sites are offset from the stream so they do not obscure symbols designating sampling sites. Several fish sampling sections weresampled in both 2008 and 2009 and they are shown as overlapping triangles that look like hourglasses. The inset map shows the entire Cherry Creek watershedincluding the 8-m-high waterfall that serves as a barrier to upstream movement, where Cherry Creek enters the Madison River, and the Cherry Creek’s locationwithin Montana. In this paper, we do not discuss fish introduced to Carpenter Creek because no movement data were available for this introduction site. CC =Cherry Creek, Pika = Pika Creek, CLC = Cherry Lake Creek, Trib = unnamed tributary, SF = South Fork, Carp = Carpenter Creek.

In all but one introduction location, we released fry in calmwater immediately downstream from the RSIs. However, fryfrom the South Fork RSIs were released in two different loca-tions: (1) just downstream from the RSIs in South Fork and (2)immediately upstream from the mouth of South Fork in main-

stem Cherry Creek, 400 m from the RSIs. We released fish fromall source populations at both of these release sites so that com-parisons of movement among source populations would not beaffected by the fact that we used two release locations for theSouth Fork introduction site.

TABLE 1. Number of Westslope Cutthroat Trout embryos introduced to each introduction site by source population.

Wild-source Hatchery source Cherry Cherry Pika Unnamed Southpopulations populations Creek Lake Creek Creek tributary Fork Total

Ray Creek 1,919 1,548 890 810 889 6,056Muskrat Creek 2,790 2,655 1,583 1,621 1,891 10,540White’s Creek 351 664 409 565 322 2,311

Sun Ranch 1,553 1,522 1,712 1,565 6,352Washoe Park 498 513 1,251 1,394 922 4,578

Total 7,111 6,902 5,845 5,955 4,024 29,837

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013


Fish sampling and identification.—We sampled juvenilefish during August and September in 2008, 2009, and 2010using backpack electrofishing crews consisting of two- to four-persons. We captured fish using Smith-Root BP-15, BP-12, andSR-24 backpack electrofishers operated at voltages in the rangeof 100–600 V, frequencies under 50 Hz, and pulse widths lessthan 2 µs to maximize the number of fish captured while min-imizing injury to fish caused by the shock (Dwyer et al. 2001).We caught immobilized fish in dip nets and held them in bucketsfilled with stream water until they were completely recovered.We then anesthetized the fish, weighed them, measured length,and removed a small portion of the pelvic fin for geneticanalysis.

We used a systematic sampling design with a nonrandomstart to capture juvenile fish (Figure 1). We began samplingwhere RSIs had been located, rather than using a nonrandomstart, because we expected the juvenile fish to be concentratednear introduction sites. We sampled 100-m sections every 300 mfor about 600 m above and from 1 to 5 km below introductionsites (Figure 1). When fish became noticeably less abundant, wedecreased our sampling frequency to sample one 100-m sectionper 500 m of stream. In some cases, such as in Cherry LakeCreek, sampled sections were more distant because we avoidedstream sections where sampling was prohibitively difficult. InCherry Lake Creek and the unnamed tributary, fish densitiesdecreased substantially when we sampled more than 500 mfrom the RSI locations, so it was unnecessary to sample as fardownstream. We continued downstream from an RSI locationuntil few or no fish were found in consecutive sections. Oursampling protocol prioritized detecting downstream movement,so our observations may not have represented the extent ofupstream movement. Because we did not sample continuouslythroughout the range of the translocated population but insteadused a systematic sampling design, we did not sample all of theindividuals that survived.

We concentrated our sampling effort to capture age-1 fish.In 2008, 2009, and 2010, we sampled the two introduction sitesused the previous year. In 2009, we were also able to sample thetwo introduction sites used in 2007. Therefore, we were able toprovide 1 year of data on age-2 individuals. In the process ofsampling for age-1 fish in 2010, we captured some age-2 andage-3 fish. We excluded these fish from further analysis becausewe did not sample throughout their population range and there-fore cannot provide a valid summary of their movement. Wedetermined the age of a captured fish by determining its par-ents; this approach worked because we knew which parent pairscontributed each year. We introduced embryos in early summerand sampled juveniles in late summer and early fall, so cap-tured individuals were approximately 14 months old (hereafterreferred to as age-1 fish) or 26 months old (hereafter referred toas age-2 fish).

We used 12 microsatellite loci to genotype each captured in-dividual and each adult that donated gametes using the labora-tory protocols of Vu and Kalinowski (2009, see full list of loci in

their Table 1). We assigned offspring to parent pairs by countingMendelian exclusions (e.g., Muhlfeld et al. 2009). We accepteda parentage assignment if an offspring had two or fewer locimismatched with only one parent pair. We excluded from fur-ther analysis any offspring that could not be matched to at leastone parent pair with two or fewer mismatches. Consequently 77offspring (5.3%) were excluded from further analysis. Another92 offspring (6.3%) were excluded because they were assignedto two or more parent pairs.

Determining movement distance.—Hereafter, we use “move-ment distance” to refer to the distance between the locationwhere a fish was released after fry emergence and the locationwhere the fish was captured at age 1 or age 2. We used handheldGPS devices to record the location of each sampling section andlocation where we released fry. We recorded the locations of acaptured individual as the lower bound of the 100-m samplingsection in which it was captured. Therefore, our calculations ofmovement distance could overestimate downstream movementand underestimate upstream movement by up to 100 m. Todetermine movement distance, we computed stream distancesbetween fry release sites, sampling sites, and a reference pointusing the national hydrography dataset plus hydrography layer(http://nhd.usgs.gov) and Network Analyst within ArcGIS ver-sion 9.3.1 (http://www.esri.com). We used negative numbers todesignate downstream movement and positive numbers to des-ignate upstream movement. Movement data were not availablefor individuals introduced to Carpenter Creek, so this creek wasexcluded from this study.

Statistical analysis.—We tested for associations betweenmovement distance and three explanatory variables—age,source population, and introduction site—and examined inter-actions between these variables. We began by fitting a saturatedANOVA model with movement distance as the response vari-able. The three-way interaction among age, source population,and introduction site was not significantly associated with move-ment distance (F4, 1073 = 1.24, P = 0.29), so we excluded it.We fit a final ANOVA model that included all of the two-wayinteractions and then completed post hoc analyses by examininginteraction plots with mean movement distance and associated95% confidence intervals. We used QQ-plots and plots of fittedversus residual values to check that the assumptions of normal-ity and homogeneity of variance were met for our final ANOVAmodel. Because the movement data were skewed and the samplesizes differed substantially among groups in some cases, theseassumptions of normality and homogeneity of variance weremildly violated. To ensure these violations did not affect theconclusions we drew, we used Kruskal–Wallis nonparametricANOVA to confirm all relationships between movement dis-tance and the three explanatory variables. In all cases, the resultsfrom the parametric and nonparametric analyses were equiva-lent, so we report only the more familiar parametric analyses inthis paper. We used an alpha level of 0.05 for all statistical tests.

We released fry in the South Fork in two different locations,whereas at the other introduction sites we released fry in only

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013

800 ANDREWS ET AL.

one location. Therefore, before comparing the movement dis-tances among source populations and among introduction sitesas described above, we looked more closely at movement dis-tance in South Fork to see whether the location of fry releaseaffected how far individuals moved. Mean movement distancesof the individuals released at the two South Fork release sites (inSouth Fork and in Cherry Creek) were equivalent (Welch’s two-sample t = 0.73, P = 0.47). Therefore, we pooled the movementdistances of the two release sites in South Fork. We could notalways determine the release location of captured individualsfrom Washoe Park Hatchery that were introduced to South Forkbecause fish from male × female crosses from Washoe ParkHatchery were sometimes split between more than one RSI.Therefore, we omitted 38 captured individuals from the WashoePark Hatchery population that were introduced to South Forkfrom all analyses.

We also omitted one individual from Ray Creek that wascaptured in the unnamed tributary. It was the only individualfrom Ray Creek captured in the unnamed tributary, which meantthat the ANOVA model could not reasonably compare the RayCreek population to other source populations in the unnamedtributary.

RESULTSWe captured and determined the population of origin of 836

age-1 Westslope Cutthroat Trout from five introduction loca-tions and 269 age-2 individuals from two introduction locations.At age 1, most individuals (90.9%) remained within 1,000 m oftheir introduction sites (Table 2), but a few individuals were cap-tured farther than 4,000 m downstream from their introductionsite (Figure 2). By age 2, the majority of individuals (58.3%)still remained within 1,000 m of their introduction site, but con-siderably more individuals had moved downstream, some morethan 6,000 m from their introduction site (Table 2; Figure 3).

We examined associations between movement distanceand three explanatory variables: age, source population, andintroduction site. Two of the three two-way interactionsamong these variables were significant. Therefore, results arepresented in the context of the interactions among variables.The pattern of differences in movement distance among sourcepopulations varied between age-1 and age-2 fish (F4, 1077 =15.45, P < 0.0001; Figure 4A), as did the pattern of differencesin movement distance among introduction sites (F41, 1077 =

11.39, P < 0.0008; Figure 4B). The pattern of differences inmovement distance among source populations did not varyamong introduction sites (F13, 1077 = 1.44, P = 0.13).

Though there were statistically significant interactions be-tween age and introduction site and between age and sourcepopulation, some patterns of movement distance were consis-tent across ages. Among source populations, individuals fromWashoe Park Hatchery and White’s Creek remained much closerto the RSI locations (represented by 0 on the y-axis in Figure 4A)than individuals from other populations at both age 1 and age2 (Figure 4A). For example, the mean movement distance ofage-1 individuals from Washoe Park Hatchery was 18 m (SD =221) downstream, while the mean movement distance of age-1individuals from Ray Creek was 778 m (SD = 873) downstream(Figure 4A). As another example, in most populations individ-uals moved farther downstream at age 2, but individuals fromWhite’s Creek were actually closer to the RSI locations at age2 than at age 1 (Figure 4A).

Among introduction sites, individuals moved farthest down-stream in Cherry Creek and moved the least in Cherry LakeCreek (Figure 4B) at both age 1 and age 2. At age 1, the meanmovement distance in Cherry Creek was 862 m (SD = 866)downstream, while the mean movement distance in Cherry LakeCreek was 4 m (SD = 249) upstream. By age 2, the mean move-ment distances were 2,161 m (SD = 1,851) downstream inCherry Creek and 835 m (SD = 1,603) downstream in CherryLake Creek.

DISCUSSIONIn this study, we observed that the movement exhibited by

translocated Westslope Cutthroat Trout varied by source pop-ulation, introduction site, and age. Overall, age-1 individualsexhibited relatively restricted movement, but there were differ-ences among introduction sites and among source populations.Among introduction sites, downstream movement farther than1,000 m from introduction sites was uncommon at all sites ex-cept one (Cherry Creek). Among source populations, individualsfrom two populations (White’s Creek and Washoe Park Hatch-ery) remained closer, on average, to their introduction sites thanindividuals from other populations. By age 2, movement wasmore extensive; more juveniles ventured farther than 1,000 mfrom their introduction site. Differences in movement betweenintroduction sites and source populations also persisted at age 2.

TABLE 2. Summary of Westslope Cutthroat Trout movement, by age. For calculations of mean distance moved, negative values were used to representdownstream movement and positive values were used to represent upstream movement.

Percent moved Percent within Mean distance Farthest distance Farthest distanceAge-group downstream 1,000 m of introduction (m) (SD) upstream (m) downstream (m)

Age 1 82.1 90.9 −402 (658) 510 4,207Age 2 84.8 58.3 −1,530 (1,857) 345 6,185Both ages 82.7 82.9 −676 (1,183) 510 6,185

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013


FIGURE 2. Sampling design and Westslope Cutthroat Trout captured at age 1 (n = 836) by introduction site and source population. Note the substantial differencein fish densities near the RSI locations (represented by 0) versus those more than 1 km downstream. Each line represents a section sampled in which no fish fromthat source population–introduction site combination were captured. Each dot represents a captured fish. Dots have been jittered horizontally and vertically toavoid overlap, but overlap still occurs in the particularly densely populated areas near the RSI sites. Distance moved equals distance between hatching location andlocation of capture at age 1. Negative values for movement distance represent downstream movement; positive values represent upstream movement. Each panelrepresents one introduction site, except South Fork and South Fork in Cherry Creek. At the South Fork introduction site, hatched fry were released in one of twolocations: at the introduction site in South Fork and in Cherry Creek immediately upstream from the mouth of South Fork. Individuals from Sun Ranch Hatcherywere not introduced to South Fork and so do not appear in these two panels of the graph. SRH = Sun Ranch Hatchery, WPH = Washoe Park Hatchery.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013

802 ANDREWS ET AL.

FIGURE 3. Sampling design and Westslope Cutthroat Trout captured at age 2 (n = 269) by introduction site and source population. Each line rep-resents a section sampled in which no fish from that source population–introduction site combination were captured. Each dot represents a capturedfish. Dots have been jittered horizontally and vertically to avoid overlap but overlap still occurs in the particularly densely populated areas near theRSI sites. Distance moved equals distance between hatching location and location of capture at age 2. Negative values for movement distance repre-sent downstream movement; positive values represent upstream movement. Each panel represents one introduction site. SRH = Sun Ranch Hatchery,WPH = Washoe Park Hatchery.

Movement after translocation could affect the success of cur-rent and future projects so it is useful to consider what may haveled to the differences in movement we observed in this study.We will do so in the context of previous research on juveniletrout movement. This study was not designed to determine whatcaused the observed differences, so the potential explanationsdescribed below should be considered as hypotheses that needto be tested in future research on juvenile trout movement aftertranslocation.

Variation in movement distance among introduction sitescould have been caused by density-dependent competition.Density-dependent survival has been observed for juveniles

of many salmonid species, and downstream displacement ofless-dominant individuals appears to be a common responseto density-dependent competition (e.g., Chapman 1966; Crisp1993; Bujold et al. 2004; Westley et al. 2008). We observedthat habitats nearest the introduction sites held the highestabundances of age-1 fish and that fish abundances declinedprecipitously as we moved away from the introduction sites(Figure 2), which suggests that density-dependent factors haveregulated how individuals moved from introduction sites. Fishmoved farthest downstream from the Cherry Creek introductionsite, where the greatest number of embryos were released(Table 1), further supporting this hypothesis. Differences in

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013


FIGURE 4. Mean movement distances of Westslope Cutthroat Trout and asso-ciated 95% confidence intervals by (A) source population and age at capture and(B) introduction site and age at capture. Zero on the y-axis represents the RSIlocation. Negative values for movement distance represent downstream move-ment; positive values represent upstream movement. Symbol shape in panel Brepresents the year of introduction: circles represent data for sites where RSIswere deployed in 2007, triangles represent data for sites in 2008, and the squarerepresents data for the site in 2009. Data for age-2 fish are only available forsites where RSIs were deployed in 2007.

movement among introduction sites also persisted at age 2 (Fig-ure 4), which is consistent with a pattern of density-dependentmovement because as fish grow they require more space (e.g.,Chapman 1966; Chapman and Bjornn 1969). Future researchshould test the hypothesis that movement after translocation isaffected by density-dependent competition.

Habitat conditions at introduction sites may also have con-tributed to the differences in movement we observed amongsites. Fish may be moving to find more ideal habitats, eitherduring the summer (e.g., Kahler et al. 2001; Gowan and Fausch2002) or during fall and early winter to find better overwinterhabitats (e.g., Bjornn 1971; Cunjak and Power 1986). We didnot systematically collect habitat data in this study, but futureresearch should test the hypothesis that fish move to find suitablehabitat after translocation.

Variation in movement distance among source populationscould be caused by heritable variation in the tendency to move.Early life movement patterns exhibited by individuals nativeto lake inlet versus lake outlet streams and from populationsabove versus below waterfalls are heritable (e.g., Northcote1962; Bowler 1975; Kaya 1989; Van Offelen et al. 1993) andselection for sedentary habits can occur very rapidly (Pearseet al. 2009).

All wild source populations used for this study occupied rel-atively short headwater reaches, but the three source streamswere different. It is plausible that selection against downstreammovement in White’s Creek contributed to the differences weobserved between individuals descended from the wild popu-lations. White’s Creek (whose progeny moved only short dis-tances) has infrequent and intermittent surface flows in its lowerreaches due to large valley-bottom alluvium, irrigation with-drawals, and mining impacts (L. Nelson, personal communica-tion). Muskrat and Ray creeks (whose progeny moved fartherdownstream than progeny from White’s Creek) both haveperennial flows that connect them to downstream habitats, andisolation of these populations probably occurred much later thanfor the White’s Creek population (L. Nelson, personal communi-cation). Given that many of the wild Cutthroat Trout populationsthat are potential source populations for translocations are se-questered above barriers to upstream movement (Shepard et al.2005) and adaptation to barriers can occur rapidly (Pearse et al.2009), future work should test the hypothesis that sequesteredpopulations of Westslope Cutthroat Trout evolve to exhibit morerestricted movement.

Differences in movement among source populations couldalso result from the fact that some individuals were descendedfrom wild populations and others were descended from popula-tions raised in captivity. Salmonids from hatchery populationscan move differently than wild fish (e.g., Bjornn and Mallet1964; Richards and Cernera 1989; Bettinger and Bettoli 2002;Baird et al. 2006). Young hatchery-origin salmonids canalso grow more quickly and behave more aggressively than

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013

804 ANDREWS ET AL.

individuals from wild populations (Rhodes and Quinn 1999;Tatara and Berejikian 2012). This could lead to hatchery-originfish outcompeting wild fish, which could lead to wild fishmoving farther than hatchery fish (Nakano 1995; Hughes 2000;Hansen and Closs 2009). In a companion study in the CherryCreek system, we observed that individuals descended from thehatchery populations (Washoe Park and Sun Ranch Hatchery)indeed grew more quickly than individuals descended from thewild populations (Andrews 2012). However, the outcomes ofthe Cherry Creek project do not support a simple relationshipbetween origin (hatchery versus wild) and movement becauseindividuals from the two hatchery populations moved verydifferently (Figure 4) despite growth rate similarities. Usinghatchery populations for translocations is not only convenient,it also avoids affecting existing wild populations. Thus, futurestudies should test the hypotheses that individuals from hatch-ery populations outgrow and outcompete individuals from wildpopulations, and that this leads to differences in movement.

Considering how juvenile trout can move after introductionwill better prepare managers to carry out and evaluate the suc-cess of translocation projects. For example, if the Cherry Creeksystem had less than 4 km of available habitat above a protectivebarrier, the project would have lost a number of individuals todownstream movement over the barrier as early as age 2. In thatcase, maintaining a genetically diverse reproducing populationmight have required recurring translocations. As anotherexample, if the project had used only individuals from the stateof Montana’s Westslope Cutthroat Trout hatchery population(Washoe Park Hatchery), we might have observed limited down-stream movement. In that case, it would have taken more timeor more introductions throughout the system to meet the conser-vation goal of establishing a reproducing population throughoutthe Cherry Creek drainage. Finally, if we had measured thesuccess of the project by sampling only age-1 individuals,we would have concluded that individuals remained close tointroduction sites, creating discrete populations with limitedranges and limiting the potential for mating between individualsintroduced to different sites. In reality, individuals exhibitedmuch more extensive movement by age 2, mitigating theseconcerns.

As is true for all management projects, the outcomes weobserved in the Cherry Creek project were affected by ourunique project design. It would be imprudent to infer that thesame patterns will be observed in translocations with differentdesigns. For instance, we introduced embryos to a fishlesssystem, which meant that age-1 individuals rarely (if ever) hadto compete with adult fish for resources. Introducing embryosto an occupied habitat could result in more extensive movementif native adults outcompete the smaller introduced juvenilesfor resources (Nakano 1995; Hughes 2000; Hansen and Closs2009). As studies of the outcomes of trout translocationsaccumulate, fisheries managers will be better able to predict thefactors that will affect the success of their unique conservationproject.

ACKNOWLEDGMENTSMajor funding for this work was provided by the National

Science Foundation (DEB 0717456). Additional funding wasprovided by Turner Enterprises, Inc. and Montana Trout Un-limited. We thank Lee Nelson, Pat Clancey, Dan Drinan, TravisLohrenz, Romie Bahram, Jake Ferguson, Jacqueline Jones, Jen-nifer Ard, Alex Hopkins, Clint Smith, Tatiana Butler, Ninh Vu,Wes Orr, Buddy Drake, Angela Smith, Mark Sweeney, ReidKoskiniemi, Mike Konsmo, Hillary Billman, and Preston De-bele for their assistance in the field. We also thank Ninh Vu,Jenn Ard, and Tatiana Butler for their assistance in the lab-oratory. Constructive feedback from internal and external re-viewers greatly improved the quality of this manuscript. TheMontana Cooperative Fishery Research Unit is jointly spon-sored by the U.S. Geological Survey, Montana Fish, Wildlifeand Parks, Montana State University, and the U.S. Fish andWildlife Service. The use of trade names or products does notconstitute endorsement by the U.S. Government. This studywas performed under the auspices of Montana State Universityinstitutional animal care and use protocol 18-07.

REFERENCESAl-Chokhachy, R., M. Peacock, L. G. Heki, and G. Thiede. 2009. Evaluating

the reintroduction potential of Lahontan Cutthroat Trout in Fallen Leaf Lake,California. North American Journal of Fisheries Management 29:1296–1313.

Alves, J., D. Krieger, and T. Nesler. 2004. Conservation plan for Rio GrandeCutthroat Trout (Oncorhynchus clarki virginalis) in Colorado. Colorado Di-vision of Wildlife, Aquatic Wildlife Section, Denver.

Andrews, T. M. 2012. Natural selection in the field and the classroom. Doctoraldissertation. Montana State University, Bozeman.

Baird, O. E., C. C. Krueger, and D. C. Josephson. 2006. Growth, movement,and catch of Brook, Rainbow, and Brown trout after stocking into a large,marginally suitable Adirondack river. North American Journal of FisheriesManagement 26:180–189.

Bettinger, J. M., and P. W. Bettoli. 2002. Fate, dispersal, and persistence ofrecently stocked and resident Rainbow Trout in a Tennessee tailwater. NorthAmerican Journal of Fisheries Management 22:425–432.

Bjornn, T. C. 1971. Trout and salmon movements in two Idaho streams as relatedto temperature, food, stream flow, cover, and population density. Transactionsof the American Fisheries Society 100:423–438.

Bjornn, T. C., and J. Mallet. 1964. Movements of planted and wild trout in anIdaho river system. Transactions of the American Fisheries Society 93:70–76.

Bouzat, J. L., J. A. Johnson, J. E. Toepfer, S. A. Simpson, T. L. Esker, and R.L. Westemeier. 2009. Beyond the beneficial effects of translocations as aneffective tool for the genetic restoration of isolated populations. ConservationGenetics 10:191–201.

Bowler, B. 1975. Factors influencing genetic control in lakeward migrations ofCutthroat Trout fry. Transactions of the American Fisheries Society 104:474–482.

Bramblett, R. G. 1998. Environmental assessment for the Madison Riverdrainage Westslope Cutthroat Trout conservation and restoration program:Cherry Creek native fish introduction. Montana Fish, Wildlife, and Parks,Bozeman.

Bujold, V., R. A. Cunjak, J. P. Dietrich, and D. A. Courtemanche. 2004. Driftersversus residents: assessing size and age differences in Atlantic Salmon (Salmosalar) fry. Canadian Journal of Fisheries and Aquatic Sciences 61:273–282.

Chapman, D. W. 1966. Food and space as regulators of salmonid populations instreams. American Naturalist 100:345–357.

Chapman, D. W., and T. C. Bjornn. 1969. Distribution of salmonids in streams,with special reference to food and feeding. Pages 153–176 in T. G. Northcote,

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013


editor. Symposium on salmon and trout in streams. Institute of Fisheries,University of British Columbia, Vancouver.

Clapp, D. F., R. D. Clark Jr., and J. S. Diana. 1990. Range, activity, and habitatof large, free-ranging Brown Trout in a Michigan stream. Transactions of theAmerican Fisheries Society 119:1022–1034.

CRCT (Colorado River Cutthroat Trout) Coordination Team. 2006. Conser-vation strategy for Colorado River Cutthroat Trout (Oncorhynchus clarkiipleuriticus) in the states of Colorado, Utah, and Wyoming. Colorado Divi-sion of Wildlife, Fort Collins.

Crisp, D. T. 1993. Population densities of juvenile trout (Salmo trutta) in fiveupland streams and their effects upon growth, survival and dispersal. Journalof Applied Ecology 30:759–771.

Cunjak, R. A., and G. Power. 1986. Winter habitat utilization by stream residentBrook Trout (Salvelinus fontinalis) and Brown Trout (Salmo trutta). CanadianJournal of Fisheries and Aquatic Sciences 43:1970–1981.

Donaghy, M. J., and E. Verspoor. 2000. A new design of instream incubator forplanting out and monitoring Atlantic Salmon eggs. North American Journalof Fisheries Management 20:521–527.

Dwyer, W. P., B. B. Shepard, and R. G. White. 2001. Effect of backpack elec-troshock on Westslope Cutthroat Trout injury and growth 110 and 250 daysposttreatment. North American Journal of Fisheries Management 21:646–650.

Fausch, K. D., B. E. Rieman, J. B. Dunham, M. K. Young, and D. P. Peterson.2009. Invasion versus isolation: trade-offs in managing native salmonids withbarriers to upstream movement. Conservation Biology 23:859–870.

GCTRT (Greenback Cutthroat Trout Recovery Team). 1998. Greenback Cut-throat Trout recovery plan. U.S. Fish and Wildlife Service, Region 6, Denver.

Gowan, C., and K. D. Fausch. 1996. Mobile Brook Trout in two high-elevationColorado streams: re-evaluating the concept of restricted movement. Cana-dian Journal of Fisheries and Aquatic Sciences 53:1370–1381.

Gowan, C., and K. D. Fausch. 2002. Why do foraging stream salmonids moveduring summer? Environmental Biology of Fishes 64:139–153.

Gresswell, R. E. 2011. Biology, status, and management of the YellowstoneCutthroat Trout. North American Journal of Fisheries Management 31:782–812.

Gresswell, R. E., and S. R. Hendricks. 2007. Population-scale movement ofCoastal Cutthroat Trout in a naturally isolated stream network. Transactionsof the American Fisheries Society 136:238–253.

Griffith, B., J. M. Scott, J. W. Carpenter, and C. Reed. 1989. Translocation as aspecies conservation tool: status and strategy. Science 245:477–480.

Hansen, E. A., and G. P. Closs. 2009. Long-term growth and movement inrelation to food supply and social status in stream fish. Behavioral Ecology20:616–623.

Harig, A. L., and K. D. Fausch. 2002. Minimum habitat requirements for es-tablishing translocated Cutthroat Trout populations. Ecological Applications12:535–551.

Harig, A. L., K. D. Fausch, and M. K. Young. 2000. Factors influencing successof Greenback Cutthroat Trout translocations. North American Journal ofFisheries Management 20:994–1004.

Hilderbrand, R. H., and J. L. Kershner. 2000. Movement patterns of stream-resident Cutthroat Trout in Beaver Creek, Idaho–Utah. Transactions of theAmerican Fisheries Society 129:1160–1170.

Hughes, N. F. 2000. Testing the ability of habitat selection theory to predict inter-annual movement patterns of a drift-feeding salmonid. Ecology of FreshwaterFish 9:4–8.

Jones, H. P. 2010. Seabird islands take mere decades to recover following rateradication. Ecological Applications 20:2075–2080.

Kaeding, L. R., and G. D. Boltz. 2004. Use of remote-site incubators to produceArctic Grayling fry of wild parentage. North American Journal of FisheriesManagement 24:1031–1037.

Kahler, T. H., P. Roni, and T. P. Quinn. 2001. Summer movement and growthof juvenile anadromous salmonids in small western Washington streams.Canadian Journal of Fisheries and Aquatic Sciences 58:1947–1956.

Kaya, C. M. 1989. Rheotaxis of young Arctic Grayling from populations thatspawn in inlet or outlet streams of a lake. Transactions of the AmericanFisheries Society 118:474–481.

Madsen, T., R. Shine, M. Olsson, and H. Wittzell. 1999. Conservation biology:restoration of an inbred adder population. Nature 402:34–35.

MDFWP (Montana Department of Fish, Wildlife and Parks). 2007. Memoran-dum of understanding and conservation agreement for Westslope CutthroatTrout and Yellowstone Cutthroat Trout in Montana. MDFWP, Helena.

Muhlfeld, C. C., S. T. Kalinowski, T. E. McMahon, M. L. Taper, S. Painter, R.F. Leary, and F. W. Allendorf. 2009. Hybridization rapidly reduces fitness ofa native trout in the wild. Biology Letters 5:328–331.

Nakano, S. 1995. Individual differences in resource use, growth and emigrationunder the influence of a dominance hierarchy in fluvial Red-Spotted Masusalmon in a natural habitat. Journal of Animal Ecology 64:75–84.

Northcote, T. G. 1962. Migratory behaviour of juvenile Rainbow Trout, Salmogairdneri, in outlet and inlet streams of Loon Lake, British Columbia. Journalof the Fisheries Research Board of Canada 19:201–270.

Pearse, D. E., S. A. Hayes, M. H. Bond, C. V. Hanson, E. C. Anderson, R. B.Macfarlane, and J. C. Garza. 2009. Over the falls? rapid evolution of ecotypicdifferentiation in steelhead/Rainbow Trout (Oncorhynchus mykiss). Journalof Heredity 100:515–525.

Peterson, D. P., B. E. Rieman, J. B. Dunham, K. D. Fausch, and M. K. Young.2008. Analysis of trade-offs between threats of invasion by nonnative BrookTrout (Salvelinus fontinalis) and intentional isolation for native WestslopeCutthroat Trout (Oncorhynchus clarkii lewisi). Canadian Journal of Fisheriesand Aquatic Sciences 65:557–573.

Pimm, S. L., L. Dollar, and O. L. Bass Jr. 2006. The genetic rescue of the Floridapanther. Animal Conservation 9:115–122.

Rhodes, J. S., and T. P. Quinn. 1999. Comparative performance of geneticallysimilar hatchery and naturally reared juvenile Coho Salmon in streams. NorthAmerican Journal of Fisheries Management 19:670–677.

Richards, C., and P. J. Cernera. 1989. Dispersal and abundance of hatchery-reared and naturally spawned juvenile Chinook Salmon in an Idaho stream.North American Journal of Fisheries Management 9:345–351.

Rodrıguez, M. A. 2002. Restricted movement in stream fish: the paradigm isincomplete, not lost. Ecology 83:1–13.

Schmetterling, D. A., and S. B. Adams. 2004. Summer movements withinthe fish community of a small montane stream. North American Journal ofFisheries Management 24:1163–1172.

Shepard, B. B., B. E. May, and W. Urie. 2005. Status and conservation ofWestslope Cutthroat Trout within the western United States. North AmericanJournal of Fisheries Management 25:1426–1440.

Shepard, B. B., R. Spoon, and L. Nelson. 2002. A native Westslope CutthroatTrout population responds positively after Brook Trout removal and habitatrestoration. Intermountain Journal of Sciences 8:193–214.

Tatara, C. P., and B. A. Berejikian. 2012. Mechanisms influencing competitionbetween hatchery and wild juvenile anadromous Pacific salmonids in freshwater and their relative competitive abilities. Environmental Biology of Fishes94:7–19.

Teuscher, D., and J. Capurso. 2007. Management plan for conservation ofBonneville Cutthroat Trout in Idaho. Idaho Department of Fish and Game,Report 07-48, Boise.

Van Offelen, H. K., C. C. Krueger, and C. L. Schofield. 1993. Survival, growth,movement, and distribution of two Brook Trout strains stocked into smallAdirondack streams. North American Journal of Fisheries Management13:86–95.

Vu, N. V., and S. T. Kalinowski. 2009. Twelve tetranucleotide microsatelliteloci for Westslope Cutthroat Trout Oncorhynchus clarki lewisi (Salmonidae).Conservation Genetics Resources 1:249–251.

Westley, P. A. H., R. Hilborn, T. P. Quinn, G. T. Ruggerone, and D. E. Schindler.2008. Long-term changes in rearing habitat and downstream movement byjuvenile Sockeye Salmon (Oncorhynchus nerka) in an interconnected Alaskalake system. Ecology of Freshwater Fish 17:443–454.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:44

12

Aug

ust 2

013



Mortality of Palmetto Bass Following Catch-and-ReleaseAnglingMatthew J. Petersen a c & Phillip W. Bettoli ba Tennessee Cooperative Fishery Research Unit and Department of Biology , TennesseeTechnological University , Box 5114, Cookeville , Tennessee , 38505 , USAb U.S. Geological Survey, Tennessee Cooperative Fishery Research Unit , TennesseeTechnological University , Box 5114, Cookeville , Tennessee , 38505 , USAc Indiana Department of Natural Resources , 5596 East State Road 46, Bloomington ,Indiana , 47401 , USAPublished online: 06 Aug 2013.

To cite this article: Matthew J. Petersen & Phillip W. Bettoli (2013) Mortality of Palmetto Bass Following Catch-and-ReleaseAngling, North American Journal of Fisheries Management, 33:4, 806-810, DOI: 10.1080/02755947.2013.812584







http://dx.doi.org/10.1080/02755947.2013.812584




MANAGEMENT BRIEF

Mortality of Palmetto Bass Following Catch-and-ReleaseAngling

Matthew J. Petersen*1

Tennessee Cooperative Fishery Research Unit and Department of Biology,Tennessee Technological University, Box 5114, Cookeville, Tennessee 38505, USA

Phillip W. BettoliU.S. Geological Survey, Tennessee Cooperative Fishery Research Unit,Tennessee Technological University, Box 5114, Cookeville, Tennessee 38505, USA

AbstractPalmetto bass (Striped Bass Morone saxatilis × White Bass M.

chrysops) have been stocked into reservoirs in the southeastern USAsince the late 1960s and have gained widespread acceptance as asport fish. These fisheries are growing in popularity and catch-and-release (CR) fishing is commonplace; however, there is a dearth ofinformation on CR mortality of palmetto bass. We experimentallyangled palmetto bass (n = 56; >373-mm TL) in a Tennessee reser-voir using traditional angling gear in water temperatures rangingfrom 13◦C to 32◦C. Ultrasonic transmitters equipped with floatswere externally attached to fish, which were released immediatelyand tracked multiple times within 10 d of release. Mortality wasnegligible (3.6%) in fall and spring at cool water temperaturesbut was high (39.3%) in summer when water temperatures ex-ceeded 26◦C. The best logistic regression model based on Akaike’sinformation criterion for small sample sizes scores relied on watertemperature alone to predict CR mortality of palmetto bass; therewas little support for other models that included all possible com-binations of the six other predictor variables we tested. Palmettobass in our study experienced lower CR mortality than StripedBass in other systems, but CR mortality rates for palmetto bassthat approach or exceed 40% during summer are still problematicif the goal is to maintain fishing quality.

Palmetto bass (Striped Bass Morone saxatilis × White BassM. chrysops) have been stocked into reservoirs in the southeast-ern USA since 1966 and have gained widespread acceptanceas a sport fish. Bishop (1968) reported that palmetto bass grewfaster, survived better, and were easier to catch than StripedBass. Until recently, Striped Bass were the predominant mo-ronid species stocked into Tennessee reservoirs; however, poor

*Corresponding author: [email protected] address: Indiana Department of Natural Resources, 5596 East State Road 46, Bloomington, Indiana 47401, USA.Received February 20, 2013; accepted June 2, 2013

water quality (warm water temperatures and low dissolved oxy-gen during summer months) in some reservoirs has promptedthe Tennessee Wildlife Resources Agency (TWRA) to reduceStriped Bass stockings. Palmetto bass, more tolerant of poorwater quality, now dominate the moronid fishery in reservoirssuch as Cherokee Lake, a tributary reservoir that experiencesStriped Bass die-offs due to a pronounced temperature and dis-solved oxygen “squeeze” during stratification (Coutant 1985).It has long been recognized that palmetto bass can be stockedin a variety of systems unsuitable for Striped Bass (e.g., Axonand Whitehurst 1985), and that trend is continuing in Tennesseeand elsewhere (Bettoli 2013).

As palmetto bass fisheries expand and become more popular,the frequency of catch-and-release (CR) fishing is likely toincrease. In the four most popular palmetto bass fisheries inTennessee, release rates ranged from 28% to 93% (P. Black,Tennessee Wildlife Resources Agency, unpublished data). Inorder for minimum length regulations or the voluntary efforts ofanglers to succeed in maintaining quality fisheries, most releasedfish need to survive. The literature is replete with studies of CRmortality of many species, including Striped Bass; however, lit-tle information exists on CR mortality of palmetto bass despitetheir growing popularity. Childress (1989) angled Striped Bassand palmetto bass, placed them in live wells for up to 1 h, andthen transferred them to net-pens where they were held for 72 h.Striped Bass in that study experienced higher CR mortality inwinter and summer (11% and 69%, respectively) than palmettobass (1% and 29%, respectively). Although those fish experi-enced the same stressors a fish would experience when hooked

806

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:45

12

Aug

ust 2

013


and landed by a recreational angler, holding fish in live wellsand pens could increase stress and bias estimates of CR mortal-ity (Carmichael et al. 1984; Donaldson et al. 2008). Confiningfish in pens is especially problematic when dealing with pelagicfishes such as Striped Bass and palmetto bass (Skomal 2007).

Biotelemetry is becoming increasingly popular as a meansto estimate CR mortality. In a review of published literature upthrough August 2005, Arlinghaus et al. (2007) found 242 CRmortality studies dating back to 1957, of which 23% (n = 55)used biotelemetry. In our study, we estimate for the first time theCR mortality rates of free-ranging palmetto bass caught usingnatural baits and artificial lures.

STUDY SITEJ. Percy Priest Lake (hereafter, JPPL) is located on the Stones

River, a tributary of the Cumberland River in middle Tennessee.Constructed in 1969, this 7,630-ha eutrophic reservoir is man-aged for recreation, flood control, and hydroelectric power bythe U.S. Army Corps of Engineers. At full pool, the reservoirhas an average depth of 9 m and an average hydraulic retentiontime of 139 d. When the reservoir is thermally stratified betweenlate spring and late fall, a pronounced temperature and dissolvedoxygen squeeze exists (Coutant 1985). In a typical summer, thehypolimnion below 6 m is anoxic (USACE 2007).

METHODSFish collection and tag implantation.—Angling was con-

ducted over a wide range of water temperatures (13–32◦C) in2011 and 2012. In order to mimic typical palmetto bass fishingtechniques, professional fishing guides or members of a localpalmetto bass club accompanied us on all sampling trips. Arti-ficial baits (umbrella rigs with one to five hooks) were trolled atspeeds ranging from 4.0 to 6.4 km/h using downriggers. Natu-ral baits (Gizzard Shad Dorosoma cepedianum, Threadfin ShadD. petenense, and Bluegill Lepomis macrochirus) were fishedon the bottom, midwater while drifting, or pulled with planerboards. Fight time, net time, and handling time were recordedfor each fish caught, as well as surface water temperature, bleed-ing status (0 = none; 1 = slight; 2 = blood flowing freely), andfishing method (artificial or natural baits). After a fish was reeledin (hook up to landing = fight time), the fish was netted and thehook was removed (landing to hook removal = net time). Iffish were deeply hooked, the line was cut and the hook was leftin place. After the hook was removed or the line was cut, thefish was placed in a 95-L cooler filled with untreated lake waterwhere it was measured for TL and the tag was attached (amountof time in cooler = handling time). Only palmetto bass longerthan 293-mm TL (∼325 g) were tagged to satisfy the “2% rule”(Winter 1983).

Sonotronics IBT-96-2-E ultrasonic tags (33 × 9 mm, 4.8 g,battery life of 3 months) were externally attached to angledfish. A float-and-tag system similar to what Osborne and Bettoli(1995) used to tag Striped Bass was used in this study. Each

tag was attached to an 80-mm-long piece of Last-A-Foam us-ing monofilament. Last-A-Foam is rigid polyether polyurethanefoam with fine closed-cell structure that resists compressionwhen submerged. Chromic gut suture (#3), which was used toattach a barrel swivel to the tag-and-float assembly, decomposedin 7–25 d (depending on water temperature) and allowed the tagto float to the surface where it could be retrieved and reused.The total assembly weighed 6.5 g in air. Using a 3/8 triangularcircle needle (size 6), Vicryl suture material was passed throughthe dorsal musculature anterior to the dorsal fin and a smallloop was tied using a surgeon’s knot; the barrel swivel was thenattached to that loop. Following tag attachment, each fish wasreleased and a GPS waypoint was taken.

A Sonotronics directional hydrophone (DH-4) and Sonotron-ics receivers (USR-5W and USR-96) were used to locate teleme-tered fish and detached tags. Fish were tracked multiple timesduring the first 10 d following release. Tracking events occurred2 out of the first 3 d postrelease to assess short-term mortal-ity and at least two additional days from 4 to 10 d postreleaseto assess delayed mortality. When tagged fish were located werecorded tag number and time, and a waypoint was taken. Allwaypoint locations were imported as shapefiles in ArcMap 10.0using DNR Garmin 5.4.1 (MNDNR 2008). A fish had to belocated at least two times before a fate was assigned. A fish wasconsidered dead if it did not move more than 100 m from itsprevious location or was discovered floating on the surface.

Data analysis.—Logistic regression was used to model theeffects of TL, fight time, net time, handling time, surface watertemperature, bleeding status, and fishing method on palmettobass CR mortality. We used Akaike’s information criterion forsmall sample sizes (AICc) to assess the relative support for lo-gistic regression models containing all possible combinations ofthe seven predictor variables. All analyses were performed usingStatistical Analysis System version 9.1 (SAS Institute 2009).

Tagging effects experiment.—To test for possible taggingeffects, palmetto bass were collected using boat-mounted DCelectrofishing gear (Smith-Root Type VI-A electrofisher) inApril and May 2012 in the Holston River in east Tennessee.Twenty-nine palmetto bass were captured and transported toEagle Bend State Fish Hatchery in Clinton, Tennessee, andplaced in a 0.04-ha hatchery pond. Fish were allowed 3–15 dto acclimate to the pond’s water temperature before initiatingthe experiment. The pond was drained on 30 April 2012, and14 palmetto bass were randomly selected for tagging usingdummy transmitters after first recording their TL. The dummytransmitters weighed 6.5 g, and the attachment method wasidentical to that used on angled fish. Fifteen fish that servedas controls were measured for TL but were not tagged. The29 fish (tagged and control) were then placed into an adjacenthatchery pond. The average water temperature in the hatcherypond over the 7-d holding period was 25.1◦C, and dissolvedoxygen concentrations measured once every other day averaged8.4 mg/L. The mean lengths of the 14 tagged fish (501-mm TL;SE = 30.4) and 15 control fish (505-mm TL; SE = 35.8) were

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:45

12

Aug

ust 2

013

808 PETERSEN AND BETTOLI

similar (t = 0.08, df = 26, P = 0.9380). When the hatchery pondwas drained after 7 d, all fish (tagged and control) were alive;thus, we concluded that our tagging procedure did not confoundour estimate of CR mortality of palmetto bass in JPPL.

RESULTSWe caught, tagged, and released 64 palmetto bass in JPPL.

Eight of the tagged fish were never located following releaseand were discarded from further analysis. The remaining 56fish were tracked for at least 4 d to determine their fate. Theyranged in TL from 379 to 682 mm (mean = 581; SE = 8.9).Fight time ranged from 40 to 228 s (mean = 104; SE = 6.1); nettime ranged from 0 to 147 s (mean 39; SE = 4.1), and handlingtime ranged from 65 to 285 s (mean 118; SE = 5.6).

Only one palmetto bass died immediately upon release. Nineof the remaining 55 palmetto bass died within 48 h of re-lease (i.e., short-term mortality = 16.4%), and two others diedafter 72 h (delayed mortality = 3.6%). During the summermonths when water temperatures exceeded 26◦C, pooled CRmortality was 39.3% (i.e., 11 of 28 fish died). During fall andspring (March, April, May, and October) at cooler temperatures(≤25◦C), CR mortality was only 3.6% (i.e., 1 of 28 fish died).

Of the 56 palmetto bass whose fate was known, 24 werecaught with artificial lures. Twenty-three of the 24 fish caughton artificial lures were hooked in the mouth and one fish washooked in the gills; that fish subsequently died. Pooled mortalityfor palmetto bass caught on artificial lures over all seasons was21%. Pooled mortality was 22% for the 32 fish caught on naturalbaits. Fifty-three percent of the fish caught on natural baitswere hooked in the mouth or jaw, and 47% were hooked inthe esophagus or swallowed the hook. The line was cut on 10of the 32 fish caught on natural bait, and three of those fishsubsequently died.

The top five models according to AICc scores all includedwater temperature as a predictor variable (Table 1), and the bestmodel contained only the water temperature variable. Therewas little support for the four next-best models that includedfight time, bleeding status, method, and net time as predictorvariables. Water temperature was the only significant predictor

TABLE 1. Rankings based on an information theoretic approach of the topfive logistic regression models predicting CR mortality of palmetto bass. Seetext for description of fight time, net time; method refers to natural or artificialbaits. Delta AICc represents the change in AIC between a given model and thetop model (i.e., the one with the lowest AIC score); Akaike weight representsthe relative likelihood of each model.

Delta AkaikeModel AICc AICc weight

Water temperature 52.2 0 0.148Water temperature, fight time 53.6 1.4 0.072Water temperature, bleeding 53.7 1.5 0.071Water temperature, method 54.1 1.9 0.058Water temperature, net time 54.2 2.0 0.054

(χ2 = 7.7, df = 1, P = 0.0055, SE = 0.0964) of hookingmortality in palmetto bass. The probability of mortality as afunction of water temperature was expressed as

Pmortality

= e−8.3325+0.2675 · temperature/(1 + e−8.3325+0.2675 · temperature).

The 90% profile likelihood confidence interval for the wa-ter temperature parameter was 0.1276–0.4497. The Hosmer–Lemeshow goodness-of-fit test indicated that the model pro-vided a good fit to the data (χ2 = 3.2, df = 6, P = 0.7827). Theprobability of CR mortality at water temperatures of 10, 20, and30◦C was 0.3, 4.8, and 42.4%, respectively.

DISCUSSIONThis study provides the first estimate of CR mortality of

free-ranging palmetto bass, and water temperature was the mostimportant factor influencing mortality. Likewise, CR mortal-ity of Striped Bass varies directly with water temperature (e.g.,Harrell 1988; Hysmith et al. 1994). Hooking mortality increaseswith water temperature in most other species as well, includingWalleye Sander vitreus (Reeves and Bruesewitz 2007), Large-mouth Bass Micropterus salmoides (Gustaveson et al. 1991;Meals and Miranda 1994), and Smallmouth Bass M. dolomieu(Cooke and Hogle 2000).

Wilde et al. (2000) compiled results from seven Striped Basshooking mortality studies in a meta-analysis and used logisticregression to model the effects of bait type and water tempera-ture on mortality of 1,275 experimentally angled Striped Bass.The predicted CR mortality of palmetto bass in JPPL was belowthe predicted mortality of Striped Bass caught on either artificialbaits or natural baits, especially at water temperatures between15◦C and 30◦C (Figure 1). The higher mortality associated

Water Temperature (oC)

0 5 10 15 20 25 30 35

Mor

talit

y (%

)

0

10

20

30

40

50

60

70

80

PalmettoBassPooled

SBNB

SBAB

FIGURE 1. Catch-and-release mortality as a function of water temperaturefor Striped Bass caught using artificial baits (SBAB) and natural baits (SBNB),and palmetto bass angled in JPPL using both natural and artificial baits. The twoStriped Bass models are from a meta-analysis reported by Wilde et al. (2000).

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:45

12

Aug

ust 2

013


with natural baits that Wilde et al. (2000) detected for StripedBass is also observed for other species (e.g., Walleye [Payeret al. 1989]; Smallmouth Bass [Weidlein 1987; Clapp andClark 1989]; Bluegill [Siewert and Cave 1990]). In the presentstudy, however, there was little support for a bait effect on CRmortality of palmetto bass, despite the fact that palmetto basscaught on natural baits were more likely to have been hookedin locations other than the jaw or buccal cavity. We cut theline on 10 palmetto bass that swallowed natural baits and sevensurvived. Cutting the line on deeply hooked fish is known toincrease survival rates for other species such as Bluegill (Fobertet al. 2009), Rainbow Trout Oncorhynchus mykiss (Mason andHunt 1967; Schill 1996; Schisler and Bergersen 1996), andWhite Seabass Atractoscion nobilis (Aalber et al. 2004).

Although palmetto bass in JPPL experienced lower hook-ing mortality than Striped Bass in other systems, losing ∼40%of palmetto bass angled and released during summer monthsrepresents cryptic exploitation and reduces the number of fishavailable for anglers to catch. Several approaches are avail-able to reduce cryptic exploitation during the summer months.Bettoli and Osborne (1998) suggested implementing a no-cullregulation for Striped Bass or closing the fishery during summermonths. A no-cull regulation prohibits anglers from catch-and-release fishing; all fish that are caught must be harvested untilthe creel limit (2 fish/d) is reached. Alternatively, Striped Bass insome Tennessee reservoirs are managed with a small size limitin the summer months (when most released fish are expected tosuccumb to CR mortality) and a high size limit in winter whenmost released fish will survive. Some fishing clubs choose toavoid fishing for palmetto bass on JPPL in summer because ofperceived high CR mortality. Their voluntary efforts to conservethe palmetto bass fishery suggest that other palmetto bass an-glers may be receptive to the idea of enacting regulations toreduce cryptic exploitation.

All palmetto bass angled with artificial baits in our studywere caught on trolled umbrella rigs. Although no statisticaldifference in survival was observed between artificial and nat-ural baits, and handling times were similar between the twotypes of baits, additional research is needed to estimate hook-ing mortality when other types of artificial baits are used (e.g.,plugs, spoons). Umbrella rigs are heavy and are trolled at fasterspeeds than single plugs or spoons and shoreline anglers castand retrieve their lures. It is unknown whether palmetto basscaught from shore or caught while trolling more “traditional”lures would experience the same hooking mortality rates we ob-served. It is also unknown whether our estimates of CR moralityare conservative given that our anglers were experienced pal-metto bass anglers and, presumably, better at handling palmettobass than generalist anglers. If more estimates of CR mortal-ity for palmetto bass become available, a meta-analysis similarto that of Wilde et al. (2000) can be performed and the moredifficult, but important, task of addressing the population levelimpacts of CR mortality on palmetto bass fisheries can be un-dertaken (Kerns et al. 2012).

ACKNOWLEDGMENTSFunding and other support for this research was provided

by the Center for the Management, Utilization, and Protec-tion of Water Resources at Tennessee Technological University;TWRA; and the U.S. Geological Survey (USGS) TennesseeCooperative Fishery Research Unit. The Tennessee Coopera-tive Fishery Research Unit is jointly sponsored by the USGS,the U.S. Fish and Wildlife Service, the Tennessee Wildlife Re-sources Agency, and Tennessee Technological University. Thismanuscript benefitted from comments provided on an earlierdraft by Thomas Roberts and David Smith. We thank MikeSmith, John Hammonds, and members of the J. Percy PriestHybrid and Striper Club for their assistance and support in cap-turing and holding palmetto bass. Reference to trade names doesnot imply endorsement by the U.S. Government.

REFERENCESAalber, S. A., G. M. Stutzer, and M. A. Drawbridge. 2004. The effects of catch-

and-release angling on the growth and survival of juvenile White Seabasscaptured on offset circle and J-type hooks. North American Journal of Fish-eries Management 24:793–800.

Arlinghaus, R., S. J. Cooke, J. Lyman, D. Policansky, A. Schwab, C. Suski,S. G. Sutton, and E. B. Thorstad. 2007. Understanding the complexity ofcatch-and-release in recreational fishing: an integrative synthesis of globalknowledge from historical, ethical, social, and biological perspectives. Re-views in Fisheries Science 15:75–167.

Axon, J. R., and D. K. Whitehurst. 1985. Striped Bass management in lakes withemphasis on management problems. Transactions of the American FisheriesSociety 114:8–11.

Bettoli, P. W. 2013. Challenges and opportunities in studying and managingStriped Bass and their hybrids in the 21st century. Pages 575–589 in C.Coutant, J. Bulak, and J. Rice, editors. Biology and management of inlandStriped Bass and hybrid Striped Bass. American Fisheries Society, Sympo-sium 80, Bethesda, Maryland.

Bettoli, P. W., and R. S. Osborne. 1998. Hooking mortality and behavior ofStriped Bass following catch and release angling. North American Journal ofFisheries Management 18:609–615.

Bishop, D. C. 1968. Evaluation of the Striped Bass (Roccus saxatilis) and WhiteBass (R. chrysops) hybrids after two years. Proceedings of the Annual Confer-ence Southeastern Association Game and Fish Commission 21(1967):245–254.

Carmichael, G. J., J. R. Tomasso, B. A. Simco, and K. B. Davis. 1984. Confine-ment and water quality-induced stress in Largemouth Bass. Transactions ofthe American Fisheries Society 113:767–777.

Childress, W. M. 1989. Hooking mortality of Striped Bass, White Bass, andhybrid Striped Bass. Texas Parks and Wildlife Department, Final ReportF-31-R-15, Austin.

Clapp, D. F., and R. D. Clark Jr. 1989. Hooking mortality of Smallmouth Basscaught on live minnows and artificial spinners. North American Journal ofFisheries Management 9:81–85.

Cooke, S. J., and W. J. Hogle. 2000. Effects of retention gear on the injury andshort-term mortality of adult Smallmouth Bass. North American Journal ofFisheries Management 20:1033–1039.

Coutant, C. C. 1985. Striped Bass, temperature, and dissolved oxygen: a specula-tive hypothesis for environmental risk. Transactions of the American FisheriesSociety 114:31–61.

Donaldson, M. R., R. Arlinghaus, K. C. Hanson, and S. J. Cooke. 2008. En-hancing catch-and-release science with biotelemetry. Fish and Fisheries 9:79–105.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:45

12

Aug

ust 2

013

810 PETERSEN AND BETTOLI

Fobert, E., P. Meining, A. Colotelo, C. O’Connor, and S. J. Cooke. 2009. Cutthe line or remove the hook? An evaluation of sublethal and lethal endpointsfor deeply hooked Bluegill. Fisheries Research 99:38–46.

Gustaveson, A. W., R. S. Wydoski, and G. A. Wedemeyer. 1991. Physiologicalresponse of Largemouth Bass to angling stress. Transactions of the AmericanFisheries Society 120:629–636.

Harrell, R. M. 1988. Catch and release mortality of Striped Bass caughtwith artificial lures and baits. Proceedings of the Annual ConferenceSoutheastern Association of Fish and Wildlife Agencies 41(1987):70–75.

Hysmith, B. T., J. H. Moczygemba, and G. R. Wilde. 1994. Hooking mortal-ity of Striped Bass in Lake Texoma, Texas–Oklahoma. Proceedings of theAnnual Conference Southeastern Association of Fish and Wildlife Agencies46(1992):413–420.

Kerns, J. A., M. S. Allen, and J. E. Harris. 2012. Importance of assessingpopulation-level impact of catch-and-release mortality. Fisheries 37:502–503.

Mason, J. W., and R. L. Hunt. 1967. Mortality rates of deeply hooked RainbowTrout. Progressive Fish-Culturist 29:87–91.

MNDNR (Minnesota Department of Natural Resources). 2008. DNR Garmin,version 5.4.1. MNDNR, St. Paul.

Meals, K. O., and L. E. Miranda. 1994. Size-related mortality of tournament-caught Largemouth Bass. North American Journal of Fisheries Management14:460–463.

Osborne, R., and P. W. Bettoli. 1995. A reusable ultrasonic tag and float as-sembly for use with large pelagic fish. North America Journal of FisheriesManagement 15:512–514.

Payer, R. D., R. B. Pierce, and D. L. Pereira. 1989. Hooking mortality ofWalleyes caught on live and artificial baits. North American Journal of Fish-eries Management 9:188–192.

Reeves, K. A., and R. E. Bruesewitz. 2007. Factors influencing the hookingmortality of Walleyes caught by recreational anglers on Mill Lacs, Minnesota.North American Journal of Fisheries Management 27:443–452.

SAS Institute. 2009. SAS/STAT user’s guide, version 9.1. SAS Institute, Cary,North Carolina.

Schill, D. J. 1996. Hooking mortality of bait-caught Rainbow Trout in an Idahostream and a hatchery: implications for special-regulation management. NorthAmerican Journal of Fisheries Management 16:348–356.

Schisler, G. J., and E. P. Bergersen. 1996. Post-release hooking mortality ofRainbow Trout caught on scented artificial baits. North American Journal ofFisheries Management 16:570–578.

Siewert, H. F., and J. B. Cave. 1990. Survival of released Bluegill, Lepomismacrochirus, caught on artificial flies, worms, and spinner lures. Journal ofFreshwater Ecology 5:407–411.

Skomal, G. B. 2007. Evaluating the physiological and physical consequences ofcapture on post-release survivorship in large pelagic fishes. Fisheries Man-agement and Ecology 14:81–89.

USACE (U.S. Army Corps of Engineers). 2007. Environmental assess-ment. Proposed master plan update. USACE, Nashville District, Nashville,Tennessee.

Weidlein, W. D. 1987. Mortality of released sublegal sized Smallmouth Bass,catch and release implications. Pages 217–228 in R. A. Barnhart and T. D.Roelofs, editors. Catch and release fishing: a decade of experience. HumboldtState University, Arcata, California.

Wilde, G. R., M. I. Muoneke, P. W. Bettoli, K. L. Nelson, and B. T. Hysmith.2000. Bait and temperature effects on Striped Bass hooking mortality infreshwater. North American Journal of Fisheries Management 20:810–815.

Winter, J. D. 1983. Underwater biotelemetry. Pages 371–395 in L. A. Nielsenand D. L. Johnson, editors. Fisheries techniques. American Fisheries Society,Bethesda, Maryland.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:45

12

Aug

ust 2

013



Sexual Dimorphism in Alligator GarD. L. McDonald a , J. D. Anderson a , C. Hurley a , B. W. Bumguardner a & C. R. Robertson ba Texas Parks and Wildlife, Coastal Fisheries Division , Perry R. Bass Marine FisheriesResearch Station , 3864 Farm Road 3280, Palacios , Texas , 77465 , USAb Texas Parks and Wildlife, Inland Fisheries Division, River Studies , Post Office Box 1685, SanMarcos , Texas , 78667 , USAPublished online: 06 Aug 2013.

To cite this article: D. L. McDonald , J. D. Anderson , C. Hurley , B. W. Bumguardner & C. R. Robertson (2013)Sexual Dimorphism in Alligator Gar, North American Journal of Fisheries Management, 33:4, 811-816, DOI:10.1080/02755947.2013.812586







http://dx.doi.org/10.1080/02755947.2013.812586




MANAGEMENT BRIEF

Sexual Dimorphism in Alligator Gar

D. L. McDonald,* J. D. Anderson, C. Hurley, and B. W. BumguardnerTexas Parks and Wildlife, Coastal Fisheries Division, Perry R. Bass Marine Fisheries Research Station,3864 Farm Road 3280, Palacios, Texas 77465, USA

C. R. RobertsonTexas Parks and Wildlife, Inland Fisheries Division, River Studies, Post Office Box 1685,San Marcos, Texas 78667, USA

AbstractThe Alligator Gar Atractosteus spatula is currently imperiled

due to habitat alterations and overharvest within much of thisspecies’ range. Recent interest in improving management for thisspecies within the USA and Mexico has spurred new creel restric-tions, spawning area closures, and stocking programs, along withincreased research on life history and population dynamics. Thesemanagement and research measures can be improved by a noninva-sive method for determining sex. Previous methods have requiredsacrificing the fish (for internal anatomy confirmation) or usingcostly and time-consuming assays from a specialized laboratory.Evidence from other gar species suggests that sex determinationis possible by examining sexually dimorphic external characters.We evaluated the utility of 13 morphological measurements fordetermining the sex of Alligator Gars of known gender (n = 117;SL range, 591–1,255 mm). Discriminate analysis identified two in-fluential variables (snout length and caudal peduncle height) assexually dimorphic. Univariate analyses identified three variablesas sexually dimorphic (head length, snout length, and anal fin baselength). Sexually dimorphic variables were used with SL to de-velop a method using serial body ratios (SL/snout length) followedby (snout length/anal fin base length) to identify sex in AlligatorGar to 93% accuracy in males and 72% accuracy in females.

The Alligator Gar Atractosteus spatula historically rangedthrough the Mississippi River and much of the southeasternUSA and northeastern Mexico, inhabiting riverine and estu-arine habitats neighboring the Gulf of Mexico from Florida toVeracruz, Mexico (Suttkus 1963; Page and Burr 1991). Over thepast 50 years, effects of habitat alterations (e.g., impoundmentof rivers) and overharvest have greatly reduced the abundanceof this species (Ferrara 2001; O’Connell et al. 2007). Thisreduction in abundance has resulted in the initiation of bothfederal and state-level management practices, such as the

*Corresponding author: [email protected] April 1, 2013; accepted June 3, 2013

implementation of reduced creel limits and closure of spawningareas in some southeastern states. Federal stocking programsin both the United States and Mexico have also been initi-ated to supplement wild populations (Mendoza et al. 2002;Perschbacher 2011).

Research concerning life history (e.g., sex ratios, spawn-ing), movement, habitat use, and broodstock collections forhatcheries has recently intensified throughout the range of theAlligator Gar (D. J. Daugherty, Texas Parks and Wildlife De-partment [TPWD], personal communication). These researchprojects may benefit from an in-the-field method to identifysexual dimorphism in Alligator Gar. In particular, the use ofmorphometrics to identify sex within a known confidence rangewould allow for the nonlethal estimation of sex ratios in wildpopulations. Such information would improve the understand-ing of population dynamics of Alligator Gar in the wild andwould provide a useful baseline for sex ratios to be targeted infuture stock enhancement operations. Current methods used toidentify sex in Alligator Gar are restricted to a noting the generalsize difference between sexes (females larger than males), sacri-ficing the fish (for gross examination of the gonads: Ferrara andIrwin 2001), or conducting time consuming assays that requirespecialized equipment and reagents (vitellogenin and estradiolconcentrations with the ELISA method: Mendoza et al. 2012).However, recent studies have documented sexual dimorphismusing morphological measurements of the head and fins in theSpotted Gar Lepisosteus oculatus (Love 2002) and LongnoseGar L. osseus (McGrath and Hilton 2012). Here, possible sexualdimorphism in Alligator Gar is evaluated using several head andfin measurements in an effort to identify whether morphologicaldifferences previously described in Lepisosteus also extend tothe genus Atractosteus.

811

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:45

12

Aug

ust 2

013

812 MCDONALD ET AL.

METHODSFish collection.—Live and freshly dead fish were collected

from three bay systems covering the midcoast of Texas duringroutine monitoring by the Texas Parks and Wildlife CoastalFisheries Division (TPWD-CF) from 2008 to 2012. Samplinglocations (Figure 1) were Cedar Lakes (n = 109), San AntonioBay (n = 3), Matagorda Bay (n = 2), and the Lower NuecesRiver before the outflow to Nueces Bay (n = 3). Sampling gearused was a 182.9-m-long, 1.2-m-deep monofilament gill netpartitioned into four 45.7-m-long sections of mesh at sizes of76, 102, 127, and 152 mm stretched mesh. Gill nets were setperpendicular to the shoreline and were set with the smallest-sized mesh closest to shore and increased in mesh size withdistance from the shoreline. Gill nets were placed within baysystems at dusk and retrieved the following morning. Fish weretransported on ice to the TPWD Perry R. Bass Marine FisheriesResearch Station, Palacios, Texas, for processing.

FIGURE 1. The Texas Gulf coast showing the collection sites of AlligatorGars (n = 117) from years 2008–2012. All Alligator Gars were collected by gillnets set overnight as part of resource monitoring for Texas Parks and Wildlife–Coastal Fisheries Division, excluding the Nueces River set nets (collected foran ongoing genetics study).

Morphological measurements.—Morphometrics were se-lected based on ease of measurement in the field by coauthorsand landmarks were identified using a beneficial paper on themorphology of gar species (Wylie 1976). Body measurements(Figure 2) included SL (snout tip to insertion of epichordallobe of caudal fin) and caudal peduncle height (CP; height ofposterior portion of the caudal peduncle at formation of thehypochordal lobe of caudal fin). Fin base measurements in-cluded the dorsal fin base length (DF) and anal fin base length(AF). Head measurements included head length (HL; center ofsnout tip to posterior margin of supratemporal bones), orbitalwidth (OW; skull width between circumorbital bones dorsal tothe orbit at top of skull), snout length (SnL; snout tip to anteriorstart of orbit, measurement made alongside snout), snout widthat center of nostrils (NW), snout width at fleshy intersection oftop and lower jaws (SwJ), opercular plate width (OpW; centerformation of opercle plate to posterior of opercle plate), orbitdiameter (OD; inside diameter of orbit), head height at forma-tion of opercular plate (HOp; top of skull at anterior formationof opercular and subopercular to base of skull), and head heightat center of orbit (HO; top of skull at center or orbit to base ofskull). A fabric measuring tape was used for body, fin base, andHL and SnL measurements to the nearest millimeter; all othermeasurements were made with digital calipers to 0.1 mm.

Statistical analyses.—A two-group descriptive discriminantanalysis (DA) (Proc DISCRIM, SAS, SAS Institute, Cary, NorthCarolina) was used to explore multivariate morphological diver-gence between male and female Alligator Gars. First, outlierswere identified by standardizing data for each variable (i.e.,mean = 0, SD = 1). Data points that resided at least 2.5 ×SD from the mean were considered outliers. For individualswith an outlying value for a single variable, data for that vari-able was coded as missing. For individuals with outliers formultiple variables, data for the entire individual was removedfrom the analysis. Each variable included in the discriminantanalysis model was tested for univariate normality. Four testsof normality were used (Shapiro–Wilk, Kolmogorov–Smirnov,Cramer–von Mise, and Anderson–Darling). Variables were con-sidered normally distributed in the event that P > 0.05 in at leastthree of four normality tests. All variables were normally dis-tributed except for OpW, which was log transformed, and HOp,which was removed from the data set. The presence of multi-collinearity in the data set was tested using Pearson correlationcoefficients. Redundant variables, those that had a significantcorrelation coefficient (i.e., r > 0.5) with any other variable,were removed from the data set. Priority of removal was givento variables that demonstrated significant correlation with multi-ple other variables. Variables included in the model were (1) DF,(2) CP, (3) SnL, (4) NW, (5) SwJ, (6) OD, (7) log-transformedOpW, and (8) HO. Both AF and OW were removed because ofmulticollinearity with numerous other variables. Also, HL wasremoved because it correlated strongly with SnL (r = 0.922,P < 0.0001). Snout length was deemed a more important vari-able than HL so that direct comparisons could be made between

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:45

12

Aug

ust 2

013


FIGURE 2. Photos of an Alligator Gar skull (lateral [upper left panel] and dorsal [upper right panel]) and lateral body (lower panel) displaying the measurementsused in this study.

this study and a previous study that also used SnL as a descrip-tive variable for Spotted Gar (Love 2002). Each variable wasdivided by SL (var/SL) to minimize the bias associated with pre-viously recognized differences in SL between the sexes. Canon-ical scores of each variable were used to derive the quadraticdiscriminant function, which was then used to classify each in-dividual into the male or female group. The squared canonicalcorrelation (Rc

2 = the ratio of between-groups sum-of-squareddeviations on CA axis to total sum-of-squared deviations onCA axis) indicates the strength of the association between themultivariate data cloud and the hypothesis of interest. The dis-criminant model was then validated in two ways, first by aresubstitution procedure of the empirical data (bootstrapping),and then by a cross-validation procedure in which individualswere removed prior to deriving the discriminant function andthen reclassified based upon the function derived using all otherindividuals (jackknifing) (McGarigal et al. 2000).

Univariate tests of sexual dimorphism were conducted byStudent’s t-tests (equal variances = pooled t, unequal variances= Satterthwaite t) on all raw (i.e., untransformed) variables be-tween sexes. A previous study of Longnose Gar demonstratedthat most body and fin measurements in gars are influencedsignificantly by TL (McGrath and Hilton 2012); thus, a secondset of t-tests were run on transformed data to account for SL(var/SL; OpW was log transformed to compel statistical normal-ity). Field-applicable sex identification techniques for AlligatorGar were explored using all sexually dimorphic morphometric

variables determined from DA analysis and t-tests. Sexually di-morphic variables were used to generate body dimension ratiosthat differed significantly between sexes using Student’s t-tests(P < 0.05). The 75th percentile values for these ratios were thenused as cut-off levels for discriminating between sexes. Accu-racy for each sex was determined by applying the resulting bodydimension ratios and cut-off values to field-collected specimens.

RESULTSOne female and three male Alligator Gars were removed

from the data set as outliers. The final data set consisted of 30female and 83 male fish, ranging in size from 591 to 1,255 mmSL. Gill nets successfully selected for similarly sized individualsfor both sexes (pooled t = −1.24, P = 0.2173; Table 1).

The canonical function derived using the discriminant analy-sis of morphometric variables was significantly correlated withsex (Rc

2 = 0.394, P < 0.0001). Examination of the standardizedcanonical coefficients suggested that two variables loadedstrongly onto the canonical axis (total canonical structure coef-ficient sij > 0.2). The influence of SnL (total canonical structurecoefficient sij = −0.92) was approximately double that of CP(sij = 0.38) in the canonical function. Females generally hadlonger snouts, but smaller caudal peduncle heights than didmales. Other variables loaded weakly on the first canonicalaxis and generally were poor predictors of sex. Using the fullcanonical function, the resubstitution procedure indicated a

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:45

12

Aug

ust 2

013

814 MCDONALD ET AL.

TABLE 1. Morphological variable measurements in female and male Alligator Gars, Student’s t-tests for each variable, and each variable weighted by SL.Sample sizes for Alligator Gars for sexes (female = 30, male = 83) were constant throughout variables. Statistically significant values (P < 0.05) are displayed inbold italic text. Variable measurements are abbreviated by length (Ln), width (Wd), and height (Ht).

Female Male Univariate test (var/SL) test

Variable Range (mm) Mean ± SD (mm) Range (mm) Mean ± SD (mm) t P t P

SL 610–1,216 879 ± 153.8 591–1,255 917 ± 141.0 −1.24 0.217Caudal peduncle Ht 45–100 66.3 ± 14.2 42–99 70.6 ± 12.4 −1.57 0.119 −1.76 0.082Dorsal fin base Ln 38–75 56.1 ± 10.1 36–80 60.1 ± 10.4 −1.83 0.070 −1.37 0.174Anal fin base Ln 36–73 53.8 ± 9.7 35–83 58.4 ± 10.8 −2.02 0.046 −2.10 0.038Head Ln 160–318 226.6 ± 39.2 149–320 225.1 ± 32.9 0.20 0.844 5.45 <0.001Orbital Wd 47.5–97.1 69.7 ± 12.7 47.2–99.4 73.0 ± 11.6 −1.32 0.190 −0.85 0.400Snout Ln 100–214 148.1 ± 26.6 95–190 144.2 ± 19.7 0.72 0.477 6.72 <0.001Snout Wd (nostril) 18.4–39.3 28.6 ± 4.9 19.2–43.7 30.5 ± 4.6 −1.82 0.071 −1.27 0.207Snout Wd (jaw) 43.1–103.2 70.8 ± 15.2 37.2–103.2 73.6 ± 13.9 −0.92 0.359 0.02 0.987Opercle plate Wd 22.6–49.1 33.5 ± 6.7 23.1–49.2 35.4 ± 5.7 −1.44 0.153 −1.07 0.289Orbit diameter 15–23.1 19.0 ± 2.1 12.2–26.6 19.9 ± 2.3 −1.98 0.050 −0.12 0.907Head Ht (Opercle) 47.4–93.4 66.8 ± 12.0 36.5–93 71.0 ± 11.6 −1.68 0.095 −1.71 0.086Head Ht (Orbit) 32.2–65.5 47.3 ± 8.5 32.7–64.6 49.7 ± 7.7 −1.42 0.158 −0.72 0.474

mean classification accuracy of 84% (male accuracy of 84%,female accuracy of 83%). The cross-validation procedureindicated a mean classification accuracy of 71% and suggestedthat classification accuracy was much greater in males (81%)than in females (62%). The finding that both the resubstitutionand cross-validation resampling procedures resulted in classi-fication accuracy > 70% suggests a relatively stable canonicalfunction, although admittedly female classification accuracyis low. While the differences in morphometrics in males andfemales suggest sexual dimorphism, it also indicates that thereis not a reliable way of conclusively determining sex in AlligatorGar using the external morphometric measurements examinedhere. The distribution of canonical scores of both males andfemales indicates a great deal of overlap in multivariate externalmorphology (Figure 3).

FIGURE 3. Canonical scores for males (black bars) and females (gray bars)from the full discriminant analysis of eight morphometric characters measuredin Alligator Gars.

Morphometric variables AF and OD were different betweensexes (P < 0.05). Anal fin base length was also significant afterbeing corrected for SL whereas the variable OD was no longersignificant after accounting for SL (t = −0.12, P = 0.9072;Table 1). Three variables were significant after correction forSL: AF (t = −2.10, P = 0.0383), HL (t = 5.45, P < 0.0001),and SnL (t = 6.72, P < 0.0001). Two sexually dimorphic bodymeasurement ratios were generated using snout length, anal finbase height, and SL: (1) SL/SnL (t = −6.65, P < 0.0001) and(2) SnL/AF (t = 5.16, P < 0.0001; Figure 4). The first ratiogenerally resulted in values ≥ 6.11 in males versus < 6.11in females; this ratio resulted in diagnostic accuracy of 63%for females (the proportion of females correctly identified asfemales) and 92% for males, with 82% meeting sex designation(proportion of accurate sex designation from total number offish examined [n = 113]). The second ratio generally resultedin values ≤ 2.66 for males and > 2.66 for females and resultedin diagnostic accuracy of 52% for females and 90% for males,with 78% meeting sex designation. The use of both ratios inseries resulted in diagnostic accuracy of 72% for females and93% for males, with 68% meeting sex designation.

DISCUSSIONOur discriminant analysis suggested that snout length and

caudal peduncle height, each of which correlated significantlywith SL, were the most influential variables separating sexof Alligator Gar. Snout length was by far the most influentialvariable in the DA model; female Alligator Gars had longersnout lengths than snout lengths of similarly sized males. Thisresult is similar to that from a sexual dimorphism study ofSpotted Gar (Love 2002) in which snout length was the only

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:45

12

Aug

ust 2

013


FIGURE 4. Box plots of body measurement ratios successful in determiningexternal identification of sex in Alligator Gars. Top panel represents body ratio= (SL / snout length) between sexes, with a graphed line displaying separationat the 75th percentile of the box plots (≥6.11 in males versus <6.11 in females).Bottom panel represents body ratio = (snout length / anal fin base length)between sexes, with a graphed line displaying separation at the 75th percentileof the box plots (≤2.66 for males and >2.66 for females).

variable included after applying a stepwise variable eliminationprocedure. Love (2002) suggested that longer snout lengths infemale Spotted Gars enhance foraging success and providesthe additional nutritional requirements necessary for viable eggproduction. There are limited data on reproductive investmentof female gars. Johnson and Noltie (1997) found that spawningfemale Longnose Gars lose 26.8% of total body weight (anaverage of 1,099 g per female) postspawn, suggesting that eggdeposition and catabolism each contribute to significant weightloss; however, snout length was not a sexually dimorphic vari-able for the Longnose Gar. McGrath and Hilton (2012) foundthat male Longnose Gars tended to have wider snouts comparedwith females. This is intriguing because the Longnose Gar isa congener of Spotted Gar (Wright et al. 2012). This resultsuggests either (1) parallel or convergent evolution in snoutdimorphism has occurred in Spotted Gar and Alligator Gar, or(2) the lack of sexual dimorphism in snout length in LongnoseGar is a derived trait. Wylie (1976) investigated the phylogenyof fossil and extant gars comparing a series of osteological mea-

surements, myological patterns, meristics, and body colorationsand concluded that the Longnose Gar is unique in the Lep-isosteus genus in that the elongated snout is an autapomorphictrait that differs from the other recent gar species. McGrath andHilton (2012) suggest that the snout morphological differencesbetween Longnose Gar and Spotted Gar had evolved in responseto differences in diet between species. Robertson et al. (2008)not only found diet dissimilarities between Longnose Gar andSpotted Gar, but also found a difference in habitat preferencebetween species (i.e., Longnose Gar preferred lotic habitatswhereas Spotted Gar and Alligator Gar preferred lentic habitats).Therefore, snout morphology may be the result of adaptation todifferent habitats and habitat-specific prey. Similar adaptationsmay be found in other gar species across both genera, and theloss of dimorphism in Longnose Gar is a derived condition.

Caudal peduncle height was also sexually dimorphic forAlligator Gars we examined, although the importance of thisvariable in discriminating sexes was inferior to that of snoutlength. Overall, male Alligator Gars had higher caudal peduncleheight compared with females. Based on spawning habits of garsit is possible that larger caudal peduncles or caudal fins may bebeneficial for competitive male Alligator Gars during spawning.Gar spawning often involves a single large female accompaniedby several males that competitively fertilize batches of dem-ersal eggs (Suttkus 1963; Tyler and Granger 1984; Boschungand Mayden 2004). McGrath and Hilton (2012) discovered thatmale Longnose Gars had larger anal fins compared with fe-males and they speculated that these larger anal fins may en-hance the ability of males to distribute milt on the batches ofeggs. Our data suggest that caudal peduncle depth and caudalfins may contribute to fertilization in much the same way. Analfin size was slightly larger in male Alligator Gars in this study,although this variable was not included in the discriminant anal-ysis because of multicollinearity with caudal peduncle height.When anal fin size was included in the discriminant analysis,its effect was similar to (and redundant with) caudal peduncleheight (data not shown). Overall larger fin sizes in male Alliga-tor Gars may afford an advantage in accessing eggs during thespawn.

The serial body ratio method developed here will be usefulfor field identification of sex in Alligator Gar. This method wasbased on three easily measured variables: SL, snout length, andanal fin base length. Anal fin base length was chosen over caudalpeduncle height for three reasons: (1) the effect of anal fin baselength was correlated with caudal peduncle size prior to DA,suggesting that their influence on sex identification is similar,(2) whereas anal fin base length was statistically differentbetween sexes from a univariate sense, caudal peduncle heightwas not, and (3) anal fin base length was easier to measure thanwas caudal peduncle height. The body ratio method was 72%accurate for females and 93% accurate for males. However,caution should be exercised with the serial body ratio methodbecause most of the Alligator Gars used in this study originatedfrom a single sample site (Cedar Lakes area) and were of a

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:45

12

Aug

ust 2

013

816 MCDONALD ET AL.

limited size range (591–1,255 mm SL). The potential for mor-phological divergence among gar populations in independentdrainages, in particular, limits the applicability of this test whenfish from outside of the sampled range are examined. It shouldbe noted, however, that six of the eight Alligator Gars from areasother than Cedar Lakes were accurately identified using theserial body ratio method, whereas the other two accurately metthe first body ratio method for males, but not the second bodyratio. It is recommended that use of the serial body ratio methodfor populations outside of the sampled area should be validatedusing the method of Ferrara and Irwin (2001). In any case, theseratios provide a simple baseline methodology for noninvasivesex identification of the Alligator Gar. Snout length, SL, andanal fin base length can all be quickly measured in the field andare (1) more accurate than identification based upon TL alone,and (2) preferred over the method of Ferrara and Irwin (2001),which requires sacrificing the fish. The described method maybe useful for generating estimates of sex ratios of AlligatorGars in wild populations when a nonlethal diagnostic ispreferred.

ACKNOWLEDGMENTSWe thank TPWD field biologists and technicians, most no-

tably E. Young and B. Karel, for assistance in Alligator Garcollection and their transportation. Assistance during sampleprocessing was provided by R. Weixelman (TPWD-CF). K.Winemiller (Texas A&M University) provided Alligator Garskull photographs. Improvements to the first draft were pro-vided by Mark Fisher (TPWD-CF), Dan Daugherty (TPWD-IF), and three anonymous reviewers. This research was fundedby Federal Aid in Sport Fish Restoration Grant F-34-M.

REFERENCESBoschung, H. T. Jr., and R. L. Mayden. 2004. Fishes of Alabama. Smithsonian

Books, Washington, D.C.Ferrara, A. M. 2001. Life-history strategy of Lepisosteidae: implications for

the conservation and management of Alligator Gar. Doctoral dissertation.Auburn University, Auburn, Alabama.

Ferrara, A. M., and E. R. Irwin. 2001. A standardized procedure for inter-nal sex identification in Lepisosteidae. North American Journal of FisheriesManagement 21:956–961.

Johnson, B. L., and D. B. Noltie. 1997. Demography, growth, and reproductiveallocation in stream-spawning Longnose Gar. Transactions of the AmericanFisheries Society 126:438–466.

Love, J. W. 2002. Sexual dimorphism in Spotted Gar Lepisosteus oculatus fromsoutheastern Louisiana. American Midland Naturalist 147:393–399.

McGarigal, K., S. Cushman, and S. Stafford. 2000. Multivariate statistics forwildlife and ecology research. Springer-Verlag, New York.

McGrath, P. E., and E. J. Hilton. 2012. Sexual dimorphism in Longnose GarLepisosteus osseus. Journal of Fish Biology 80:335–345.

Mendoza, R., C. Aguilera, G. Rodrıguez, M. Gonzalez, and R. Castro. 2002.Morphophysiological studies on Alligator Gar (Atractosteus spatula) larvaldevelopment as a basis for their culture and repopulation of their naturalhabitats. Reviews in Fish Biology and Fisheries 12:133–142.

Mendoza, R., O. Santillan, A. Revol, C. Aguilera, and J. Cruz. 2012. AlligatorGar (Atractosteus spatula, Lacepede 1803) vitellogenin: purification, char-acterization and establishment of an enzyme-linked immunosorbent assay.Aquaculture Research 43:649–661.

O’Connell, M. T., T. D. Shepherd, A. M. U. O’Connell, and R. A. Myers.2007. Long-term declines in two apex predators, Bull Sharks (Carcharhinusleucas) and Alligator Gar (Atractosteus spatula), in Lake Pontchartrain, anoligohaline estuary in southeastern Louisiana. Estuaries and Coasts 30:567–574.

Page, L. M., and B. M. Burr. 1991. A field guide to freshwater fishes: NorthAmerica north of Mexico. Houghton Mifflin, Boston.

Perschbacher, P. W. 2011. Effects of structure, forage, and stocking densityon juvenile production of Alligator Gars in outdoor pools. North AmericanJournal of Aquaculture 73:21–23.

Robertson, C. R., S. C. Zeug, and K. O. Winemiller. 2008. Associations betweenhydrological connectivity and resource partitioning among sympatric garspecies (Lepisosteidae) in a Texas river and associated oxbows. Ecology ofFreshwater Fish 17:119–129.

Suttkus, R. D. 1963. Order Lepisostei. Pages 61–68 in H. B. Bigelow, editor.Fishes of the western North Atlantic, part 3, soft-rayed bony fishes. SearsFoundation for Marine Research, Yale University, New Haven, Connecticut.

Tyler, J. D., and M. N. Granger. 1984. Notes on food habits, size, and spawningbehavior of Spotted Gar in Lake Lawtonka, Oklahoma. Proceedings of theOklahoma Academy of Science 64:8–10.

Wright, J. J., S. R. David, and T. J. Near. 2012. Gene trees, species trees, andmorphology converge on a similar phylogeny of living gars (Actinopterygii:Holostei: Lepisosteidae), an ancient clade of ray-finned fishes. MolecularPhylogenetics and Evolution 63:848–856.

Wylie, E. O. 1976. The phylogeny and biogeography of fossil and recent gars(Actinopterygii: Lepisosteidae). University of Kansas Museum of NaturalHistory Miscellaneous Publication 64.D

ownl

oade

d by

[D

epar

tmen

t Of

Fish

erie

s] a

t 23:

45 1

2 A

ugus

t 201

3



Spatiotemporal Predictive Models for JuvenileSouthern Flounder in Texas EstuariesBridgette. F. Froeschke a e , Philippe Tissot b , Gregory W. Stunz c & John T. Froeschke da Florida Center for Community Design and Research , University of South Florida , 4202 EastFowler Avenue, HMS 301, Tampa , Florida , 33620-8340 , USAb Texas A&M University–Corpus Christi , 6300 Ocean Drive, Corpus Christi , Texas ,78412-5869 , USAc Harte Research Institute for Gulf of Mexico Studies , 6300 Ocean Drive, Corpus Christi ,Texas , 78412-5869 , USAd Gulf of Mexico Fishery Management Council , 2203 North Lois Avenue, Suite 1100, Tampa ,Florida , 33607 , USAe Janicki Environmental, Inc. , 1727 Dr. Martin Luther King Jr. Street North, St. Petersburg ,Florida , 33704 , USAPublished online: 06 Aug 2013.

To cite this article: Bridgette. F. Froeschke , Philippe Tissot , Gregory W. Stunz & John T. Froeschke (2013) SpatiotemporalPredictive Models for Juvenile Southern Flounder in Texas Estuaries, North American Journal of Fisheries Management, 33:4,817-828, DOI: 10.1080/02755947.2013.811129







http://dx.doi.org/10.1080/02755947.2013.811129




ARTICLE

Spatiotemporal Predictive Models for Juvenile SouthernFlounder in Texas Estuaries

Bridgette. F. Froeschke*1

Florida Center for Community Design and Research, University of South Florida,4202 East Fowler Avenue, HMS 301, Tampa, Florida 33620-8340, USA

Philippe TissotTexas A&M University–Corpus Christi, 6300 Ocean Drive, Corpus Christi, Texas 78412-5869, USA

Gregory W. StunzHarte Research Institute for Gulf of Mexico Studies, 6300 Ocean Drive, Corpus Christi,Texas 78412-5869, USA

John T. FroeschkeGulf of Mexico Fishery Management Council, 2203 North Lois Avenue, Suite 1100, Tampa,Florida 33607, USA

AbstractSouthern Flounder Paralichthys lethostigma supports a multimillion dollar commercial and recreational fishery

in the Gulf of Mexico. Despite its economic importance, the Southern Flounder population has been declining fordecades. To improve the management of this fishery, both population trends and changes in environmental conditionsneed to be considered. Using two different statistical modeling techniques, boosted regression tree (BRT) and artificialneural network (ANN), a 29-year fisheries-independent record of juvenile Southern Flounder abundance in Texaswas examined to illustrate how environmental factors influence the temporal and spatial distribution of juvenileSouthern Flounder. Boosted regression trees show the presence of juvenile Southern Flounder is closely associatedwith relatively low temperatures, low salinity levels, and high dissolved oxygen concentrations. Both ANN and BRTmodels resulted in high predictive performance with slight spatial differences in predicted distribution. Both modelssuggested high probability of occurrence in Galveston Bay and East Matagorda Bay. The ANN accurately predictedhigher probability of occurrence in Sabine Lake compared with the BRT model. Our results will provide tools forfisheries managers to enhance management and sustainability of the Southern Flounder population. Moreover, theseresults also identify a predictive framework for proactive approaches to ecosystem management by providing moredata to identify essential habitat features and understanding relationships between abiotic and biotic factors withinthose habitats.

Declines in abundance and extensive exploitation of theworld’s fisheries and marine habitats have caused concernamong many researchers (Jackson et al. 2001; Pauly et al. 2002;Hilborn et al. 2003; Halpern et al. 2008). Human impacts havealtered the distribution, quantity, and quality of marine habitats

*Corresponding author: [email protected] address: Janicki Environmental, Inc., 1727 Dr. Martin Luther King Jr. Street North, St. Petersburg, Florida 33704, USA.Received October 16, 2012; accepted May 22, 2013

(Pyke 2004; Lotze et al. 2006). They have contributed to thedepletion of more than 90% of estuarine species, degraded wa-ter quality, and accelerated species invasions, and have reducedseagrass and wetland habitat among estuaries and coastal seasby 65% (Lotze et al. 2006). Seventy-five percent of fisheries

817

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013

818 FROESCHKE ET AL.

worldwide are over exploited or fully exploited (NMFS 2002).Within the USA, 17% of fisheries are subject to overfishing and24% are overfished (NMFS 2008). Impacts from recreational(Coleman et al. 2004) and commercial fishing and bycatch canbe significant contributors to the decline of fisheries (Jacksonet al. 2001; Pauly et al. 2002; Hilborn et al. 2003).

In the Gulf of Mexico, the Southern Flounder Paralichthyslethostigma supports a multimillion dollar commercial andrecreational fishery, but declines in this stock (Froeschke et al.2011) have led to reduced recreational and commercial catches.Southern Flounder populations have been on the decline fordecades and are currently at all-time lows in Texas (Froeschkeet al. 2011). Management efforts for Southern Flounder in Texashave focused on implementing regulations for recreational andcommercial fisheries, yet the population remains in decline,suggesting that factors other than fishing may be negatively in-fluencing the Southern Flounder population. Time-series analy-sis indicated that both juvenile and adult Southern Flounder aredeclining in Texas estuaries (Froeschke et al. 2011). Juvenile re-cruitment is decreasing 1.3% per year (1977–2007), whereas theadult population is decreasing at a rate of 2.5% per year (1975–2008; Froeschke et al. 2011). Moreover, abundance trends ofjuvenile and adult Southern Flounder are independent, particu-larly with high mortality rates of postjuvenile flounder that oc-curred during the 29-year study period (Froeschke et al. 2011).Stunz et al. (2000) demonstrated that a reduced proportion ofSouthern Flounder is reaching the age of maturity. To addressthese concerns, the Southern Flounder population may benefitfrom a shift towards an ecosystem-based approach incorporat-ing interactions among physical and biological components ofthe system management (Pikitch et al. 2004; Marasco et al.2007; Crowder et al. 2008). Within this perspective, fisheriesmanagement includes ecological factors that identify essentialfish habitat (EFH) including both abiotic and biotic componentsof the environment.

Spatiotemporal models provide valuable information thatmay enhance management and ensure sustainability of not onlythe Southern Flounder fishery in particular but other fisheriesas well. The use of boosted regression trees (BRT) is relativelynew in ecological applications but has proven to be an effectivemethod to identify relationships between fish distribution pat-terns and environmental predictors (Leathwick et al. 2006, 2008;Froeschke et al. 2010; Froeschke and Froeschke 2011). More-over, BRT can be effective in predicting the occurrence of ju-venile Southern Flounder to determine EFH within the AransasBay, Texas, complex (Froeschke et al., in press). The artificialneural network (ANN) model is a well-established method foridentifying complex hydrographical patterns associated with theabundance and dynamics of different phases in the life cycle offish (Suryanarayana et al. 2008). Many researchers have usedANNs to predict fish recruitment (Kusakabe et al. 1997; En-gelhard and Heino 2002) and age of fish (Potter et al. 1993;Robertson and Morison 1999; Engelhard et al. 2003) from ex-planatory variables (Suryanarayana et al. 2008).

The goal of this study was to provide additional informa-tion that can be used for the management of Southern Flounderby using statistical modeling techniques to explain how envi-ronmental factors influence the temporal and spatial patternsof juvenile fish. Additionally, this study compared a relativelynew modeling technique (BRT) with a well-accepted technique(ANN). Specifically, this study (1) determined the relationshipbetween temporal (month, year), spatial (distance to the inlet),and physical (temperature, turbidity, dissolved oxygen, salinity,and depth) variables with the occurrence of juvenile SouthernFlounder; (2) used BRT and ANN models to make spatial pre-dictions of the probability of presence in Texas bays; and (3)compared the predictive power and predicted spatial distributionof the trained and tested BRT and ANN.

METHODSStudy area.—The study was conducted in nine major bays

along the Texas coast, located along the northwestern Gulfof Mexico (Figure 1). The Texas coast is 563 km in lengthand contains five barrier islands that stretch approximately161 km. There are six consistently open pathways for waterexchange and animal transport between adult and nursery habi-tat in the nearshore bays and the Gulf of Mexico (http://goliath.cbi.tamucc.edu/TexasInletsOnline/TIO%20Main/index.htm).

Data collection.—Data were provided courtesy of the TexasParks and Wildlife Department (TPWD) and were collected aspart of their Resource and Sport Harvest Monitoring Programtargeting juvenile finfish and shellfish. Sampling has occurredsince 1977 for juveniles in nine bays along the Texas coast(1977–2007, n = 18,078; Figure 1). All sampling followedprotocols detailed in the Marine Resource Monitoring Oper-ations Manual (Martinez-Andrade et al. 2009). Juvenile South-ern Flounder (age < 2 years, 11–290 mm TL; Stokes 1977;Etzold and Christmas 1979; Stunz et al. 2000) were sampledmonthly using a randomized, stratified sampling design alongthe shoreline of each bay with 18.3 × 1.8-m bag seines. Thebag seines used in this study were designed to sample juve-nile estuarine fish populations (Martinez-Andrade 2009). Whileformal gear selection studies were not performed, previous stud-ies on this species have shown this to be an effective gear forsampling juvenile Southern Flounder (Nanez-James et al. 2009;Froeschke et al., in press). Bag seines were deployed perpen-dicular to the shoreline and were carried parallel to shore for15.2 m. Twenty bag seines were deployed each month in SabineLake, Galveston Bay, West Matagorda Bay, San Antonio Bay,Aransas Bay, Corpus Christi Bay, upper Laguna Madre, andlower Laguna Madre, and 10 bag seines were deployed eachmonth in East Matagorda Bay. Months with high abundance ofjuvenile recruitment (January–May) were used in the models(Nanez-James et al. 2009; Froeschke et al. 2011).

Patterns of eight environmental variables relevant to fishwere examined coast-wide to investigate relationships be-tween environmental conditions and juvenile Southern Flounder

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013

PREDICTIVE MODELS FOR JUVENILE SOUTHERN FLOUNDER 819

FIGURE 1. Bag seine sampling locations (circles, n = 18,078) for the TPWD Resource and Sport Harvest Monitoring Program from January through May1979–2007 (each site was sampled once over the course of the study).

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013


FIGURE 2. Input factors for boosted regression trees and artificial neuralnetworks to identify probability of presence for juvenile Southern Flounderalong the Texas coast.

distributions (Figure 2). Salinity (psu), temperature (◦C), tur-bidity (NTU), and dissolved oxygen (mg O2/L) were collectedin the surface waters (0–15 cm) for each sampling event. Tur-bidity readings were processed in the laboratory within 24 husing a calibrated turbidimeter. Water depth, sampling time,and location were also recorded for each sample. All variableswere measured during each sampling (i.e., all years and bays).The increase of variance estimates of the estimated regressioncoefficient for each variable from collinearity was tested usingthe variance inflation factor (VIF; Table 1).

To examine potential relationships between the distributionof juvenile Southern Flounder and connection to the Gulf ofMexico, distance from each sampling location to the nearesttidal connection to the Gulf of Mexico (Figure 1) was calculatedusing the cost–distance function in the ArcGIS software packagewith the spatial analyst extension (ESRI), using the shoreline asa barrier (Whaley et al. 2007; Froeschke et al. 2010; Froeschkeand Froeschke 2011). The cost–distance function calculates theshortest distance between two points while accounting for ge-ographic boundaries (i.e., land) to provide more accurate rela-tive distance estimates than Euclidian (straight-line) techniques(Froeschke et al. 2010; Froeschke and Froeschke 2011). ForCorpus Christi Bay, two distance matrices were calculated. One

TABLE 1. Variance inflation factors (VIF) of the variables included in theboosted regression trees and artificial neural network models indicated nocollinearity of the variables.

Variables VIF

Year 1.0515092Month 2.8223845Salinity (psu) 1.2764295Temperature (◦C) 2.8601381Dissolved oxygen (mg O2/L) 1.2642134Turbidity (NTU) 1.0531671Depth (m) 1.1686556Distance to the inlet (cost-distance units) 1.0250715

distance matrix was developed without the Packery Channel in-let and applied to all samples collected before the opening ofthis channel. A second matrix was calculated including PackeryChannel and the distance estimates were applied to all samplingevents after June 2005.

Boosted regression trees.—Relationships of juvenile South-ern Flounder with physical, spatial, and temporal variables weredetermined using a forward fit, stage-wise, binomial BRT model(De’ath 2007). Analyses were conducted in R (version 2.9; RDevelopment Core Team 2009) using the ‘gbm’ library supple-mented with functions from Elith et al. (2008). The adjustablemodel parameters for BRT are tree complexity (tc), learning rate(lr), and bag fraction (bf ), where tc controls whether interactionsare fitted, lr weights the contribution of each tree to the growingmodel, and bf specifies the proportion of data selected at eachstep (Elith et al. 2008). The model was fit to allow interactionsusing a tree complexity that had a value of 5 and a learning rateof 0.01 to minimize predictive deviance and maximize predic-tive performance. Tenfold cross validation of training data (n =12,651) was used to determine the optimal number of trees.

The BRT technique is an ensemble method and is a combi-nation of techniques between statistical and machine learningtraditions that has the power to (1) accept different types of pre-dictor variables, (2) accommodate missing values through theuse of surrogates, (3) resist effects of outliers, and 4) fit inter-actions between predictors (Elith et al. 2006, 2008; Leathwicket al. 2006, 2008). This is a relatively new method to addressecological questions but can be effective to identify relationshipsbetween fish distribution patterns and environmental predictors(Leathwick et al. 2006, 2008; Froeschke et al. 2010; Froeschkeand Froeschke 2011).

Unlike traditional regression techniques, BRT combines thestrength of two algorithms, regression trees and boosting, tocombine large numbers of relatively simple tree models insteadof a single “best” model (Elith et al. 2006, 2008; Leathwicket al. 2006, 2008). Each individual model consists of a simpleregression tree assembled by a rule-based classifier that par-titions observations into groups having similar values for theresponse variable based on a series of binary splits constructedfrom predictor variables (Friedman 2001; Leathwick et al. 2006;Elith et al. 2008). The BRTs often have a higher predictive per-formance than single tree methods due to the inherent strengthsof regression trees and the robustness of model averaging thatimproves predictive performance. While overfitting can occur,this is minimized by incorporating 10-fold cross validation intothe model-fitting process (Elith et al. 2006, 2008; Leathwicket al. 2006, 2008).

Artificial neural network.—A sigmoidal–sigmoidal, mul-tilayer feed-forward ANN model with back-propagationLevenberg–Marquardt learning algorithm was used to predictthe presence and absence of juvenile Southern Flounder alongthe Texas coast. The model used the same eight predictor vari-ables as the BRT model (Figure 3), one hidden layer with fourhidden neurons, and the output layer with the presence and ab-sence of Southern Flounder as the target (n = 12,651; Figure 3).

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013


FIGURE 3. Input factors for the sigmoidal–sigmoidal, multilayer, feed-forward artificial neural network model with back-propagation learning algorithm,consisting of eight inputs, one hidden layer, four hidden neurons, and one output layer.

The number of hidden neurons was determined by comparingareas under the curve (AUCs) for each receiver operating char-acteristic (ROC) curve while varying the number of hidden neu-rons, and a validation, training, and testing set was used to avoidoverfitting. Analyses were conducted using the nprtool packagein MATLAB (2010b, The MathWorks, Natick, Massachusetts).

Artificial intelligence neural network models do not haveassumptions of linearity, normality, or homogeneity (Campbellet al. 2007), can model multivariate and nonlinear data with dis-continuous regions, and do not require transformation of data(Suryanarayana et al. 2008). Therefore, an ANN model providesan appropriate technique to approximate nonlinear relationshipsand has been suggested as one of the best choices for model-ing spatiotemporal patterns of fish (Suryanarayana et al. 2008).Artificial neural networks consist of neurons (processing units)with weights and biases (parameters) fitted by training a modelover a portion of the data set. The result is a model that maps aset of given values (inputs) to an associated set of targets (out-put; Salia 2005; Zuur et al. 2007). Model weights are trained bypassing through a pair set of inputs and outputs and adjustingprogressively to the weights to minimize the error between theanswer predicted by the ANN and the true answer provided inthe training set (Zuur et al. 2007). All inputs are individuallyweighted and combined prior to being transformed in a hiddenlayer (consisting of a variable number of neurons) that performsa nonlinear transformation of the derived linear value (Zuuret al. 2007). Values of predictor variables varied widely and thesigmoid function of the neural network was used as it is moreresistant to the effects of extreme values than regression-basedmodels (Campbell et al. 2007).

Model selection.—Prior to model fitting, data (n = 18,078)were randomly split into training (70%, n = 12,651) and inde-pendent testing sets (30%, n = 5,427). Model performance andcomparison of BRT and ANN models was assessed for predic-

tions computed for the independent testing set. For each model,AUC calculated from the ROC performance metric was deter-mined (Wilks 2006). To identify spatial patterns of recruitmentthe probability of capture was predicted for the study area usinga form of logistic regression based on the fitted BRT and ANNmodels (Elith et al. 2008). Predictions were computed basedon the probability that a species occurs (y = 1) at a locationwith covariates X and P(y = 1|X) using the logit : logit[P(y =1|X)] = f (X) scale. Suites of environmental conditions weredeveloped for each month (January–May) based on environ-mental variables measured during each month included in theanalysis using ordinary kriging (Saveliev et al. 2007). The BRTand ANN model outputs were then used to predict probabilityof capture coastwide during these specific seasonal conditions.To evaluate the performance of the mapped probability ofoccurrence for each model (ANN and BRT), probability ofoccurrence at each sampling location was compared with theindependent testing data set (i.e., not used in model building).

RESULTS

Physicochemical ConditionsSouthern Flounder were captured in 1,255 of 12,651 samples

(frequency of occurrence = 10%) from January to May in thetraining data set and in 550 of 5,427 samples in the independent(testing) data set. On the Texas coast, physical conditionsvaried widely among bay systems. Salinity increased withdecreasing latitude from hyposaline positive (Sabine Lake andGalveston Bay) to moderate (15–35 psu) along the centralcoast, and hypersaline negative estuaries (>35 psu) in thesouthernmost upper and lower Laguna Madre. Over the courseof the study salinity ranged from 0 to 64 psu (mean = 21 psu).Mean sea surface temperature also increased slightly fromnorth to south along the coast, and water temperatures ranged

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013


FIGURE 4. Receiver operating characteristic curve obtained from the trainedboosted regression tree (BRT) model (AUC = 0.828) and trained artificial neuralnetwork (ANN) model (AUC = 0.707) indicating strong predictive power.

from 1.4◦C to 36.5◦C with a mean of 20.9◦C during samplingevents. Dissolved oxygen concentrations (range, 0–28.00 mgO2/L; mean = 8.21 mg O2/L; Figure 4C), turbidities (range,0–999 NTU; mean = 33.16 NTU), and sampling depths (range,0–6.6 m; mean = 0.44 m) were similar among bay systems.

Boosted Regression TreesThe BRT model evaluation suggested good predictive perfor-

mance based on the result predictions to independent data (n =5,427, AUC = 0.76; Table 2). Evaluation of the training modelalso indicated good predictive performance (AUC = 0.83; Fig-ure 4). The BRT model calibrated over the training set and allvariables provided insight into the relationship between the spa-tial, physical, and temporal input variables and the distributionof juvenile Southern Flounder. The variable months (17.5%)and distance to inlet (16.7%) explained the greatest proportionof deviance. Probability of occurrence increased from Januaryto March and declined after March to May. Year of capture(15.2%) predicted the highest probability of occurrence in thelate 1980s and mid-1990s. Overall, the probability of occurrence

of juvenile Southern Flounder has been decreasing since 1997(Figure 5). Temperature (14.8%) explained the most deviance ofthe physical variables considered, followed by salinity (11.5%),turbidity (11.5%), dissolved oxygen (7.6%), and depth (5.1%)(Figure 5). The fitted functions from the BRT model indicatedthat the highest occurrence rates of juvenile Southern Flounderwere in March, were closest to the inlet, and had water temper-ature greater than 10◦C but lower than 20◦C, salinity less than40 psu, turbidity of around 200 NTU and greater than 300 NTU,and depth greater than 1 m (Figure 5).

The independent testing set (n = 5,427) was used to ex-amine spatial predictions for the presence of Southern Floun-der along the coast. For locations that were sampled multipletimes through the course of the study the mean probability ofoccurrence for these sampling events was determined by aver-aging all data points for the site. The resulting data set usedfor the calibration of the spatial prediction model consisted ofn = 3,375 predictions. Spatially explicit models predicted theprobability of capture based on the BRT model output for eachmonth (January–May) by making predictions of the fitted BRTmodel to an interpolated surface of environmental variables.The spatial model exhibited good predictive performance basedon independent data (AUC = 0.719; Figure 6). Capture prob-ability increased each month from January to March, declinedslightly in April, and was low during May (Figure 7). Probabil-ity of capture began increasing first in Galveston Bay and EastMatagorda Bay in February (Figure 7B). In March, probabilityof capture was the highest near the tidal inlets from Galve-ston Bay to Corpus Christi and between Galveston and EastMatagorda Bay (Figure 7C). In April, probability of occurrencestarted to slightly decrease between East Matagorda Bay andCorpus Christi Bay (Figure 7D). Overall, probability of captureincreased within areas with low salinities, cooler temperatures,and closest to tidal inlets.

Artificial Neural NetworkThe “best” neural network model based on AUC for

predicting the presence and absence of Southern Flounderconsisted of eight inputs and four hidden neurons (Figure 3).Model evaluation demonstrated good predictive performanceto independent data (n = 5,427). Furthermore, evaluation ofthe training model also exhibited good predictive performance(mean square error = 0.08, AUC = 0.707; Figure 4).

TABLE 2. Predictive performance of boosted regression tree models for juvenile Southern Flounder; tc = tree complexity, lr = learning rate, bf = bag fraction,nt = number of trees, ROC = receiver operating characteristic curve.

Area under the receiver operating characteristicPercentage deviance explained curve (AUC)

Cross Total ROC cross ROC cross-tc lr bf nt validation Training deviance Independent validation validation SE Train

5 0.01 0.6 1,550 9.70% 19.30% 0.647 0.757 0.735 0.004 0.828

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013


FIGURE 5. Functions fitted for the eight predictor variables by a boosted regression tree (BRT) model relating the probability of capture of juvenile SouthernFlounder to the environment in order to identify the probability of capture along the Texas coast. Y-axes are on the logit scale with mean of zero. Tic marks atinside top of plots show distribution of data across that variable, in deciles. X-axes parameters: month (1 = January, 2 = February, 3 = March, 4 = April, 5 =May), distance to the nearest inlet (DI; cost-distance units), year, temperature (temp; ◦C), salinity (psu), turbidity (NTU), dissolved oxygen (DO; mg O2/L), anddepth (m).

Spatially explicit model predictions of probability of capturefrom the ANN model were determined for each month (January–May) by making predictions of the fitted ANN model to theinterpolated surface of environmental variables. The spatialmodel showed good predictive performance on the independentdata (AUC = 0.69; Figure 6). Capture probability increasedeach month from January to March, declined slightly in April,and was low during May (Figure 8). Spatial patterns were alsoevident. Probability of capture began increasing first in SabineLake, Galveston Bay, and East Matagorda Bay in February(Figure 8B). In March, probability of capture was the highestnear the tidal inlets (Figure 8C). However, there was a relativelymoderate to high probability of occurrence among all of the bays(Figure 8C). In April, probability of occurrence started toslightly decrease between East Matagorda Bay and CorpusChristi Bay, but remained relatively high in Sabine Lake,Galveston Bay, and the lower edge of lower Laguna Madre(Figure 8D). In May, probability of occurrence consisted ofa similar pattern as observed in January but with a moderateprobability of occurrence still prevalent for Sabine Lake and

Galveston Bay (Figure 8E). Overall, probability of captureincreased in areas with low salinities, cooler temperatures, andareas closest to tidal inlets.

DISCUSSIONDistribution and occurrence rates of juvenile Southern Floun-

der were influenced by temporal, physical, and spatial variables.Occurrence patterns exhibited strong seasonal variation, andsampling month was the most influential variable in the BRTmodel. This study demonstrated the importance of incorporatingtemporal, physical, and spatial variables and their interactions inspecies habitat models to identify frequency of occurrence pat-terns of juvenile Southern Flounder. Probability of occurrenceincreased in February and March before a slight decrease inApril and May. Overall, juvenile recruitment patterns observedwere consistent with seasonality of recruitment reported pre-viously (Froeschke et al. 2011). Peak abundances of juvenileSouthern Flounder have been reported from February to Maywith a peak in March along the Texas coast (Froeschke et al.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013


FIGURE 6. Receiver operating characteristic curve obtained from the spatialtested data set against the trained boosted regression tree (BRT) model (AUC =0.71) and tested artificial neural network (ANN) model (AUC = 0.69) indicatinggood predictive power to an independent data set.

2011). Gunter (1945) reported Southern Flounder recruitment inDecember and from February to April, whereas Stokes (1977)reported presence of juveniles starting in January with a peakin February. Simmons and Hoese (1959) reported recruitmentfrom March to May, peaking in April. Rogers and Herke (1985)reported recruitment from January to March, and peaks occurredfrom February to March.

Distance to the nearest inlet was the second most importantpredictor of occurrence with the highest probability of occur-rence closest to the inlets. Many estuarine species (includingSouthern Flounder) spawn offshore and juveniles recruit intoestuaries via tidal inlets. As a result, juvenile abundance is oftengreatest near inlets (Whaley et al. 2007; Froeschke et al. 2010).

Essential fish habitat for age-0 Southern Flounder in AransasBay and Copano Bay, Texas, was suggested to occur in highsalinity, vegetated habitats (seagrass and marsh edge) that occurclosest to the tidal inlet between Aransas Bay and the Gulf ofMexico (Nanez-James et al. 2009). Overall, the current studyconsidered inlets with a variety of habitat types nearby and sug-gests that inlet proximity remains an important feature of habitatquality across biotic habitat types.

Sampling year was the third most important variable demon-strating increasing probability of occurrence until 1990 and thena large decline followed by an increase in 1996 before a steadydecline until the end of the study period in 2007. Results areconsistent with reported time-series analysis demonstrating along-term decline in recruitment of this species in Texas bays(Froeschke et al. 2011).

With respect to environmental variables, temperature was themost important predictor of occurrence, and the highest occur-rence was observed at temperatures less than 20◦C. These resultsindicate temperatures less than 20◦C are optimal for recruitmentof juvenile Southern Flounder. Previous work has shown thatthe optimum recruitment temperature range of Southern Floun-der is 16–16.2◦C (Stokes 1977). However, juvenile SouthernFlounder in Texas bays have been captured in water temper-atures between 14.5◦C and 21.6◦C (Gunter 1945). A studyon juvenile Southern Flounder in the Aransas Bay complex(Mission–Aransas National Estuarine Research Reserve) in-dicated the highest probability of occurrence was at temper-atures less than 15◦C (Froeschke et al., in press). Due to apreference of cooler temperatures, projected sea temperatureincreases are of potential concern for this species. Seawatertemperature is projected to increase by 4◦C in the 21st century(Thuiller 2007). Both Applebaum et al. (2005) and Fodrie et al.(2010) have previously reported rising sea temperatures withinthe Gulf of Mexico. Additional predicted increases in tempera-ture could have substantial effects on the temporal and spatialrecruitment patterns and ultimately population size of SouthernFlounder.

FIGURE 7. Spatial prediction of juvenile Southern Flounder from the “best” boosted regression tree (BRT) model indicating the highest probability of collectionwould occur in March in Galveston Bay, East Matagorda Bay, and areas closest to the inlets. Spatial predictions from BRT of juvenile Southern Flounder capturefor the months of (A) January, (B) February, (C) March, (D) April, and (E) May. [Figure available in color online.]

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013


FIGURE 8. Spatial prediction of juvenile Southern Flounder from the “best” artificial neural network (ANN) model indicating the highest probability of collectionwould occur in March in Sabine Lake, Galveston Bay, East Matagorda Bay, and areas closest to the inlets. Spatial predictions from ANN of juvenile SouthernFlounder capture for the months of (A) January, (B) February, (C) March, (D) April, and (E) May. [Figure available in color online.]

Salinity was also an important predictor of frequency ofoccurrence; Southern Flounder frequency of occurrence washighest at salinities less than 10 psu and decreased at salinitiesgreater than 40 psu. Spatial predictions from both the BRTand ANN models indicated the highest probability of juve-nile Southern Flounder occurred in Sabine Lake and Galve-ston Bay and the lowest probability of occurrence was in theupper and lower Laguna Madre. Along the Texas coast, salin-ity increases with decreasing latitude from hyposaline positive(Sabine Lake and Galveston Bay) to moderate (15–35 psu) alongthe central coast, and hypersaline negative estuaries (>35 psu)in the southernmost upper and lower Laguna Madre. South-ern Flounder are euryhaline (Deubler 1960), but survivorshipand growth rates increase in lower salinity waters (Hickman1968; Stickney and White 1974). This study supports these pre-vious findings as Southern Flounder were more prevalent inthe low salinity and cooler water temperature environments ofSabine Lake and Galveston Bay. This result illuminates poten-tial ramifications of reduced freshwater inflow into these baysystems as historic inflows are increasingly diverted for humanuse.

Turbidity and dissolved oxygen were less important predic-tors of occurrence. This is consistent with Froeschke et al. (inpress) who did not find an effect of turbidity on the probabilityof occurrence of juvenile Southern Flounder in the Mission–Aransas National Estuarine Research Reserve. While dissolvedoxygen levels can influence the distribution, abundance, anddiversity of organisms (Breitburg 2002; Vaquer-Sunyer andDuarte 2008; Montagna and Froeschke 2009), this primarilyoccurs at low oxygen levels (i.e., <2 mg O2/L). In this study,few samples were taken in low dissolved oxygen conditions,but low dissolved oxygen events (e.g., hypoxia) are increasingin frequency and spatial extent in Texas estuaries (Applebaumet al. 2005; Montagna and Froeschke 2009). These data suggestthat oxygen levels influence the distribution and abundance ofSouthern Flounder.

Southern Flounder spawning and recruitment success maybe directly influenced by estuarine conditions, highlighting theimportance of high quality habitat necessary to support impor-tant fishery species. The interaction between habitat quantityand quality can affect the survivorship of flatfish, in which thelargest recruitment potential occurs in areas with high habitatquantity and quality and smallest recruitment potential in areaswith low habitat quantity and quality (Gibson 1994). Biologicalvariables such as prey abundance, predators, habitat structure,water depth, and physical factors such as temperature, salinity,dissolved oxygen, and hydrodynamics affect growth and sur-vival of flatfish (Gibson 1994; Allen and Baltz 1997; Stoneret al. 2001; Glass et al. 2008).

Both models indicated higher probability of occurrence nearthe tidal inlets from Galveston Bay to Corpus Christi and be-tween Galveston and East Matagorda Bay. Overall, probabilityof capture for both models increased with decreasing salinities,cooler temperatures, and proximity to tidal inlets. Althoughoverall accuracy of the ANN model was slightly lower thanthe BRT spatially tested model, the ANN correctly predicteda higher probability of occurrence in Sabine Lake whereas theBRT did not. Based on the biology of the species, we suggest thatthe high probability of occurrence in Sabine Lake is accurate.Moreover, BRT and ANN models both displayed good predic-tive performance of spatial predictions to an independent dataset. The ANN consisted of a similar number of observed andpredicted occurrences than did the BRT. However, the BRT hada higher predictive performance for the training set comparedwith the ANN model and a higher percentage correct for the pre-diction of presence–absence of juvenile Southern Flounder. TheANN and BRT models were similar with regard to the numberof observed and predicted fish. The primary difference betweenthe overall percentage correct between the two models for thetraining and testing sets was the number of fish predicted andnot observed, suggesting that the ANN model and possibly theBRT could be overfitting, a common feature of correlation-based

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013


predictive models including ANN (Zuur et al. 2007). Despitethis limitation, ANN remains a powerful tool for prediction andoften outperforms other methods (Suryanarayana et al. 2008).These results also suggest the need to evaluate a variety of po-tential methods to identify the most robust modeling approachfor a particular application, which is difficult or impossible toidentify a priori given the complexity of large multivariate datasets typically used to guide management of natural resources.

Mapped distribution patterns permit rapid identification anddelineation of important areas in a spatiotemporal context,which is essential for ecosystem based management approaches(Pikitch et al. 2004). Predicted distribution patterns were verysimilar between the spatiotemporal models. For both models,capture probability increased each month from January toMarch and declined slightly starting in April. Salinity levels inSabine Lake are the lowest among the Texas bays, suggestingthat the high predicted frequency of occurrence determinedfrom the ANN is consistent with salinity preference seenamong juvenile Southern Flounder (Hickman 1968; Stickneyand White 1974; Froeschke et al., in press).

Despite the utility of our modeling approaches, there aresome limitations to both methodologies. While model evalu-ation indicated good performance of both the BRT and ANNat predicting the independent testing cases, substantial unex-plained deviance remained in the models. This suggests thatsome important variables in the habitat usage of these specieswere available in the study data set. For example, biotic compo-nents such as spawning location, prey and predator density, andmovement patterns of individuals were not considered in thisstudy. The methods used in this study allowed the considerationof several variables simultaneously and provided timely infor-mation for conservation and management of Southern Flounder.Spatially explicit models permit applications that are not feasiblewith other approaches (e.g., prediction of distribution patternsrelated to dynamic environmental patterns).

Construction of spatiotemporal models for juvenile SouthernFlounder along the Texas coast addresses state and national es-tuarine and coastal resource management issues because it pro-vides information on the spatial distribution and nursery habitatrequirements for this fishery species. Our results provide toolsfor fisheries managers to promote sustainability of the SouthernFlounder fishery. For example, the effect of increased salinitydue to changes in precipitation or urban water diversion couldbe evaluated in this context as grids of environmental conditionswere developed for predictive purposes. A range of scenarioscould be explored and the change in occurrence or distributionof Southern Flounder could be evaluated. This study providesa predictive framework for proactive approaches to ecosystemmanagement where the effects on environmental conditions ona population can be considered and incorporated into harveststrategies. Hidalgo et al. (2011) demonstrated that stock deple-tion can enhance the impact on environmental forcing on fishpopulations. These data suggest that this species is present evenif the best biotic habitat (e.g., seagrass meadow) is available and

if the physical environment (e.g., temperature, salinity) is notwithin the tolerable range for that species. Thus, decreases infreshwater inflow could have a major impact on the distributionof juvenile Southern Flounder. The modeling approaches em-ployed in this study provide a predictive framework from whichchanges in environmental conditions or management measurescould be evaluated to promote development of sustainable man-agement strategies for Southern Flounder in Texas.

ACKNOWLEDGMENTSWe thank Texas Parks and Wildlife Department, especially

Science Director M. Fisher and F. Martinez-Andrade, for pro-viding access and insight into the Southern Flounder monitoringdata. Additionally, we thank the Mission–Aransas National Es-tuarine Research Reserve Fellowship Program, and the HarteResearch Institute for the Gulf of Mexico Studies for fund-ing and support. Furthermore, we thank B. Sterba-Boatwright,J. Fox, and L. McKinney for their assistance with and commentson the manuscript.

REFERENCESAllen, R. L., and D. M. Baltz. 1997. Distribution and microhabitat use by

flatfishes in a Louisiana estuary. Environmental Biology of Fishes 50:85–103.

Applebaum, S., P. A. Montagna, and C. Ritter. 2005. Status and trends ofdissolved oxygen in Corpus Christi Bay, Texas, U.S.A. Environmental Mon-itoring and Assessment 107:297–311.

Breitburg, D. 2002. Effects of hypoxia, and the balance between hypoxia andenrichment, on coastal fishes and fisheries. Estuaries 25:767–781.

Campbell, N., K. Mackenzie, A. F. Zuur, E. N. Ieno, and G. M. Smith. 2007.Fish stock identification through neural network analysis of parasite fauna.Pages 449–462 in A. F. Zuur, E. N. Ieno, and G. M. Smith, editors. Analysingecological data. Springer, New York.

Coleman, F. C., W. F. Figueira, J. S. Ueland, and L. B. Crowder. 2004. Theimpact of U.S. recreational fisheries on marine fish populations. Science305:1958–1960.

Crowder, L. B., E. L. Hazen, N. Avissar, R. Bjorkland, C. Latanich, and M. B.Ogburn. 2008. The impacts of fisheries on marine ecosystems and the transi-tion to ecosystem-based management. Annual Review of Ecology, Evolution,and Systematics 39:259.

De’ath, G. 2007. Boosted trees for ecological modeling and prediction. Ecology88:243–251.

Deubler, E. E. Jr. 1960. Salinity as a factor in the control of growth and survivalof postlarvae of the Southern Flounder, Paralichthys lethostigma. Bulletin ofMarine Science of the Gulf and Caribbean 10:338–345.

Elith, J., C. H. Graham, and R. P. Anderson. 2006. Novel methods improveprediction of species’ distributions from occurrence data. Ecography 29:129–151.

Elith, J., J. R. Leathwick, and T. Hastie. 2008. A working guide to boostedregression trees. Journal of Animal Ecology 77:802–813.

Engelhard, G. H., U. Dieckmann, and O. R. Godo. 2003. Age at maturationpredicted from routine scale measurements in Norwegian spring-spawningherring (Clupea harengus) using discriminant and neural network analyses.ICES Journal of Marine Science 60:304–313.

Engelhard, G. H., and M. Heino. 2002. Maturation characteristics in Norwegianspring-spawning herring before, during, and after a major population col-lapse. International Council for the Exploration of the Sea, C.M. 2002/Y:10,Copenhagen.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013


Etzold, D. J., and J. Y. Christmas. 1979. A Mississippi marine finfish manage-ment plan. Mississippi-Alabama Sea Grant Consortium, MASGP-78-046,Ocean Springs, Mississippi.

Fodrie, F. J., K. L. Heck, S. P. Powers Jr., W. M. Graham, and K. L. Robin-son. 2010. Climate-related, decadal scale assemblage change of seagrass-associated fishes in the northern Gulf of Mexico. Global Change Biology16:48–59.

Friedman, J. H. 2001. Greedy function approximation: a gradient boostingmachine. Annals of Statistics 29:1189–1232.

Froeschke, J. T., and B. F. Froeschke. 2011. Spatio-temporal predictive modelbased on environmental factors for juvenile Spotted Seatrout in Texas estu-aries using boosted regression trees. Fisheries Research 111:131–138.

Froeschke, B. F., B. Sterba-Boatwright, and G. W. Stunz. 2011. AssessingSouthern Flounder (Paralichthys lethostigma) long-term population trends inthe northern Gulf of Mexico using time series analyses. Fisheries Research108:291–298.

Froeschke, B. F., G. W. Stunz, M. M. R. Robillard, J. Williams, and J. T.Froeschke. In press. A modeling and field approach to identify essentialfish habitat for juvenile Bay Whiff (Citharichthys spilopterus) and SouthernFlounder (Paralichthys lethostigma) within the Aransas Bay Complex, TX.Estuaries and Coast. DOI: 10.1007/s12237-013-9600-9.

Froeschke, J. T., G. W. Stunz, and M. L. Wildhaber. 2010. Environmentalinfluences on the occurrence of coastal sharks in estuarine waters. MarineEcology Progress Series 401:279–292.

Gibson, R. N. 1994. Impact of habitat quality and quantity on the recruitmentof juvenile flatfishes. Netherlands Journal of Sea Research 32:191–206.

Glass, L. A., J. R. Rooker, R. T. Kraus, and G. J. Holt. 2008. Distribution,condition, and growth of newly settled Southern Flounder (Paralichthyslethostigma) in the Galveston Bay Estuary, TX. Journal of Sea Research59:259–268.

Gunter, G. A. 1945. Studies of marine fishes of Texas. Publication of the Instituteof Marine Science University of Texas 1.

Halpern, B. S., S. Walbridge, K. A. Selkoe, C. V. Kappel, F. Micheli, C.D’Agrosa, J. F. Bruno, K. S. Casey, C. Ebert, H. E. Fox, R. Fujita, D. Heine-mann, H. S. Lenihan, E. M. P. Madin, M. T. Perry, E. R. Selig, M. Spalding,R. Steneck, and R. Watson. 2008. A global map of human impact on marineecosystems. Science 319:948–952.

Hickman, C. P. Jr. 1968. Glomerular filtration and urine flow in the euryhalineSouthern Flounder, Paralichthys lethostigma, in seawater. Canadian Journalof Zoology 46:427–437.

Hidalgo, M., T. Rouyer, J. C. Molinero, E. Massuti, J. Moranta, B. Guijarro,and N. C. Stenseth. 2011. Synergistic effects of fishing-induced demographicchanges and climate variation on fish population dynamics. Marine EcologyProgress Series 426:1–12.

Hilborn, R., T. A. Branch, B. Ernst, A. Magnussum, C. V. Minte-Vera, M.D. Scheurell, and J. L. Valero. 2003. State of the world’s fisheries. AnnualReview of Environment and Resources 28:359–399.

Jackson, J. B. C., M. X. Kirby, W. H. Berger, K. A. Bjorndal, L. W. Botsford, B. J.Bourque, R. H. Bradbury, R. Cooke, J. Erlandson, J. A. Estes, T. P. Hughes,S. Kidwell, C. B. Lange, H. S. Lenihan, J. M. Pandolfi, C. H. Peterson, R. S.Steneck, M. J. Tegner, and R. R. Warner. 2001. Historical overfishing and therecent collapse of coastal ecosystems. Science 293:629–638.

Kusakabe, T., T. Komatsu, T. Tamaki, M. Nakajima, and I. Aoki. 1997. Pre-diction of the recruit of Sand Eel in the eastern Seto Inland Sea using aneural network. Bulletin of the Japanese Society of Fisheries Oceanography61:375–380.

Leathwick, J. R., J. Elith, W. L. Chadderton, D. Rowe, and T. Hastie. 2008. Dis-persal, disturbance, and the contrasting biogeographies of New Zealand’sdiadromous and non-diadromous fish species. Journal of Biogeography35:1481–1497.

Leathwick, J. R., J. Elith, M. P. Francis, T. Hastie, and P. Taylor. 2006. Variationin demersal fish species richness in the oceans surrounding New Zealand:an analysis using boosted regression trees. Marine Ecology Progress Series321:267–281.

Lotze, H. K., H. S. Lenihan, B. J. Bourque, R. H. Bradbury, R. G. Cooke, M.C. Kay, S. M. Kidwell, M. X. Kirby, C. H. Peterson, and J. B. C. Jackson.2006. Depletion, degradation, and recovery potential of estuaries and coastalseas. Science 312:1806–1809.

Marasco R. J., D. Goodman, C. B. Grimes, P. W. Lawson, A. E. Punt, and T.J. Quinn II. 2007. Ecosystem-based fisheries management: some practicalsuggestions. Canadian Journal of Fisheries and Aquatic Sciences 64:928–939.

Martinez-Andrade, F., M. Fisher, B. Bowling, and B. Balboa. 2009. Marine re-source monitoring operations manual. Texas Parks and Wildlife Department,Austin.

Montagna, P. A., and J. T. Froeschke. 2009. Long-term biological effects ofcoastal hypoxia in Corpus Christi Bay, Texas, USA. Journal of ExperimentalMarine Biology and Ecology 381:S21–S30.

Nanez-James, S. E., G. W. Stunz, and S. Holt. 2009. Habitat use patterns of newlysettled Southern Flounder, Paralichthys lethostigma, in Aransas–Copano Bay,Texas. Estuaries and Coast 32:350–359.

NMFS (National Marine Fisheries Service). 2002. Annual report to Congresson the status of U.S. fisheries 2001. NMFS, Silver Spring, Maryland.

NMFS (National Marine Fisheries Service). 2008. Annual report to Congress onthe status of U.S. fisheries 2007. NMFS, Silver Spring, Maryland. Available:http://www.nmfs.noaa.gov/sfa/statusoffisheries/2007/2007StatusofUSFisheries Report to Congress.pdf. (July 2013).

Pauly, D., V. Christensen, S. Guenette, T. J. Pitcher, U. R. Sumaila, C. J. Walters,R. Watson, and D. Zeller. 2002. Towards sustainability in world fisheries.Nature 418:689–695.

Pikitch, E. K., C. Santora, E. A. Babcock, A. Bakun, R. Bonfil, D. O. Conover,P. Dayton, P. Doukakis, D. Fluharty, B. Heneman, E. D. Houde, J. Link, P. A.Livingston, M. Mangel, M. K. McAllister, J. Pope, and K. J. Sainsbury.2004. Ecology: ecosystem-based fishery management. Science 305:346–347.

Potter, E. C. E., L. Kell, and D. G. Reddin.1993. The discrimination of NorthAmerican and European salmon using a genetic algorithm and by neutral net-work. International Council for the Exploration of the Sea, C.M. 1993/M:18,Copenhagen.

Pyke, C. R. 2004. Habitat loss confounds climate change impacts. Frontiers inEcology and Environment 2:178–182.

R Development Core Team. 2009. R: A language and environment for statisti-cal computing. R Foundation for Statistical Computing, Vienna. Available:http://www.R-project.org. (July 2013).

Robertson, S. G., and A. K. Morison. 1999. A trial of artificial neural networksfor automatically estimating the age of fish. Marine and Freshwater Research50:73–82.

Rogers, B. D., and W. H. Herke. 1985. Temporal patterns and size characteristicsof migrating juvenile fishes and crustaceans in Louisiana marsh. LouisianaState University, Agricultural Experiment Station, Research Report 81, BatonRouge.

Salia, S. B. 2005. Neural networks used in classification with emphasis onbiological populations in stock identification methods: applications in fisheryscience. Academic Press, Amsterdam.

Saveliev, A. A., S. S. Mukharamova, N. A. Chizhikova, R. Budgey, and A. F.Zuur. 2007. Spatially continuous data analysis and modeling. Pages 341–372in A. F. Zuur, E. N. Ieno, and G. M. Smith, editors. Analysing ecologicaldata. Springer, New York.

Simmons, E. G., and H. D. Hoese. 1959. Studies on the hydrography and fishmigration of Cedar Bayou, a natural tidal inlet on the central Texas coast.Publications of the Institute of Marine Science 6:56–80.

Stickney, R. R., and D. B. White. 1974. Effects of salinity on the growthof Paralichthys lethostigma postlarvae reared under aquaculture conditions.Proceedings of the Annual Conference of the Southeastern Association ofGame and Fish Commissioners 27(1973):532–540.

Stokes, G. G. 1977. Life history studies of Southern Flounder (Paralichthyslethostigma) and Gulf Flounder (P. albigutta) in the Aransas Bay area ofTexas. Texas Parks and Wildlife Department, Technical Series 25, Austin.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013


Stoner, A. W., J. P. Manderson, and J. P. Pessutti. 2001. Spatially explicit analy-sis of estuarine habitat for juvenile Winter Flounder: combining generalizedadditive models and geographic information systems. Marine EcologyProgress Series 213:253–272.

Stunz, G. W., T. L. Linton, and R. L. Colura. 2000. Age and growth of SouthernFlounder in Texas waters, with emphasis on Matagorda Bay. Transactions ofthe American Fisheries Society 129:119–125.

Suryanarayana, I., A. Braibanti, R. Sambasiva Rao, V. Anantha Ramam, D.Sudarsan, and G. Nageswara Rao. 2008. Neural networks in fisheries research.Fisheries Research 92:115–139.

Thuiller, W. 2007. Climate change and the ecologist. Nature 448:550–552.

Vaquer-Sunyer, R., and C. M. Duarte. 2008. Thresholds of hypoxia for marinebiodiversity. Proceedings on the National Academy of Sciences of the USA105:15452–15457.

Whaley, S. D., J. J. Burd, and B. A. Robertson. 2007. Using estuarine land-scape structure to model distribution patterns in nekton communities andin juveniles of fishery species. Marine Ecology Progress Series 330:83–99.

Wilks, D. S. 2006. Statistical methods in the atmospheric sciences, 2nd edition.Academic Press, Amsterdam.

Zuur, A. F., E. N. Ieno, and G. M. Smith, editors. 2007. Analysing ecologicaldata. Springer, New York.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:46

12

Aug

ust 2

013



Field and Laboratory Evaluation of Dam Escapement ofMuskellungeMax H. Wolter a , Corey S. DeBoom a & David H. Wahl aa Kaskaskia Biological Station , Illinois Natural History Survey , 1235 County Road 1000N,Sullivan , Illinois , 61951 , USAPublished online: 06 Aug 2013.

To cite this article: Max H. Wolter , Corey S. DeBoom & David H. Wahl (2013) Field and Laboratory Evaluationof Dam Escapement of Muskellunge, North American Journal of Fisheries Management, 33:4, 829-838, DOI:10.1080/02755947.2013.812585







http://dx.doi.org/10.1080/02755947.2013.812585




ARTICLE

Field and Laboratory Evaluation of Dam Escapementof Muskellunge

Max H. Wolter,* Corey S. DeBoom, and David H. WahlKaskaskia Biological Station, Illinois Natural History Survey, 1235 County Road 1000N,Sullivan, Illinois 61951, USA

AbstractMuskellunge Esox masquinongy occur in many Midwestern reservoirs where dam escapement is often reported.

Because densities of Muskellunge in many reservoirs are low, escapement is a concern. Little is known regarding thefactors that influence rates of Muskellunge dam escapement or the proportion of reservoir populations that escapeannually. We used controlled laboratory experiments to examine how juvenile Muskellunge interact with flow overa barrier at varying levels of turbidity, flow rate, habitat availability, and periods in the diel cycle. In the field weinserted PIT tags into juvenile and adult Muskellunge, monitored their escapement over a dam with an antenna array,and then compared escapement among demographic groups and described escapement in relation to precipitationevents, water temperature, and water clarity. Both laboratory and field studies found Muskellunge were more likelyto escape during the day than at night. We estimated that 25% of a reservoir Muskellunge population escaped withinthe 1-year period of this study, with escapement occurring during late spring but not during fall. Adults were morelikely to escape than juveniles, and both sexes escaped at equal rates. Methods developed here can be used to provideuseful information to managers and develop mitigation practices to limit escapement in situations where it is notdesirable.

Passage of fish over dams and spillways is highly variableand unpredictable, and of significant interest to fish managers(Hergenrader and Bliss 1971; Axon and Whitehurst 1985; Wahl1999; Paller et al. 2006). In many situations passage of fish overdams is desired, such as both upstream and downstream dampassage of salmonids, which is often accommodated (Raymond1988; Chapman et al. 1997; Connor et al. 2000). In the Midwest-ern United States “dam escapement,” the permanent emigrationof fish past the impounding barrier of a reservoir, often detractsfrom the goal of establishing and maintaining sport fish popula-tions (Louder 1958; Wahl 1999). Factors thought to contributeto dam escapement of sport fishes include movement relatedto spawning or foraging, spillway design, habitat preference,and amount of overflow (Louder 1958; Lewis et al. 1968; Palleret al. 2006). Previous research has highlighted the magnitude ofdam escapement by a large proportion of Largemouth Bass Mi-cropterus salmoides stocked into a new impoundment escaping

*Corresponding author: [email protected] March 19, 2013; accepted June 3, 2013

within a year (31%: Lewis et al. 1968). In a similar study an esti-mated 10,000 fish escaped from a 65-ha reservoir in Illinois overa 23-month period (Louder 1958). Losses of fish over spillwayscanbe species-specific and vary between reservoirs (Lewis et al.1968; Paller et al. 2006). Size or life stage-specific losses havealso been identified in some species where adults were moreprone to escapement than juveniles (Lewis et al. 1968; Navarroand McCauley 1993; Paller et al. 2006).

In some tailwaters of reservoirs high-density fisheries can becreated as a result of dam escapement (Jacobs and Swink 1983).When sufficient outflow creates consistent riverine conditions,large-bodied fishes often thrive (Harrison and Hadley 1979)and escapement has been described as essentially an annualstocking program for these systems (Jacobs and Swink 1983;Trammell et al. 1993; Schultz et al. 2003). The influx of un-wanted and often nonnative fish carries a risk of negative effectson downstream resident fish communities (Martinez et al. 1994;

829

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:47

12

Aug

ust 2

013

830 WOLTER ET AL.

Spoelstra et al. 2008). Negative consequences for escapees canalso be high as habitat, prey availability, and thermal conditionsin the outflow of smaller impoundments are often not adequateto support large-bodied fish (various biologists, Illinois Depart-ment of Natural Resources, personal communications). In manysituations losses due to escapement are costly when reservoirpopulations are maintained through stocking.

Muskellunge Esox masquinongy are often stocked into reser-voirs to create recreational fishing opportunities. Muskellungeescapement over spillways frequently occurs and is reportedanecdotally across the Midwestern United States (Storck andNewman 1992; Wahl 1999). Because these fish are stocked inlow numbers, Muskellunge have limited potential for naturalreproduction in many environments (Dombeck et al. 1984), andpreventative barriers are often infeasible or ineffective at highflows (Plosila and White 1970), escapement could be one of theprimary factors limiting development of higher-density reser-voir Muskellunge populations (Wahl 1999). At this time we donot have a clear understanding of the mechanisms and mag-nitude of Muskellunge spillway escapement. Such knowledgewould be useful in developing and implementing mitigationefforts and making management decisions. Important informa-tion on the conditions (season, flow, diel period, temperature,water clarity, spillway design) associated with escapement andon fish traits (sex, size, maturity) that are the most suscepti-ble to escapement are needed. Estimates of the proportion of apopulation escaping annually from reservoirs will aid in mak-ing management recommendations and can be used to justifyspecific remedial actions. To address these issues we conductedlaboratory and field evaluations to quantify conditions underwhich Muskellunge escapement occurs and described the traitsand proportion of Muskellunge escaping from a reservoir.

METHODSLaboratory experiment.—In the laboratory we examined a

variety of environmental variables for their effect on juvenile

Muskellunge movement with flow over a simulated barrier. Ina 460 × 100 × 50-cm fiberglass tank we created a spillway byblocking off one end with a notched (2 × 18 cm) board, sim-ulating a dam (Figure 1). Pumps (1/4 hp, 0.76 L/s) were usedto move water from below the spillway back to the other endof the tank creating a closed loop. The upper meter of the tankwas partitioned off with a net so that fish in the trial arenacould not encounter the area where pumps discharged water.Previous field studies have shown that spillway overflow heightis an important determinant of escapement (Lewis et al. 1968;Paller et al. 2006). Resulting overflow heights in this systemwere greater than the body depth (1.25–1.75 cm) of a juvenileMuskellunge, and water velocity at the face of the simulateddam (6 cm/s) was comparable with velocities observed at themouth of the spillway at Lake Sam Dale, Illinois, at similarlevels of overflow (4 cm/s). Water depth was 34 cm in the trialarena when pumps were running, whereas the water level in thecatch basin was 24 cm, which prevented fish from moving backinto the test arena after escaping. The trial arena was 378 L involume and 198 cm long, an approximate distance of 15 bodylengths for an average fish used in the trials. Three identicaltanks were used for replication and were housed indoors whenused for these experiments. Tanks were surrounded by opaquecurtains to prevent fish from being disturbed during trials. Ajuvenile Muskellunge (100–180 mm in length) was randomlyselected from one of four pools of fish (∼200 individuals) andwas placed into the arena and allowed to acclimate for 1 h undertrial conditions with no flow over the barrier. After acclimation,the pumps were turned on to begin the 2-h trial. Diel periodtreatments included day (fish acclimated in light, trial run inlight) and night (fish acclimated in dark, trial run in dark) vari-ables. Flow rate treatments included “no flow” (2 cm height,0 m/s), “low flow” (1 cm height, 6 cm/s), and “high flow” (2 cmheight, 10 cm/s) variables. The “no flow” treatment was in-cluded to test whether observed escapement was a response toflow or the result of random movements of fish throughout the

FIGURE 1. Top-view diagram of the experimental setup used to test the effects of light, flow rate, habitat, and turbidity on Muskellunge interaction with flowover a spillway. Approximate tank dimensions are 4.6 m long × 1 m wide × 0.5 m deep.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:47

12

Aug

ust 2

013

DAM ESCAPEMENT OF MUSKELLUNGE 831

tank. In these treatments the tank was filled to 2 cm above damheight so that fish could cross the dam in the absence of flow.Escapement in “no flow” trials was determined through directand video-recorded observations since fish had the potentialto cross back over the dam after escaping. Habitat treatmentsincluded either presence or absence of simulated vegetation (20-cm strands of yellow nylon rope attached to a 60 × 60-cm wirescreen with a density of 60 strands/m2) placed in the center ofthe tank. Turbidity was altered by adding Bentonite clay withtreatments including clear (0–1 NTU) and turbid (15–30 NTU)levels. After a trial, fish were temporarily separated from thetrial pool of Muskellunge to ensure that the same fish did notexperience another trial within 48 h. Water temperature and tur-bidity measurements were made at the beginning of each trial.Due to the large number of treatment combinations (n = 24)and limited number of tanks (n = 3), it was not possible to con-duct a complete replicate of all treatments in a single day. Wetherefore conducted each replicate experiment over a 4-d period(block) within which we randomly assigned treatment combi-nations to day or night (hereafter referred to as diel periods).All 24 treatment combinations were replicated 20 times overthe summers of 2010 and 2011. Fish were classified as havingeither “escaped” or “not escaped” during the 2-h trials resultingin a binomial response variable. Logistic multiple regressionanalysis was used to identify the importance of each variableto escapement rate using PROC GENMOD (SAS version 9.2).Dummy variable coding was used to create design variables foreach factor level. Reference levels for design variables includedabsence of habitat and turbidity, day periods, and zero flow. Incases of significant main effects, the difference in log odds ofeach level within a factor was compared using linear contrasts(PROC GENMOD; Littell et al. 2002).

We regressed occurrence of escapement against each pre-dictor variable and tested for significance relative to a modelcontaining only the intercept using the likelihood ratio test (alsoknown as the G-statistic or deviance; Kutner et al. 2004). Vari-ables with P-values less than 0.25 were retained as candidatesfor the multivariate model (Quinn and Keough 2002). A fullmodel including the factors selected from univariate analysisand all possible interactions was then fit and compared againstall possible subset models using likelihood ratio tests. Fit of ourfinal model was evaluated with the Hosmer–Lemeshow statis-tic (Quinn and Keough 2002). Performance of the fitted modelwas tested using standard cross validation where a single ob-servation is removed from the data set, the model is fit with theremaining data, and the probability of escapement for the deletedobservation is then predicted from the model. Probabilities of0.5 or greater were used as a positive prediction of escape-ment. Classification rates from cross validation were comparedwith a random expectation using Cohen’s kappa statistic (Tituset al. 1984). These model building and fitting procedures wereconducted with SAS software (SAS Institute, version 9.2, andJMP, version 9). Significance of all tests was determined usingP ≤ 0.05.

Reservoir experiment.—Lake Sam Dale located in southernIllinois was selected for the field portion of the study based ona history of Muskellunge escapement determined from conver-sations with local biologists and Muskellunge angler groups.Lake Sam Dale is a 38-ha impoundment with a “drop-box”style spillway design (dimensions: 5.5 × 2.4 m) that receivesoverflow from the reservoir on three sides, and abuts the shore-line on the fourth side. Spillway overflow typically results fromincreases in water surface elevation caused by precipitation asthe dam is not operated to provide power, irrigation, or drink-ing water. Overflowing water is routed under an earthen damthrough two large concrete chutes before descending down aspillway outflow structure into a small creek below. The wa-tershed that feeds Lake Sam Dale is 1,754 ha and is primarilycropland (IEPA 2007). The reservoir has an established popula-tion of adult Muskellunge and receives an annual fall stockingof approximately 200–300 fingerlings.

Prior to monitoring escapement, the Muskellunge populationin Lake Sam Dale was surveyed in the spring using overnightfyke-net sets and nighttime shoreline electrofishing. Fyke netswere set at 10 different stations throughout the lake every 2 d.Fyke nets had 3.8-cm bar mesh and frames were 1.2 × 1.8 mwith six 0.75-m hoops and leads that were 15 m in length.Pulsed DC electrofishing was conducted at night with one personnetting. Electrofishing followed the shoreline and between 3 and4 km were fished each night on randomly selected transects.

Length (mm), weight (g), sex, and maturity of each fish weredetermined at the time of sampling. Sex was determined byextrusion of gametes and observation of the urogenital open-ing (Lebeau and Pageau 1989; 92–98% accuracy for juveniles,100% for adults). Maturity was determined by extrusion of ga-metes during the peak spawning season or by length at othertimes of the year, and fish > 700 mm were considered mature(most fish over this size extruded gametes). Most Muskellungein the population were of known age based on unique marksfrom liquid nitrogen freeze branding that occurred at the timeof stocking. Ages of the few fish captured without easily iden-tifiable freeze brands were determined by counting annuli onscales at the time of sampling with scales from freeze-brandedfish used as a reference set.

Muskellunge of all sizes were marked with a uniquely codedPIT tag (half duplex, 23 mm, Texas Instruments), which wasinserted into the dorsal musculature (Wagner et al. 2007; Younket al. 2010), as well as a caudal fin clip. Tagging began onFebruary 21, 2011. On all subsequent sampling dates each fishwas examined for a caudal clip to determine whether it was arecapture and then scanned with a hand scanner to record the tagcode and date of initial capture. Fish with intact caudal fins wereprocessed and implanted with a PIT tag. All fish were taggedbetween February 21 and March 25, 2011 (temperature range,7–11◦C), which was before the primary spring precipitationand overflow events. After a 10-min recovery from tagging,each fish was released into the middle portion of the reservoir.We assumed that relocation from the site of capture would not

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:47

12

Aug

ust 2

013

832 WOLTER ET AL.

affect the potential for recapture or facilitate escapement. Anyfish that was observed to be in poor condition was either nottagged or was removed from all analyses, but these instanceswere few. A Peterson mark–recapture population estimate wasconducted with the marking period occurring between February21 and March 4 (temperature range, 7–8◦C) using only fykenetting. Tagged fish were then allowed to redistribute throughoutthe population over a 9-d period. Recapture efforts using fykenets and electrofishing occurred between March 14 and March25. We used a Peterson model with the Chapman modification(Chapman 1951) to calculate population size (equation 1).

N = (M + 1)(n + 1)

(m + 1)− 1, (1)

where N is the estimated population size, M is the numberof marked fish from the first sample that were returned to thepopulation, n is the number of fish in the second sample, and mis the number of marked fish in the second sample. Based on theproportion of marked fish in the recapture sample, a binomialconfidence interval for this population estimate was calculated(Seber 1982) using equation (2).

N ± Z × SE mn, (2)

where SE is standard error, m is the number of marked fish inthe second sample, and n is the number of fish in the secondsample.

Recaptured fish were easily identified using caudal fin clipsand PIT tags. To satisfy the assumptions associated with the Pe-terson mark–recapture population estimate we used the uniquePIT tags to identify the first occurrence of recapture and onlyused these data to calculate the population estimate. Addition-ally, we were able to satisfy the assumption that the populationwas essentially closed during the time of the population esti-mate because the PIT tag antenna was actively monitoring theoutflow of the reservoir and none of the tagged fish were of legalsize for harvest. We used caudal clips as a double mark to assess1-year PIT tag retention in the population during sampling inthe spring of 2012 using the same sampling methods.

A PIT tag antenna (Zydlewski et al. 2006) was used to col-lect tag information from fish as they passed over the spillwayof the reservoir. The PIT antenna at Lake Sam Dale was in-stalled directly on the spillway to receive and record the PITtag numbers of escaping Muskellunge. The antenna consistedof a single rectangular loop (Connolly et al. 2008) of thermo-plastic high heat-resistant nylon-coated (THHN) 12-gauge wirethat spanned the width (490 cm) of the spillway outflow struc-ture and had an approximate height of 61 cm. The antennawas suspended vertically 15 cm above the concrete face of thespillway outflow structure to avoid interference from rebar andother imbedded metals (Connolly et al. 2008). The antennawas positioned on the spillway outflow at a downstream pointwhere tagged fish, once they started to pass, could not physically

return to the reservoir or reencounter the antenna. The antennawas attached to high-strength, low-stretch tech cord (EnduraBraid, New England Ropes) and wrapped in electrical tape toreduce drag during high flow. The antenna wire was connectedto a half-duplex interrogator that linked to a data logger (com-ponents from OregonRFID) and a 12-V power source. The PITtag antenna scanned for the presence of tags 10 times per secondand was activated whenever there was flow over the spillway.Tags were readily detected either when fish passed through theloop or were in close proximity to the loop including the smallgaps below and on either side of the antennae adjacent to thespillway base and walls. The coverage of the antenna was testedat installation and periodically throughout the study period bypassing tags through by hand with no gaps in antenna coveragedetected at any time.

Several variables thought to be related to escapement in-cluded Muskellunge demographics, time of day, water temper-ature, water clarity (Secchi disk depth), and precipitation. De-mographic characteristics including mean length and age werecompared between the escaped and tagged population using atwo-tailed t-test. A chi-square test was used to examine differ-ences in rates of escapement between adult males and females,and between juveniles and adults. Sunrise and sunset times wereused to classify daytime (one-half hour before sunrise) throughnighttime (one-half hour after sunset). We compared observeddaytime escapement rates with those expected if escapementwas random using a chi-square test. We expected that the otherenvironmental variables measured might have been highly cor-related (see Results) making it impossible to isolate the effectsof individual variables on the probability of Muskellunge es-capement. A temperature logger placed at 0.5 m depth at themiddle of the reservoir was used to obtain daily temperaturevalues. Daily turbidity values were interpolated from weeklySecchi disk depth measurements (centimeters) taken near thespillway. Daily precipitation (centimeters) was obtained from anearby U.S. Geological Survey weather station in Johnsonville,Illinois. The sum of precipitation for the day of escapement andthe 2 d prior was used to relate to escapement as overflow valueswere difficult to determine due to their flashy and sporadic na-ture. When possible, overflow height was characterized using astaff gauge mounted at the face of the spillway. Values for eachindividual variable were stratified by days with and without es-capement during the spring period (February 21–May 15) whenthere was a minimum of 5 cm of overflow at the dam. The influ-ence of each variable on escapement of Muskellunge was testedusing a two-tailed t-test (assuming unequal variance) compar-ing values of that variable on days when Muskellunge escapedversus days when Muskellunge did not escape. Significance forall analyses was determined at P ≤ 0.05.

Correction factor for tag detection rate.—Determining thedetection efficiency of a remote sensing array is important whenthe objectives are to obtain population estimates or informa-tion on survival and movement (Horton et al. 2007). Correctionfactors can be developed to estimate total fish passage at an

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:47

12

Aug

ust 2

013


antenna when detection efficiency is less than 100%. We usedan outflow structure from an experimental pond facility to de-velop a correction factor for detection of downstream passage offish in a spillway setting from the PIT tag antenna. Height andvelocity of water passing the antennae could be adjusted usingoutflow valves from several 0.4-ha experimental ponds. Becausethey were more readily available we made the assumption thatLargemouth Bass (N = 8, 250–325 mm) would be an accept-able surrogate for Muskellunge. Largemouth Bass used in thesetrials were tagged in the dorsal musculature similar to Muskel-lunge. These fish were placed in a slack-water area above theantennae where they inevitably passed downstream carried bythe current. In most instances trial fish maintained equilibriumin the current and exhibited limited upstream bursts. High watervelocity and tag orientation relative to the field of detection aretwo important determinants of detection efficiency (Zydlewskiet al. 2006; Aymes and Rives 2009). Thirty-six trials were con-ducted at each of three outflow water velocities (25, 50, and90 cm/s, N = 108 total) to simulate variability and intensity offlow that would be present in an actual spillway. For each fishpassage, observations of body orientation in relation to the an-tennae were made and tag detection was noted. Trial fish wereallowed to rest for a brief period of time between trials and anyfish exhibiting fatigue or irregular behavior was removed fromthe experimental system and allowed to recover. Detection effi-ciency was calculated as the proportion of passages resulting ina successful tag capture during these trials. A correction factorfor the detection efficiency of our system was then calculatedas the inverse of detection efficiency. The correction factor wasthen used to estimate the total number of escaping fish in thefield (i.e., number of detected fish escaping × correction factor= estimated total number of fish escaping).

RESULTS

Laboratory ExperimentUnivariate analyses using the likelihood ratio test indicated

that single-variable models including block, fish length, flowrate, and diel period were significantly different than the modelcontaining only the intercept. The final model included the termsblock, flow rate, and diel period (Table 1). Trial fish ranged from100 to 180 mm TL (mean, 133 mm; SD, 20), but length hadonly a marginal contribution to the model (P = 0.07) and wastherefore not included in the final model. Of the main effectsexamined, diel period had the greatest effect on the probabilityof escapement (Figure 2). The odds of a fish escaping duringthe night trials were 6.25 times lower than during the day (Ta-ble 1). Flow rate also had a significant effect on the probabilityof escapement with both high and low flow rates having signifi-cantly greater odds of escapement than no flow (odds ratios: low/ none = 1.96; high / none = 2.71; linear contrasts: all P < 0.01)(Figure 2). There was no difference in the odds of escapementbetween low and high flow rates (odds ratio: high / low = 1.38;linear contrast: P = 0.20). Blocking through time accounted for a

TABLE 1. Summary of the logistic regression model built to test factors con-trolling the probability of juvenile Muskellunge escapement over a simulatedspillway. The level of each design variable is shown with its correspondingregression coefficient, 95% CI, odds ratio, and SE. The G-statistic for the like-lihood ratio test of the model was 52.20 (df = 4, P < 0.0001). The low flowvariable did not contribute significantly to the model (P = 0.21).

Regression coefficient OddsVariable (95% Wald CI) SE ratio

Intercept 0.08 (−0.34 to 0.50) 0.21Block −0.05 (−0.02 to −0.09) 0.02 0.95Diel period (night) −0.92 (−0.71 to −1.13) 0.11 0.16Flow (high) 0.44 (0.15 to 0.73) 0.15 2.71Flow (low) 0.12 (–0.17 to 0.40) 0.15 1.96

significant negative probability of Muskellunge escapement, andthe odds declined slightly with each subsequent block (Table 1).The Hosmer–Lemeshow test produced a chi-square statistic of12.28 (P = 0.14, df = 8), indicating a good fit of the model. Thecorrect cross-validation classification of Muskellunge escape-ment was 68% at a probability level of 0.5. The kappa statisticwas 0.38 (95% CI = 0.30–0.47, P < 0.001), indicating that ourmodel correctly classified Muskellunge escapement at a rate38% greater than that by chance.

Field ExperimentTagged fish (N = 118) ranged in length from 415 to 964 mm

and were composed of 16 age-1, 15 age-2, 53 age-3, and 34age-4 individuals. Muskellunge were recaptured with fyke nets(N = 22 recaptures) and electrofishing gear (N = 2 recaptures)after receiving PIT tags and a caudal fin clip. Based on individ-ual PIT tags we determined that no fish were recaptured morethan once after tagging. Mark–recapture methods estimated 186(95% CI = 142–257) Muskellunge were present in the reservoirat the time of sampling. One tagged fish was discovered deadand two fish escaped during the mark–recapture period and wereaccounted for in calculations of population size and escapementestimates. Long-term (1 year) tag retention was 100% (N = 10),similar to rates described in the literature (Younk et al. 2010).

In the detection efficiency experiments, 50% of fish had theaxis of their body oriented at a 70–90◦ angle from the antenna(swimming parallel to the flow either upstream or downstream).Fish passing at this orientation had an associated 86% detec-tion efficiency. Another 31% of fish passed the antennae at anorientation of 21–69◦ with an associated detection efficiency of81%. Only 19% of fish passed with their body axis oriented at a0–20◦ angle, and as anticipated based on limitations of the tech-nology, the detection efficiency of the antenna was lower (71%)for these fish. Detection efficiency actually increased with ve-locity (72% at 25 cm/s, 82% at 50 cm/s, and 92% at 90 cm/s),which can be attributed to fish being more likely to have a bodyorientation that was parallel to the direction of flow as velocityincreased. We estimated an 81.6% overall detection efficiency

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:47

12

Aug

ust 2

013

834 WOLTER ET AL.

A

Day Night

Esc

apem

ent

%

0

20

40

60

80

100B

Turbid Clear

C

None Low High

Esc

apem

ent

%

0

20

40

60

80

100D

Present Absent

z

Diel Period Turbidity

Flow Rate Habitat Availability

y

z

yy

FIGURE 2. Rates of escapement of Muskellunge from a simulated dam and spillway under varying (A) diel period, (B) turbidity, (C) flow rate, and (D) habitatavailability. Turbidity levels were none (0 NTU) and moderate (15–30 NTU). Flow rates were none (0 cm/s), low (4 cm/s), and high (6 cm/s). Diel periods wereday and night, and habitat availability was presence or absence of simulated vegetation. Significant differences between levels were assessed using linear contrasts.Significance was determined at P ≤ 0.05 and significantly different statistical groupings are denoted with different letters.

of fish passing downstream across varying water velocities. Bydividing the probability of complete capture success (100%)by the detection efficiency determined in our trials (81.6%) weobtained a correction factor of 1.23.

The PIT tag antenna and data logger were activated on Febru-ary 22 when flow first passed over the spillway. In the spring, 24individual tags were detected by the antennae between March10 and May 3 (Figure 3). The actual number of tags detectedaccounts for 20.3% of the tagged population. By applying thecorrection factor for antenna efficiency we estimate escapementof the tagged population at 25.0% (i.e., 20.3% × 1.23). By ap-plying this rate to the estimated population size we estimate that47 (95% CI = 36–64) Muskellunge escaped from Lake SamDale during the spring of 2011.

The mean length and age of escaping fish (811 ± 32 mm,3.3 ± 0.25 years) were significantly higher than those for thetagged population as a whole (744 ± 26 mm, 2.9 ± 0.17 years;P = 0.03 and 0.04, respectively; Figure 4). None of the taggedage-1 fish (N = 16, 400–450 mm) were detected escaping the

reservoir (Figure 4), and there was a disproportionately higherescapement of adults compared with juvenile fish (chi-square =4.22, P = 0.04). The sex ratio of escaping fish (11 females :13 males) was similar to the ratio of the tagged population as awhole (53 females : 49 males; chi-square = 0.04, P ≥ 0.05).

Precipitation events in the area typically resulted in an in-crease in overflow at the spillway within 24 h, but the durationof overflow varied. Duration and maximum height of overflowwere variable and presumably related to rainfall intensity, dura-tion, ground saturation, and delayed runoff from previous events.From late February to mid-May there was an almost continu-ous baseline flow of water over the spillway (∼5 cm overflowheight) between pulses from specific precipitation events. InMarch, two fish escaped on days that were not associated with aspecific precipitation event (cumulative precipitation < 0.1 cmfor 3 d prior; Figure 3). The majority of escapement (22 of 24fish) followed two events in early and late April that had 2 and5.5 cm of daily rainfall, respectively, at their peak (Figure 3). Ex-act peak overflow heights were difficult to determine, but these

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:47

12

Aug

ust 2

013


Date

02/20/11 03/06/11 03/20/11 04/03/11 04/17/11 05/01/11 05/15/11

Esc

apin

g F

ish

(N)

0

2

4

6

8

Pre

cipi

tati

on (

cm)

0

1

2

3

4

5

6

7

Tem

pera

ture

(C

)

4

8

12

16

20

24

28

FIGURE 3. Daily precipitation (solid line) and water temperature (dottedline) in the spring and fall of 2011 at Lake Sam Dale, Illinois (top panel). Dailynumber of fish escaping over the dam is shown as vertical bars (bottom panel).Escapement was determined by tag detections on a PIT tag antennae coveringthe lower portion of the spillway. Date is given as month/day/year.

two precipitation events led to >13 and >25 cm of overflowheight, respectively.

The water level of Lake Sam Dale dropped several centime-ters below normal pool level during the summer of 2011 due toevaporative processes, which resulted in no summer days withspillway overflow. Precipitation throughout the fall gradually

Length Class (mm)

400-

450

451-

500

501-

550

551-

600

601-

650

651-

700

701-

750

751-

800

801-

850

851-

900

901-

950

951-

1000

Pro

port

ion

0

5

10

15

20

25

30

Tagged PopulationEscaped Fish

FIGURE 4. Length frequency distribution of the tagged and escaped portionof the Muskellunge population in Lake Sam Dale, Illinois, in 2011.

Time of Day00

00-02

00

0200

-0400

0400

-0600

0600

-0800

0800

-1000

1000

-1200

1200

-1400

1400

-1600

1600

-1800

1800

-2000

2000

-2200

2200

-2400

Num

ber o

f Esc

apin

g Fi

sh

0

1

2

3

4

5

FIGURE 5. Ordinal timing of escapement for Muskellunge leaving Lake SamDale, Illinois, in the spring of 2011. Escapement timing was determined by firstdetection of PIT tags by an antenna covering the spillway below the dam. Timeof day is given in 2-h intervals beginning at midnight (0000 hours).

raised the water level until late November when several daysof overflow occurred. A single precipitation event of >4 cmresulted in an overflow height of >15 cm. However, during thisperiod no tagged Muskellunge escaped (Figure 3). As such, allMuskellunge escapement detected by the antenna occurred nearwhat has been observed as the spawning season in Tennessee(Parsons 1959).

A majority of fish escaped during daylight hours (19 of 24),with peak escapement happening in the afternoon and evening(Figure 5). The observed numbers of escaping fish during day-light hours was significantly higher than that expected if es-capement occurred randomly throughout the diel cycle (chi-square = 5.12, P = 0.03). Water clarity (Secchi disk depth)values were highly correlated to precipitation (Pearson corre-lation coefficient r = 0.34, P < 0.01) and temperature (r =0.39, P < 0.01), whereas precipitation and temperature valueswere marginally correlated to one another (r = 0.20, P = 0.06).Because of the collinearity of these variables it is difficult todetermine the influence of each variable independently. Escape-ment of Muskellunge occurred on 11 d in the spring and noescapement was observed on 73 d. Mean Secchi disk depth wassignificantly lower on days when escapement of Muskellungeoccurred (0.32 cm) than on days when escapement did not occur(0.44 cm; P < 0.01). Similarly, daily precipitation values werehigher (3.9 cm for that day and 2 d prior) when escapementoccurred than when escapement did not occur (1.0 cm; P <

0.01). Finally, mean daily temperature was higher on days whenescapement occurred (15.8◦C) than on days when escapementdid not occur (13.1◦C; P = 0.04).

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:47

12

Aug

ust 2

013

836 WOLTER ET AL.

DISCUSSIONIn the laboratory, we used small, sexually immature Muskel-

lunge due to the size of the experimental system. Even thoughwe did not observe high rates of escapement of immature fish inthe field, the laboratory experiment should still provide relativeassessment of the factors most important in determining escape-ment. We found greater escapement of Muskellunge during theday than at night in both the laboratory and field. We expectedthat if fish were being passively swept over dams in high flowsduring periods of low activity, then escapement would be higherat night. In our laboratory experiment, escapement was signifi-cantly higher during the day, and 80% of escapement in the fieldexperiment was observed during daylight hours with the high-est levels of escapement occurring in the morning and evening.Studies of radio-tagged Muskellunge have found they are ac-tive primarily during the day, and activity peaks in the earlymorning and late evening hours (Reynolds and Casterlin 1979;Miller and Menzel 1986; Wagner and Wahl 2011), suggestingthat escapement is an active and nonrandom behavior.

Muskellunge escapement appeared to be strongly linked toprecipitation events and the resulting overflow. However, be-cause precipitation was strongly correlated to water clarity andtemperature it was difficult to separate these effects in the field.The laboratory experiment allowed us to separate and test theinfluence of these variables that are almost always correlatedin the field, and also to examine habitat effects that are logis-tically difficult to examine in the field. In the laboratory study,Muskellunge escapement was not influenced by either waterclarity or habitat manipulations. We did find a positive associa-tion to flow, suggesting that escapement is not likely a result ofrandom movement of fish that happen to encounter a spillway.Muskellunge can alter habitat selectivity in response to changesin flow, selecting shallower habitat as flow increases (Harrisonand Hadley 1979; Brenden et al. 2006), which may contribute toa higher likelihood of encountering spillways. Muskellunge ap-pear to be primarily adapted to riverine environments (Crossman1986), which could make flow a natural and important deter-mining factor in seasonal movements. The apparent behavioralassociation between flow and movement in Muskellunge maybe what makes them particularly susceptible to escapement.

Behavioral differences between juvenile and adult Muskel-lunge were evident in the field. Stocked juvenile Muskellungemovement is high during the first several weeks after stocking,but after that point fish are relatively immobile (Hanson andMargenau 1992). Similarly, juvenile Muskellunge in an Illi-nois reservoir had comparatively smaller home ranges than didadults (Wagner and Wahl 2011). We hypothesize that juvenileMuskellunge in our reservoir study did not move great distanceswhereas adults exhibiting spawning behavior did, increasing thelikelihood of spillway encounter. During summer months adultMuskellunge have a defined home range, but spawning activityoften forces movement outside of home ranges and congregatesthe population (Miller and Menzel 1986; Crossman 1990). Es-capement may be an active behavior related to spawning or

postspawning behavior (Louder 1958), a hypothesis that is sup-ported by our observations of escaping Muskellunge at LakeSam Dale. Also, Weeks and Hansen (2009) hypothesized thatstocked Muskellunge with no homing instinct move more ran-domly during the spawning season, which in reservoirs mayincrease the chance of encountering spillways in the spring. Thecoincidence of Muskellunge spawning season and high springflows from precipitation may explain the timing of escapementobserved in this study and in general why dam escapement ofMuskellunge appears to be common in many reservoirs.

Annual emigration of Muskellunge may be one of the pri-mary factors structuring reservoir populations (Wahl 1999). Wefound escapement of adult fish to be 25% of the population in asingle year. We estimated that between 36 and 64 Muskellungeescaped Lake Sam Dale in 2011. Cumulated over many yearsthese levels would account for a high proportion of populationmortality (population estimate at the time of this study was190 Muskellunge). Reductions in abundance in the reservoir re-sulting from dam escapement are obvious, but escapement ofprimarily adult fish would also affect size structure. We predictsmaller mean lengths and size distributions (percent size dis-tribution) for Muskellunge populations in reservoirs that havehigh annual reductions in adult standing stock from escapement.In contrast, escapement could be expected to annually restockdownstream systems with adult fish. In instances where ther-mal habitat and prey resources are adequate to support tailwaterfisheries, escapement may be considered beneficial (Jacobs andSwink 1983). However, in many instances downstream systemsmay not have appropriate environmental conditions to sustainadult Muskellunge through summer months if fish cannot moveto other habitats.

High flow rates and tag orientation can compromise readingefficiency of PIT tag antennas (Connolly et al. 2008), whichhas implications for work on spillways where flow is flashy andoften intense, and fish movements are largely unpredictable. Be-cause outflow was typically extremely turbid during our study,observations of fish orientation and passing speed could notbe made in the field, and thus detection efficiency during highflow is unknown. To calculate detection efficiency of our PITtag system we used Largemouth Bass as a surrogate speciesfor Muskellunge. Orientation to flow could vary among speciesbased on body morphology, but other studies have found an-tennae efficiency for a range of fish species (70–100%) to besimilar to those we found (82%) for downstream-passing fish(Connolly et al. 2008; Aymes and Rives 2009). Future studiesusing similar remote sensing arrays should make identificationof detection efficiency an early priority for system development.

Possible solutions to dam escapement include barrier nets(Stober et al. 1983) or barrier bars (Powell and Spencer 1979;Schultz et al. 2003). However, physical barriers are costly andinfeasible in many scenarios where they can become compro-mised by debris and extreme flows (Plosila and White 1970).Sound, light, and bubble barriers can be effective at limitingmovement of other species (Patrick et al. 1985) and should be

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:47

12

Aug

ust 2

013


evaluated as a method to limit dam escapement of Muskellungeand other reservoir sport fish. A PIT tag detection system, suchas the one used in this study, can be employed to conduct suchevaluations in the field. The data generated by our study can alsobe used to develop strategies to mitigate escapement. Adjust-ment of stocking rates may be used to compensate for numbersof escaping fish, but the effects of stocking rate and populationsize on escapement is unclear and would be factors to considerin future evaluations. In systems where water discharge is regu-lated by gates or other mechanisms, peak times for escapement(both seasonally and daily) can be identified and avoided bymodifying dam discharge schedules (Jacobs and Swink 1983).We found that 80% of Muskellunge escapement happened dur-ing daylight hours and similar patterns have been found forLargemouth Bass (Lewis et al. 1968). In situations with flex-ibility in the timing of water discharges, preferential releasesof water at night may limit escapement rates of Muskellunge.However, other species that are active at night, such as catfishes,may have increased rates of escapement at these times (Lewiset al. 1968). Our laboratory study indicates that habitat manip-ulations may not be effective in limiting Muskellunge interac-tions with spillways, but this should be evaluated in the field.Estimating dam escapement from reservoirs using a PIT tag de-tecting antenna can aid in understanding population dynamics ofboth reservoir and tailwater fish populations, determining mag-nitude of effects of escaped fish on downstream populations,coordinating fish rescues from tailwater areas, and evaluatingmitigation structures and strategies. Expanding knowledge onfish escapement from reservoirs will be valuable in light of theoverall magnitude, persistence, and potential cost of unwantedemigration.

ACKNOWLEDGMENTSFunding for this study was provided by the Hugh C. Becker

Foundation of Muskies Inc., the Illinois Chapter of the Amer-ican Fisheries Society Larimore Student Research Grant, andfrom Federal Aid in Sportfish Restoration Act Project F-151-Radministered through the Illinois Department of Natural Re-sources (IDNR). We thank L. Dunham and S. Pallo for coordi-nating activities within the IDNR. V. Tranquilli and W. Leech(OregonRFID) provided helpful technical assistance. M. Diana,E. Geibelstien, D. Schermerhorn, C. Salzmann, J. Mulhollem, G.Gaulke, J. Dub, J. Tompkins, J. Maxwell, J. English, J. Wisher,and S. Lewandowski of the Kaskaskia and Sam Parr Biologi-cal Stations provided assistance in the field. We also thank M.Garthaus (IDNR) for use of equipment and the staff of LakeSam Dale State Park for cooperation and technical assistance.

REFERENCESAxon, J. R., and D. K. Whitehurst. 1985. Striped Bass management in lakes with

emphasis on management problems. Transactions of the American FisheriesSociety 114:8–11.

Aymes, J. C., and J. Rives. 2009. Detection efficiency of multiplexed pas-sive integrated transponder antennas is influenced by environmental condi-tions and fish swimming behaviour. Ecology of Freshwater Fish 18:507–513.

Brenden, T. O., B. R. Murphy, and E. M. Hallerman. 2006. Effect of discharge ondaytime habitat use and selection by Muskellunge in the New River, Virginia.Transactions of the American Fisheries Society 135:1546–1558.

Chapman, D. 1951. Some properties of the hypergeometric distribution withapplications to zoological sample censuses. University of California Press,Berkeley.

Chapman, D., C. Carlson, D. Weitkamp, G. Matthews, J. Stevenson, and M.Miller. 1997. Homing in Sockeye and Chinook salmon transported aroundpart of their smolt migration route in the Columbia River. North AmericanJournal of Fisheries Management 17:101–113.

Connolly, P. J., I. G. Jezorek, K. D. Martens, and E. F. Prentice. 2008. Measuringthe performance of two stationary interrogation systems for detecting down-stream and upstream movement of PIT-tagged salmonids. North AmericanJournal of Fisheries Management 28:402–417.

Connor, W. P., R. K. Steinhorst, and H. L. Burge. 2000. Forecasting survivaland passage of migratory juvenile salmonids. North American Journal ofFisheries Management 20:651–660.

Crossman, E. J. 1986. The noble Muskellunge: a review. Pages 1–13 inG. E. Hall, editor. Managing muskies: a treatise on the biology and propaga-tion of Muskellunge in North America. American Fisheries Society, SpecialPublication 15, Bethesda, Maryland.

Crossman, E. J. 1990. Reproductive homing in Muskellunge, Esoxmasquinongy. Canadian Journal of Fisheries and Aquatic Sciences 47:1803–1812.

Dombeck, M. P., B. W. Menzel, and P. N. Hinz. 1984. Muskellunge spawninghabitat and reproductive success. Transactions of the American FisheriesSociety 113:205–216.

Hanson, D. A., and T. L. Margenau. 1992. Movement, habitat selection, be-havior, and survival of stocked Muskellunge. North American Journal ofFisheries Management 12:474–483.

Harrison, E. J., and W. F. Hadley. 1979. Biology of Muskellunge (Esoxmasquinongy) in the upper Niagara River. Transactions of the AmericanFisheries Society 108:444–451.

Hergenrader, G. L., and Q. P. Bliss. 1971. The White Perch in Nebraska. Trans-actions of the American Fisheries Society 100:734–738.

Horton, G. E., T. L. Dubreuil, and B. H. Letcher. 2007. A model for estimatingpassive integrated transponder (PIT) tag antenna efficiencies for interval-specific emigration rates. Transactions of the American Fisheries Society136:1165–1176.

IEPA (Illinois Environmental Protection Agency). 2007. Skillet Fork watershedTMDL report. IEPA, Bureau of Water, IEPA/BOW/07-009, Springfield.

Jacobs, K. E., and W. D. Swink. 1983. Fish abundance and population stabilityin a reservoir tailwater and an unregulated headwater stream. North AmericanJournal of Fisheries Management 3:395–402.

Kutner, M. H., C. Nachtsheim, and J. Neter. 2004. Applied linear regressionmodels, 4th edition. McGraw-Hill, New York.

Lebeau, B., and G. Pageau. 1989. Comparative urogenital morphology and exter-nal sex determination in Muskellunge, Esox masquinongy Mitchill. CanadianJournal of Zoology 67:1053–1060.

Lewis, W. M., R. Heidinger, and M. Konikoff. 1968. Loss of fishes over thedrop box spillway of a lake. Transactions of the American Fisheries Society97:492–494.

Littell, R. C., W. W. Stroup, and R. J. Freund. 2002. SAS R© for linear models,4th edition. SAS Institute, Cary, North Carolina.

Louder, D. 1958. Escape of fish over spillways. Progressive Fish-Culturist20:38–41.

Martinez, P. J., T. E. Chart, M. A. Trammell, J. G. Wullschleger, and E. P.Bergersen. 1994. Fish species composition before and after construction ofa main stem reservoir on the White River, Colorado. Environmental Biologyof Fishes 40:227–239.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:47

12

Aug

ust 2

013

838 WOLTER ET AL.

Miller, M. L., and B. W. Menzel. 1986. Movements, homing, and home rangeof Muskellunge, Esox masquinongy, in West Okoboji Lake, Iowa. Environ-mental Biology of Fishes 16:243–255.

Navarro, J. E., and D. J. McCauley. 1993. Fish escapement from two storagereservoirs in Michigan. Rivers 4:36–47.

Paller, M. H., D. E. Fletcher, M. M. Standora, T. B. Grabowski, T. A. Jones, S. A.Dyer, and J. J. Isely. 2006. Emigration of fish from two South Carolina coolingreservoirs. North American Journal of Fisheries Management 26:976–982.

Parsons, J. W. 1959. Muskellunge in Tennessee streams. Transactions of theAmerican Fisheries Society 88:136–140.

Patrick, P. H., A. E. Christie, D. Sager, C. Hocutt, and J. Stauffer Jr. 1985.Responses of fish to a strobe light/air-bubble barrier. Fisheries Research3:157–172.

Plosila, D. S., and B. D. White. 1970. A swinging vertical screen for fish barrierdams. Progressive Fish-Culturist 32:178–179.

Powell, D. H., and S. L. Spencer. 1979. Parallel-bar barrier prevents fish lossover spillways. Progressive Fish-Culturist 41:174–175.

Quinn, G. P., and M. J. Keough. 2002. Experimental design and data analysisfor biologists. Cambridge University Press, Cambridge, UK.

Raymond, H. L. 1988. Effects of hydroelectric development and fisheries en-hancement on spring and summer Chinook Salmon and steelhead in theColumbia River basin. North American Journal of Fisheries Management8:1–24.

Reynolds, W. W., and M. E. Casterlin. 1979. Thermoregulatory rhythm in ju-venile Muskellunge (Esox masquinongy): evidence of diel shift in the lowerset-point. Comparative Biochemistry and Physiology 63A:523–525.

Schultz, A. A., O. E. Maughan, S. A. Bonar, and W. J. Matter. 2003. Effectsof flooding on abundance of native and nonnative fishes downstream froma small impoundment. North American Journal of Fisheries Management23:503–511.

Seber, G. A. F. 1982. The estimation of animal abundance and related parame-ters. Edward Arnold, London.

Spoelstra, J. A., R. A. Stein, J. A. Royle, and E. A. Marschall. 2008. Movementof reservoir-stocked riverine fish between tailwaters and rivers. Transactionsof the American Fisheries Society 137:1530–1542.

Stober, Q. J., R. W. Tyler, and C. E. Petrosky. 1983. Barrier net to reduceentrainment losses of adult kokanee from Banks Lake, Washington. NorthAmerican Journal of Fisheries Management 3:331–354.

Storck, T. W., and D. L. Newman. 1992. Contribution of Tiger Muskel-lunge to the sport fishery of a small, centrarchid-dominated im-poundment. North American Journal of Fisheries Management 12:213–221.

Titus, K., J. A. Mosher, and B. K. Williams. 1984. Chance-corrected classi-fication for use in discriminant analysis: ecological applications. AmericanMidland Naturalist 111:1–7.

Trammell, M. A., E. P. Bergersen, and P. J. Martinez. 1993. Evaluationof an introduction of Colorado Squawfish in a main stem impound-ment on the White River, Colorado. Southwestern Naturalist 38:362–369.

Wagner, C. P., M. J. Jennings, J. M. Kampa, and D. H. Wahl. 2007. Survival,growth, and tag retention in age-0 Muskellunge implanted with passive in-tegrated transponders. North American Journal of Fisheries Management27:873–877.

Wagner, C. P., and D. H. Wahl. 2011. Movement, home range and habitat se-lection of stocked juvenile Muskellunge, Esox masquinongy, in Forbes Lake,Illinois: exploring the effects of latitudinal origin. Fisheries Management andEcology 18:482–496.

Wahl, D. H. 1999. An ecological context for evaluating the factors influencingMuskellunge stocking success. North American Journal of Fisheries Man-agement 19:238–248.

Weeks, J. G., and M. J. Hansen. 2009. Walleye and Muskellunge movement inthe Manitowish Chain of lakes, Vilas County, Wisconsin. North AmericanJournal of Fisheries Management 29:791–804.

Younk, J. A., B. R. Herwig, and B. J. Pittman. 2010. Short- and long-termevaluation of passive integrated transponder and visible implant elastomertag performance in Muskellunge. North American Journal of Fisheries Man-agement 30:281–288.

Zydlewski, G. B., G. Horton, T. Dubreuil, B. Letcher, S. Casey, and J. Zydlewski.2006. Remote monitoring of fish in small streams: a unified approach usingPIT tags. Fisheries 31:492–502.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:47

12

Aug

ust 2

013



Evaluation of Aging Structures for Silver Carp fromMidwestern U.S. RiversJustin R. Seibert a b & Quinton E. Phelps aa Missouri Department of Conservation , Big Rivers and Wetlands Field Station , 3815 EastJackson Boulevard, Jackson , Missouri , 63755 , USAb Spokane Tribal Fisheries , 6290 D Ford-Wellpinit Road, Wellpinit , Washington , 99040 , USAPublished online: 06 Aug 2013.

To cite this article: Justin R. Seibert & Quinton E. Phelps (2013) Evaluation of Aging Structures for Silver Carp fromMidwestern U.S. Rivers, North American Journal of Fisheries Management, 33:4, 839-844, DOI: 10.1080/02755947.2013.815670







http://dx.doi.org/10.1080/02755947.2013.815670




MANAGEMENT BRIEF

Evaluation of Aging Structures for Silver Carpfrom Midwestern U.S. Rivers

Justin R. Seibert*1 and Quinton E. PhelpsMissouri Department of Conservation, Big Rivers and Wetlands Field Station,3815 East Jackson Boulevard, Jackson, Missouri 63755, USA

AbstractTo combat the potential deleterious effects that Silver Carp Hy-

pophthalmichthys molitrix have on native populations, managementof this species is essential. Before developing population-level mod-els, a determination of which aging structure for estimating the ageof Silver Carp is needed. To our knowledge, no consensus has beenreached on which structure should be used for estimating SilverCarp ages. We collected 120 Silver Carp from the Illinois, Missis-sippi, Missouri, and Ohio rivers via electrofishing to evaluate agingstructures. Removal time, processing time, and discernible annuliwere evaluated for scales, opercles, vertebrae, pectoral fin rays,postcleithra, and asterisci and lapilli otoliths. Asteriscus otolith,opercle and scale annuli were difficult to discern and not evaluatedfurther. Total processing times for postcleithra (246.1 s) andlapilli (251.2 s) were the most time-efficient; pectoral fin rays andvertebrae were more time intensive. Between-reader precision andagreement rates resulted in lapilli being the most precise, followedby postcleithra, pectoral fin rays, and vertebrae. Comparisons ofstructures with lapilli revealed that pectoral fin rays exhibited 78%agreement, 49% agreement with postcleithra, and 53% agreementfor vertebrae. In terms of agreement ± 1 year to lapilli, pectoralfin ray, postcleithrum, and vertebra resulted in high agreement(>85%). Age bias plots revealed that these discrepancies consis-tently underestimated ages compared with lapilli. Discrepanciesmay be attributed to erosion of the central lumen of fin rays andpostcleithra, while locating the first annulus on vertebrae mayhave led to this disparity. Based on previous studies, evaluation ofoverall processing times, assessment of between-reader precision,between-reader agreement rates, and bias that may be involvedwith alternative structures, we recommend lapilli otoliths be usedfor estimating age of Silver Carp. Future efforts should focus onvalidating accuracy of lapilli for estimating Silver Carp ages.

To combat the potential deleterious effects that Silver CarpHypophthalmichthys molitrix have on native fishes, manage-ment of this species is essential (Kolar et al. 2005; Irons et al.2007; Conover et al. 2007). Developing population-level models

*Corresponding author: [email protected] address: Spokane Tribal Fisheries, 6290 D Ford-Wellpinit Road, Wellpinit, Washington 99040, USA.Received March 1, 2013; accepted June 5, 2013

that enhance our ability to reduce or eradicate stocks can min-imize their effects on native species (Sakai 2001). To developthese models, a thorough understanding of their demographics(e.g., age, growth, and mortality) is needed. Accurate ages arerequired for age-structure analysis, growth analysis, and mor-tality rate estimation, all of which are key parameters in popu-lation modeling (Campana 2001). The effects of inaccurate ageestimates cannot be overstated (Bradford 1991; Richards et al.1992; Morison et al. 1998); models built on erroneous age dataresult in overly optimistic estimates of population dynamics,particularly when age is underestimated (Campana 2001).

Despite the need for accurately estimated ages, researchersfocus on using time-efficient structures while still being ableto acquire adequate representations of population demograph-ics (Isermann et al. 2003; Maceina et al. 2007). Few studieshave formally evaluated the amount of time it takes to removeand process different structures used to estimate age (Isermannet al. 2003). Many studies have offered anecdotal information onremoval and processing times and have requested further eval-uation of effort (Boxrucker 1986; Welch et al. 1993; Kocovskyand Carline 2000; Buckmeier et al. 2002; Koch and Quist 2007;Stolarski and Hartman 2008). As such, removal and process-ing times are undoubtedly important components of estimatingages, especially in cases where accuracy and precision may besacrificed.

Despite the necessity for an efficient structure that exhibitsan accurate and precise age estimate, no consensus has beenformed for Silver Carp. Shefler and Reich (1977) used scalesto estimate their age for growth within Lake Kinnerect, Israel.Kamilov (1984) established that the first ray of the pectoral fins,the pterygiophore of the first ray of the dorsal fins, and vertebraewere suitable for aging, whereas scales, opercles, and otolithswere not. However, the structures were not compared within that

839

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:48

12

Aug

ust 2

013

840 SEIBERT AND PHELPS

study. Other structures such as the postcleithra have also beenreported suitable for aging (Johal et al. 2000). More recently,Williamson and Garvey (2005) used pectoral fin rays to esti-mate Silver Carp ages. The central theme of these demographicstudies suggests multiple structures are used for estimating agesand that there is a need for a comparative study.

Because agencies may have concern with the times associ-ated with aging structures, our first objective was to compareremoval and processing times associated with several structures.Due to a lack of consensus and because there has been no formalcomparative evaluation of Silver Carp structures, we evaluatedscales, opercles, vertebrae, pectoral fin rays, postcleithra, andasterisci and lapilli otoliths, all commonly used for estimatingages of freshwater fishes. Our second objective was to deter-mine if these structures contained discernible annuli. We hadno known-age Silver Carp, so validation of age was impossible.Thus, we determined between-reader precision of each struc-ture. Lastly, we tested for any potential bias associated withthese structures.

METHODSFish collection.—During fall 2011, 120 Silver Carp were col-

lected from the Illinois; Missouri; Ohio; and upper, middle, andlower Mississippi rivers (20 fish from each river or section) us-ing daytime electrofishing. Total length (mm) and weight (g) ofeach fish were measured; scales, opercles, vertebrae, pectoral finrays, postcleithra, and asterisci and lapilli otoliths were removedand placed in appropriate containers.

Removal and processing of aging structures.—Scales wereremoved from an area directly behind the tip of the pectoral finand above the lateral line (Shefler and Reich 1977). Scales wereplaced in coin envelopes, air dried, and 10 indiscriminatelyselected scales were subsequently pressed onto acetate slidesusing a roller press. Scale impressions on acetate slides wereevaluated under a microfiche reader.

The right opercle and opercular assembly was removed fromeach fish. Skin was removed from all sides. Opercles were placedin individually marked bags and frozen to avoid decompositionor desiccation. Prior to counting annuli, opercles were boiled forapproximately 1 min to remove any excess tissue. Occasionally,a bristled brush was used to scrub off excess soft tissue. Opercleswere then viewed under transmitted fluorescent light with thenaked eye as described by Phelps et al. (2007).

The first hard ray of the pectoral fin was removed from the leftside of every fish when possible; the right side was used when notpossible. Tissue around the pectoral fin rays was removed andthree 0.8-mm-thick subsections were removed from the anteriorportion of the pectoral fin ray with a Buehler Isomet low-speedsaw and then secured on microscope slides. Sections were ex-amined under a dissecting microscope (10–40 × magnification)with transmitted light (as described by Phelps et al. 2007).

The first vertebra of each fish was extracted (methods ofDeters et al. 2011) and placed in boiling water to remove

tissue. Vertebrae were subsequently air dried and placed into125-mL plastic bottles filled with 2% sodium hypochlorite so-lution, sealed, shaken, allowed to soak for 1 h, cleaned, soakedfor 15–20 min in distilled water, air dried for 24 h, and then readwith a dissecting microscope (4–10 × magnification).

The postcleithrum was removed by dissecting tissue alongeach side of this structure. Once extracted any excess muscle tis-sue was removed with a scalpel. Three transverse sections werethen taken from the middle of the postcleithrum using a fine jew-eler’s saw (Johal et al. 2000). These sections were ground andpolished using a carborundum stone and fine-ground glass to athickness of 0.3–0.5 mm using water as a lubricant and mountedon glass slides in DPX glue to be viewed under a dissecting mi-croscope (10–40 × magnification) with transmitted light (Johalet al. 2000).

Asterisci otoliths were removed from the lagena ventrally (asper Secor et al. 1991). Asterisci were cleaned, air dried, mountedin clear epoxy, and transversely sectioned along the dorsoventralplane through the nucleus with a low-speed saw. To ensure thatthe nucleus was included, several 0.4-mm-thick sections withinthis area were cut. Sections of asterisci were examined under adissecting microscope (10–40 × magnification) with transmit-ted light as documented for Common Carp Cyprinus carpio(Brown et al. 2004).

Lapilli otoliths were removed by sectioning through thesupraoccipital bone using a hacksaw. The cut was made in linewith the gap between the preopercle and the opercle. Lapilliwere removed with forceps from the posterior portion of theskull. These removal procedures were similar to lapilli removalin Channel Catfish Ictalurus punctatus as explained by Buck-meier et al. (2002) and Long and Stewart (2010). Specifically,we found that using forceps to grip lapilli instead of epoxyingthem to a slide provided a more efficient method. Lapilli weresanded with 400-grit wetted sandpaper from the anterior sidein order to get close to the nucleus. At that point, fine wettedsandpaper (1,000-grit) was used to reach the nucleus. The lapil-lus’s reading surface was burnt golden brown with a candle.Next, the lapillus was placed posterior side down in putty and adrop of immersion oil was applied. Annuli were viewed undera dissecting microscope (4–10 × ) with a fiber optic filament(1-mm-diameter tip) connected to a light source. The movablefiber optic filament facilitated identification of all annular rings.

For each structure, we recorded the time required to removethe structure (hereafter referred to as removal time). We alsorecorded the time required to process each structure for estimat-ing age (hereafter referred to as processing time). For removaland processing times, we used times from two individuals tosimulate variability due to differences in skill level among per-sonnel (Isermann et al. 2003). Overall processing times werecalculated by summing the times required to remove and pro-cess each individual structure for each fish. In order to estimateage for structures, annuli were recorded independently by twoexperienced readers that had no knowledge of fish length, esti-mated age of other structures, or source river. The age estimated

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:48

12

Aug

ust 2

013


by each reader was recorded and later analyzed for between-reader precision. If age estimates differed between readers foran individual structure, both readers viewed the structure to-gether until a consensus was reached. If consensus betweenreaders could not be reached, the structure was removed fromfurther analysis.

Statistical analyses.—Differences in removal, processing,and overall processing times among aging structures were exam-ined using a one-way analysis of variance (ANOVA) and pair-wise comparisons were evaluated with Tukey’s honestly signif-icant difference (HSD) test (all pairwise procedures; α = 0.05).Following the evaluation of processing times, precision betweenreaders was determined via CV (i.e., 100 × SD/mean; Chang1982). Furthermore, we calculated between-reader agreementrates (exact agreement and agreement ± 1 year) for each struc-ture, under the supposition that structures displaying easily dis-cernible annuli would result in high agreement rates betweenreaders. To determine if the CV calculated between-readers var-ied among and between structures we used Kruskal–Wallis k-sample tests (Welch et al. 1993). Percent agreement betweenstructures was used to compare age determinations (Phelps elal. 2007). To determine if bias occurred between structures,age-bias plots were generated (Campana et al. 1995; Buckmeieret al. 2002).

RESULTSThe additional four structures (i.e., lapilli, postcleithra,

pectoral fin rays, and vertebrae) had apparent annular marks.Removal times were significantly different among structureswith discernible annuli (ANOVA: F3, 476 = 47.91, P <

0.01). Furthermore, Tukey’s HSD indicated several pairwisedifferences (Table 1). Pectoral fin rays required the shortesttime for removal (39.6 s), while lapilli (46.1 s) and postcleithra(46.0 s) showed similar results, and vertebrae required thelongest time (56.2 s; all pairwise comparisons P < 0.05).

TABLE 1. Mean removal, processing and total processing times associatedwith the use of lapilli, postcleithra, pectoral fin rays, and vertebrae to estimate theage of Silver Carp collected from the Illinois, Missouri, Mississippi, and Ohiorivers during the fall 2011 via boat electrofishing. Standard errors are reported inparentheses. Values for a given statistic with different letters indicate significantdifferences between structures (Tukey’s HSD test; all pairwise procedures, P <

0.05).

Elapsed time (s)

Mean Mean Totalremoval processing processing

Structure (SE) (SE) (SE)

Lapilli 46.1 (1.22) y 251.2 (4.77) z 297.2 (5.04) zPostcleithra 46.0 (1.30) y 246.1 (5.57) z 292.1 (5.92) zPectoral fin

rays39.6 (0.79) z 352.9 (7.54) y 389.8 (7.64) y

Vertebrae 56.2 (1.18) x 537.0 (11.23) x 593.1 (11.51) x

TABLE 2. Mean coefficients of variation between reader age assignments andreader agreement rates associated with the use of lapilli, postcliethra, pectoralfin rays, and vertebrae to age Silver Carp collected from the Illinois, Missouri,Mississippi, and Ohio rivers during the fall 2011. Mean coefficients of variationwith different letters indicate significant differences between structures (allpairwise procedures, P < 0.05).

ReaderAverage Exact reader agreement ± 1

Structure CV (SE) agreement (%) year (%)

Lapilli 4.22 (0.83) z 76 98Postcleithra 9.73 (1.27) y 58 91Pectoral fin

rays10.22 (1.56) y 57 93

Vertebrae 15.73 (2.56) y 51 85

Processing times were significantly different among structures(ANOVA: F3, 476 = 312.32, P < 0.01), and several pairwisedifferences existed. Processing times were similar betweenlapilli (251.2 s) and postcleithra (246.1 s); however, pectoralfin rays (352.9 s) and vertebrae (537.0 s) resulted in longertimes, the latter representing the greatest time (all pairwisecomparisons; P < 0.05). Overall, total processing times weresignificantly different among structures (ANOVA: F3, 476 =315.14, P < 0.01). Lapilli (297.2 s) and postcleithra (292.1 s)required the least time for overall processing, while pectoral finrays (389.8 s) and vertebrae (593.1 s) overall processing timeswere greater overall (all pairwise comparisons; P < 0.05).

Between-reader precision was significantly different amongstructures (Kruskal–Wallis; χ2 = 21.37, df = 3, 476, P <

0.01) and between structures (all comparisons P < 0.05) withdiscernible annuli. Lapilli resulted in higher between-readerprecision than postcleithra, pectoral fin rays, and vertebrae, asindicated by the significantly lower CV between readers (allpairwise comparisons P < 0.05; Table 2). Of estimates fromlapilli, 76% agreed exactly among readers, and 98% of theestimates were ± 1 year (Table 2). The percent exact agreementamong readers for postcleithra, pectoral fin rays, and vertebraewas less than 60%, but 85–91% of the estimates agreed within1 year (Table 2).

Lapilli resulted in an estimated age range of 2–10 years,pectoral fin rays ranging 2–9 years, postcleithra 2–8 years, andvertebra 2–7 years. Because of the significantly lower between-reader precision and the potential drawbacks of pectoral fin rays,postcleithra, and vertebrae, all between-structure age compar-isons were evaluated using lapilli as the primary age determinant(Table 2; Figure 1). These comparisons revealed that lapilli ex-hibited 78% agreement with pectoral fin rays, 49% agreementwith postcleithra, and 53% agreement with vertebrae. In termsof agreement ± 1 year to lapilli, percentages were high for pec-toral fin rays (97%), postcleithrum (91%), and vertebra (87%).However, age-bias plots for all structures revealed that these

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:48

12

Aug

ust 2

013


FIGURE 1. Aging structures of a Silver Carp (#106) collected via boat electrofishing from the Illinois River during the fall 2011, showing (A) lapillus and 1-yeardisagreements (i.e., underestimate) between lapillus and the alternative aging structures: (B) pectoral fin ray, (C) postcleithrum, and (D)vertebra. N = nucleus ofthe aging structure, X = individual annuli on the aging structure. Note erosion of the central lumen for the pectoral fin ray and postcleithrum, and the missing firstannuli on the vertebra.

discrepancies consistently had lower estimates of Silver Carpage than did lapilli (Figure 2).

DISCUSSIONOtoliths are known as valid and precise structures for accu-

rately aging multiple freshwater fish species and are typicallypreferred over scales, fin sections, and vertebrae (Heidinger andClodfelter 1987; Welch et al. 1993; Buckmeier et al. 2002;Isermann et al. 2003; Maceina et al. 2007; Phelps et al. 2007).However, agencies have continued to use alternative aging struc-tures (i.e., not otoliths) because of supposedly faster total pro-cessing times and no required sacrifice of fish (Isermann et al.2003; Maceina et al. 2007). Over the course of our evaluation,total processing times were similar for lapilli and postcleithra,but pectoral fin rays and vertebrae required significantly more

time. Other studies have also found that otoliths had similar(Buckmeier et al. 2002; Isermann et al. 2003) or faster overallprocessing times than alternate aging structures (Kocovsky andCarline 2000; Stolarski and Hartman 2008).

We found that pectoral fins rays, postcleithra, vertebrae,and lapilli contained discernible annuli. Prior to this study,Silver Carp otoliths were considered fragile and opaque, andtherefore difficult to read (Kamilov 1984; Johal et al. 2000).This corresponds to our findings for asteriscis. In contrast,annuli on lapilli were clearly distinguishable and quick toprocess (Figure 2a). Similar outcomes have been reported forvalidated Channel Catfish lapilli otoliths (Buckmeier et al.2002; Long and Stewart 2010).

Lapilli had higher between-reader precision than alterna-tive aging structures. Also between-reader agreement rates forlapilli resulted in a higher percentage of exact agreement and

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:48

12

Aug

ust 2

013


FIGURE 2. Age bias plots for pectoral fin rays, postcleithra, and vertebraeages (in years) compared with lapilli for Silver Carp collected in fall 2011from the Illinois, Missouri, Mississippi, and Ohio rivers via boat electrofishing.Dashed line indicates a 1:1 line (i.e., perfect agreement) and numbers indicatethe frequency of that observation.

agreement ± 1 year than alternative aging structures. This seemsto be a common theme within the literature, in which otoliths aremore precise than alternative aging structures (Welch et al. 1993;Isermann et al. 2003; Maceina et al. 2007; Phelps et al. 2007).

Pectoral fin rays and vertebrae tended to underestimate agemore than otoliths (Welch et al. 1993; Isermann et al. 2003;Maceina et al. 2007; Phelps et al. 2007); however, to our knowl-edge this is the first age comparison for postcleithra. Similar tothese studies, we found each of these structures underestimatedthe age of Silver Carp more than lapilli; however, the differencesin age estimates were typically ± 1 year of age. We noted that

even though discrepancies existed between lapilli and alterna-tive aging structures, the differences were typically ± 1 yearof lapilli age. However, we did not have any known-age fish inthis study. Pectoral fin rays had the higher agreement (i.e., ± 1year) than lapilli. Other studies also documented the precisionof pectoral fin rays for other fish species such as the closelyrelated Bighead Carp Hypophthalmichthys nobilis (Nuevo et al.2004) and Common Carp (Phelps et al. 2007). Fin rays exhib-ited high agreement ( ± 1 year) with lapilli. However, age wastypically underestimated with fin rays, perhaps due to erosionof the central lumen, thereby confounding identification of thefirst annulus in the fin section (Figure 1b; Muncy 1959; Mayhew1969; Isermann et al. 2003).

To our knowledge no studies have compared postcleithrawith other aging structures for estimating age of Silver Carp. Asimple evaluation was completed by Johal et al. (2000), whichnoted discernible rings using postcleithrum. In our study, post-cleithrum also showed high agreement with lapilli (i.e., ± 1year) but typically displayed underestimated ages more thanlapilli. This discrepancy of 1 year could possibly be explainedby erosion of the central lumen (similar to pectoral fin rays),which was documented for the postcleithrum in this study (Fig-ure 1c). Vertebrae age discrepancies (compared to lapilli) mayhave been a result of difficulty locating the first annuli (Figure1d). Similarly, two independent fisheries studies noted discrep-ancies over the identification of the first annulus when usingvertebrae (Francis and Stevens 2000; Natanson et al. 2002).Difficulties in locating the first annulus for alternative structuresmay be compounded due to the protracted spawning exhibitedby Silver Carp, potentially leading to even more difficulty inlocating the first annulus (Lohmeyer and Garvey 2009).

To our knowledge, this study is the first to use lapilli forestimating age of Silver Carp. Alternatively, if sacrifice of thisspecies is undesired, we recommend that pectoral fin rays shouldsuffice based on fairly high agreement with lapilli, especially± 1 year (Sikstrom 1983; Nuevo et al. 2004; Phelps et al. 2007).However, caution should be taken to ensure proper managementdecisions using pectoral fin rays. Similar discrepancies existedbetween lapilli and the lethal alternative aging structures (i.e.,postcleithra and vertebrae). However, the accuracy of lapilliotoliths for the use of estimating ages of Silver Carp is yet to bevalidated. Nevertheless, taking into account previous fish-agingstudies, evaluation of overall processing times, assessment ofbetween-reader precision, between-reader agreement rates, andbias involved with alternative structures, we recommend thatlapilli otoliths should be used to estimate age of Silver Carp.

ACKNOWLEDGMENTSWe would like to thank Kasey Yallaly, Ron Brooks, Sara

Tripp, Nick Keeton, Chris Hickey, Paul Rister, Neil Jackson,Ryan Kausing, Levi Solomon, Andrew Friedunk, and NathanRedecker for field collection assistance of Silver Carp. Alsowe would like to thank Kasey Yallaly and Andrew Niebuhr for

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:48

12

Aug

ust 2

013


removal and laboratory assistance of aging structures. We thankDave Herzog for support in completing this project. Fundingfor this project was provided by the Missouri Department ofConservation, U.S. Geological Survey, and the U.S. Army Corpsof Engineers.

REFERENCESBoxrucker, J. 1986. A comparison of the otolith and scale methods for aging

White Crappies in Oklahoma. North American Journal of Fisheries Manage-ment 6:122–125.

Bradford, M. J. 1991. Effects of ageing errors on recruitment time series es-timated from sequential population analysis. Canadian Journal of Fisheriesand Aquatic Sciences 48:555–558.

Brown, P., C. Green, K. P. Sivakumaran, D. Stoessel, and A. Giles. 2004.Validating otolith annuli for annual age determination of Common Carp.Transactions of the American Fisheries Society 133:190–196.

Buckmeier, D. L., E. R. Irwin, R. K. Betsill, and J. A. Prentice. 2002. Validityof otoliths and pectoral spines for estimating ages of Channel Catfish. NorthAmerican Journal of Fisheries Management 22:934–942.

Campana, S. E. 2001. Accuracy, precision and quality control in age determi-nation, including a review of the use and abuse of age validation methods.Journal of Fish Biology 59:197–242.



Conover, G., R. Simmonds, and M. Whalen, editors. 2007. Management andcontrol plan for Bighead, Black, Grass, and Silver carps in the United States.Asian Carp Working Group, Aquatic Nuisance Species Task Force, Wash-ington, D.C. Available: www.asiancarp.org/Documents/Carps ManagementPlan.pdf. (April 2010).

Deters, J. E. 2011. Removal and processing of Bigheaded Carp (Hypoph-thalmichthys spp.) vertebrae for use in age and growth studies. U.S. Fishand Wildlife Service, Final Report, Washington, D.C.

Francis, M. P., and J. D. Stevens. 2000. Reproduction, embryonic development,and growth of the Porbeagle Shark, Lamna nasus, in the southwest PacificOcean. U.S. National Marine Fisheries Service Fishery Bulletin 98:41–63.

Heidinger, R. C., and K. Clodfelter. 1987. Validity of the otolith for determiningage and growth of Walleye, Striped Bass, and Smallmouth Bass in power plantcooling ponds. Pages 241–251 in R. C. Summerfelt and G. E. Hall, editors.Age and growth of fish. Iowa State University Press, Ames.

Irons, K. S., G. G. Sass, M. A. McClelland, and J. D. Stafford. 2007. Reducedcondition factor of two native fish species coincident with invasion of non-native Asian carps in the Illinois River, U.S.A.: is this evidence for competitionand reduced fitness? Journal of Fish Biology 71(Supplement D):258–273.

Isermann, D. A., J. R. Meerbeek, G. D. Scholten, and D. W. Willis. 2003.Evaluation of three different structures used for Walleye age estimation withemphasis on removal and processing times. North American Journal of Fish-eries Management 23:625–631.

Johal, M. S., H. R. Esmaeili, and K. K. Tandon. 2000. Postcleithrum of SilverCarp, Hypophthalmichthys molitrix (Val. 1844), an authentic indicator forage determination. Current Science 79:945–946.

Kamilov, B. G. 1984. Morphology of growth structures in Silver Carp Hypoph-thalmichthys molitrix, in relation to estimation of age and growth rate. Journalof Ichthyology 25:49–59.

Koch, J. D., and M. C. Quist. 2007. A technique for preparing fin rays andspines for age and growth analysis. North American Journal of FisheriesManagement 27:782–784.

Kocovsky, P. M., and R. F. Carline. 2000. A comparison of methods for esti-mating ages of unexploited Walleyes. North American Journal of FisheriesManagement 20:1044–1048.

Kolar, C. S., D. C. Chapman, W. R. Courtenay Jr., C. M. Housel, J. D. Williams,and D. P. Jennings. 2005. Asian carps of the genus Hypophthalmichthys(Pisces, Cyprinidae)—a biological synopsis and environmental risk assess-ment. U.S. Fish and Wildlife Service, Washington, D.C.

Lohmeyer, A. M., and J. E. Garvey. 2009. Placing the North American invasionof Asian carp in a spatially explicit context. Biological Invasions 11:905–916.

Long, J. M., and D. R. Stewart. 2010. Verification of otolith identity used byfisheries scientists for aging Channel Catfish. Transactions of the AmericanFisheries Society 139:1775–1779.

Maceina, M. J., J. Boxrucker, D. L. Buckmeier, R. S. Gangl, D. O. Lucchesi,D. A. Isermann, J. R. Jackson, and P. J. Martinez. 2007. Current status andreview of freshwater fish aging procedures used by state and provincial fish-eries agencies with recommendations for future directions. Fisheries 32:329–340.

Mayhew, J. K. 1969. Age and growth of Flathead Catfish in the Des MoinesRiver, Iowa. Transactions of the American Fisheries Society 98:118–121.

Morison, A. K., P. C. Coutin, and S. G. Robertson. 1998. Age determination ofBlack Bream, Acanthopagrus butcheri (Sparidae), from the Gippsland Lakesof south-eastern Australia indicates slow growth and episodic recruitment.Marine and Freshwater Research 49:491–498.

Muncy, R. J. 1959. Age and growth of Channel Catfish from the Des MoinesRiver, Boone County, Iowa, 1955 and 1956. Iowa State Journal of Science34:127–137.

Natanson, L. J., J. J. Mello, and S. E. Campana. 2002. Validated age and growthof the Porbeagle Shark (Lamna nasus) in the western North Atlantic Ocean.U.S. National Marine Fisheries Service Fishery Bulletin 100:266–278.

Nuevo, M., R. J. Sheehan, and R. C. Heidinger. 2004. Accuracy and precision ofage determination techniques for Mississippi River Bighead Carp Hypoph-thalmichthys nobilis (Richardson 1845) using pectoral spines and scales.Archiv fur Hydrobiologie 160:45–56.

Phelps, Q. E., K. R. Edwards, and D. W. Willis. 2007. Precision of five structuresfor estimating age of Common Carp. North American Journal of FisheriesManagement 27:103–105.

Richards, L. J., J. T. Schnute, A. R. Kronlund, and R. J. Beamish. 1992. Statisticalmodels for the analysis of ageing error. Canadian Journal of Fisheries andAquatic Sciences 49:1801–1815.

Sakai, A. K., F. W. Allendorf, J. S. Holt, D. M. Lodge, J. Molofsky, K. A. With,S. Baughman, R. J. Cabin, J. E. Cohen, N. C. Ellstrand, D. E. McCauley,P. O’Neil, I. M. Parker, J. N. Thompson, and S. G. Weller. 2001. The popula-tion biology of invasive species. Annual Review of Ecology and Systematics32:305–332.

Secor, D. H., J. M. Dean, and E. H. Laban. 1991. Manual for otolith removal andpreparation for microstructural examination. Electric Power Research Insti-tute and Belle W. Baruch Institute for Marine Biology and Coastal Research,Columbia, South Carolina.

Shefler, D., and K. Reich. 1977. Growth of Silver Carp (Hypophthalmichthysmolitrix) in Lake Kinneret, Israel, in 1969–1975. Bamidgeh 29(1):3–16.

Sikstrom, C. B. 1983. Otolith, pectoral fin ray, and scale age determinations forArctic Grayling. Progressive Fish-Culturist 45:220–223.

Stolarski, J. T., and K. J. Hartman. 2008. An evaluation of the precision of finray, otolith, and scale age determinations for Brook Trout. North AmericanJournal of Fisheries Management 28:1790–1795.

Welch, T. J., M. J. Van Den Avyle, R. K. Betsill, and E. M. Driebe. 1993.Precision and relative accuracy of Striped Bass age estimates from otoliths,scales, and anal fin rays and spines. North American Journal of FisheriesManagement 13:616–620.

Williamson, C. J., and J. E. Garvey. 2005. Growth, fecundity, and diets of newlyestablished Silver Carp in the middle Mississippi River. Transactions of theAmerican Fisheries Society 134:1423–1430.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:48

12

Aug

ust 2

013



An Evaluation of Harvest Control Rules for Data-PoorFisheriesJohn Wiedenmann a b , Michael J. Wilberg a & Thomas J. Miller aa Chesapeake Biological Laboratory, University of Maryland Center for EnvironmentalSciences, Post Office Box 38 , Solomons , Maryland , 20688 , USAb Institute of Marine and Coastal Sciences, Rutgers, State University of New Jersey, 71Dudley Road , New Brunswick , New Jersey , 08901 , USAPublished online: 08 Aug 2013.

To cite this article: John Wiedenmann , Michael J. Wilberg & Thomas J. Miller (2013) An Evaluation of Harvest Control Rulesfor Data-Poor Fisheries, North American Journal of Fisheries Management, 33:4, 845-860, DOI: 10.1080/02755947.2013.811128







http://dx.doi.org/10.1080/02755947.2013.811128




ARTICLE

An Evaluation of Harvest Control Rulesfor Data-Poor Fisheries

John Wiedenmann,*1 Michael J. Wilberg, and Thomas J. MillerChesapeake Biological Laboratory, University of Maryland Center for Environmental Sciences,Post Office Box 38, Solomons, Maryland 20688, USA

AbstractFor federally managed fisheries in the USA, National Standard 1 requires that an acceptable biological catch be

set for all fisheries and that this catch avoid overfishing. Achieving this goal for data-poor stocks, for which stockassessments are not possible, is particularly challenging. A number of harvest control rules have very recently beendeveloped to set sustainable catches in data-poor fisheries, but the ability of most of these rules to avoid overfishinghas not been tested. We conducted a management strategy evaluation to assess several control rules proposed fordata-poor situations. We examined three general life histories (“slow,” “medium,” and “fast”) and three exploitationhistories (under-, fully, and overexploited) to identify control rules that balance the competing objectives of avoidingoverfishing and maintaining high levels of harvest. Many of the control rules require information on species lifehistory and relative abundance, so we explored a scenario in which unbiased knowledge was used in the control ruleand one in which highly inflated estimates of stock biomass were used. Our analyses showed that no single controlrule performed well across all scenarios, with those that performed well in the unbiased scenario performing poorlyin the biased scenarios and vice versa. Only the most conservative data-poor control rules limited the probability ofoverfishing across most of the life history and exploitation scenarios explored, but these rules typically required veryconservative catches under the unbiased scenarios.

In many fisheries, management actions are based on esti-mates of stock biomass and management targets (biologicalreference points [BRPs]) produced from stock assessment mod-els. Such models typically require long time series of catch andrelative abundance by age and often life history information,and stocks for which there is such information are considered“data rich.” For many stocks, however, this information is lack-ing, preventing the use of a data-driven assessment model. Suchstocks are considered “data poor,” and they pose a challenge tofisheries managers.

In the USA, fisheries managers are now confronting this chal-lenge due to the Magnuson–Stevens Fishery Conservation andManagement Reauthorization Act (MSFCMRA). The act re-quires that the Statistical and Scientific Committees of eachof the eight regional fisheries management councils recom-

*Corresponding author: [email protected] address: Institute of Marine and Coastal Sciences, Rutgers, State University of New Jersey, 71 Dudley Road, New Brunswick, New

Jersey 08901, USA.Received June 22, 2012; accepted May 22, 2013

mend acceptable biological catch (ABC) levels for all stocksunder a fisheries management plan. National Standard 1 of theMSFCMRA further requires that the ABC prevent overfishing(i.e., when the fishing mortality rate exceeds that which producesthe maximum sustainable yield, or FMSY), while still attemptingto achieve optimum yield for the fishery. To prevent overfish-ing, the ABC must have a probability of overfishing (POF) thatdoes not exceed 50%. Scientific uncertainty must also be con-sidered in the selection of an ABC, with the goal of achieving aspecific, acceptable probability of overfishing. Importantly, theABCs constrain the council’s annual catch limits, which maynot exceed the ABC.

For data-rich stocks, approaches have been developed forselecting a catch level that is expected to achieve a specifiedprobability of overfishing, or P* (Shertzer et al. 2008). Although

845

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013

846 WIEDENMANN ET AL.

National Standard 1 does not mandate the use of the P* ap-proach, many councils have adopted some variant of thistechnique for setting ABCs (e.g., Prager and Shertzer 2010;Ralston et al. 2011). The challenge in setting an ABC with theP* approach lies in determining whether scientific uncertaintyhas been adequately accounted for in estimating stock biomassand BRPs.

For data-poor stocks, however, implementation of the P* ap-proach is impossible, and setting ABCs that prevent overfishingfor these stocks is challenging (Wetzel and Punt 2011). Recently,a number of approaches for setting ABCs for data-poor stockshave been developed. These approaches are called harvest con-trol rules, as they specify a rule or set of rules for setting harvestsin response to various factors, such as stock abundance (Derobaand Bence 2008). Data-poor harvest control rules were reviewedand ranked by Berkson et al. (2011), who recommend using adepletion-based stock reduction analysis (DB-SRA; Dick andMacCall 2011) when a catch series spanning the entire historyof the fishery is available. If such catch data are not available,Berkson et al. (2011) recommend using a depletion-correctedaverage catch analysis (DCAC; MacCall 2009). MacCall (2009)advises that DCAC only be used for stocks with low natural mor-tality rates (M) and values of FMSY at or below M. In cases inwhich DCAC is not appropriate, Berkson et al. (2011) recom-mend using a general framework they developed called the onlyreliable catch series (ORCS) approach.

The rankings described above were not based on a formalevaluation of how these control rules performed with respect topreventing overfishing. Wetzel and Punt (2011) conducted a sim-ulation analysis to explore how well DB-SRA and DCAC esti-mated the catch that achieves FMSY (called the overfishing limit,or OFL) for species with life histories typical of groundfishes,principally flatfishes (order Pleuronectiformes) and members ofthe genus Sebastes, found off the western USA. They foundthat both DB-SRA and DCAC generally produced estimatesof the OFL at or below the true values. However, Wetzel andPunt (2011) did not look at the long-term effects of applyingeach control rule to the population. Although Wetzel and Punt(2011) showed that DCAC and DB-SRA can be effective atlimiting overfishing, these control rules cannot be applied in allsituations due to the limitations described above. Therefore, abroader examination of data-poor control rules is needed.

In this study, we used simulation testing (also called man-agement strategy evaluation) to explore the performance of asuite of data-poor harvest control rules over a 20-year periodfor a range of fishing pressures and species’ life histories. Wecalculated different performance measures associated with eachcontrol rule but focused on identifying control rules that wererobust at preventing overfishing across the range of scenarioswe explored. Our analysis included the control rules recom-mended by Berkson et al. (2011) as well as other rules becausewe wanted to evaluate a broad spectrum of potential data-poorapproaches to provide quantitative advice in managing fisheries.To our knowledge no formal approach for updating the control

rules has been proposed, but we reapplied control rules sequen-tially over the 20-year period, as this allowed us to evaluate howthey perform when updated with new information.

METHODS

Model StructureOur simulation study included an operating model and a

management model. The operating model represented the truepopulation dynamics of the stock, whereas the managementmodel determined the annual catch harvested from the stock byapplying a particular control rule. Each model run represented a60-year period divided into two phases. During the first 40 years,an unregulated fishery harvested the population. The remaining20 years represented the data-poor management phase, in whichcontrol rules were applied every 4 years to determine the ABCfor the stock (Figure 1). An amount equivalent to the ABC wasthen harvested from the stock each year (if sufficient biomasswas available), which in turn affected stock size in subsequentyears. At the end of each run, the performance of the controlrule was summarized over the 20-year period. The simulationwas repeated 1,000 times for each control rule.

Operating model.—We used an age-structured populationmodel to generate population dynamics. The equations govern-ing the population dynamics are defined in Table 1 and variabledefinitions are provided in Table 2. Hereafter, the equations usedin the model are referenced by their number in Table 1, suchthat, for example, the numerical abundance at age is referred toas equation T1.1. The annual abundances of a recruited age wasdetermined from the abundance of that cohort in the previousyear, as decreased by continuous natural and fishing mortal-ity (equation T1.1). Recruitment to the population followed theBeverton–Holt stock–recruit relationship, with bias-correctedlognormal stochasticity (equation T1.2). The parameters for theBeverton–Holt model were derived from the unfished spawningbiomass, the unfished recruitment, and the steepness parame-ter (equation T1.3), where steepness represents the fraction ofunfished recruitment that results when the spawning biomass isreduced to 20% of the unfished level. Total spawning biomassin a given year was calculated by summing the product of the

FIGURE 1. Timeline of events in the simulations. For each scenario explored,this timeline was repeated 1,000 times for each control rule to account forthe variability in the population dynamics. The abbreviation ABC stands foracceptable biological catch.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013

HARVEST CONTROL RULES FOR DATA-POOR FISHERIES 847

TABLE 1. Equations characterizing the age-structured population and fishing dynamics in the operating model (see Quinn and DeRiso 1999 for more details).

Equation Description

Population dynamics

1 N (a, t) ={

R(t) a = aR

N (a − 1, t − 1)e[−M−s(a−1)F(t−1)] aR < a ≤ amaxNumerical abundance at age

2 R(t) = S(t−aR )α+βS(t−aR ) e

θR−0.5σ2R Stock–recruit relationship

3α = S0(1−h)

4h R0

β = 5h−14h R0

Stock–recruit parameters

4 S(t) = ∑amaxa=aR

m(a)w(a)N (a, t) Spawning biomass

Life history5 L(a) = L∞(1 − e−k(a−a0)) Length at age

6 w(a) = bL(a)c Weight at length

7 m(a) = 1

1 + e−(

a−m50%mslope

)Maturity at age

Fishing dynamics

8 s(a) = 1

1 + e−(

a−s50%sslope

)Selectivity at age in the fishery

9 C(t) = ∑a

s(a)F(t)M+s(a)F(t)w(a)N (a, t)(1 − e−M−s(a)F(t)) Total catch

proportion mature, weight at age, and abundance at age overall recruited age-classes (equation T1.4). Weight at age wasan allometric function of length at age, which followed a vonBertalanffy growth function (equations T1.5 and T1.6). The pro-portion mature at age was calculated using a logistic function(equation T1.7). Length, weight, and maturity at age were fixedfor a given species’ life history.

The model contained a single fishery, with selectivity at agecalculated using a logistic function (equation T1.8). Because weassumed that both natural (M) and fishing mortality (F) occurredcontinuously throughout the year, catch was calculated usingthe Baranov catch equation (Quinn and Deriso 1999; equationT1.9). We initialized the population in an unfished state. Theunfished abundance at age was calculated using equation T1.1assuming that the abundance of recruits was the unfished equi-librium recruitment (R0; specified in Table 2), and abundanceof older cohorts was calculated assuming a stable age distri-bution under F = 0. The unfished spawning biomass (S0) wascalculated using equation T1.4 using estimates of unfished abun-dance. Subsequently, F increased linearly for the first 20 years,and then reached a plateau for the remaining years (Figure 2).We applied three levels of fishing pressure during this initial40 years: light, moderate, and heavy. These different fishingpressures resulted in median population sizes in year 41 (whenthe control rules were first applied) of 165, 96, and 40% of thespawning stock biomass at maximum sustainable yield (SMSY),corresponding to underexploited, fully exploited, and overex-ploited populations (Table 2).

In year 41 a particular control rule was applied to estimatethe ABC, and in most cases it was reapplied every 4 yearsthroughout the final 20-year period. In a few cases a fixed catchwas applied for the entire time period, or catches estimatedfrom projections were applied across years. Target catches wereconverted to fishing mortality rates by solving equation T1.9numerically. The true catch series (i.e., no error in determiningannual catch) was used for each control rule. The ABC wasfixed at the estimated value for the four-year interval before thereapplication of the control rule. It is possible for control rulesin our model to set the ABC above the exploitable biomass. Toprevent a control rule from removing all exploitable biomassfrom the population in a year, we set the achieved catch to 75%of the exploitable biomass in that year in cases in which theABC exceeded the exploitable biomass.

We ran the model for three different life histories, which welabeled “slow,” “medium,” and “fast.” The definitions for theslow, medium, and fast life histories were based on the char-acteristics of the Spiny Dogfish Squalus acanthias, SummerFlounder Paralicthys dentatus, and Butterfish Peprilus triacan-thus, respectively, species of importance in the Mid-AtlanticBight of the USA. The slow life history had slow growth, latematuration, and low productivity. In contrast, the fast life his-tory had rapid growth, early maturation, and high productivity.The medium life history was intermediate between the slowand fast life histories. The parameter values used to repre-sent these life histories were obtained from the most recentstock assessments (NEFSC 2006, 2008, 2010) and FISHBASE

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


TABLE 2. Parameter values for the slow, medium, and fast life histories modeled in the data-poor simulation. Important quantities derived from these parametersthat were used in the analyses are also listed. The slow, medium, and fast life histories are parameterized roughly using values for Spiny Dogfish (NEFSC 2006),Summer Flounder (NEFSC 2008), and Butterfish (NEFSC 2010). Some parameters were not available for every species (like steepness) or were modified slightlyfrom the assessment values. Therefore, reference points do not exactly match those reported in the assessments. In cases in which parameter values could not beobtained from the stock assessment, we used FISHBASE (www.fishbase.org), although for steepness we used Myers et al. (1999) as a guide.

Life history

Parameter Description Slow Medium Fast

SpecifiedaR Age at recruitment (to population) 1 1 1amax Maximum age 30 15 7M Natural mortality rate 0.1 0.25 0.65R0 Unfished recruitment 1 × 106 1 × 106 1 × 106

h Steepness 0.45 0.65 0.85σR Standard deviation of recruitment variability 0.4 0.4 0.4a0 Age at length = 0 −2 0 −0.214L∞ Maximum length 105 90 19.19k Growth rate 0.15 0.35 0.4b Length–weight scalar 2.98 × 10−7 3.5 × 10−6 4.27 × 10 −5

c Length–weight exponent 3.6 3.15 2.8m50 Age at 50% maturity 5 2 1.4mslope Slope of maturity function 0.5 0.36 0.1s50 Age at 50% selectivity 6 2.5 1sslope Slope of selectivity function 0.5 0.34 0.2

DerivedS0 Unfished spawning biomass 2,045,034 717,120 6,825SMSY Spawning biomass that produces MSY 821,080 259,329 2,054SMSY/S0 Ratio of SMSY to S0 0.4 0.36 0.3FMSY Fishing mortality that produces MSY 0.07 0.25 0.75MSY Maximum sustainable yield 50,520 48,026 1,065FMSY/M Ratio of FMSY to M 0.7 1 1.15Finit/FMSY Maximum F in the initial period to FMSY for the under-,

fully, and overexploited runs, respectively0.5, 1.02, 2.0 0.45, 1.01, 2.2 0.54, 1.25,2.65

(http://www.fishbase.org; Table 2). Steepness values for SpinyDogfish and Butterfish were based on the meta-analysis ofMyers et al. (1999). We set steepness for the medium life his-tory between the those for the slow and fast life histories, as thevalues for flatfishes from Myers et al. (1999) were similar to thatfor Butterfish. Recruitment variability was fixed across species(σR = 0.4). In addition, we used the same R0 for each species,which resulted in BRPs for each stock that are not comparable tothe assessment-estimated values. We calculated the maximumsustainable yield (MSY)–based BRP (Table 2) for each stockfollowing the standard yield-per-recruit and spawning biomass-per-recruit approach (Shepherd 1982; NEFSC 2002).

Management model.—The control rules that we tested var-ied greatly in the level of information required to estimate theABC (Table 3). Some control rules only required a catch se-ries (which need not cover the entire history of the fishery),while others required additional information on life history andrelative abundance. Given that these control rules are to be ap-plied to data-poor stocks, relative abundance and general life

history characteristics may be difficult to ascertain. For relativeabundance, some control rules required that a stock be classifiedin broad categories (e.g., under-, fully, or overexploited), whileothers required a more exact measure of depletion. Like thosefor life history traits, some control rules required a species to bebroadly categorized based on productivity (e.g., low, medium,or high), while others required specific values for parameters.These requirements pose a challenge for scientists, and expertopinion is often required.

In our application of the control rules, we explored two sce-narios. In the first scenario, unbiased (i.e., perfect) informationon stock biomass and life history was used in the control rule.In the second scenario, we again used unbiased informationon life history, but stock biomass was assumed to be inflated(biased upward). We call these scenarios the unbiased andbiased runs, respectively. We describe the specifics of these runswhen discussing each control rule, as using inflated estimatesof abundance required different approaches for different controlrules.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


FIGURE 2. Fishing mortality rates (F) and catch histories during the initial period in the simulations for the slow, medium, and fast life histories. The valuesshown are for model runs with no stochasticity in recruitment during the initial phase.

The control rules that we explored vary with respect to thecatch that is estimated, with some explicitly estimating the OFLand others estimating the ABC. In some cases, the control ruleis used to estimate a sustainable catch (not necessarily the OFL

TABLE 3. Level of catch and life history knowledge required and how stockabundance is determined for the control rules used in the analyses. Completecatch series refers to whether or not an entire catch history (starting in theunfished state) is needed for the control rule. Life history and stock abundancerefer to the level of knowledge of the species’ life history and current abundance,with basic meaning that a species is broadly categorized (e.g., as being of low orhigh productivity for life history, and under- or overexploited for abundance);detailed indicates that specific values (in the form of values drawn from specifiedprobability distributions) are required as inputs. For the control rules, ORCSrefers to the only reliable catch series, DCAC to the depletion-corrected averagecatch, and DB-SRA to depletion-based stock reduction analysis.

Complete Life history StockControl rule catch series? knowledge abundance

Mean/median catch No None NonePercent of the

mean/median catchNo None None

ORCS No Basic BasicRestrepo No None BasicDCAC No Detailed DetailedDB-SRA Yes Detailed Detailed

or ABC). For each rule described below, we explicitly statewhether we assumed that the catch was the OFL or the ABC.

Although councils are required to annually specify ABCs,the 4-year interval between updates to each control rule may beunrealistically short. The data-poor control rules we are testingwere not necessarily designed to be updated on a semiannualbasis unless new information becomes available to justify anupdate in the catch limit. We selected an interval of 4 yearsfor two reasons. First, while longer intervals may make moresense given the limited data, there will likely be pressure fromstakeholders to update catch targets that are deemed too lowor too high, even in cases in which data are not available tosupport their claims. Second, we wanted to explore how thecontrol rules performed under a best-case scenario, and havingcontrol rules updated with the true information every fourth yearseemed reasonable, as updates for data-poor stocks would notlikely occur more frequently even with data available to informthe control rule. In updating the data-poor control rules, we areillustrating how these rules would perform if updated and arenot providing insight into how one would go about updatingthem.

Harvest Control RulesSummary catch statistic.—In many situations, only catch data

are available for a particular stock, and it is not possible for

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


scientists or managers to determine the status of the stock. Insuch cases, a summary catch statistic may be used to set futurecatches. A summary catch statistic can be any metric calculatedfrom the catch history. For our analyses, we used the medianof the catch history as well as 50% and 75% of the median asour summary statistics and set these as the ABC. Because thesummary catch statistic does not use any information regardingstock size, there is no effect of having a biased perception ofstock abundance.

An important issue when calculating a summary catch statis-tic is whether or not to update the catch history using the catchesproduced from a particular control rule, as using an updatedcatch series can have a ratcheting effect. For example, if thecontrol rule is to use 50% of the median catch as the ABC andthe catch history used to calculate the median is updated, themedian catch will decrease over time. Therefore, the ABC that isestimated using 50% of the median catch will also decline overtime. This effect also works in the other direction if the ABCis greater than the median catch. For analyses that required asummary catch statistic, we did not update the catch series, aswe felt that the ratcheting was unrealistic.

Only reliable catch series (ORCS) approach.—An extensionof the summary catch statistic control rule is to adjust the statisticbased on the available evidence. For example, if a populationis believed to be underexploited, the ABC may be set at somemultiple of the median catch. Berkson et al. (2011) proposed ageneral framework for developing a control rule to be used whenadditional information is available, but it is not possible to usemore information-intensive control rules (i.e., DCAC and DB-SRA). This framework, called the ORCS approach, is extremelyflexible in its application, and we note that we are only testinga few possibilities.

The ORCS framework requires that stock abundance first beput into into three broad categories: (1) underexploited, (2) fullyexploited, and (3) overexploited. Berkson et al. (2011) suggestthat stocks with a spawning biomass below 19% of S0 be clas-sified as overexploited, that stocks with a spawning biomassmore than 65% of S0 be classified as lightly exploited, and thatstocks in between be classified as fully exploited. The justifi-cation for these biomass classification thresholds comes fromHilborn (2008), who showed that between these thresholds a“pretty good yield” (≥80% of MSY) could be sustainably re-moved. We used these classification groups, but other groupsor different boundaries may also be appropriate. Given a clas-sification of abundance for a stock, the next step in the ORCSapproach is to adjust the summary catch statistic (e.g., the meanor median, denoted C) to generate an estimate for the OFL.Berkson et al. (2011) recommend multipliers of 0.5, 1, or 2for the over-, fully, and underexploited stocks, respectively, togenerate an OFL estimate, such that

OFL =

⎧⎪⎨⎪⎩

2C S(t) ≥ S65%

C S19% ≤ S(t) < S65%

0.5C S(t) < S19%

. (1)

TABLE 4. Overfishing limit (the catch at FMSY) buffering options (θ) inrelation to risk presented used with the only reliable catch series (ORCS) controlrule.

Risk level (productivity) θ1 θ2 θ3

Low risk (high productivity) 0.9 0.8 0.7Moderate risk

(moderate productivity)0.8 0.65 0.5

High risk (low productivity) 0.7 0.5 0.3

The ABC is then calculated by multiplying the OFL by a scalar(θ ≤ 1) based on the perceived level of “risk” for the stock(ABC = θ × OFL). The risk for a stock may be based on itsassumed productivity (i.e., how fast biomass can recover) or sus-ceptibility to the fishery (i.e., how easily a stock can be affectedby a fishery), or both (see Patrick et al. 2009 for an example ofclassifying stocks using productivity–susceptibility analysis).Berkson et al. (2011) suggest that the amount of buffer shoulddepend on the life history of the stock, with a high-productivitystock receiving a smaller buffer than a low-productivity stocks.They also provide a range of different possible values for θ basedon risk. Berkson et al. (2011) do not explicitly state which sum-mary catch statistic to choose but mention the median and the75th percentile of the catch history as two possibilities. There-fore, we applied the ORCS approach using both the medianand the 75th percentile of the catch, and we explored three θs(denoted θ1, θ2, and θ3; Table 4). For the unbiased model run,the stocks are correctly classified according to the categories inequation (1). For the biased model run, the stocks are incorrectlyclassified into the adjacent-less-depleted category. That is, if thestock is overexploited it is classified as being fully exploited,and if it is fully exploited it is classified as underexploited. Thebiased model run is meant to be a worst-case scenario, and in thisinstance the population is never assumed to be overexploited.

The Restrepo rule.—Restrepo et al. (1998) developed guid-ance for specifying catch limits in data-poor situations for the1996 reauthorization of the Magnuson–Stevens Act (we refer totheir approach as the Restrepo rule). Like the ORCS approach,their approach requires classifying a stock into under-, fully, andoverexploited categories (although the biomass thresholds defin-ing these categories are slightly different, using the overfishedand recovery thresholds of 0.5 × SMSY and SMSY, respectively)and adjusting a summary catch statistic to calculate the ABC:

ABC =

⎧⎪⎨⎪⎩

0.75C S(t) ≥ SMSY

0.5C 0.5SMSY ≤ S(t) < SMSY

0.25C S(t) < 0.5SMSY

. (2)

A key difference between the Restrepo and ORCS approachesis in the adjustment of C , with the Restrepo rule being muchmore conservative as it decreases C , even for an underexploitedstock. Restrepo et al. (1998) recommended using recent stablecatches to estimate C , so we used recent stable catches as well as

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


the median catch in our exploration of the Restrepo rule. Recentstable catches are defined in our model as those occurring in themost recent 5-year period in which the coefficient of variation(CV) of the catch series is at or below 5%. If no such periodexisted, we increased the CV to 10%, 15%, and so on. If therewas no period with a CV below 30%, we used the mean catchin the most recent 5-year period.

Depletion-corrected average catch (DCAC).—MacCall(2009) developed DCAC as a way to calculate a sustainableyield in data-poor situations using a catch time series andassumptions about the life history parameters and relative statusof the stock. The formula for calculating the target catch usingDCAC requires a catch series (Cobs) spanning some periodbetween tfirst and tlast and assumptions about the relative declinein biomass over this period (� = [S{tfirst} – S{tlast}]/S0), M,the ratio of FMSY to M, and the ratio of SMSY to S0. In practice,assumptions can be made about the two ratios (Thorson et al.2012b; Zhou et al. 2012) and M can be estimated from longevityinformation. Information on current depletion can be based onan index of abundance or expert opinion. The generic formulafor calculating the target catch C(t) in year t (>tlast) usingDCAC is

C(t) =∑tlast

tfirstCobs(t)

n + �(SMSY

S0

)(FMSY

M

)M

, (3)

where n is the number of years of catch data. Given the inherentuncertainty in these inputs, particularly for data-poor stocks,MacCall (2009) suggests that DCAC be calculated using aMonte Carlo approach. For the Monte Carlo simulation wedrew values for M, FMSY/M, and SMSY/S0 from a lognormaldistribution with means equal to the true values and CVs of 50,25, and 15%, respectively. For the unbiased model run, � wasdrawn from a normal distribution (allowing for negative values)with the mean equal to the true value and a CV of 30%. For thebiased model run, � was drawn from a normal distribution witha mean calculated assuming that S(tlast) is 50% greater than thetrue value (i.e., [S{tfirst} – 1.5 × S{tlast}]/S0) and a CV of 30%.We set a maximum value of 0.98 for �. The CVs assumed forM, FMSY/M, and � are consistent with the values suggestedby MacCall (2009), and our CV for SMSY/S0 matched thatused by Wetzel and Punt (2011). MacCall (2009) recommendsthat DCAC not be used for species with an M above 0.2 andan FMSY/M above 1. The constraint on M was based on thederivation of DCAC and the concept of a “windfall harvest”(the amount needed to reduce the population from S0 to SMSY

in a single year). At high values of M, the sustainable yield ap-proaches the windfall harvest, such that the depletion correctionbecomes small (equations 3–7 in MacCall 2009). The constrainton FMSY/M was based on the observations of the ratios for otherstocks (e.g., Zhou et al. 2012) and the fact that in the absence ofinformation it would be prudent to assume an FMSY/M less than1. The true M is greater than 0.2 for the medium and fast life his-

tories, and FMSY/M is above 1 for the fast life history (Table 2).Despite its being above the limits, we applied DCAC to theselife histories to see how it performed in such cases and refer tothis as the base DCAC run. We also applied DCAC with a fixedlevel of � = 0.6 in all years, representing a decline in biomassfrom S0 to 40% of S0, which we call the fixed DCAC run. Thepurpose of this run was to explore the effects of an assumeddepletion in cases in which a mean � could not be decided.Each time DCAC was used, we conducted 1,000 parameterdraws to create a catch distribution. The catch distribution is notan explicit estimate of the OFL or the ABC, so we assumed thatthe median of the distribution was the ABC. For the base runthe depletion level was updated every 4 years and a new ABCwas calculated. For the fixed run, DCAC was only used once (att = 41) and the ABC was fixed for the remainder of the period.

Depletion-based stock reduction analysis (DB-SRA).—Although DCAC is a way of adjusting the average catch basedon current depletion and stock life history, DB-SRA (a com-bination of DCAC and stock reduction analysis [Walters et al.2006] proposed by Dick and MacCall 2011) is a method forobtaining a distribution for the OFL in data-poor situations. Itrequires nearly the same inputs as DCAC, with assumptionsbeing made about M, FMSY/M, SMSY/S0, and �. An importantdistinction between DB-SRA and DCAC is that DB-SRA re-quires a complete catch history, starting at the unfished state.Thus, for DB-SRA � = 1 – S(t)/S0 and can only range between0 and 1 (whereas � can be negative in DCAC).

Like DCAC, DB-SRA was applied using a Monte Carloapproach, with draws for each required quantity. Given a par-ticular set of values, we then calculated FMSY = FMSY/M × Mand the harvest fraction UMSY = (FMSY/[FMSY + M]) × (1 –exp[−M − FMSY]). Next, we iteratively solved for the unfishedbiomass, S0, that resulted in the assumed current level of de-pletion, given the catch history and a modified Pella–Tomlinsonproduction model (see Dick and MacCall 2011 for the reasonsfor using a modified production function). With an estimate ofS0, we then estimated SMSY = S0 × SMSY/S0 and S(t) = S0 (1 –�), and finally the OFL with C(t) = UMSY × S(t). The modifiedproduction model can be used to project the population biomassfor a number of years by fishing at UMSY, producing a timeseries of OFL estimates. The Monte Carlo approach producesa distribution for the OFL (in one or many years), such thatone could take some percentile (e.g., ≤50%) to buffer the ABCaway from the median OFL estimate.

As with DCAC, we drew values for M, FMSY/M and SMSY/S0

from a lognormal distribution with mean equal to the true valueand CVs of 50, 25, and 15%, respectively. Values for � weredrawn from a beta distribution (constraining it between 0 and1) with a CV of 30%, a mean equal to the true value in theunbiased run, and the mean calculated with S(t) being 50%greater than the true value in the biased run. We set a maximumvalue of 0.98 for �. We explored three permutations of DB-SRA, each of which used 1,000 parameter sets. Under the firstpermutation (called the base run), DB-SRA was applied every

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


4 years and the ABC for the 4-year interval was fixed at themedian of the OFL distribution for the current year. For thesecond permutation (called the projected run), DB-SRA wasonly applied in the first year (t = 41), and we generated adistribution of OFL over the remaining period by projecting thepopulation biomass assuming that U(t) = UMSY for all years. Forthe final permutation (called the fixed run), DB-SRA was onlyapplied in the first year, with an assumed � = 0.6 correspondingto an assumed biomass of 40% of S0. This assumption was basedon Dick and MacCall (2011), who noted that DB-SRA generallyperformed well when an assumed � = 0.6 was used, regardlessof the population size of the stocks they explored. For each yearwe set the ABC equal to the median of the OFL distribution forthat year (i.e., there was no buffering).

Performance MeasuresIt is important to understand how each control rule performs

with respect to objectives that are of interest to both managersand stakeholders, and how performance varies according to thelife history and exploitation history of a population. We calcu-lated performance measures for each control rule for each run(of the 1,000 total) for a given scenario (life history, exploita-tion history, and level of bias in estimated abundance). Managersmust prevent overfishing under the MSFCMRA and revised Na-tional Standard 1, so we calculated the probability of overfish-ing (POF) across the 20-year period over which the control ruleswere applied (between years 41 and 60) as the proportion ofyears in which F(t) exceeded FMSY. Catches with a POF greaterthan 0.5 are more likely than not to result in overfishing and arenot allowed under National Standard 1. Managers also want toprevent stocks from becoming overfished (or rebuild those thatare), so we computed the ratio of the mean spawning biomassover the final 10 years to SMSY. We used the final 10 years in thecalculation to reduce the potential effect of transient dynamicsin the initial years. Stakeholders are interested in the yield to thefishery, so we computed the mean ratio of the observed catchto the true MSY over the entire 20-year period the control rulewas applied.

RESULTSControl rules had variable performance across the life his-

tory and exploitation history scenarios, and no control rule wasbest across all scenarios. For the scenarios in which unbiasedinformation was used when assessing stock abundance and lifehistory (the unbiased runs), some patterns in harvest rates ap-peared across the range of life histories and exploitation historieswe explored. For a given life history, many of the control rulesresulted in the general trend of an increasing POF going fromthe underexploited population to the overexploited population(Figure 3; Table A.1 in the appendix). Many of the control ruleswe explored were extremely variable across exploitation sce-narios, with the median POF near zero in the under- and fullyexploited cases and above 0.5 in the overexploited cases (Fig-ure 3; Table A.1). In addition, in the overexploited case POF was

generally higher for the slow life history than for the mediumand fast life histories for many of the control rules. An exceptionto these patterns was DB-SRA, which showed the greatest con-sistency in POF across life history and exploitation scenarios,with a median value between approximately 0.2 and 0.6 and thevariation in the estimates increasing from the fast to slow lifehistories.

Control rule performance differed among scenarios with bi-ased information. For the model runs in which the stock abun-dance was inflated, the POF for control rules using only a sum-mary catch statistic (the median catch and 50% of the mediancatch) did not differ from those of the unbiased run, as theserules do not use any information on stock abundance (Figure 4).Control rules requiring classification of stock abundance gen-erally resulted in higher POF than in the unbiased scenarios(Table A.2). For the fast life history, the difference in the POF

for most of the control rules was small between the scenarioin which the population was fully exploited (but assumed tobe underexploited) and overexploited (but assumed to be fullyexploited). For the medium life history, and particularly for theslow life history, most control rules resulted in higher POF whenthe population was overexploited than when it was fully ex-ploited (Figure 4). The exception again was DB-SRA, whichresulted in similar or slightly lower values of POF for the differ-ent life histories that were overexploited. For the medium andslow life histories, many of the control rules performed poorly,with many having a POF in excess of 0.8.

Estimates of POF only provide information on the frequencyof overfishing, not the magnitude of the catches. Fishing closeto FMSY will result in catches being above, close to, or belowMSY for under-, fully, and overexploited populations, respec-tively. Thus, a control rule that performed well (fishing closeto FMSY) would result in decreases in the ratio of mean catchto MSY going from the underexploited to the overexploitedscenarios. However, we only observed this pattern for DB-SRA(Figure 5). For the underexploited runs, most control rulesresulted in the catch being below MSY (Figure 5), causingthe population biomass to be two to three times above SMSY

(Figure 6). In contrast, DB-SRA resulted in catches rangingbetween 100% and 150% of MSY and biomass between 50%and 150% of SMSY. For the fully exploited scenarios, both themedian catch and DB-SRA produced the highest catches foreach life history, with the catch-to-MSY ratio generally closeto 1 (Figure 5). The remaining control rules produced moreconservative catches, resulting in biomass between 100% and200% of SMSY (Figures 5, 6). For the overexploited scenariothere was less variation in catches despite considerable variationin POF (Figure 4) and stock size (Figure 6). Some control rulesmaintained high catches by allowing severe overfishing anddriving biomass towards zero (the median catch and DCACfor the slow and medium life histories), while others weremore conservative and allowed the population to rebuild, thusproducing similarly higher catches (ORCS, Restrepo, andDB-SRA).

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


FIGURE 3. Probability of overfishing in the unbiased run across control rules and life histories for the three exploitation histories explored (underexploited,fully exploited, and overexploited). The life histories (fast, medium, and slow) are shown from left to right for each control rule and are distinguished by shading(none, light, and dark, respectively), although in some instances the shading is not evident because the range of outcomes was relatively small. The lower andupper edges of each box represent the first and third quartiles, respectively, with the solid horizontal line within each box representing the median. The whiskersrepresent ±1.5× the interquartile range, and the circles represent outliers. See the text for descriptions of the control rules.

For many of the alternative control rule variants we explored,the behavior was predictable from the reduced subset of controlrules (Tables A.1, A.2). For example, across scenarios the moreconservative ORCS option (θ1) resulted in lower catch, higherbiomass, and smaller POF, while using 75% of the mediancatch resulted in catch, biomass, and POF between the medianand 50% of the median catch control rules (Tables A.1, A.2).For other variants, particularly for DCAC and DB-SRA, theresponses were not as predictable. In addition to the base explo-rations of DCAC and DB-SRA, we explored situations in whichbiomass was assumed to be 40% of S0 and a variant of DB-SRAin which the ABC in all years was based on projections done inthe initial year. When projections were used for DB-SRA, therewere instances in which this approach was more conservativethan the base run of DB-SRA as well as instances in which it wasless conservative. For the DB-SRA runs in which biomass wasassumed fixed at 40% of S0, catch estimates were generally moreconservative than the base DB-SRA run when the true popula-tion was higher (the underexploited scenario); estimates werecomparable when the population size was similar (the fully ex-ploited scenario) and much greater when the true population sizewas less than the assumed value. This trend did not hold for the

fast life history, however, where the fixed DB-SRA run resultedin higher instances of overfishing than in the base run acrossthe exploitation scenarios (Table A.1). In contrast, assuming afixed depletion for DCAC resulted in much smaller differencesin performance measures across exploitation scenarios than inthe base DCAC run. These differences were smaller for themedium and fast life histories, the result of smaller depletioncorrections for high values of M and FMSY/M (equation 3).

Both DCAC and DB-SRA produce a distribution of the catchestimate, and we selected the median value for all scenarios asthe estimate of the ABC. Different percentiles of the distribu-tion could be selected to reduce the probability of overfishingin cases in which these control rules resulted in values of POF

greater than 0.5. For each year we calculated the percentile ofthe catch distribution that would achieve the OFL and then cal-culated the mean percentile across years for a particular scenario(Table 5). For DCAC, there were many instances in whichthe catch distribution barely overlapped (i.e., high or low per-centiles) or did not overlap at all with the true OFL. In contrast,DB-SRA (for the base and projected runs) produced narrowerranges of percentiles between the 33rd and 87th percentiles(with most estimates between the 46th and 60th percentiles;

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


FIGURE 4. Probability of overfishing across control rules and life histories for the fully exploited and overexploited exploitation histories in the model run usinginflated estimates of stock abundance in the control rules that require such information (called the biased scenario). The terms fully exploited and overexploitedrefer to the true abundance of the stock. See Figure 3 for additional details.

FIGURE 5. Mean catch relative to the maximum sustainable yield (MSY) across control rules for the different life histories and exploitation histories explored.See Figure 3 for additional details.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


TABLE 5. The mean percentile of the catch distributions produced by DCAC (depletion corrected average catch) and DB-SRA (depletion-based stock reductionanalysis) that achieves the true overfishing limit (OFL) across exploitation and life history runs. Values are presented as decimals, such that a value of 0.5 representsthe 50th percentile of the distribution. A value of 0 indicates that the all values of the catch distribution were above the true OFL, while a value of 1 indicatesall values were below the true OFL. Control rules were not applied to underexploited populations in the biased run, and we did not apply DCAC and DB-SRAassuming a fixed depletion in the biased run.

Unbiased Biased

Harvest pressure Control rule Fast Medium Slow Fast Medium Slow

Underexploited DCAC 1.00 1.00 1.00DCAC (fixed) 1.00 1.00 1.00DB-SRA 0.46 0.52 0.53DB-SRA (projected) 0.33 0.55 0.60DB-SRA (fixed) 0.19 0.88 0.95

Fully exploited DCAC 0.80 0.91 0.84 0.75 0.85 0.58DCAC (fixed) 0.79 0.86 0.79DB-SRA 0.48 0.51 0.49 0.23 0.16 0.13DB-SRA (projected) 0.39 0.52 0.48 0.13 0.07 0.04DB-SRA (fixed) 0.01 0.30 0.47

Overexploited DCAC 0.42 0.08 0.01 0.40 0.05 0.00DCAC (fixed) 0.42 0.04 0.00DB-SRA 0.59 0.64 0.51 0.30 0.20 0.15DB-SRA (projected) 0.87 0.76 0.49 0.59 0.28 0.10DB-SRA (fixed) 0.00 0.00 0.00

FIGURE 6. Mean spawning biomass (S) relative to the spawning biomass that produces the MSY (SMSY) across control rules for the different life histories andexploitation histories explored. Each ratio was calculated using the mean spawning biomass from the final 10 years of the model run. See Figure 3 for additionaldetails.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


Table A.1). However, the catch distributions for DB-SRA as-suming a fixed depletion (� = 0.6) were less precise, partic-ularly for the overexploited scenarios in which the distributionwas always higher than the true OFL.

DISCUSSIONAlthough it is important to consider a wide range of per-

formance measures in our analysis of harvest control rules, theMSFCMRA mandates that U.S. federal fisheries managementavoid overfishing. National Standard 1 further requires that ABCcontrol rules be developed in conjunction with the regional fish-ery management councils to achieve specific probabilities ofoverfishing (≤0.5). Our analysis of data-poor control rules forspecifying ABCs indicated that no single control rule performedbest across all of the scenarios we explored. In the scenarios inwhich unbiased information on stock abundance was used, onlythe most conservative ORCS approach (option θ1 using the me-dian catch and the 75th percentile of the catch) and the Restreporule using recent stable catches resulted in a median POF lessthan 0.5 for all the exploitation and life history scenarios weexplored. A number of rules resulted in POF values greater than0.5 only for the overexploited species with a slow life history.For the scenarios in which biased (inflated) estimates of stockabundance were used in the control rules, no control rule re-sulted in a POF less than 0.5 for all of the exploitation and lifehistory combinations we explored (Tables A.1, A.2).

Minimizing the risk of overfishing is not the only objectiveof management. Fishery managers must also try to achieve opti-mum sustainable yield. If a control rule were able to balance thetrade-offs between a sizeable harvest and a low risk of overfish-ing, we would expect to see POF values somewhat lower than 0.5across scenarios. Across exploitation scenarios, DB-SRA (boththe base and projected runs) resulted in some of the highestcatches but also in POF values above 0.5 in some scenarios, par-ticularly when the stock was assumed to be in better conditionthan it actually was. However, DB-SRA performed well whenunbiased information on stock abundance was used. An addi-tional benefit to DB-SRA is that it produced accurate estimatesof the OFL for the heavily exploited stocks across life histories.Greater accuracy for more depleted stocks was noted by Dickand MacCall (2011), and it is an extremely useful property, asit allows for more risk-averse harvesting in high-risk cases.

If a data-poor stock can be classified into the different lifehistory and population status categories we explored, our re-sults suggest it is possible to select a preferred control rule thatperforms consistently well. Our analyses, however, did not con-sider the full range of potential scenarios. For example, there areslower or faster life histories than those we used. Also, we didnot explore all ranges of potential depletion, and stocks may bemore or less depleted within a particular exploitation category.However, given that many of the control rules performed worsefor the overexploited populations with the slow life history, theapplication of such control rules to populations with slower life

histories that are more severely overfished will likely result inoverfishing. Finally, our analyses did not consider a broad rangeof effort dynamics leading up to the use of a control rule, as weused the straightforward effort dynamics shown in Figure 2.

Many other potential catch histories could result in the differ-ent population sizes we explored (Vasconcellos and Cochrane2005). Different dynamics would produce different catch his-tories, which could impact the performance of many of thecontrol rules. For example, a constant catch that resulted in anoverexploited stock might have a very different median catchthan those resulting from the catch histories we used (Figure 2).Alternative catch histories may also influence DCAC and DB-SRA, as noted by Wetzel and Punt (2011). Therefore, a broaderanalysis of the effects of different catch histories on control ruleperformance is warranted.

Care is also needed when trying to select an appropriate con-trol rule for a given stock, as there is the potential for error inclassifying the stock into the different categories. Misclassifi-cation works in both directions (i.e., assumed biomass can belower or higher than the true value, or the assumed life historycan be slower or faster than assumed), but we only exploredthe effects of assuming that the population was higher than thetrue value, as we considered this a worst-case scenario. Many ofthe control rules that performed well when the true populationabundance and life history were used performed poorly whenthe stock was assumed to be in better condition than it was. Hav-ing broad classification categories for a control rule (as in theORCS or Restrepo rules) may reduce the likelihood of misclas-sifying a stock (as opposed to requiring a specific value, albeitfrom a distribution, for relative stock abundance, as in DCACand DB-SRA), but there is still potential for error. Assuminga stock is more abundant than it actually is can have a large,negative impact on the population, as many control rules thatwere applied with inflated estimates resulted in large declines inbiomass for the medium and slow life histories, in some casesdriving the population to near extinction. Dick and MacCall(2011) showed, however, that for stocks with biomasses nearS0 the model performed better when biomass was assumed tobe lower than the true value, making it problematic to assumethat a stock is very lightly exploited. Thus, following Dick andMacCall’s (2011) recommendation and assuming a greater de-pletion level for a stock that is believed to be near the unfishedlevel would reduce or negate the effects of this misspecification.

A few control rules resulted in large declines in biomass whenunbiased information was used for the species with the slow lifehistory (the median catch, 50% of the median, and DCAC;Figure 6). Given these findings, it is important to identify whencatches from a control rule are driving population biomass tosuch low levels. A formal analysis of methods to detect thestatus of data-poor stocks (e.g., Scandol 2003; Cope and Punt2009) is beyond the scope of this paper, but it may be possibleto use the trends in catches (Costello et al. 2012; Thorson et al.2012a). For example, dramatic declines in catches may be abellwether for a poorly performing control rule (in the absence

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


of dramatic behavioral or regulatory changes). Another potentialsign of a declining population is the inability of the fishery toreach the quota, although using this metric might result in anoverreporting of the catch to prevent declines in the catch limit.

Various approaches have been proposed to obtain informa-tion on stock abundance and life history, such as using expertopinion based on similar species or trophic groups or from sim-ilar fisheries (MacCall 2009; Berkson et al. 2011). In cases inwhich it is not possible to classify a stock, one could assume thatthe population is in the worst possible category (i.e., overfished)or only some measure of the catch history may be used (e.g.,the median or 50% of the median catch). We did not explorethe effects of assuming a worst-case population in our analysis,but we did explore control rules that do not require abundanceinformation. Such control rules are beneficial in that they makeno assumptions about the current abundance of the stock. Weexplored three catch statistics: the median catch, 75% of themedian, and 50% of the median as well as using DCAC andDB-SRA with a fixed estimate of stock abundance. None ofthese control rules resulted in POF < 0.5 in all of the scenar-ios explored, with the summary catch statistics often producinghigh POF values for the overexploited scenario. Thus, using veryconservative catch statistics (e.g., ≤50% of the median) may besuitable when no information is available if one chooses to erron the side of caution. It is also possible to select lower assumedlevels of depletion for DB-SRA (�40% of S0), but this assump-tion may result in poor model performance in some cases, asour runs assuming a fixed depletion for the fast life history per-formed poorly in all the exploitation scenarios explored (TablesA.1, A.2). For DCAC, misspecification of the abundance hadless of an effect on estimates than DB-SRA (Tables A.1, A.2),and this property of a control rule is beneficial. Assuming a fixedvalue for stock abundance often resulted in target catches thatwere below the OFL for underexploited stocks and above it foroverexploited stocks. As a result, care is needed when applyingthis control rule, especially for overexploited populations.

Berkson et al. (2011) recommend using DB-SRA, if possi-ble, followed by DCAC and then their own approach, which wecalled the ORCS approach. In addition to these rules we alsoevaluated the Restrepo rule, on which the ORCS approach isbased. While DB-SRA and DCAC were the top-recommendedcontrol rules by Berkson et al. (2011) when stock status couldbe reliably determined, both resulted in POF > 0.5 for somescenarios (although the scenarios in which this occurred dif-fered between them). Although DB-SRA resulted in overfish-ing in some scenarios, lower percentiles of the OFL distribu-tion could be selected, resulting in a lower POF in such cases.Our analyses showed that choosing between the 30th and 40thpercentiles would result in a POF below 0.5 for the unbiasedmodel run but that much lower percentiles (between the 10thand 30th) were needed when inflated stock abundance was used(Table 5). Both the Restrepo rule and more conservative ORCSapproaches we explored (θ1 and θ2) resulted in no or at mostone scenario in which POF > 0.5 (the slow life history that was

overexploited). For the biased scenarios, the Restrepo rule re-sulted in the fewest scenarios with POF greater than 0.5 becauseit used the largest buffer for the catch across all exploitation sce-narios. If avoiding overfishing is the most important objective,the Restrepo rule or the more conservative variants of the ORCSrule might be appropriate.

In general, DCAC performed well (POF less than 0.5) in theunder- and fully exploited scenarios when unbiased informationwas used, but it resulted in high harvest rates and occurrencesof overfishing for the overexploited stocks, particularly for themedium and slow life histories. MacCall (2009) recommendsusing DCAC only for species with M less than 0.2 and FMSY/Mless than 1, but we applied it to populations with larger values(the medium and fast life histories). The DCAC approach didnot result in POF exceeding 0.5 for these life histories whenthey were under- or fully exploited, but it did when the popula-tions were overexploited. Our analysis suggests that restrictionson using DCAC be based on perceived population abundance(instead of life history), with DCAC use limited to populationsbelieved to be under- or fully exploited. A caveat to using DCAC(for all life histories) is the effect of having a long time seriesof catches, as the total catch is adjusted based on the number ofyears (n) and the depletion correction factor (equation 3). Forthe same depletion correction factor, the overall effect on theestimated catch is less for larger values of n.

Our analysis of DCAC and DB-SRA generally agreed withWetzel and Punt (2011), who evaluated the ability of these con-trol rules to estimate the OFL. In their analysis, Wetzel andPunt (2011) estimated the OFL for two stocks with differing lifehistories (both with M at or below 0.2) for a range of scenar-ios, including catch histories, current abundance, and parametermisspecification. Our work differed in that we looked at theeffects of the repeated application of each control rule over a20-year period (whereas they compared OFL estimates with thetrue value in 1 year only), and we explored a broader range oflife histories. As with our work, they showed that DCAC over-estimated the OFL when the population was overfished and thatthe error was higher for the species with the slower life history(Wetzel and Punt 2011). For DB-SRA, Wetzel and Punt foundinstances of both over- and underestimation of the OFL and thatthe estimation was better for the slower life history.

In summary, no control rule performed best across the rangeof life histories, population abundance, and misclassificationscenarios we explored. Stocks with slow life histories that hada history of overexploitation were particularly challenging un-less an unbiased estimate of stock abundance could be obtained.While DB-SRA produced relatively accurate estimates of theOFL, it was sensitive to misspecification of current stock abun-dance. Selecting lower percentiles of the OFL distribution pro-duced by DB-SRA may reduce the effect of assuming inflatedstock abundance in many cases. Approaches that only requireclassification of abundance into broad categories (i.e., ORCSand Restrepo) may be a useful alternative, as the potential to in-correctly specify stock abundance is reduced. We only explored

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


a few potential options for these rules, but the more conserva-tive ORCS and Restrepo options were generally robust acrossscenarios, although the resulting ABCs were very conservativefor the populations that had a history of underexploitation. Fur-ther refinement of these control rules (DB-SRA, ORCS, andRestrepo) is warranted based on our results, particularly refine-ment of the catch adjustment factors based on abundance andlife history (ORCS and Restrepo approaches) and refinement ofthe methods of classifying current stock abundance (DB-SRA).Such refinements might result in a single control rule that isrobust across the range of scenarios that may be encountered.

ACKNOWLEDGMENTSWe thank the members of this research project’s steering

committee (Lee Anderson, John Boreman, Liz Brooks, JessicaCoakley, Rick Robins, and Rich Seagraves) for helpful discus-sions on formulating this work. We also thank Jim Berkson,Andre Punt, Chris Legault, Doug Vaughan, and the membersof the Mid-Atlantic Fisheries Management Council’s Scientificand Statistical Committee for comments on an earlier version.In addition, we thank E. J. Dick for discussions on the DB-SRA control rule, Alec MacCall, as well as the associate editor,and two anonymous reviewers for comments on earlier drafts.Funding was provided by the Mid-Atlantic Fisheries Manage-ment Council. This is contribution 4750 of the University ofMaryland Center for Environmental Science.

REFERENCESBerkson, J., L. Barbieri, S. X. Cadrin, S. L. Cass-Calay, P. Crone, M.

Dorn, C. Friess, D. Kobayashi, T. J. Miller, W. S. Patrick, S. Pautzke,S. Ralston, and M. Trianni. 2011. Calculating acceptable biologicalcatch for stocks that have reliable catch data only (only reliable catchstocks–ORCS). NOAA Technical Memorandum NMFS-SEFSC-616. Avail-able: http://www.pifsc.noaa.gov/tech/NOAA Tech Memo SEFSC 616.pdf.(February 2011).

Cope, J. M., and A. E. Punt. 2009. Length-based reference points for data-limited situations: applications and restrictions. Marine and Coastal Fisheries:Dynamics, Management, and Ecosystem Science [online serial] 1:169–186.

Costello, C., D. Ovando, R. Hilborn, S. G. Gaines, O. Deschenes, and S. E.Lester. 2012. Status and solutions for the world’s unassessed fisheries. Science338:517–520.

Deroba, J. J., and J. R. Bence. 2008. A review of harvest policies: understandingrelative performance of control rules. Fisheries Research 94:210–223.

Dick, E. J., and A. D. MacCall. 2011. Depletion-based stock reduction analysis:a catch-based method for determining sustainable yields for data-poor fishstocks. Fisheries Research 110:331–341.

Hilborn, R. 2008. Pretty good yield and exploited fishes. Marine Policy 34:193–196.

MacCall, A. D. 2009. Depletion-corrected average catch: a simple formula forestimating sustainable yields in data-poor situations. ICES Journal of MarineScience 66:2267–2271.

Myers, R. A., K. G. Bowen, and N. J. Barrowman. 1999. Maximum reproductiverates of fish at low population sizes. Canadian Journal of Fisheries and AquaticSciences 56:2404–2419.

NEFSC (Northeast Fisheries Science Center). 2002. Final report of theworking group on re-evaluation of biological reference points for NewEngland groundfish. NEFSC, Reference Document 02-04, Woods Hole,

Massachusetts. Available: http://www.nefsc.noaa.gov/nefsc/publications/.(July 2010).

NEFSC (Northeast Fisheries Science Center). 2006. 43rd Northeast re-gional stock assessment workshop (43rd SAW), stock assessment re-view committee (SARC), consensus summary of assessments. NEFSC,Reference Document 06-25, Woods Hole, Massachusetts. Available:http://www.nefsc.noaa.gov/nefsc/publications/. (April 2011).

NEFSC (Northeast Fisheries Science Center). 2008. 47th Northeast re-gional stock assessment workshop (47th SAW), assessment report. NEFSC,Reference Document 08-12a, Woods Hole, Massachusetts. Available:http://www.nefsc.noaa. gov/nefsc/publications/. (October 2011).

NEFSC (Northeast Fisheries Science Center). 2010. 49th Northeast re-gional stock assessment workshop (49th SAW), assessment report.NEFSC, Reference Document 10-03, Woods Hole, Massachusetts. Availablehttp://www.nefsc.noaa. gov/nefsc/publications/. (March 2012).

Patrick, W. S., P. Spencer, O. Ormseth, J. Cope, J. Field, D. Kobayashi, T.Gedamke, E. Cortes, K. Bigelow, W. Overholtz, J. Link, and P. Lawson.2009. Use of productivity and susceptibility indices to determine stock vul-nerability, with example applications to six U.S. fisheries. NOAA Techni-cal Memorandum NMFS-F/SPO-101. Available: http://www.pifsc.noaa.gov/library/pubs/NOAA TM NMFS F-SPO 101.pdf. (November 2012).

Prager, M. H., and K. W. Shertzer. 2010. Deriving acceptable biologicalcatch from the overfishing limit: implications for assessment models. NorthAmerican Journal of Fisheries Management 30:289–294.

Quinn, T., and R. Deriso. 1999. Quantitative fish dynamics, Oxford UniversityPress, New York.

Ralston, S., A. E. Punt, O. S. Hamel, J. D. DeVore, and R. J. Conser. 2011. Ameta-analytic approach to quantifying scientific uncertainty in stock assess-ments. U.S. National Marine Fisheries Service Fishery Bulletin 109:217–231.

Restrepo, V. R., G. G. Thompson, P. M. Mace, W. L. Gabriel, L. L. Low, A. D.MacCall, R. D. Methot, J. E. Powers, B. L. Taylor, P. R. Wade, and J. F. Witzig.1998. Technical guidance on the use of precautionary approaches to imple-menting National Standard 1 of the Magnuson–Stevens Fishery Conservationand Management Act. NOAA Technical Memorandum NMFS-F/SPO-31.

Scandol, J. P. 2003. Use of cumulative sum (CUSUM) control charts of landedcatch in the management of fisheries. Fisheries Research 64:19–36.

Shepherd, J. G. 1982. A versatile new stock–recruitment relationship for fish-eries, and the construction of sustainable yield curves. Journal du ConseilInternational pour l’Exploration de la Mer 40:67–75.

Shertzer, K., M. Prager, and E. Williams. 2008. A probability-based approachto setting annual catch levels. U.S. National Marine Fisheries Service FisheryBulletin 106:225–232.

Thorson, J. T., T. A. Branch, and O. P Jensen. 2012a. Using model-basedinference to evaluate global fisheries status from landings, location, and lifehistory data. Canadian Journal of Fisheries and Aquatic Sciences 69:645–655.

Thorson, J. T., J. M. Cope, T. A. Branch, and O. P Jensen. 2012b. Spawningbiomass reference points for exploited marine fishes, incorporating taxo-nomic and body size information. Canadian Journal of Fisheries and AquaticSciences 69:1556–1568.

Vasconcellos, M., and K. Cochrane. 2005. Overview of world status of data-limited fisheries: inferences from landings statistics. Pages 1–20 in G. H.Kruse, V. F. Gallucci, D. E. Hay, R. I. Perry, R. M. Peterman, T. C. Shirley,P. D. Spencer, B. Wilson, and D. Woodby, editors. Fisheries assessment andmanagement in data-limited situations. Alaska Sea Grant College Program,University of Alaska, Fairbanks.

Walters, C. J., S. J. D. Martell, and J. Korman. 2006. A stochastic approach tostock reduction analysis. Canadian Journal of Fisheries and Aquatic Sciences63:212–223.

Wetzel, C. R., and A. E. Punt. 2011. Model performance for the determination ofappropriate harvest levels in the case of data-poor stocks. Fisheries Research110:342–355.

Zhou, S., S. Yin, J. T. Thorson, A.D. M. Smith, and M. Fuller. 2012. Linkingfishing mortality reference points to life history traits: an empirical study.Canadian Journal of Fisheries and Aquatic Sciences 69:1292–1301.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


Appendix: Detailed Results

TABLE A.1. Means and coefficients of variation (in parentheses) for the performance measures calculated for all control rules for the unbiased run across all ofthe life history and exploitation history scenarios explored. The control rules are as follows: ORCS = the only reliable catch series (with θ1, θ2, and θ3 referringto the buffering options shown in Table 4), DCAC = the depletion-corrected average catch, and DB-SRA = the depletion-based stock reduction analysis. For theperformance measures, C/MSY is the ratio of the mean catch to the true maximum sustainable yield and S/SMSY is the ratio of the mean spawning biomass overthe final 10 years in the simulation to the true spawning biomass that produces MSY.

Fast Medium Slow

Harvest Overfishing Overfishing Overfishingpressure Control rule probability C/MSY S/SMSY probability C/MSY S/SMSY probability C/MSY S/SMSY

Underexploited Median 0.02 (4.50) 0.75 (0.09) 1.95 (0.25) 0.00 (14.95) 0.72 (0.09) 1.79 (0.19) 0.00 (31.62) 0.86 (0.09) 1.58 (0.13)75% of median 0.00 (∞) 0.57 (0.09) 2.39 (0.17) 0.00 (∞) 0.54 (0.09) 2.04 (0.15) 0.00 (∞) 0.64 (0.09) 1.73 (0.12)50% of median 0.00 (∞) 0.38 (0.09) 2.74 (0.14) 0.00 (∞) 0.36 (0.09) 2.27 (0.13) 0.00 (∞) 0.43 (0.09) 1.86 (0.11)ORCS θ1 (median) 0.08 (1.16) 0.82 (0.12) 1.73 (0.23) 0.00 (5.21) 0.78 (0.18) 1.71 (0.12) 0.00 (13.66) 0.85 (0.24) 1.58 (0.08)ORCS θ1 (75th %) 0.19 (0.82) 0.90 (0.11) 1.43 (0.27) 0.03 (2.25) 0.86 (0.16) 1.56 (0.14) 0.03 (2.32) 0.95 (0.22) 1.51 (0.08)ORCS θ2 (median) 0.05 (1.52) 0.77 (0.13) 1.88 (0.19) 0.00 (∞) 0.69 (0.18) 1.84 (0.12) 0.00 (∞) 0.69 (0.23) 1.69 (0.09)ORCS θ2 (75th %) 0.13 (1.02) 0.86 (0.12) 1.59 (0.25) 0.00 (6.74) 0.77 (0.17) 1.73 (0.12) 0.00 (∞) 0.79 (0.24) 1.64 (0.08)ORCS θ3 (median) 0.01 (2.53) 0.71 (0.13) 2.03 (0.17) 0.00 (∞) 0.58 (0.17) 1.99 (0.13) 0.00 (∞) 0.46 (0.20) 1.84 (0.10)ORCS θ3 (75th %) 0.07 (1.32) 0.80 (0.13) 1.78 (0.22) 0.00 (∞) 0.66 (0.18) 1.88 (0.12) 0.00 (∞) 0.54 (0.21) 1.79 (0.10)Restrepo (median) 0.00 (31.62) 0.56 (0.09) 2.4 (0.17) 0.00 (∞) 0.54 (0.09) 2.05 (0.15) 0.00 (∞) 0.64 (0.09) 1.72 (0.12)Restrepo (stable) 0.01 (5.66) 0.62 (0.23) 2.28 (0.23) 0.00 (15.29) 0.6 (0.22) 1.96 (0.19) 0.00 (∞) 0.70 (0.20) 1.70 (0.13)DCAC (base) 0.01 (6.98) 0.71 (0.07) 2.06 (0.22) 0.00 (∞) 0.63 (0.09) 1.92 (0.15) 0.00 (∞) 0.62 (0.11) 1.73 (0.11)DCAC (fixed) 0.05 (7.78) 0.70 (0.07) 2.09 (0.22) 0.00 (∞) 0.59 (0.08) 1.97 (0.15) 0.00 (∞) 0.51 (0.07) 1.80 (0.12)DB-SRA (base) 0.52 (0.24) 0.89 (0.11) 1.00 (0.22) 0.44 (0.57) 1.01 (0.17) 1.00 (0.20) 0.37 (1.04) 1.25 (0.23) 1.26 (0.09)DB-SRA

(projected)0.65 (0.64) 0.96 (0.13) 0.85 (0.85) 0.37 (1.16) 1.13 (0.14) 1.07 (0.54) 0.26 (1.51) 1.31 (0.19) 1.32 (0.15)

DB-SRA (fixed) 0.73 (0.39) 0.94 (0.13) 0.56 (0.83) 0.01 (5.75) 0.81 (0.09) 1.63 (0.20) 0.00 (∞) 0.62 (0.08) 1.73 (0.12)Fully exploited Median 0.40 (0.98) 0.86 (0.10) 1.19 (0.60) 0.55 (0.73) 0.95 (0.12) 0.9 (0.52) 0.99 (0.05) 1.37 (0.09) 0.69 (0.28)

75% of median 0.03 (4.06) 0.68 (0.08) 2.10 (0.23) 0.04 (3.36) 0.75 (0.09) 1.49 (0.25) 0.65 (0.61) 1.03 (0.09) 0.96 (0.20)50% of median 0.00 (31.62) 0.46 (0.09) 2.60 (0.15) 0.00 (∞) 0.5 (0.09) 1.93 (0.16) 0.00 (17.73) 0.69 (0.09) 1.22 (0.15)ORCS θ1 (median) 0.19 (0.87) 0.86 (0.12) 1.47 (0.27) 0.09 (2.13) 0.83 (0.12) 1.36 (0.21) 0.42 (0.99) 0.96 (0.09) 1.02 (0.18)ORCS θ1 (75th %) 0.41 (0.53) 0.90 (0.12) 1.15 (0.35) 0.42 (0.86) 0.94 (0.11) 1.04 (0.34) 0.92 (0.24) 1.16 (0.08) 0.86 (0.23)ORCS θ2 (median) 0.11 (1.1) 0.82 (0.12) 1.65 (0.23) 0.01 (5.91) 0.73 (0.14) 1.59 (0.16) 0.00 (15.14) 0.69 (0.11) 1.21 (0.15)ORCS θ2 (75th %) 0.26 (0.77) 0.88 (0.12) 1.35 (0.31) 0.08 (2.28) 0.82 (0.13) 1.38 (0.22) 0.10 (2.53) 0.84 (0.09) 1.11 (0.17)ORCS θ3 (median) 0.06 (1.37) 0.77 (0.13) 1.83 (0.21) 0.00 (∞) 0.62 (0.16) 1.78 (0.12) 0.00 (∞) 0.43 (0.16) 1.40 (0.13)ORCS θ3 (75th %) 0.15 (0.99) 0.84 (0.12) 1.53 (0.27) 0.00 (6.77) 0.69 (0.15) 1.65 (0.14) 0.00 (∞) 0.52 (0.13) 1.35 (0.14)Restrepo (median) 0.01 (6.00) 0.65 (0.09) 2.15 (0.20) 0.00 (11.61) 0.71 (0.11) 1.60 (0.20) 0.17 (1.41) 0.89 (0.16) 1.07 (0.12)Restrepo (stable) 0.04 (2.78) 0.68 (0.19) 2.03 (0.26) 0.04 (3.31) 0.73 (0.19) 1.53 (0.23) 0.27 (1.28) 0.90 (0.21) 1.06 (0.15)DCAC (base) 0.23 (1.38) 0.84 (0.08) 1.44 (0.45) 0.11 (2.14) 0.84 (0.08) 1.33 (0.27) 0.06 (3.10) 0.86 (0.09) 1.10 (0.15)DCAC (fixed) 0.25 (1.35) 0.83 (0.09) 1.43 (0.46) 0.14 (1.90) 0.83 (0.08) 1.32 (0.30) 0.11 (2.34) 0.85 (0.07) 1.09 (0.17)DB-SRA (base) 0.51 (0.24) 0.86 (0.12) 0.96 (0.22) 0.45 (0.59) 0.94 (0.13) 1.00 (0.18) 0.53 (0.76) 1.02 (0.19) 0.98 (0.10)DB-SRA

(projected)0.61 (0.65) 0.92 (0.13) 0.83 (0.75) 0.47 (0.87) 1.00 (0.15) 0.96 (0.49) 0.54 (0.79) 1.00 (0.17) 0.98 (0.16)

DB-SRA (fixed) 0.99 (0.04) 0.85 (0.12) 0.33 (0.19) 0.75 (0.46) 0.96 (0.17) 0.68 (0.69) 0.57 (0.73) 0.99 (0.07) 0.98 (0.20)Overexploited Median 0.72 (0.49) 0.76 (0.14) 0.79 (0.80) 0.98 (0.09) 0.48 (0.35) 0.13 (1.50) 1.00 (0.00) 0.69 (0.21) 0.03 (0.50)

75% of median 0.23 (1.23) 0.66 (0.09) 1.88 (0.36) 0.76 (0.42) 0.58 (0.29) 0.53 (0.97) 1.00 (0.00) 0.77 (0.23) 0.07 (0.93)50% of median 0.03 (2.07) 0.45 (0.09) 2.58 (0.16) 0.20 (1.15) 0.49 (0.13) 1.38 (0.33) 1.00 (0.03) 0.78 (0.11) 0.31 (0.52)ORCS θ1 (median) 0.15 (0.94) 0.79 (0.12) 1.52 (0.27) 0.26 (1.00) 0.68 (0.15) 1.11 (0.29) 0.97 (0.08) 0.69 (0.20) 0.45 (0.18)ORCS θ1 (75th %) 0.35 (0.6) 0.85 (0.12) 1.15 (0.35) 0.75 (0.31) 0.75 (0.21) 0.66 (0.49) 1.00 (0.00) 0.79 (0.14) 0.30 (0.41)ORCS θ2 (median) 0.10 (1.11) 0.75 (0.13) 1.67 (0.24) 0.06 (1.96) 0.59 (0.16) 1.36 (0.22) 0.64 (0.3) 0.60 (0.21) 0.56 (0.15)ORCS θ2 (75th %) 0.23 (0.81) 0.82 (0.12) 1.35 (0.31) 0.40 (0.74) 0.71 (0.14) 0.97 (0.33) 0.961 (0.08) 0.67 (0.19) 0.46 (0.18)ORCS θ3 (median) 0.06 (1.45) 0.70 (0.14) 1.84 (0.20) 0.01 (3.90) 0.49 (0.18) 1.61 (0.16) 0.02 (4.14) 0.40 (0.17) 0.72 (0.17)ORCS θ3 (75th %) 0.15 (0.97) 0.78 (0.13) 1.52 (0.27) 0.07 (1.81) 0.60 (0.14) 1.35 (0.24) 0.41 (0.61) 0.52 (0.17) 0.63 (0.17)Restrepo (median) 0.01 (3.81) 0.59 (0.10) 2.18 (0.20) 0.01 (3.28) 0.57 (0.14) 1.45 (0.20) 0.626 (0.31) 0.60 (0.21) 0.56 (0.15)Restrepo (stable) 0.03 (2.87) 0.58 (0.23) 2.15 (0.27) 0.07 (2.85) 0.52 (0.28) 1.50 (0.27) 0.50 (0.82) 0.56 (0.29) 0.58 (0.25)DCAC (base) 0.61 (0.61) 0.76 (0.13) 0.99 (0.70) 0.93 (0.21) 0.55 (0.35) 0.27 (1.34) 1.00 (0.00) 0.83 (0.16) 0.19 (0.69)DCAC (fixed) 0.65 (0.55) 0.76 (0.13) 0.92 (0.76) 0.96 (0.15) 0.51 (0.36) 0.19 (1.52) 1.00 (0.00) 0.81 (0.21) 0.11 (0.88)DB-SRA (base) 0.44 (0.25) 0.73 (0.13) 0.96 (0.22) 0.24 (0.80) 0.67 (0.15) 1.04 (0.16) 0.42 (0.83) 0.55 (0.20) 0.60 (0.16)DB-SRA

(projected)0.15 (1.88) 0.88 (0.13) 1.69 (0.39) 0.22 (1.53) 0.77 (0.16) 1.17 (0.35) 0.56 (0.73) 0.53 (0.19) 0.60 (0.19)

DB-SRA (fixed) 0.99 (0.01) 0.77 (0.13) 0.33 (0.20) 1.00 (0.01) 0.44 (0.27) 0.08 (0.84) 1.00 (0.00) 0.75 (0.24) 0.06 (0.98)

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


TABLE A.2. Means and coefficients of variation (in parentheses) for the performance measures calculated for all control rules for the biased run across all ofthe life history and exploitation history scenarios explored. Results are omitted for control rules that did not require stock classification (the median catch, 75% ofthe median, and 50% of median) as well as for those that had an assumed fixed stock abundance across years (DCAC and DB-SRA with assumed biomass of 40%of S0). See Table A.1 for additional details.

Fast Medium Slow

Harvest Overfishing Overfishing Overfishingpressure Control rule probability C/MSY S/SMSY probability C/MSY S/SMSY probability C/MSY S/SMSY

Fully exploited ORCS θ1 (median) 0.91 (0.12) 0.85 (0.15) 0.45 (0.55) 0.99 (0.02) 0.80 (0.23) 0.16 (0.85) 1.00 (0.00) 1.55 (0.13) 0.37 (0.35)ORCS θ1 (75th %) 0.98 (0.06) 0.83 (0.14) 0.34 (0.39) 1.00 (0.00) 0.72 (0.20) 0.1 (0.51) 1.00 (0.00) 1.57 (0.14) 0.20 (0.56)ORCS θ2 (median) 0.83 (0.19) 0.87 (0.15) 0.6 (0.56) 0.95 (0.14) 0.94 (0.20) 0.41 (0.70) 0.99 (0.05) 1.33 (0.1) 0.70 (0.28)ORCS θ2 (75th %) 0.95 (0.09) 0.84 (0.15) 0.39 (0.53) 0.99 (0.02) 0.81 (0.23) 0.17 (0.84) 1.00 (0.00) 1.49 (0.12) 0.51 (0.30)ORCS θ3 (median) 0.70 (0.28) 0.89 (0.14) 0.80 (0.51) 0.50 (0.72) 0.95 (0.10) 0.98 (0.38) 0.08 (2.82) 0.83 (0.08) 1.11 (0.17)ORCS θ3 (75th %) 0.88 (0.15) 0.86 (0.15) 0.52 (0.56) 0.87 (0.26) 0.97 (0.17) 0.57 (0.61) 0.56 (0.75) 1.00 (0.08) 0.98 (0.20)Restrepo (median) 0.02 (3.76) 0.68 (0.08) 2.08 (0.21) 0.04 (3.45) 0.75 (0.09) 1.50 (0.24) 0.65 (0.60) 1.03 (0.08) 0.96 (0.20)Restrepo (stable) 0.13 (2.01) 0.71 (0.19) 1.88 (0.37) 0.15 (2.01) 0.77 (0.19) 1.43 (0.34) 0.61 (0.73) 1.06 (0.20) 0.92 (0.26)DCAC (base) 0.29 (1.18) 0.85 (0.09) 1.33 (0.49) 0.19 (1.54) 0.88 (0.09) 1.21 (0.28) 0.42 (0.96) 0.97 (0.10) 1.01 (0.16)DB-SRA

(projected)0.87 (0.32) 0.84 (0.14) 0.51 (0.91) 0.94 (0.20) 0.84 (0.20) 0.31 (1.23) 1.00 (0.00) 1.53 (0.16) 0.45 (0.41)

DB-SRA (base) 0.70 (0.15) 0.90 (0.13) 0.70 (0.25) 0.93 (0.09) 0.95 (0.15) 0.54 (0.26) 1.00 (0.00) 1.32 (0.16) 0.64 (0.13)Overexploited ORCS θ1 (median) 0.87 (0.16) 0.79 (0.17) 0.49 (0.59) 0.99 (0.04) 0.56 (0.35) 0.17 (1.03) 1.00 (0.00) 0.79 (0.20) 0.08 (0.89)

ORCS θ1 (75th %) 0.97 (0.06) 0.77 (0.15) 0.35 (0.41) 1.00 (0.00) 0.44 (0.28) 0.08 (0.50) 1.00 (0.00) 0.70 (0.19) 0.03 (0.43)ORCS θ2 (median) 0.76 (0.23) 0.83 (0.16) 0.64 (0.54) 0.93 (0.13) 0.68 (0.29) 0.37 (0.76) 1.00 (0.00) 0.82 (0.15) 0.27 (0.46)ORCS θ2 (75th %) 0.93 (0.11) 0.79 (0.16) 0.42 (0.51) 1.00 (0.01) 0.52 (0.35) 0.13 (0.93) 1.00 (0.00) 0.80 (0.20) 0.09 (0.86)ORCS θ3 (median) 0.60 (0.36) 0.84 (0.15) 0.83 (0.51) 0.64 (0.45) 0.74 (0.17) 0.77 (0.43) 0.84 (0.18) 0.66 (0.21) 0.50 (0.15)ORCS θ3 (75th %) 0.84 (0.19) 0.81 (0.17) 0.53 (0.59) 0.95 (0.09) 0.67 (0.31) 0.33 (0.77) 1.00 (0.00) 0.74 (0.16) 0.39 (0.25)Restrepo (median) 0.07 (1.72) 0.63 (0.10) 2.10 (0.22) 0.42 (0.77) 0.65 (0.15) 1.01 (0.38) 1.00 (0.00) 0.81 (0.14) 0.29 (0.46)Restrepo (stable) 0.13 (1.90) 0.61 (0.23) 2.05 (0.34) 0.35 (1.16) 0.56 (0.27) 1.12 (0.52) 0.95 (0.17) 0.73 (0.22) 0.34 (0.60)DCAC (base) 0.62 (0.57) 0.76 (0.14) 0.98 (0.70) 0.95 (0.16) 0.55 (0.37) 0.24 (1.31) 1.00 (0.00) 0.85 (0.17) 0.18 (0.69)DB-SRA

(projected)0.45 (0.94) 0.76 (0.12) 1.20 (0.64) 0.77 (0.43) 0.68 (0.22) 0.59 (0.82) 1.00 (0.03) 0.78 (0.18) 0.36 (0.32)

DB-SRA (base) 0.61 (0.17) 0.86 (0.13) 0.73 (0.25) 0.90 (0.11) 0.81 (0.17) 0.61 (0.23) 1.00 (0.00) 0.68 (0.18) 0.44 (0.15)

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013



A Standardized Technique to Back-Calculate Length atAge from Unsectioned Walleye SpinesJonathan R. Meerbeek a & Kimberly A. Hawkins aa Iowa Department of Natural Resources , 122 252nd Avenue, Spirit Lake , Iowa , 51360 , USAPublished online: 08 Aug 2013.

To cite this article: Jonathan R. Meerbeek & Kimberly A. Hawkins (2013) A Standardized Technique to Back-CalculateLength at Age from Unsectioned Walleye Spines, North American Journal of Fisheries Management, 33:4, 861-868, DOI:10.1080/02755947.2013.812583







http://dx.doi.org/10.1080/02755947.2013.812583




MANAGEMENT BRIEF

A Standardized Technique to Back-Calculate Length at Agefrom Unsectioned Walleye Spines

Jonathan R. Meerbeek* and Kimberly A. HawkinsIowa Department of Natural Resources, 122 252nd Avenue, Spirit Lake, Iowa 51360, USA

AbstractThe primary assumption of using back-calculated length-at-age

(BCL) methods to estimate fish growth is that the body length andhard-part radius relationship is linear. Therefore, standardizedtechniques are required for BCL methods to be comparable overtime. Recent studies have found that unsectioned dorsal spines ofSander spp. improve efficiency and provide accurate age estimates.We evaluated a simple, standardized method of using unsectionedWalleye S. vitreus (N = 162) dorsal spines to estimate age andBCL (direct-proportion and Fraser–Lee methods) by comparingthese estimates with those of a sectioned spine technique. We val-idated the technique by comparing observed growth (measuredvia a mark–recapture study) with BCL growth estimated via sec-tioned and unsectioned techniques. The time required to process,prepare, view and perform BCL for each of the two techniqueswas also compared. Both reader age estimates and exact readeragreement rates were significantly better using the sectioned spinetechnique. Age frequencies varied significantly among readers andtechniques as differences ranged up to 8 years. No difference wasdetected among the slopes of the back-calculation methods amongtechniques, but there were significant differences in the mean BCLestimates for ages < 10 years old. The relationships between ob-served growth and growth estimated via BCL methods were signif-icant; however, BCL methods significantly overestimated Walleyegrowth. Total processing time for the unsectioned technique wassignificantly less, but based on our poor ability to replicate ages,we do not suggest the use of our standardized unsectioned spinetechnique to estimate the age or BCL older Walleye. In addition,no BCL method accurately estimated annual growth regardless oftechnique or reader experience. We conclude that more researchis necessary to validate dorsal spine-based BCL estimates on Wall-eye populations. Once validated, further research evaluating newor modifying our technique for unsectioned Walleye spines needsto be explored.

Accurate fish age and growth information is necessary toproperly assess the health of a fishery (e.g., estimating growth,recruitment, and mortality); hence, managers are constantly insearch of the most efficient technique to estimate the age of

*Corresponding author: [email protected] February 14, 2013; accepted June 2, 2013

numerous fish. In general, otoliths are the preferred structureused for age estimation for many fish species because they ac-curately reflect the fish’s age (Heidinger and Clodfelter 1987;Buckmeier and Howells 2003; Brown et al. 2004; DeCicco andBrown 2006) and are quicker to process and view than manyother structures (Kocovsky and Carline 2000; Isermann et al.2003). The drawback to using otoliths is that fish sacrifice isrequired. Dorsal spines offer a nonlethal alternative to otolithsfor estimating fish age (Campbell and Babaluk 1979; Erickson1983; Borkholder and Edwards 2001), but preparation time andthe need for specialized equipment have limited the use ofspines even though age estimates closely mimic those of otoliths(Campbell and Babaluk 1979; Belanger and Hogler 1982; Logs-don 2007). Recently, Buckmeier et al. (2002) introduced a tech-nique for preparing and viewing the pectoral spines of ChannelCatfish Ictalurus punctatus that involved sanding the proximalend of the cut spine and viewing under a microscope with sideillumination as described by Heidinger and Clodfelter (1987).This technique negated the need for mounting and sectioningthe spine, and did not require the use of specialized equip-ment (i.e., low-speed saw, microscope slide, mounting mediumfor processing). Logsdon (2007) modified this technique foruse on the dorsal spines of Walleye Sander vitreus and foundthat unsectioned spines closely replicated otolith-based esti-mates of population age structure. Furthermore, Williamson andDirnberger (2010) compared precision of age estimates and pro-cessing times of sectioned (as described by Beamish and Chilton1977) and unsectioned dorsal spines of Sauger S. canadensis,and found that while both techniques yielded similar age es-timates, unsectioned dorsal spines took one-half as much pro-cessing time. In addition, side illumination of the unsectionedspine increased visibility of the crowded outer annuli on olderfish as compared with that of the sectioned technique wherethose same annuli appeared split or discrete when viewed as athin section (Logsdon 2007; Williamson and Dirnberger 2010).

861

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013

862 MEERBEEK AND HAWKINS

The use of unsectioned dorsal spines to estimate Walleye agehas improved efficiency and has provided an additional meansof age estimation without sacrificing fish; however, many stateand federal agencies often use bony structures to back-calculatelength-at-age (BCL) estimates to (1) increase sample size oflength-at-age data to fit growth curves; (2) estimate length atage for ages that are rarely observed; (3) compare growth ratesbetween sexes, cohorts, or populations of the same species; and(4) examine growth rates in populations where annual sam-pling is not feasible (Francis 1990; Maceina et al. 2007). Hence,many BCL methods have been developed for various bony struc-tures since the early 1900s for these purposes (Dahl 1907; Lea1910; Fraser 1916; Lee 1920; Whitney and Carlander 1956;Weisberg 1986; Campana 1990; Ricker 1992). The two mostcommonly used BCL methods used in fisheries surveys todayare the direct-proportional method (DP; Lea 1910) and Fraser–Lee method (FL; Fraser 1916; Lee 1920; Maceina et al. 2007).The primary assumption of using BCL to estimate the fish lengthat age x is that the body length and hard-part radius relationshipis linear. These methods also rely on the following conditions:(1) the radius of yearly marks formed on the bony structuresmust remain constant from the time of formation; (2) the markis a true annulus, formed on a yearly basis; and (3) the formulaused to back-calculate growth accurately describes the body :bony structure relation (Francis 1990). To this end, each newstructure should be validated with fish of known age (Beamishand McFarlane 1983). Dorsal spine BCL estimates have neverbeen validated using fish of known age, but Schram (1989) vali-dated annulus formation for Walleye dorsal spines in a multiyearmark–recapture study and since then, dorsal spines have beenwidely adopted as a reliable structure to estimate Walleye age.

Borkholder and Edwards (2001) were the first to compareestimates of BCLs from scales and dorsal spine sections forWalleye along various transects and found that dorsal spineBCL estimates compared favorably with those of scales. Con-sequently, many agencies have made adjustments to incorporatedorsal spines as a valid technique to estimate age and growthparameters for nonlethal fish surveys. The unsectioned spinetechnique proposed by Logsdon (2007) and later found to be abetter and quicker technique as compared with sectioned spines(Williamson and Dirnberger 2010) has become an attractivealternative to estimate the age of numerous fish, but no BLCtechnique has been described or validated for this technique,thus limiting its use in traditional fisheries surveys. Our primaryobjective was to develop a standardized technique to performBCL on unsectioned dorsal spines that could be used with thetwo most common BCL methods (i.e., DP and FL). To validateour technique, we (1) compared the relative precision of ageestimates between sectioned and unsectioned dorsal spines, (2)compared BCL estimates obtained from measurements of sec-tioned and standardized unsectioned spine techniques for eachBCL method, and (3) compared observed growth increments(via a mark–recapture study) with those obtained from mea-surements of sectioned and standardized unsectioned spine tech-

niques for each BCL method. Lastly, we compared total process-ing time (i.e., removal, cleaning, processing, viewing, and per-forming back-calculation measurements) associated with sec-tioned or the standardized unsectioned dorsal spine technique.

METHODSWalleye (n = 2,794) were collected via gill nets (6 ft ×

320 ft × 2.5-in bar mesh) in April 2011 from three naturallakes in northwest Iowa and were transported to a hatchery fa-cility where they were measured to the nearest 0.1 in (TL),sexed (determined by extrusion of gametes), and individuallytagged (Meerbeek et al. 2013). Additionally, the first two dor-sal spines of each Walleye were removed (for age estimation)from the point of attachment with a pair of side cutters beforebeing released back to the lake of capture. In April 2012 thesemethods were repeated; tag number and TL was recorded foreach recaptured fish, and the third dorsal spine was removed forage estimation. Detached spines during both years were storedin individually labeled coin envelopes for at least 2 months andwere not cleaned prior to storage.

In the laboratory, a scalpel was used to remove dried skinfrom dorsal spines. Three to five cross sections (0.039 in) weretaken from the dorsal spine with a Dremel tool fitted with a den-tal saw blade (Margenau 1982). The Dremel tool was fastenedto a custom-built stand equipped with a sliding metric ruler. Theremaining section of spine from each Walleye was placed backinto coin envelopes (for processing time comparisons). Crosssections were placed on a microscope slide and viewed undera dissecting microscope (30× magnification) with transmittedlight. Digital images were captured using an Olympus DP70camera mounted on an Olympus SZ6045 microscope, and crosssections were viewed using Image-Pro Plus software calibratedat 30× magnification. Two readers (always the same individ-uals) of varying experience levels (i.e., >10-years experienceor <5-years experience at estimating fish age) independentlyestimated the age of each Walleye using a dorsal spine crosssection of their choice. Each reader identified the focus and mea-sured the distance from the focus to each annuli and the spineedge along the horizontal compressed transect (Borkholder andEdwards 2001) via the Image-Pro Plus software program. Theremaining segment of dorsal spine for each Walleye was againaffixed to the Dremel tool stand and cut to a length of 0.39 in(standardized dorsal spine section). This section was placed ina small block of modeling clay used to hold the distal end ofthe spine flush with the base of the microscope stage plate forviewing at a standardized distance of 0.39 in. We calibrated theImage-Pro Plus software at a distance of 0.39 in using a mi-crometer at 40× magnification. The proximal end of the stan-dardized dorsal spine was coated with mineral oil and examinedunder the Olympus SZ6045 microscope with side illumination(Buckmeier et al. 2002). To ensure that the focal plane was notdistorted from mineral oil beading on the surface of the spine,mineral oil was finely brushed on the surface of the spine via an

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


artist brush. Side illumination was provided by a 150-W halogenilluminator light box and was transmitted across the proximalend of the spine with a 45◦ fiber optic light guide. The light guidewas firmly secured, and digital images were captured, viewed,and read using the same techniques as described above.

The coefficient of variation (CV) was used to measure preci-sion of age estimates from spines between paired assignmentsof age and was calculated as

CV j = 100 ×√∑R

i=1(Xij−X j )2

R−1

X j,

where R is the number of times the age of each fish was esti-mated, Xj is the mean age estimated for the jth fish, and Xij is theith age for the jth fish (Chang 1982). Mean CVs for sectionedand standardized unsectioned spine techniques were comparedusing a paired t-test (SAS Institute 2001). Reader agreementrates (exact agreement and agreement within ± 1 year) forboth techniques were compared via chi-square analysis (SASInstitute 2001). To test each reader’s ability to replicate agesamong the two techniques, technique agreement rates (exactagreement and agreement within ± 1 year) were comparedalso using chi-square analysis. Age frequencies were comparedamong readers for each technique using a Kolmogorov–Smirnovtest (SAS Institute 2001).

Simple linear regression was used to examine the relationshipbetween body length and dorsal spine radius among readersand techniques (SAS Institute 2001). Two measures of back-calculation were calculated for both techniques and for eachreader, expressed as

DP (Lea 1910): Li = Si

ScLc,

and

FL (Fraser 1916; Lee 1920): Li + (Lc − a)Si

Sc,

where Li is the back-calculated length of the fish when the ithincrement was formed, Lc is the fish length at capture, Sc isthe radius of spine at capture, Si is the radius of the spine atthe ith increment, and a is the biological intercept for the de-pendent variable (i.e., mean back-calculated TL) based on thegeneral linear model procedure (Isely and Grabowski 2007).We used ANCOVA (SAS Institute 2001) to test the slope of thelength-at-age regression between the two techniques (sectionedand standardized unsectioned) for each BCL method (Isely andGrabowski 2007). Simple linear regression was used to exam-ine the relationship between observed growth measurements andlast increment growth estimated via BCL methods (SAS Insti-tute 2001). We used ANOVA followed by Tukey’s post hoc testto compare mean annual growth increments since the last annu-lus formation from the BCL methods (for each reader and for

each technique) with those from TL measurements obtained viaconsecutive annual recaptures (SAS Institute 2001). This proce-dure was also used to test for differences in mean BCL estimates(ages 1–12) for each reader using the four BCL methods (DP-sectioned, DP-unsectioned, FL-sectioned, FL-unsectioned).

Two workers randomly selected 200 spine samples and inde-pendently measured the time (min) to open the coin envelope,separate the spines, and clean the spine to be inspected using ascalpel using groups of 10 fish (N = 10 groups of 10 fish perindividual). The workers also estimated the time to cut stan-dardized dorsal spine sections (0.39 in; i.e., process time) usinggroups of 10 fish (N = 20 groups of 10 fish per individual)from spines used in this study and those from an independentsample of Walleye containing similar sizes and sex ratios. Eachworker also estimated viewing time for both sectioned and stan-dardized spine section techniques using groups of 10 fish (N =10 groups of 10 fish per individual) by recording the time toestimate the age; measure the focus, annuli, and spine edge; andcapture digital images of each structure for each 10-fish group.Total processing time for standardized dorsal spine preparationwas derived for 10-fish groups by adding the times requiredto remove (1.5 min, SE = 0.07; Isermann et al. 2003), clean,process, and view dorsal spines. Total processing time for dor-sal spine sections were calculated for each 10-fish group byadding removal and processing times (Isermann et al. 2003) tomean cleaning and viewing time. We compared viewing andtotal processing time between sectioned and standardized spinesection techniques using t-tests (SAS Institute 2001). Alpha wasmaintained at 0.05 for all statistical procedures.

RESULTSOne hundred sixty-two Walleye (mean TL = 21.3 in;

SE = 0.13) were recaptured in April 2012, and their TLsranged from 17.7 to 29.5 in. More males (N = 111; mean TL =20.7 in; SE = 0.10) were recaptured than females (N = 51,mean TL = 22.7 in, SE = 0.27; Figure 1). Mean CV in readerage estimates was significantly less using sectioned spines(CV = 6.30) than unsectioned dorsal spines (CV = 8.13; t <

−2.08, pooled df = 322, P = 0.04). Exact reader agreementrate was also significantly higher with sectioned spines (50.6%)than with unsectioned spines (37.0%; χ2 = 6.07, df = 1, P =0.01). However, we found no significant difference in rates ofreader agreement within ± 1 year between the two techniques(sectioned = 82.1%; unsectioned = 73.5%; χ2 = 3.50, df = 1,P = 0.06). Differences in age assignments of up to 4 years wereobserved with the sectioned technique and up to 8 years withthe unsectioned technique. We also noted significant differencesin reader 1 and reader 2’s ability to exactly replicate ages usingthe two techniques (χ2 = 20.06, df = 1, P < 0.0001). Exactreader agreement in assigned ages among the two techniquesfor reader 1 was 49.4% (80 of 162), whereas reader 2’s exactagreement among techniques was 25.3% (41 of 162). Likewise,there was a significant difference in each reader’s ability to

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


FIGURE 1. Length frequency distribution (TL, in) of recaptured Walleye(N = 162) collected from natural lakes in Iowa during April 2012.

obtain age estimates within ±1 year among the two techniques(84.5% and 59.3%; χ2 = 25.69, df = 1, P < 0.0001).

Significant differences in age frequency assignmentsamong readers were observed for both the sectioned spine(Kolmogorov–Smirnov asymptotic test statistic = 1.78; df =162, 162; P = 0.004) and unsectioned spine techniques(Kolmogorov–Smirnov asymptotic test statistic = 1.61; df =162, 162; P = 0.011; Table 1). The youngest estimated age was

age 4, and only 10.5–25.3% of age assignments for each tech-nique and reader were between 4 and 7 years old. Conversely,44.4–73.5% of age assignments for each technique and readerwere aged at ≥ 10 years old (Table 1).

Linear models used to examine the relationship betweenbody length and dorsal spine radius among readers and tech-niques were all significant (Table 2). There was no difference inthe slope of the length-at-age regression between the two tech-niques (sectioned and standardized unsectioned) for each BCLmethod (ANCOVA: df = 1, P ≥ 0.14). All linear models exam-ining the relationship between observed growth measurementsfrom consecutive Walleye recaptures and last year’s growth es-timated via BCL methods were significant (P < 0.001; Table 3),but the magnitude of growth estimated via BCL methods from2011 to 2012 was significantly higher compared with those fromobserved recaptures (i.e., known growth rates; ANOVA: df =8, P < 0.05; Table 4). Average growth estimated via sectionedspines (both DP and FL BCL methods) for reader 2 was signif-icantly higher than all other growth estimates and nearly threetimes higher than observed growth. Combined, only 16.3% (317of 1,944) of BCL growth increments were within ± 0.25 infrom observed measurements; 13.4% (261 of 1,944) of the BCLmeasurements were ≥± 1.0 in from observed measurements.There were significant differences in estimated BCL for Walleye< 10 years old among techniques and BCL methods (Table 5).

TABLE 1. Mean TL (in) at age for each reader as estimated from sectioned and unsectioned dorsal spines of recaptured Walleyes (n = number examined)collected from natural lakes in Iowa during April 2012.

Sectioned Unsectioned

Reader 1 Reader 2 Reader 1 Reader 2

Age n TL (SE) n TL (SE) n TL (SE) n TL (SE)

1234 2 21.1 (1.50) 1 19.65 5 20.1 (0.51) 9 20.1 (0.42) 7 20.1 (0.38) 2 21.2 (0.20)6 10 21.5 (0.52) 9 20.4 (0.24) 8 20.9 (0.54) 7 20.4 (0.64)7 15 22.8 (0.33) 23 22.5 (0.35) 16 22.6 (0.43) 8 22.1 (0.58)8 17 21.4 (0.48) 18 21.4 (0.40) 8 22.2 (0.62) 13 22.3 (0.51)9 18 20.2 (0.22) 31 20.7 (0.23) 27 21.4 (0.38) 13 21.2 (0.43)

10 26 21.4 (0.38) 35 21.3 (0.30) 22 21.1 (0.49) 17 21.5 (0.64)11 51 21.1 (0.21) 27 21.4 (0.36) 51 20.9 (0.15) 72 21.0 (0.17)12 13 21.7 (0.53) 7 21.7 (0.81) 14 21.2 (0.29) 12 21.4 (0.55)13 2 21.3 (1.20) 2 21.8 (0.60) 4 23.3 (0.97) 7 21.1 (0.46)14 3 21.3 (0.98) 4 21.9 (0.25)15 1 22.7 4 21.9 (0.78)16 2 21.6 (0.40) 1 22.017 1 22.018 1 22.019 2 22.0 (0.75)20

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


TABLE 2. Linear regression statistics relating Walleye (n = 162) body lengthand dorsal spine radius among readers (R1 = reader 1; R2 = reader 2) and back-calculation techniques.

Technique Reader Slope y-intercept r2 P

Sectioned R1 0.0020 0.0042 0.36 <0.0001R2 0.0018 0.0075 0.26 <0.0001

Unsectioned R1 0.0019 0.0056 0.35 <0.0001R2 0.0021 −0.0004 0.33 <0.0001

Mean DP BCL estimates were always less than FL estimateson fish ≤10 years old. Reader 2’s sectioned dorsal spine BCLestimates for ages ≤7 using the DP method were significantlylower than most DP or FL mean BCL estimates (Table 5).

On average, it required 8.6 min (SE = 0.25) to clean dorsalspines for a 10-fish group. Unsectioned spines (7.2 min; SE =0.21) took less time to process than sectioned spines (12.5 min,SE = 0.28; Table 6). There was no significant difference inviewing time among techniques (t < −0.40, pooled df = 38,P = 0.69). Total processing time was less for unsectioneddorsal spines than sectioned spines (t < 4.06, pooled df = 38,P < 0.001; Table 6). Unsectioned spines took approximately4.8 min less for a 10-fish group, saving about 8 h/1,000structures when compared with using sectioned spines.

DISCUSSIONHigh reader agreement rates (i.e., low CV) indicate the ease

of annuli identification (Campana et al. 1995) and are essential instudies evaluating new or modified age estimation techniques.The study results indicate that the standardized unsectioneddorsal spine technique was less precise than that of the sec-tioned spine technique for Walleye collected from natural lakesin northwest Iowa. Reader agreement rates in this study associ-ated with standardized unsectioned Walleye dorsal spines (37%)were not as high as those observed with sectioned spines (51%)nor were they better than those of any other study that usedsectioned spine-based techniques to estimate the age of Walleye

TABLE 3. Linear regression statistics relating BCL last year’s growth of 162Walleye via DP and FL methods to observed growth using both the sectionedand unsectioned dorsal spine age estimation techniques.

BCLType method Reader Slope y-intercept r2 P

Sectioned DP R1 0.08 0.54 0.26 <0.0001DP R2 0.32 0.06 0.19 <0.0001FL R1 0.57 0.08 0.26 <0.0001FL R2 0.34 0.06 0.19 <0.0001

Unsectioned DP R1 0.43 0.12 0.24 <0.0001DP R2 0.34 0.21 0.15 <0.001FL R1 0.45 0.11 0.24 <0.0001FL R2 0.37 0.20 0.14 <0.0001

TABLE 4. Mean 2011–2012 Walleye growth estimates derived from observedTL measurements (i.e., caught and measured two consecutive years) and lastyear’s BCL estimates via the FL and DP methods for each technique (unsec-tioned and sectioned) for both readers (R1 = reader 1; R2 = reader 2). Differentlowercase letters represent significant differences (P < 0.05).

Mean 2011–2012 BCLGrowth estimate growth method

Observed 0.54 (0.04) yR1 unsectioned 0.93 (0.05) z FLR1 sectioned 0.80 (0.04) z FLR2 unsectioned 0.90 (0.04) z FLR2 sectioned 1.41 (0.06) x FLR1 unsectioned 0.98 (0.05) z DPR1 sectioned 0.84 (0.04) z DPR2 unsectioned 0.96 (0.05) z DPR2 sectioned 1.49 (0.06) x DP

(47–81%; Campbell and Babaluk 1979; Olson 1980; Erickson1983; Isermann et al. 2003). In addition, our standardized un-sectioned dorsal spine reader agreement rates were much lowerthan those observed for unsectioned dorsal spines of Walleye intwo Minnesota lakes (70% and 95%; Logsdon 2007) and Saugerin the lower Missouri River (73%; Williamson and Dirnberger2010). This was somewhat expected since standard broodstocksampling in natural lakes in Iowa typically target only large,adult Walleye, thus strongly influencing age structure of thesample. Studies that have recorded high reader agreement rates(>70%) for sectioned or unsectioned dorsal spine techniqueswere on Walleye populations dominated by younger individ-uals (Campbell and Babaluk 1979; Erickson 1983; Logsdon2007; Williamson and Dirnberger 2010). Our age analysis es-timated that 44–74% of recaptured Walleye used in our studywere ≥10 years old and our reader agreement rates were poor(≤51%). However, reader agreement rates using the unsectionedspine technique for Mille Lacs Lake Walleye (≥19.7 in) in which61% of the sample was estimate to be ≥10 years old was 70%(Logsdon 2007).

Several factors other than age structure of the sample may beresponsible for the reduced reader agreement rates we observed.Reduced reader agreement rates may have been due in part todifferences in reader experience levels. Although both readers inthis study were experienced at estimating Walleye age derivedfrom dorsal spines, reader 1 had considerably more experiencethan reader 2 in using the unsectioned technique. We felt it wasimportant to include this variability in reader experience to pro-vide unbiased and more realistic estimates of reader agreementrates. In this case, we found that there were significant differ-ences among the readers’ ability to replicate ages derived fromsectioned and unsectioned dorsal spines with readers of varyingexpertise levels. However, we also found that the most experi-enced reader’s assigned ages only agreed on 49% of the aged

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


TABLE 5. Mean BCL (in) for each reader as estimated from sectioned and unsectioned dorsal spines via the DP and FL BCL methods for recaptured Walleyescollected from natural lakes in Iowa. Different lowercase letters represent significant differences (P < 0.05).

Method

DP FL

Sectioned Unsectioned Sectioned Unsectioned

Reader 1 Reader 2 Reader 1 Reader 2 Reader 1 Reader 2 Reader 1 Reader 2Age TL (SE) TL (SE) TL (SE) TL (SE) TL (SE) TL (SE) TL (SE) TL (SE)

1 5.0 (0.12) wx 4.3 (0.11) v 4.7 (0.13) vw 4.3 (0.10) v 5.8 (0.11) z 5.2 (0.10) xy 5.5 (0.12) yz 5.4 (0.10) xyz2 9.2 (0.15) xyz 8.6 (0.15) w 8.8 (0.16) xyz 7.7 (0.15) v 9.8 (0.14) z 9.3 (0.14) yz 9.4 (0.15) yz 8.6 (0.14) wx3 11.8 (0.17) yz 11.3 (0.16) yz 11.6 (0.18) yz 10.3 (0.16) w 12.3 (0.16) z 11.9 (0.16) yz 12.0 (0.17) yz 11.0 (0.15) wz4 13.8 (0.20) z 13.5 (0.18) yz 13.5 (0.20) yz 12.3 (0.18) x 14.1 (0.19) z 13.9 (0.17) z 13.9 (0.19) z 12.9 (0.17) yz5 15.5 (0.21) z 15.4 (0.20) yz 15.3 (0.21) yz 14.2 (0.20) x 15.8 (0.20) z 15.7 (0.19) z 15.6 (0.20) z 14.6 (0.19) yx6 17.0 (0.22) z 16.9 (0.21) yz 16.7 (0.22) yz 15.7 (0.21) x 17.2 (0.22) z 17.1 (0.20) z 16.9 (0.21) yz 16.0 (0.20) yx7 18.3 (0.20) z 18.3 (0.21) z 18.0 (0.21) yz 17.1 (0.21) x 18.4 (0.20) z 18.5 (0.21) z 18.2 (0.21) yz 17.4 (0.20) yx8 19.0 (0.17) yz 19.0 (0.18) yz 18.8 (0.18) yz 18.3 (0.19) y 19.1 (0.16) z 19.1 (0.17) z 18.9 (0.18) yz 18.5 (0.18) yz9 19.7 (0.16) yz 19.8 (0.17) yz 19.7 (0.17) yz 19.1 (0.17) y 19.8 (0.15) yz 19.9 (0.17) z 19.8 (0.17) yz 19.2 (0.17) yz

10 20.6 (0.18) z 20.6 (0.23) z 20.3 (0.16) z 20.0 (0.18) z 20.6 (0.17) z 20.7 (0.22) z 20.3 (0.16) z 20.1 (0.18) z11 21.0 (0.19) z 21.1 (0.31) z 20.8 (0.14) z 20.6 (0.16) z 21.0 (0.19) z 21.1 (0.31) z 20.8 (0.14) z 20.6 (0.16) z12 21.3 (0.43) z 21.3 (0.62) z 21.2 (0.35) z 20.5 (0.31) z 21.3 (0.43) z 21.4 (0.62) z 21.2 (0.34) z 20.6 (0.30) z13 20.8 (0.60) z 21.1 (0.77) z 21.9 (0.81)z 20.7 (0.29) z 20.8 (0.58) z 21.2 (0.73) z 21.9 (0.80) z 20.7 (0.28) z14 21.1 (0.64) z 20.5 20.8 (0.84) z 21.1 (0.37) z 21.1 (0.63) z 20.6 20.9 (0.83) z 21.1 (0.36) z15 21.6 (0.58) z 21.4 20.2 21.3 (0.51) z 21.6 (0.57) z 21.4 20.3 21.4 (0.50) z16 21.6 (0.40) z 22.0 21.0 21.1 (0.51) z 21.6 (0.40) z 22.0 21.0 21.1 (0.50) z17 21.7 21.5 (0.51) z 21.7 21.5 (0.50) z18 22.0 21.5 (0.71) z 22.0 21.5 (0.71) z19 22.0 21.5 (0.71) z 22.0 21.5 (0.71) z20 22.0 (0.75) z 22.0 (0.75) z

structures among techniques, thus indicating additional factorsthat may be confounding our results.

Although Belanger and Hogler (1982) observed high ageagreement for second and third dorsal spine, they also foundthat distal sections of the spine yielded progressively lower ageestimates. Our age estimation technique required that multiplesections of the spine (three to five cross sections, one 0.39-in section) be taken so that BCL growth estimates could becompared with that of observed growth. We always cut crosssections from the proximal edge of the spine before standardized

TABLE 6. Processing, viewing, and total processing time (min) per 10-fishgroup associated with the use of unsectioned and sectioned methods to ageand back-calculate Walleye dorsal spines. Removal and processing time forsectioned spines was provided by Isermann et al. (2003); different lowercaseletters represent significant differences (P < 0.05).

Processing Viewing Total processingtime time time

Unsectioned 7.2 (0.21) 30.0 (0.71) y 47.3 (0.81) ySectioned 12.5 (0.28) 29.5 (1.02) y 52.1 (0.86) z

sections were cut, thus potentially introducing additional biasesinto the study (i.e., lower age assignments for the unsectionedspine technique). Although this could have influenced our ageestimates and reader agreement rates, based on age frequenciesfor both techniques, readers did not consistently have lower ageassignments using the unsectioned spine technique.

Since all fish used in this study were from a mark–recaptureevaluation used for broodstock monitoring, a large proportionof the sample was not only large but was also predominantlymale. Sexual dimorphism in Walleye after onset of maturity hasbeen well documented, males having slower growth rates andsmaller asymptotic size than females (Henderson et al. 2003;Rennie et al. 2008). These characteristics make estimating agefrom male Walleye more difficult than from females because theannuli become compressed along the outer edge of the structure,causing inconsistencies in a reader’s ability to interpret annuli.Many male Walleye in this study were estimated to be olderthan age 6, and we often observed compressed annuli nearthe edge of dorsal spines while using each technique. BothLogsdon (2007) and Williamson and Dirnberger (2010) foundthat crowded annuli were much more visible for the unsectioned

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


technique because of flexibility in lighting conditions whenusing the fiber optic light guide. However, our standardizedunsectioned technique required the fiber optic light guide tobe affixed to a permanent stand so BCL measurements couldbe taken at a standard distance, thus reducing the flexibility ofthe light source and resolution of outer annuli. The inability toconsistently identify outer annuli with our modified techniquewith large, adult male Walleye may be partly responsible forreductions in reader agreement rates we found.

The preferred method to validate BCL estimates is to com-pare them with observed measurements of individuals withina population (Francis 1990). However, these types of valida-tion studies are lacking in the literature. Those that have beenconducted have compared scale-based BCL estimates with thatof observed lengths via mark–recapture studies and have foundgood agreement (within ± 10%) among BCL lengths and ob-served lengths for various types of BCL methods (Davies andSloane 1986; Klumb et al. 1999a, 1999b). Although Borkholderand Edwards (2001) found that BCL estimates derived fromWalleye dorsal spine sections compared favorably with thoseestimated via scales, no study has directly validated dorsal spine-based BCL methods via a mark–recapture study or known-agefish. Both BCL methods (DP or FL) examined in this studyeither substantially under- or overestimated last year’s growth,regardless of the viewing technique. In comparison to otherBCL validation studies, only 2.5–9.3% of BCL estimates us-ing either BCL method or viewing technique were within ±10% of the observed measurement. Back-calculation of lengthsfrom structures relies on recognition of annuli to calculate anestimated body length associated with each annulus. Some ofthe same factors described above could be responsible for thepoor agreement we observed with known growth incrementscompared with those estimated via BCL methods. Regardlessof those inconsistencies, the results of our study do suggest thatdorsal spine-based BCL methods do not perform well in predict-ing recent Walleye growth patterns in populations with an olderage structure. Morita and Matsuishi (2001) found that growthrates of slow-growing fish were substantially overestimated us-ing otolith-based BCL methods compared with observed growthrates and attributed this bias to changes in somatic growth andotolith size ratios. Our Walleye sample consisted mostly ofolder, slower-growing fish, and BCL methods predicting lastyear’s growth may have been compromised by similar changesin somatic growth and dorsal spine size. Since spine-based BCLmethods are gaining popularity for multiple state and federalagencies, further mark–recapture studies examining patterns inBCL growth derived from spine-based techniques are neces-sary to validate these techniques for Walleye populations withvarious age structures and growth rates.

Although we found that our unsectioned Walleye dorsal spineage estimation technique and BCL did not perform as well as thesectioned spine technique on adult Walleye, the time required toprocess, prepare, view, and perform back-calculations on struc-tures was significantly less. Williamson and Dirnberger (2010)

observed the same relationship, but they also found good readeragreement for Sauger dorsal spines when comparing techniques.Most of the differences in total processing times among the twotechniques were with time required to process. Unsectionedspines required only one standardized cut, thus nearly reducingprocessing time in half. However, the fiber optic light guidestand did not increase speed of viewing time, likely due to thetime required to apply mineral oil, place the spine in the mod-eling clay and align in viewing space, and, if necessary, lightlywet-sand the proximal end of the spine prior to viewing. Mod-ifications to the standardized technique, such as improvementsin the stand or fiber optic light guide, may improve the read-ability of each structure and improve time efficiency. This studyand the study conducted by Williamson and Dirnberger (2010)found that unsectioned dorsal spines are a promising techniqueto improve the efficiency of processing numerous structures,but further research is needed to examine new BCL techniques,modifications to our BCL technique, or both for use on bothyoung and old Walleye populations.

The large size, older age structure, and slow growth ratesof recaptured Walleye made it difficult to consistently obtainsimilar age estimates among readers, regardless of the age es-timation technique. Even with relatively high reader agreementrates (70%) for an older Walleye population, Logsdon (2007)recommended that unsectioned spines were most effective atreplicating otolith ages of individuals younger than age 7. Basedon our poor ability to replicate ages, we do not suggest the use ofour standardized unsectioned dorsal spine technique to estimatethe age or BCL older Walleye. In addition, no BCL methodaccurately estimated annual growth regardless of technique orreader experience. The results of this study also strongly suggestthat more research is necessary to validate dorsal spine-basedBCL estimates on both old and young Walleye populations.

ACKNOWLEDGMENTSThe Sport Fish Restoration Act provided funding for the

project. We would like to thank the crew at the Spirit LakeFish Hatchery and management and research Iowa Departmentof Natural Resources staff that assisted with fish and data col-lection. We would also like to thank all of the reviewers whocontributed comments to this manuscript.

REFERENCESBeamish, R. J., and D. Chilton. 1977. Age determination of Lingcod (Ophiodon

elongates) using dorsal fin rays and scales. Journal of the Fisheries ResearchBoard of Canada 34:1305–1313.

Beamish, R. J., and G. A. McFarlane. 1983. The forgotten age requirement infisheries biology. Transactions of the American Fisheries Society 112:735–743.

Belanger, S. E., and S. R. Hogler. 1982. Comparison of five ageing method-ologies applied to Walleye (Stizostedion vitreum) in Burt Lake, Michigan.Journal of Great lakes Research 8:666–671.

Borkholder, B. D., and A. J. Edwards. 2001. Comparing the use of dorsalfin spines with scales to back-calculate length-at-age estimates in Walleyes.North American Journal of Fisheries Management 21:935–942.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013


Brown, P., C. Green, K. P. Sivakumaran, D. Stoessel, and A. Giles. 2004.Validating otoliths annuli for annual age determination of Common Carp.Transactions of the American Fisheries Society 133:190–196.

Buckmeier, D. L., and R. G. Howells. 2003. Validation of otoliths for estimatingages of Largemouth Bass to 16 years. North American Journal of FisheriesManagement 23:590–593.

Buckmeier, D. L., E. R. Irwin, R. K. Betsill, and J. A. Prentice. 2002. Validity ofotoliths and pectoral spines for estimating ages of Channel Catfish to 16 years.North American Journal of Fisheries Management 22:934–942.

Campana, S. E. 1990. How reliable are growth back-calculations based onotoliths? Canadian Journal of Fisheries and Aquatic Sciences 47: 2219–2227.


Campbell, J. S., and J. A. Babaluk. 1979. Age determination of Walleyes,Stizostedion vitreum vitreum (Mitchill), based on the examination of eightdifferent structures. Canada Fisheries and Marine Service Technical Report849.


Dahl, K. 1907. The scales of the herring as a means of determining age, growthand migration. Report on Norwegian Fishery and Marine-Investigations 2(6),Bergen.

Davies, P. E., and R. D. Sloane. 1986. Validation of aging and length back-calculation in Rainbow Trout, Salmo gairdneri Rich., from Dee Lagoon,Tasmania. Australian Journal of Marine and Freshwater Research 37:289–295.

DeCicco, A. L., and R. J. Brown. 2006. Direct validation of annual growthincrements on sectioned otoliths from adult Arctic Grayling and a comparisonof otoliths and scale ages. North American Journal of Fisheries Management26:580–586.

Erickson, C. M. 1983. Age determination of Manitoban Walleyes using otoliths,dorsal spines, and scales. North American Journal of Fisheries Management3:176–181.

Francis, R. I. C. C. 1990. Back-calculation of fish length: a critical review.Journal of Fish Biology 36:883–902.

Fraser, C. M. 1916. Growth of the spring salmon. Transactions of the PacificFisheries Society 1916:29–39.

Heidinger, R. C., and K. Clodfelter. 1987. Validity of the otolith for determiningage and growth of Walleyes, Striped Bass, and Smallmouth Bass in powerplant cooling ponds. Pages 241–251 in R. C. Summerfelt and G. E. Hall,editors. Age and growth of fish. Iowa State University Press, Ames.

Henderson, B. A., N. Collins, G. E. Morgan, and A. Vaillancourt. 2003. Sexualsize dimorphism of Walleye (Stizostedion vitreum vitreum). Canadian Journalof Fisheries and Aquatic Sciences 60:1345–1352.

Isely, J. J., and T. B. Grabowski. 2007. Age and growth. Pages 187–228 inC. S. Guy and M. L. Brown, editors, Analysis and interpretation of freshwaterfisheries data. American Fisheries Society, Bethesda, Maryland.

Isermann, D. A., J. R. Meerbeek, G. D. Scholten, and D. W. Willis. 2003.Evaluation of three different structures used for Walleye age estimation withemphasis on removal and processing times. North American Journal of Fish-eries Management 23:625–631.

Klumb, R. A., M. A. Bozek, and R. V. Frie. 1999a. Proportionality of bodyto scale growth: validation of two back-calculation models with individuallytagged and recaptured Smallmouth Bass and Walleyes. Transactions of theAmerican Fisheries Society 128:815–831.

Klumb, R. A., M. A. Bozek, and R. V. Frie. 1999b. Validation of the Dahl–Leaand Fraser–Lee back-calculation models by using oxytetracycline-markedBluegills and Bluegill × Green Sunfish hybrids. North American Journal ofFisheries Management 19:504–514.

Kocovsky, P. M., and R. F. Carline. 2000. A comparison of methods for esti-mating ages of unexploited Walleyes. North American Journal of FisheriesManagement 20:1044–1048.

Lea, E. 1910. On the methods used in the herring investigations. Publications deCirconstance Conseil Permanent International pour l’Exploration de la Mer53:7–25.

Lee, R. 1920. A review of the methods of age and growth determination in fishesby means of scales. Fishery Investigations Series II Marine Fisheries GreatBritain Ministry of Agriculture Fisheries and Food 4(2).

Logsdon, D. E. 2007. Use of unsectioned dorsal spines for estimating Wall-eye ages. North American Journal of Fisheries Management 27:1112–1118.

Maceina, M. J., J. Boxrucker, D. L. Buckmeier, R. S. Gangl, D. O. Lucchesi,D. A. Isermann, J. R. Jackson, and P. J. Martinez. 2007. Current status and re-view of freshwater fish aging procedures used by state and provincial fisheriesagencies with recommendations for future directions. Fisheries 32:329–340.

Margenau, T. L. 1982. Modified procedure for ageing Walleyes by using dorsalspine sections. Progressive Fish-Culturist 44:204.

Meerbeek, J. R., J. G. Larscheid, M. J. Hawkins, and G. D. Scholten. 2013.Retention of large-format, soft visible implant alphanumeric tags in Walleye.North American Journal of Fisheries Management 33:26–31.

Morita, K., and T. Matsuishi. 2001. A new model of growth back-calculationincorporating age-effect based on otoliths. Canadian Journal of Fisheries andAquatic Sciences 58:1805–1811.

Olson, D. E. 1980. Comparison of marks on scales and dorsal spine sectionsas indicators of Walleye age. Minnesota Department of Natural ResourcesSection of Fisheries Investigational Report 371.

Rennie, M. D., C. R. Purchase, N. Lester, N. C. Collins, B. J. Shuter, and P. A.Abrams. 2008. Lazy males? Bioenergetic differences in energy acquisitionand metabolism help to explain sexual size dimorphism in percids. Journalof Animal Ecology 77:916–926.

Ricker, W. E. 1992. Back-calculation of fish lengths based on proportionalitybetween scales and length increments. Canadian Journal of Fisheries andAquatic Sciences 49:1018–1026.

SAS Institute. 2001. SAS/STAT user’s guide, version 8.0. SAS Institute, Cary,North Carolina.

Schram, S. T. 1989. Validating dorsal spine readings of Walleye age. WisconsinDepartment of Natural Resources, Fish Management Report 138, Madison.

Weisberg, S. 1986. A linear model approach to back-calculation of fish length.Journal of the American Statistical Association 81:922–927.

Whitney, R. R., and K. D. Carlander. 1956. Interpretation of body-scale re-gression for computing body length of fish. Journal of Wildlife Management20:21–27.

Williamson, C. W., and R. R. Dirnberger. 2010. A comparison of techniques us-ing dorsal spines to estimate Sauger age. North American Journal of FisheriesManagement 30:1016–1019.

Dow

nloa

ded

by [

Dep

artm

ent O

f Fi

sher

ies]

at 2

3:49

12

Aug

ust 2

013

Date post:	24-Aug-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times