Biological Control 70 (2014) 65–72

Contents lists available at ScienceDirect

Biological Control

journal homepage: www.elsevier .com/locate /ybcon

Habitat effects on second-order predation of the seed predator Harpalusrufipes and implications for weed seedbank management

1049-9644/$ - see front matter � 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.biocontrol.2013.12.004

⇑ Corresponding author.E-mail addresses: sonja.birthisel@maine.edu (S.K. Birthisel), gallandt@maine.edu

(E.R. Gallandt), rjabbour@uwyo.edu (R. Jabbour).1 Present address: University of Wyoming, Department of Plant Sciences, Dept.

3354, 1000 E. University Avenue, Laramie, WY 82071, USA.

Sonja K. Birthisel ⇑, Eric R. Gallandt, Randa Jabbour 1

University of Maine, School of Food and Agriculture, 5722 Deering Hall, Orono, ME 04469-5722, USA

h i g h l i g h t s

� Assays were used to quantify second-order predation of weed seedpredator Harpalus rufipes.� Motion-sensing wildlife cameras

were used to capture images ofsecond-order predators.� Second-order predation was greater

in complex habitats than simple.� A simulation model predicts second-

order predation substantiallyincreases seedbank inputs.

g r a p h i c a l a b s t r a c t

a r t i c l e i n f o

Article history:Received 22 July 2013Accepted 4 December 2013Available online 11 December 2013

Keywords:Second-order predationSeed predationFood webEcosystem serviceHarpalus rufipesCarabidae

a b s t r a c t

Seed predators provide a valuable ecosystem service to farmers by reducing densities of weed seeds, and,in turn, densities of weed seedlings they must manage. The predominant invertebrate weed seed pred-ator in Maine, USA, agroecosystems is the carabid beetle Harpalus rufipes DeGeer. Pitfall trapping hasshown that H. rufipes prefers sites with vegetative cover to fallow sites, preference speculated to be dri-ven by predator avoidance behavior. To test this hypothesis, ‘second-order predation assays’ were devel-oped, in which live H. rufipes prey were presented to second-order predators. Field experiments wereconducted to determine foremost if H. rufipes was subject to second-order predation, and secondly,whether (a) vegetative cover affords H. rufipes protection from second-order predators, and (b) high ratesof second-order predation correspond with decreased invertebrate seed predation rates. Two 72-h exper-iments were conducted (mid August and September 2012) at crop and non-crop sites across a 28 hadiversified farm in Stillwater, ME, USA.

Second-order predation was 2.8% per day. Based on images from motion-sensing cameras, H. rufipes’predators included birds and small mammals. Neither a relationship between second-order predationand vegetative treatment, nor an empirical relationship between second-order predation and inverte-brate seed predation were detected. However, a simulation model predicted that 2.8% per day second-order predation could increase the number of seeds entering the seedbank by more than 17% annually.Additionally, complex habitats supported higher rates of second-order predation than did simplehabitats.

� 2013 Elsevier Inc. All rights reserved.

1. Introduction

Seed predation is a significant source of weed mortality in agro-ecosystems (Gallandt, 2006). Seed predators thus provide an eco-system service to farmers: by reducing weed seedbank inputs,seed predators reduce the weed pressure farmers experience insubsequent growing seasons. Seed predation shows promise as a

Fig. 1. Illustration of tethering methods: location of tether stake relative to soilsurface (dotted line) (a), two-half-hitch knot (b) and tethered Harpalus rufipes (c).

66 S.K. Birthisel et al. / Biological Control 70 (2014) 65–72

component in multi-tactic ecologically based weed managementstrategies (Bohan et al., 2011; Gallandt, 2006; Liebman andGallandt, 1997; Mirsky et al., 2010; Westerman et al., 2005). How-ever, seed predation rates can be highly variable (Booman et al.,2009; Meiners et al., 2000), and lack of knowledge about the fac-tors supporting and limiting seed predation hampers its usefulnessin the biological control of weeds.

Carabid beetles are common seed predators in agroecosystems(Tooley and Brust, 2002), and have been much studied for their po-tential as biological control agents (e.g., Hawthorne et al., 1998;Saska et al., 2010; Shearin et al., 2007). In Maine, USA agroecosys-tems, the carabid Harpalus rufipes DeGeer is the predominantinvertebrate seed predator (Birthisel, 2013; Gallandt et al., 2005;Zhang et al., 1997). H. rufipes shows a clear preference for habitatthat provides vegetative cover: in a mark-recapture study, H. rufi-pes were twice as likely to be re-captured in their starting plots ifreleased in vegetated vs. fallow plots (Shearin et al., 2008). Weedseed predation, too, is positively associated with vegetative cover(reviewed in Meiss et al., 2010; but see Jacob et al., 2006). We spec-ulated that H. rufipes prefers vegetated environments because theyafford protection from second-order predators (Shearin et al.,2008). However, this hypothesis has not been tested. Indeed, pre-dation of H. rufipes has not, to our knowledge, been documented.

Predation of carabids is ‘‘one of the least researched areas ofcarabidology’’ (Holland, 2002), and what little research has beendone offers mixed results. By placing beetles in individual plasticcontainers in an outdoor insectary, Luff (1980) estimated a lowH. rufipes mortality rate of 8.5% month�1. It is unclear from thesemethods, however, whether second-order predators were able toaccess H. rufipes in the insectary, or whether this figure representsnon-predation mortality only. Conversely, total carabid abundancein a shrub-steppe ecosystem was 111% higher within rodent exclo-sures than without after two years of study (Parmenter andMacmahon, 1988). In a UK grassland system, however, a similarstudy found no exclosure effect on carabid abundance (Churchfieldet al., 1991).

Carabid predators common in temperate agroecosystemsinclude rodents and other small mammals (Larochelle, 1975a),birds (Larochelle, 1980, 1975b), amphibians, reptiles (Larochelle,1975c), and invertebrates including members of the Araneae,Carabidae, and Formicidae (Thiele, 1977). Many of these predatorsare omnivorous, consuming seeds in addition to invertebrate prey.The tremendous diversity of omnivorous seed feeders potentiallyexisting within a single farmscape (Evans et al., 2011) providesthe potential for a myriad of yet-unstudied trophic interactionsbetween and amongst seed feeding guilds.

We know of no research that has quantitatively tested theeffects of second-order predation of carabids on weed seed preda-tion. Our group conducted field experiments to document andmeasure second-order predation of H. rufipes, and identify theresponsible predators. This study sought further to quantify the ef-fect of vegetative cover on second-order predation of H. rufipes, andthe effect of second-order predation on invertebrate weed seedpredation. Hypotheses were as follows:

1. Second-order predation is a source of H. rufipes mortality.2. Vegetative cover affords H. rufipes protection from second-order

predators.3. Second-order predation decreases invertebrate seed predation.

2. Methods

2.1. Second-order predation field study

Field experiments to measure second-order predation of H. ruf-ipes were conducted at the University of Maine Rogers Farm in

Stillwater, ME, USA. Two 72-h experiments were conducted, begin-ning on August 8 and September 10, 2012.

2.1.1. Sample sitesTo maximize the chance of detecting second-order predation,

sites at least 100 m distant from each other, and located in threedistinct habitat types (crop, forest ecotone, and mowed grass) werechosen for sampling. In August, 15 sites were sampled; in Septem-ber, 12 different sites. At each site, in addition to habitat type, up tofour plant taxa representing the majority of plant biomass presentin a 3-m2 area were recorded. An AccuPAR LP-80 ceptometer(Decagon Devices, Pullman, WA, USA) was used to quantify leafarea index (LAI). LAI readings represent an average of 5 above-can-opy and 5 below-canopy readings. Visual estimates of percentground cover, average vegetation height, presence/absence of plantresidue, and presence/absence of recent disturbance (mowing ortillage within past 2 weeks) were recorded at each site.

2.1.2. Second-order predation assaysSecond-order predation assays were constructed using

55 � 43 � 7 cm plastic greenhouse flats filled with Metro-Mix�

560 SUN-COIR™ soilless medium (Sun Gro Horticulture, Agawam,MA, USA). Assays received one of two treatments: bare fallow orvegetated. Bare fallow assays consisted of only medium, with nofurther modification. Vegetated assays were seeded with8.8 ± 0.2 g organic oat seeds (Johnny’s Select Seeds, Fairfield, ME,USA), representing a sowing density of 280 kg ha�1. Oat seedlingswere grown in a greenhouse to a height of approximately 23 cm.

Five paper clips were bent into loops and inserted into the med-ium on each second-order predation assay to serve as stakes towhich H. rufipes were tethered (Fig. 1a). One live H. rufipes wastethered to each stake, for a total of five H. rufipes assay�1, using2.7 kg Trilene� XL fishing line (Berkley, Spirit Lake, IA, USA) se-cured behind the first pair of legs with a two-half-hitch knot(Fig. 1b) and stop knot (not shown). Tether length was 10 cm(Fig. 1c).

Vegetated and fallow second-order predation assays were pre-sented in pairs at each site. The greenhouse flats were buried flushwith the surrounding soil surface (Fig. 2). At a subset of five sites,additional control second-order predation assays were includedduring the September experiment. Control assays consisted of fal-low assays covered with 1-mm mesh exclosure cages to excludesecond-order predators. Presence/absence of tethered H. rufipeswas recorded at dusk and dawn to separate nocturnal from diurnalsecond-order predation. Predated and dead beetles were replaced.Non-predation mortality was 9.3 ± 1.2%. The majority of theselosses (93%) occurred when beetles failed to burrow into the soilduring the day and, presumably, desiccated.

Fig. 2. Photo of a sample site, showing paired vegetated and fallow second-orderpredation assays, and motion-sensing wildlife camera.

S.K. Birthisel et al. / Biological Control 70 (2014) 65–72 67

Percent second-order predation was calculated using a modi-fied form of Abbott’s (1945) formula:

%Predation ¼ Zp � C � Zr

Zp � C� 100 ð1Þ

where Zp was the number of H. rufipes presented on a second-orderpredation assay, Zr the number of H. rufipes remaining at the end ofthe measurement period, and C the proportion of H. rufipes retainedon control second-order predation assays. The control values fromSeptember were used to adjust both August and September sec-ond-order predation values.

Wildlife cameras were used to capture images of potential sec-ond-order predators. Four Reconyx� PM75, four Reconyx� PC85(Reconyx, Inc., Holden, WI, USA), and two Bushnell� 119436E(Bushnell Outdoor Products, Overland Park, KS, USA) wildlife cam-eras, set to motion-capture, were focused on tethered H. rufipes tocapture images and video of second-order predators (Fig. 2). Carni-vores or omnivores photographed during time periods coincidingwith second-order predation were considered likely second-orderpredators.

2.1.3. Invertebrate activity-densityInvertebrate activity-density was measured with pitfall ‘live’

traps. Traps were fabricated from recycled 500 mL soda bottles,the tops of which were cut off and inverted to form funnels intothe traps. Two traps were installed flush with the soil surface ateach site, 2 m from second-order predation assays, and no less than2 m apart. Traps were checked at dawn and dusk, and the numberof captured H. rufipes, other Carabidae, Formicidae, and Gryllidaerecorded. Trapped H. rufipes were collected and stored at 4 �C foruse in subsequent trials. Other invertebrates were released at least2 m from traps to prevent immediate re-capture.

2.1.4. Invertebrate seed predationInvertebrate seed predation was estimated using seed assays

(Brust and House, 1988; Gallandt et al., 2005). Seed assays wereconstructed from inverted 100 � 15 mm Petri dish bottoms, cov-ered with Grafix Double Tack mounting film. Weed seeds weregently adhered to the film: 30 Setaria lutescens Hubb. (yellow fox-tail), 30 Amaranthus retroflexus L. (redroot pigweed), and 30 Dacty-lis glomerata L. (orchard grass) seeds, for a total of 90 seeds assay�1.Fine sand was sifted over the seed assays so that predators wouldnot stick to the film. This thin layer of sand did not obscure seedsfrom view. S. lutescens seeds were collected from Rogers Farm inStillwater, ME, USA. A. retroflexus seeds were purchased from

Herbiseed, Twyford, England; D. glomerata seeds from Seeds Trust,Inc., Littleton, CO, USA.

Two seed assays were installed at each site, at least 2 m fromany other site feature. Each seed assay was installed such thatthe surface of the assay was flush with the surrounding soil sur-face. Seed assays were covered with vertebrate exclosures con-structed from 1-cm hardware cloth. To estimate seed loss fromabiotic factors, additional control seed assays, covered by 1-mmmesh exclosures designed to exclude all seed predators, werelocated at a randomly chosen subset of five sites trial�1. Seed as-says were exposed in the field for the duration of each 72-h trialperiod.

Mean seed loss was calculated by averaging seed loss from thetwo seed assays at each site. Percent seed predation was calculatedas modified from (Abbott, 1945):

%Seed predation ¼ Si � C � Sf

Si � C� 100 ð2Þ

where Si was the initial number of seeds, Sf the final number ofseeds, and C the proportion of seeds retained on control seed assays.

2.2. Statistical analyses

All analyses were performed in R (R-Development Core Team,2011), using non-parametric tests appropriate to count data(Crawley, 2013).

To test the effect of month (August, September) on second-or-der predation, a 2-sample test for equality of proportions withYates continuity correction (which reduces the error in assuminga continuous v2 distribution) was used. To test the effects ofsample period (day, night) and vegetative treatment (bare fallow,vegetated) on second-order predation, Pearson’s v2 tests wereused.

To test for effects of habitat variables on second-order preda-tion, 2- and 3-sample tests for equality of proportions with Yatescontinuity correction were used for categorical variables (habitattype, presence/absence of plant residue, presence/absence ofdisturbance), and Kendall’s s tests for continuous variables (LAI,percent ground cover, average vegetation height). Pearson’s v2

was used to test for a sample site effect.To test for effects of categorical variables (month, seed species)

on percent seed predation, Pearson’s v2 tests were used. Kendall’s stests were used to test for effects of continuous variables (H. rufipesactivity-density, second-order predation) on seed predation, and totest for correlation between second-order predation and H. rufipesactivity-density.

2.3. Modeling the effect of second-order predation on the weedseedbank

A simulation model was constructed in R (R-Development CoreTeam, 2011) to estimate the impact of second-order predation onthe number of seeds entering the weed seedbank during August–September in temperate agroecosystems (Appendix).

The model was comprised of three scalar functions, which cal-culated absolute density of H. rufipes, number of seeds availableon the soil surface, and number of seeds entering the weed seed-bank at each time step, respectively. Each simulation included 61time steps, corresponding to the number of days in August andSeptember, and calculated an estimate of the total number of seedsentering the seedbank over this time frame.

The absolute density of H. rufipes (B), or number of beetles m�2,was a function of prior absolute density, rate of loss to second-or-der predation (z), and rate of new adult emergence (n):

Btþ1 ¼ Bt � z � Bt þ n � Bt ð3Þ

68 S.K. Birthisel et al. / Biological Control 70 (2014) 65–72

H. rufipes larvae mature into adults in late summer, emergingfrom metamorphosis July–August (Zhang, 1993). Because carabideggs and larvae may be subject to high mortality (Heessen, 2013;Heessen and Brunsting, 1981), emergence was considered the bestindicator of new beetles entering the system. Negligible immigra-tion, emigration, and mortality to sources other than second-orderpredation were assumed. Given that carabid populations arethought to be relatively stable over time (Luff, 1982; Thomaset al., 2001), we think these assumptions reasonable.

The number of seeds available on the soil surface (A) was afunction of seed dispersal (D), seed burial rate (b), vertebrate seedpredation rate (v), per-beetle seed consumption rate (p), and abso-lute density of H. rufipes (B).

Atþ1 ¼ At þ D� b � At � v � At � p � B ð4Þ

Function (4) was constructed such that H. rufipes seed preda-tion, p * B, was not seed density dependent. This was done becauseinvertebrate seed predators typically do not respond numericallyto seed density (Baraibar et al., 2012; Marino et al., 2005;Westerman et al., 2008; but see Bohan et al., 2011). For the sakeof model simplicity, the effect of seed species on H. rufipes seedpredation (Harrison and Gallandt, 2012; Zhang, 1993) wasassumed to be negligible. Further, seed predation by invertebratesother than H. rufipes was assumed to be negligible. Our pitfall trapdata showed that H. rufipes comprised a vast majority (79%) of cap-tured invertebrate seed feeders in this system, and we are notaware of data that would allow us to estimate a rate of invertebrateseed predation due to non-H. rufipes seed predators only. As aresult of this last assumption, however, the model likely underes-timates the total impact of invertebrate seed predators on seedsentering the seedbank.

The number of seeds entering the seedbank (S) was a functionof seed burial rate and seeds available on the soil surface:

Stþ1 ¼ St þ b � At ð5Þ

While seed burial is in reality a complex process, subject to var-iation caused by weather and other disturbances (Benvenuti, 2007;Westerman et al., 2009), a constant seed burial rate was assumedfor the sake of model simplicity.

2.3.1. Parameter estimatesThe model was parameterized using values from our data and

the literature (Table 1).Few estimates of H. rufipes absolute density, B, are reported in

the literature. Briggs (1965) recorded densities ranging from 0.2to 13.5 m�2 across a UK farmscape; Hamon et al. (1990) recordeddensities of 0.97–3.43 m�2 in UK field beans. We think it likely,however, that H. rufipes absolute densities can exceed these esti-mates in Maine, USA agroecosystems. In a landscape-level studyconducted in Dixmont, ME, USA in August–September 2012, ourgroup recorded an H. rufipes pitfall catch rate of (mean ± SE)

Table 1Variables and parameter estimates for second-order predation simulation model.

Parameter Variable Estimate

Seeds entering seedbank S 0 at t0

Seeds available on surface A 0 at t0

Seeds dispersed D 111–2622 seeds m�2

Beetle absolute density B 0–20 beetles m�2

Seed burial rate b 0.04 seeds seed�1 daVertebrate seed predation v 0.066 seeds seed�1 dSeeds predated beetle�1 p 23.3–38.8 seeds beetNew beetle emergence rate n 0–0.04 beetles beetleSecond-order predation rate z 0–0.042 beetles beet

13.2 ± 1.7 beetles trap�1 3 days�1 (Birthisel, 2013). This exceedsby more than a factor of ten the corresponding estimate of 1.1H. rufipes trap�1 3 days�1 extrapolated from Hamon et al.(1990). Although pitfall traps have been criticized as inexact(Lang, 2000; Thomas et al., 1998), carabid absolute densities arethought to be strongly correlated with activity-density (Baars,1979; Lang, 2000; Luff, 1982). Thus, the high activity-densitiesfound in Maine agroecosystems may reflect H. rufipes absolutedensities significantly exceeding published estimates. To deter-mine the sensitivity of seeds entering the seedbank, S, to absolutedensity, B, simulations were run over a range of B values from 0to 20 beetles m�2.

Rate of new H. rufipes emergence, n, was estimated from pitfalldata collected in Maine potato agroecosystems by Zhang (1993).Newly emerged and total adult pitfall catch numbers were usedto calculate weekly values of the proportion of adults newlyemerged. These values were used to construct a vector of dailyemergence rates for August–September (Appendix).

Second-order predation rate, z, was estimated using second-or-der predation assay data. Simulations were run at values rangingfrom 0 to 0.042 predated beetles beetle�1 day�1 to determine thesensitivity of seeds entering the seedbank to z. The maximumvalue corresponded to 1.5 times our measured estimate of 2.8%second-order predation.

Seed rain data from Davis and Raghu (2010, Fig. 2) was used toestimate the seed dispersal parameter, D. The time frame, latitude,and weed species utilized in this study was more relevant to oursystem than those presented in other studies. Mean seed rain m�2 -week�1 was calculated, and used to construct a vector of daily seedrain values (Appendix).

Seed burial rate, b, was estimated from Westerman et al. (2009,Fig. 5) b was calculated as mean seed burial day�1 across surrogateseed sizes, cropping systems, and years.

Vertebrate seed predation day�1, v, was estimated from seedpredation data collected at Peacemeal Farm in Dixmont, ME(Birthisel, 2013). Daily seed consumption beetle�1 in August andSeptember was estimated from Saska et al. (2010, Fig. 1b) and tem-perature means measured during second-order predation experi-ments at the Bangor Airport weather station in Bangor, ME. Avector of daily seed consumption rates beetle�1 was constructedfrom these estimates (Appendix).

3. Results

3.1. Second-order predation of H. rufipes

Daily second-order predation rate was 2.8% (Table 2). Controlbeetle loss was low; two total beetles went missing from controlassays. Second-order predation was statistically constant acrossmonths (v2 = 0.098, P = 0.754). Thus, months were considered to-gether in subsequent analyses unless otherwise noted.

Source

––

day�1 Davis and Raghu (2010)Briggs (1965), Hamon et al. (1990) and Birthisel (2013)

y�1 Westerman et al. (2009)ay�1 Birthisel (2013)le�1 day�1 Saska et al. (2010)s�1 day�1 Zhang (1993)les�1 day�1 Birthisel (2013)

Table 2Predation of Harpalus rufipes by second-order predators in August, September, andacross months.

N H. rufipes predated (#) H. rufipes predated day�1 (%)

August 150 23 2.5September 120 21 3.2Total 270 44 2.8

S.K. Birthisel et al. / Biological Control 70 (2014) 65–72 69

Based on images from motion-sensing cameras, likelysecond-order predators included mice (Peromyscus spp.), Easternchipmunk (Tamias striatus), American red squirrel (Tamiasciurushudsonicus), common raccoon (Procyon lotor), and red-wingedblackbird (Agelaius phoeniceus) (Fig. 3).

Second-order predation varied among the 27 sampled locationsat Rogers Farm (v2 = 183.045, P = 0.000) (Fig. 4a), and was greaterat night than during the day (v2 = 5.818, P = 0.016) (Fig. 4b).

Fig. 4. Second-order predation at each of 27 sample sites: sites 1–15 are from theAugust experiment, sites 16–27 from September (a), diurnal vs. nocturnal second-order predation (b) and second-order predation in vegetated vs. fallow assays (c).

3.2. Effects of vegetative treatment and habitat on second-orderpredation

Second-order predation did not differ between bare fallow andvegetated assays (v2 = 0.8182, P = 0.366) (Fig. 4c). However, sec-ond-order predation differed significantly between habitat types(v2 = 73.552, P = 0.000), being greater in forest edge than in cropor mowed grass habitats (forest edge > crop: v2 = 63.636,P = 0.000; forest edge > mowed grass: v2 = 20.260, P = 0.000; crop� mowed grass: v2 = 0.000, P = 1.000). Second-order predationwas positively correlated with leaf area index (s = 0.208,P = 0.001), percent ground cover (s = 0.142, P = 0.031), and averagevegetation height (s = 0.181, P = 0.005). Additionally, second-orderpredation was greater at sites with plant residue present than atsites without plant residue (v2 = 32.206, P = 0.000), and greater atundisturbed sites than disturbed sites (v2 = 13.571, P = 0.000).

Fig. 3. Photos of likely second-order predators. From top left: mouse (Peromyscus sp.),blackbird (Agelaius phoeniceus).

3.3. Relationships between seed predation, H. rufipes activity-density,and second-order predation

Mean percent seed predation was greater in August than inSeptember (Fig. 5) (v2 = 25.638, P = 0.000). Seed predation wasnot affected by seed species in either trial period (August:v2 = 4.345, P = 0.114; September: v2 = 0.005, P = 0.998).

Eastern chipmunk (Tamias striatus), common raccoon (Procyon lotor), red-winged

Fig. 5. Seed predation (mean ± 1 SE) of three seed species (AMARE = Amaranthusretroflexus, SETLU = Setaria lutescens, DACTGL = Dactylis glomerata) in August andSeptember.

Table 3Estimated number of seeds entering the seedbank m-2 (S) at varying Harpalus rufipesbeetle densities (B) and second-order predation rates (z).

Seeds entering the seedbank m-2 (S)

B = 5 B = 10 B = 15 B = 20

z = 0.0% 10,984 5813 641 -4,530z = 1.4% 12,432 8707 4,983 1,258z = 2.8% 13,376 10,596 7,816 5,036z = 4.2% 14,011 11,866 9,721 7,576

70 S.K. Birthisel et al. / Biological Control 70 (2014) 65–72

There was no correlation between seed predation and second-order predation of H. rufipes (s = 0.028, P = 0.669), nor was therea relationship between second-order predation and H. rufipespitfall catch (s = �0.063, P = 0.377). However, there was a strongpositive correlation between seed predation and H. rufipes activ-ity-density (s = 0.201, P = 0.001).

3.4. Predicted effect of second-order predation on the weed seedbank

A simulation model predicted the number of weed seeds enter-ing the weed seedbank in August–September (Table 3). In the ab-sence of H. rufipes, second-order predation was predicted to haveno impact on number of seeds entering the seedbank (estimated

Fig. 6. Simulations of number of seeds entering the seedbank m�2 at startingHarpalus rufipes absolute densities of 0 (a), 5 (b), 10 (c), and 20 (d) beetles m�2 andsecond-order predation rates of 0% (—), 1.4% (� � �), 2.8% (---), and 4.2% (-�-) day�1.

to be 16,000 seeds m�2) (Fig. 6a). When H. rufipes were present,the number of seeds entering the seedbank was predicted to in-crease with second-order predation (Fig. 6). The magnitude ofthe second-order predation effect was strongly dependent on H.rufipes absolute density (e.g. Fig. 6b vs. Fig. 6d).

4. Discussion

4.1. Second-order predation of H. rufipes

Field studies demonstrated that second-order predation was asource of H. rufipes mortality (Table 2), providing support forHypothesis 1. The motion capture camera images represent thefirst documentation of vertebrates likely responsible for second-or-der predation of H. rufipes (Fig. 3). The list of photographed second-order predators enumerated in Section 3.1 is unsurprising, giventhat Red-winged blackbird (Larochelle, 1975b), common raccoon,and several species of mice and squirrels are known predators ofother carabids (Larochelle, 1975a).

The finding that second-order predation was greater at nightthan during the day was expected, as H. rufipes are primarilynight-active (Thiele, 1977). The majority of likely second-orderpredators captured on camera were likewise nocturnal.

4.2. Effects of vegetated treatment and habitat on second-orderpredation

Based on Shearin et al.’s (2008) suggestion that H. rufipes pref-erence for vegetative cover may be driven by predator avoidancebehavior, we expected to find higher rates of second-order preda-tion on fallow assays than vegetated assays (Hypothesis 2). Theresults did not support this hypothesis (Fig. 4c). Failure to detecta vegetative treatment effect may have been due to the close prox-imity of vegetated and fallow assays to one another (Fig. 2). Forinstance, beetles in the bare fallow assay may have drawn sec-ond-order predators close enough to the vegetated assay to findand prey upon those H. rufipes that would otherwise have beenshielded from view by the vegetation. Alternatively, H. rufipes pref-erence for vegetative cover may be driven by factors other thansecond-order predation. For example, vegetative cover may pro-mote a favorable microclimate (Magura et al., 2001), or providehabitat for carabid prey (Hawthorne and Hassall, 1995).

The observed variation in second-order predation rates be-tween sites (Fig. 4a) is partially attributable to habitat differences.Specifically, the high rates of second-order predation in the forestecotone habitat, coupled with the positive relationships betweensecond-order predation and leaf area index, ground cover, vegeta-tion height, presence of residue, and absence of disturbance, pro-vide strong indication that complex habitats support higher ratesof second-order predation than do simple habitats. This is consis-tent with existing literature: complex habitats and ecotones gener-ally promote a favorable habitat for predators (Denno et al., 2005;Langellotto and Denno, 2004), including many invertebrates(Thiele, 1977) and small mammals (Larochelle, 1975a) that mayprey upon carabids.

4.3. Effects of second-order predation on seed predation and H. rufipesactivity-density

The results did not support Hypothesis 3: contrary to expecta-tion, second-order predation was uncorrelated with invertebrateseed predation. However, the strong positive correlation foundbetween invertebrate weed seed predation and H. rufipes activ-ity-density (Section 3.3) suggests that, as expected, second-orderpredation has the potential to negatively impact invertebrate seed

S.K. Birthisel et al. / Biological Control 70 (2014) 65–72 71

predation rates. We believe the predicted relationship may bedetectible with larger sample sizes.

4.4. Predicted effect of second-order predation on the weed seedbank

The simulation model allowed prediction of the sensitivity ofseeds entering the seedbank to both H. rufipes absolute densityand second-order predation rate (Fig. 6; Table 3). As second-orderpredation (z) increased, the rate of increase in seeds entering theseedbank (S) was less than proportional to the increase in z (Ta-ble 3). Thus, even a moderate second-order predation rate of1.4% day�1 (half our experimental estimate) in a system with aconservative initial H. rufipes absolute density (B) of 5 beetles m�2

may cause a greater than 10% increase in number of seeds enteringthe seedbank.

Conversely, the effect of increasing H. rufipes absolute densityon S was more than proportional to the rate of increase (Table 3).At z = 1.4%, our model predicted that S would decrease by nearlytenfold if B increased by a factor of four, from 5 to 20 beetlesm�2. Thus, the model suggests that supporting large H. rufipespopulations may provide substantial weed control benefits tofarmers, even in the face of considerable second-order predation.This corroborates the empirical finding that H. rufipes activity-den-sity appears to significantly drive invertebrate seed predation rates(Section 3.3), and highlights H. rufipes potential as a biological con-trol agent.

4.5. Considerations for future research

It is possible that by restricting the motion of otherwise highlymobile H. rufipes (Zhang, 1993) and presenting them via inherentlyunnatural assays, our methods made beetles more vulnerable tosecond-order predators that they would otherwise be. Tetheringseems a better reflection of natural conditions than pinning baitin place (e.g., Lundgren et al., 2007); however, further studies onthe effectiveness of tethering as a means of assessing predationof mobile species such as H. rufipes are recommended.

This study allowed identification of likely vertebrate second-order predators of H. rufipes (Fig. 3). Specific invertebrate second-order predators, however, were not identified, nor were the effectsof vertebrate vs. invertebrate second-order predation separated.Davis and Raghu (2010) found invertebrate seed predation to benegatively correlated with Araneae abundance, illustrating thepotential for invertebrate second-order predators to impact seedpredation. Future work to further explore the impact of predatoryinvertebrates on carabid communities and invertebrate seedpredation rates is recommended.

The finding that second-order predation is greater in complexhabitats could have implications for work aimed at supportingcarabid-mediated ecosystem services via agri-environmentalschemes (Gaines and Gratton, 2010; Guerrero et al., 2010;Woodcock et al., 2013). We suggest that the effects of habitatcomplexity on second-order predation be more thoroughly quanti-fied in future work.

5. Conclusions

This study used a ‘tethering’ method to provide documentationthat important seed predator H. rufipes falls prey to second-orderpredators. A model simulating the effect of second-order predationon seeds entering the weed seedbank predicted that moderaterates of second-order predation have the potential to substantiallyincrease the number of weed seeds entering seedbanks in temper-ate agroecosystems each fall. The results further suggest that

habitat complexity is an important driver of second-order preda-tion rates.

Acknowledgments

Many thanks to Frank Drummond for his experimental designassistance, Bill Halteman for statistical guidance, and DavidHiebeler for modeling advice. Thanks also to Krissy Birthisel,Bourcard Nesin, Dan Blanton, Conor McKaig, and Nick Innis fortheir assistance in lab and field. This study was made possiblethrough a Grant from, and is Publication Number 3345 of, theMaine Agricultural and Forest Research Station.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.biocontrol.2013.12.004.

References

Abbott, W.S., 1945. A method of computing the effectiveness of an insecticide. J.Econ. Entomol. 18, 265–267.

Baars, M.A., 1979. Catches in pitfall traps in relation to mean densities of carabidbeetles. Oecologia 46, 25–46.

Baraibar, B., Daedlow, D., De Mol, F., Gerowitt, B., 2012. Density dependence of weedseed predation by invertebrates and vertebrates in winter wheat. Weed Res. 52,79–87.

Benvenuti, S., 2007. Natural weed seed burial: effect of soil texture, rain and seedcharacteristics. Seed Sci. Res. 17, 211–219.

Birthisel, S.K., 2013. Sources of variability in agronomic weed seed predation: Time,space, habitat, and hyperpredation (Masters Thesis). University of Maine,Orono, ME, USA.

Bohan, D.A., Boursault, A., Brooks, D.R., Petit, S., 2011. National-scale regulation ofthe weed seedbank by carabid predators. J. Appl. Ecol. 48, 888–898.

Booman, G.C., Laterra, P., Comparatore, V., Murillo, N., 2009. Post-dispersalpredation of weed seeds by small vertebrates: interactive influences ofneighbor land use and local environment. Agric. Ecosyst. Environ. 129, 277–285.

Briggs, J.B., 1965. Biology of some ground beetles (col., carabidae) injurious tostrawberries. East Malling Res. Stn. Rep. 1964, 79–93.

Brust, G.E., House, G.J., 1988. Weed seed destruction by arthropods and rodents inlow-input soybean agroecosystems. Am. J. Alternative Agr. 3, 19–25.

Churchfield, S., Hollier, J., Brown, V.K., 1991. The effects of small mammal predatorson grassland invertebrates, investigated by field exclosure experiment. Oikos60, 283–290.

Crawley, M.J., 2013. The R Book, second ed. Wiley, Chichester, England.Davis, A.S., Raghu, S., 2010. Weighing abiotic and biotic influences on weed seed

predation. Weed Res. 50, 402–412.Denno, R.F., Finke, D.H., Langellotto, G.A., 2005. Direct and indirect effects of

vegetation structure and habitat complexity on predator–prey and predator–predator interactions. In: Castellanos, I., Barbosa, P. (Eds.), Ecology of Predator–Prey Interactions. Oxford University Press, New York, NY, USA, pp. 211–239.

Evans, D.M., Pocock, M., Brooks, J., Memmott, J., 2011. Seeds in farmland food-webs:resource importance, distribution and the impacts of farm management. Biol.Conser. 144, 2941–2950.

Gaines, H.R., Gratton, C., 2010. Seed predation increases with ground beetlediversity in a Wisconsin (USA) potato agroecosystem. Agric. Ecosyst. Environ.137, 329–336.

Gallandt, E.R., 2006. How can we target the weed seedbank? Weed Sci. 54, 588–596.Gallandt, E.R., Molloy, T., Lynch, R.P., Drummond, F.A., 2005. Effect of cover-

cropping systems on invertebrate seed predation. Weed Sci. 53, 69–76.Guerrero, I., Martínez, P., Morales, M.B., Oñate, J.J., 2010. Influence of agricultural

factors on weed, carabid and bird richness in a Mediterranean cereal croppingsystem. Agric. Ecosyst. Environ. 138, 103–108.

Hamon, N., Bardner, R., Allen-Williams, L., Lee, J.B., 1990. Carabid populations infield beans and thier effect on the population dynamics of Sitona lineatus (L.).Ann. Appl. Biol. 117, 51–62.

Harrison, S., Gallandt, E.R., 2012. Behavioural studies of harpalus rufipes de geer: animportant weed seed predator in northeastern us agroecosystems. Int. J. Ecol.2012, 1–6.

Hawthorne, A., Hassall, M., Sotherton, N., 1998. Effects of cereal headlandtreatments on the abundance and movements of three species of carabidbeetles. Appl. Soil Ecol. 9, 714–722.

Hawthorne, A., Hassall, M., 1995. The effect of cereal headland treatments oncarabid communities. In: Toft, S., Riedel, W. (Eds.), Arthropod Natural Enemiesin Arable Land I. Aarhus University Press, Aarhus, Denmark, pp. 185–198.

Heessen, H.J.L., Brunsting, A.M.H., 1981. Mortality of larvae of pterostichusoblongopunctatus (Fabricus) Col., Carabidae and Philonthus Decorus(Gravenhorst) Col Staphylinidae. Neth. J. Zool. 31, 729–745.

72 S.K. Birthisel et al. / Biological Control 70 (2014) 65–72

Heessen, H.J.L., 2013. Egg Mortality in Pterostichus Oblongopunctatus (Fabricius)(Col., Carabidae). Oecologia 50, 233–235.

Holland, J.M., 2002. Carabid beetles: their ecology, survival and use inagroecosystems. In: Holland, J.M. (Ed.), The Agroecology of Carabid Beetles.Intercept, Andover, England, pp. 1–40.

Jacob, H.S., Minkey, D.M., Gallagher, R.S., Borger, C.P., 2006. Variation inpostdispersal weed seed predation in a crop field. Weed Sci. 54, 148–155.

Lang, A., 2000. The pitfalls of pitfalls: a comparison of pitfall trap catches andabsolute density estimates of epigeal invertebrate predators in Arable Land. J.Pest Sci. 73, 99–106.

Langellotto, G.A., Denno, R.F., 2004. Responses of invertebrate natural enemies tocomplex-structured habitats: a meta-analytical synthesis. Oecologia 139, 1–10.

Larochelle, A., 1980. A list of birds of Europe and Asia as predators of carabid beetlesincluding Cicindelini (Coleoptera: Carabidae). Cordulia 6, 1–19.

Larochelle, A., 1975a. A list of mammals as predators of carabidae. Carabidologia 3,95–98.

Larochelle, A., 1975b. A list of North American birds as predators of carabid beetles.Carabologia 4, 153–163.

Larochelle, A., 1975c. A list of amphibians and reptiles as predators of NorthAmerican carabidae. Carabologia 3, 99–103.

Liebman, M., Gallandt, E.R., 1997. Many little hammers: ecological approaches formanagement of crop–weed interactions. In: Jackson, L.E. (Ed.), Ecology inAgriculture. Academic Press, San Diego, CA, USA, pp. 291–330.

Luff, M.L., 1982. Population dynamics of carabidae. Ann. Appl. Biol. 101, 164–170.Luff, M.L., 1980. The biology of the ground beetle Harpalus Rufipes in a strawberry

field in Northumberland. Ann. Appl. Biol. 94, 153–164.Lundgren, J.G., Shaw, J.T., Zaborski, E.R., Eastman, C.E., 2007. The influence of organic

transition systems on beneficial ground-dwelling arthropods and predation ofinsects and weed seeds. Renew. Agric. Food Syst. 21, 227–237.

Magura, T., Tóthmérész, B., Molnár, T., 2001. forest edge and diversity : carabidsalong forest-grassland transects. Biodivers. Conserv. 10, 287–300.

Marino, P.C., Westerman, P.R., Pinkert, C., van der Werf, W., 2005. Influence of seeddensity and aggregation on post-dispersal weed seed predation in cereal fields.Agric. Ecosyst. Environ. 106, 17–25.

Meiners, S., Handel, S., Pickett, S., 2000. Tree seedling establishment under insectherbivory: edge effects and inter-annual variation. Plant Ecol. 151, 161–170.

Meiss, H., Lagadec, L.L., Munier-Jolain, N., Waldhardt, R., Petit, S., 2010. Weed seedpredation increases with vegetation cover in perennial forage crops. Agric.Ecosyst. Environ. 138, 10–16.

Mirsky, S.B., Gallandt, E.R., Mortensen, D.A., Curran, W.S., Shumway, D.L., 2010.Reducing the germinable weed seedbank with soil disturbance and cover crops.Weed Res. 50, 341–352.

Parmenter, R.R., Macmahon, J.A., 1988. Factors influencing species composition andpopulation sizes in a ground beetle community (Carabidae): predation byrodents. Oikos 3, 350–356.

R-Development Core Team, 2011. R: A Language and Environment for StatisticalComputing. R Foundation for Statistical Computing, Vienna, <http://www.r-project.org/>.

Saska, P., Martinkova, Z., Honek, A., 2010. Temperature and rate of seedconsumption by ground beetles (Carabidae). Biol. Control 52, 91–95.

Shearin, A.F., Reberg-Horton, C.S., Gallandt, E.R., 2007. Direct effects of tillage on theactivity density of ground beetle (Coleoptera: Carabidae) weed seed predators.Environ. Entomol. 36, 1140–1146.

Shearin, A.F., Reberg-Horton, C.S., Gallandt, E.R., 2008. Cover crop effects on theactivity-density of the weed seed predator Harpalus rufipes (Coleoptera:Carabidae). Weed Sci. 56, 442–450.

Thiele, H.-U., 1977. Carabid Beetles and Their Environments: A Study on HabitatSelection by Adaptations in Physiology and Behaviour. Farner, D.S., Hoar, W.S.,Hoelldobler, B., Langer, H., Lindauer, M. (Eds.)Springer-Verlag, New York, NY,USA.

Thomas, C.F.G., Parkinson, L., Griffiths, G.L.K., Fernandez Garcia, A., Marshall, E.P.J.,2001. Aggregation and temporal stability of carabid beetle distributions in fieldand hedgerow habitats. J. Appl. Ecol. 38, 100–116.

Thomas, C.F.G., Parkinson, L., Marshall, E.J.P., 1998. Isolating the components ofactivity-density for the carabid beetle Pterostichus Melanarius in farmland.Oecologia 116, 103–112.

Tooley, J., Brust, G.E., 2002. Weed seed predation by carabid beetles. In: Holland,J.M. (Ed.), The Agroecology of Carabid Beetles. Intercept, Andover, England, pp.215–230.

Westerman, P.R., Dixon, P.M., Liebman, M., 2009. Burial rates of surrogate seeds inarable fields. Weed Res. 49, 142–152.

Westerman, P.R., Borza, J.K., Andjelkovic, J., Liebman, M., Danielson, B., 2008.Density-dependent predation of weed seeds in maize fields. J. Appl. Ecol. 45,1612–1620.

Westerman, P.R., Liebman, M., Menalled, F.D., Heggenstaller, A.H., Hartzler, R.G.,Dixon, P.M., 2005. Are many little hammers effective? Velvetleaf (Abutilontheophrasti) population dynamics in 2- and 4-year crop rotation systems. WeedSci. 53, 382–392.

Woodcock, B., Savage, J., Bullock, J.M., Nowakowski, M., Orr, R., Tallowin, J.R.B.,Pywell, R.F., 2013. Enhancing beetle and spider communities in agriculturalgrasslands: the roles of seed addition and habitat management. Agric. Ecosyst.Environ. 167, 79–85.

Zhang, J., 1993. Biology of Harpalus rufipes DeGeer (Coleoptera: Carabidae) in Maineand Dynamics of Seed Predation (Masters Thesis). University of Maine, Orono,ME, USA.

Zhang, J., Drummond, F.A., Liebman, M., Hartke, A., 1997. Insect Predation of Seedsand Plant Population Dynamics. University of Maine, Orono, ME, USA.