AMNWR 06/08

MAPPING DISTRIBUTION AND RELATIVE DENSITY OF NESTING LEAST AND CRESTED AUKLETS AT SEGULA ISLAND, ALASKA, MAY-JUNE 2006

Heather M. Renner1 and Joel H. Reynolds2

Key Words: Aethia cristatella, Aethia pusilla, Aleutian Islands, colony mapping, crested auklet, least auklet, Segula Island

U.S. Fish and Wildlife Service

1Alaska Maritime National Wildlife Refuge 95 Sterling Highway, Suite 1

Homer, Alaska 99603

2Division of Natural Resources National Wildlife Refuge System

1011 E. Tudor Rd, MS 221 Anchorage, Alaska 99503

September 2006

_______

Cite as: Renner, H.M. and J.H. Reynolds. 2006. Mapping distribution and relative density of nesting least and crested auklets at Segula Island, Alaska, May-June 2006. U.S. Fish and Wildlife Service Report AMNWR 06/08. Homer, AK.

EXECUTIVE SUMMARY The Alaska Maritime National Wildlife Refuge was established, in part, to conserve marine bird populations. This requires the ability to detect large changes in their abundance. Monitoring populations of auklets and other crevice-nesting seabirds has proven problematic even though numerous methods have been attempted since the mid 1960’s. Quantifying changes in geographical size of auklet colonies may be a useful alternative to directly measuring population size. Anecdotal evidence suggests several large colonies have decreased recently in both extent and abundance, simultaneous with vegetation encroachment and succession. In May-June 2006 we employed a recently developed standardized method for colony mapping, using a randomized systematic grid survey, on Segula Island. Additionally, we improved the method to support fitting patch occupancy models to account for less than perfect detection (Mackenzie et al. 2002). These models simultaneously estimate detection rates and the occupancy rate, allowing for unbiased estimates of occupancy and colony area. Quantitatively mapping all large auklet colonies is logistically feasible using this method and could provide an important monitoring baseline for the status of auklet colonies through time. INTRODUCTION During May-June 2006, we visited Segula Island (52° 02’00” N, 178° 09’00” E) in the Aleutian Island Unit of the Alaska Maritime National Wildlife Refuge. As part of the Refuge’s seabird monitoring program, we systematically surveyed the least and crested auklet (Aethia pusilla and A. cristatella) colony near Gula Point. To provide baseline data on geographic extent and relative density of auklets in the colony, we mapped the colonies following Renner et al. (2006) with the addition of slight amendments described below.

Two-dimensional area and geographic extent of the Gula Point auklet colony was estimated using three different methods. Each method determined the presence or absence of nesting auklet evidence on a sample of cells from an overlain grid of cells; cells were approximately 100 m x 100 m squares.

Method 1 - One 16 m2 survey plot per grid cell was visited following Renner et al. (2006) to estimate colony area without correcting for lack of perfect detection, i.e., this method is biased and underestimates colony area.

Method 2 - Five survey plots on a random sample of 80 grid cells were visited to get unbiased estimates of detection probability, patch occupancy and their associated uncertainties.

Method 3 - Each cell’s occupancy was subjectively assessed by having the field observer search the cell area exhibiting the most suitable habitat. This was done strictly for comparison to Method 2 as this approach was expected to provide the highest detection probability possible and thus highest colony area estimates, though it is impossible to standardize the method or to estimate its associated uncertainty. An additional objective was to estimate relative density by counting feathers and droppings on sampling plots visited in Method 1. Why use patch occupancy models to estimate colony area?

Presence-absence data obtained from a single visit to a sample of sites provides biased estimates of the proportion of sites (or sampled area) occupied (occupancy rate) because of confounding with less than perfect detection. The absent sites are a mixture of sites where the

2

item of interest was truly absent and sites where the item was present but undetected. This leads to underestimates of occupancy rates and, for long-term monitoring, confounding of changes in occupancy with changes in (the unknown) detection rate.

Less than perfect detection can occur for at least three reasons: due to only observing a subset of the full cell (spatial subsampling), observer error (deficient searching), or loss of evidence due to weather (temporal subsampling). The main approach presented here, Method 2, accounts for the first factor, which we feel is dominant in this application. It also reduces the second source by choice of observational unit (small enough for thorough searching) and standardized guidelines; the third factor may be partially controlled by field work timing.

Occupancy models overcome the lack of perfect detection by simultaneously estimating detection rates and the occupancy rate, allowing for unbiased occupancy estimates. Fitting these models, which are built from standard mark-recapture models, requires multiple independent sampling events at each site (Mackenzie 2002, 2003).

STUDY AREA The colony was described in 1979 (Early et al. 1980) and more thoroughly surveyed in 1995 (Thomson 1995a). However, methods for detecting auklet presence were not standardized nor were survey points georeferenced. The auklet colony lies south of Gula Point (Thomson 1995) and consists of a densely overgrown lava flow that is 200-300 years old (Nelson 1959). Much of the area used by auklets is covered in tall grass tussocks where nesting crevices only occur in very low densities. It is impossible to assess habitat change since 1995, but the colony appears to be rapidly overgrowing. One set of three observation plots used for attendance counts in 1995 was completely overgrown by 2006 and gave no evidence of any potential habitat or attending birds. METHODS The survey was conducted between 24 May and 6 June 2006 (early incubation period) at the Gula Point auklet colony on the north end of Segula Island. Weather in this period was unusually dry for summer in the Aleutians, with no rain. There is no regular reference for judging this year’s apparently low auklet attendance on Segula, but during our survey period attendance was reported to be extremely low on Buldir Island (approximately 10% of ‘normal’ until around 7 June; Ian L. Jones, pers. comm..) but not on Kasatochi Island (Brie Drummond, pers. comm.). Nesting chronology for auklets was about a week later than normal at both sites.

Survey methods, detailed below, followed Renner et al. (2006) with slight variation due to the additional objectives described above (‘Method 2’). The need for multiple independent samples within each grid cell required dividing each grid cell into equal-sized sample units which fit evenly into a cell, e.g. partitioning the grid cell into squares or hexagons. In contrast, Renner et al. (2006) just randomly selected the center of a circular observation plot within each grid cell. Our choice of hexagonal sample units led to slightly off-square grid cells formed from rectangular arrays of hexagonal sample units.

The circular observation unit was retained due to its simple implementation in the field. Since the largest circular observation unit that could be inscribed in a sample unit did not fully cover the hexagon sample unit (area covered = 91%) or square sample unit (78.3%), there was an added component in the probability of detection attributable to observing only a subset of each

3

sample unit. Effectively, this is just incorporated into the overall probability of detection. The choice of hexagonal sample units was driven by the desire to minimize this added component. Sampling Design

The observational unit was a 16 m2 circular plot (Figure 1) inscribed in the sample unit, a hexagon with radius 2.26 m and area 17.58 m2 (91% of the sample unit was observed). A grid of sample units was overlain across the whole island, partitioned for logistical convenience into subregions (approximately 1 km x 1 km) composed of squarish cells (approximately either 100m x 100m or 50m x 50m) of sample units.

Final dimensions of logistically convenient collections of sample units were: Cells: extent North - South = 108.3 m; extent East - West = 101.6 m, i.e., a 26 row x 24 column array of sample unit hexagons. Cell area was 10969.92 m2. Subregions: extent North - South = 1083.2 m; extent East - West = 1016.3 m, i.e., a 10 x 10 array of cells.

The counting unit for the colony area estimate was the cell. A cell was considered occupied if any evidence of occupancy was detected within the cell. That is, the analysis ignored the fact that some cells might not have been completely filled with potential habitat.

Stage 1. [before heading to the field] Create and overlay a grid of cells across the island. Stage 2. [1 day] Define the initial sample frame (collection of relevant cells) by

exploring the colony perimeter via auditory and visual cues of occupancy. On the first day in the field we walked the approximate perimeter of the colony. The sample frame was defined as all cells crossed by or contained within the approximate perimeter, along with a 1-2 cell thick boundary around the approximate perimeter.

Stage 3. [6 days, two people working independently] Visit an unaligned systematic sample of one sample unit per cell in the sample frame. The observational unit was centered on the center of the selected sample unit and investigated for any evidence of auklet presence: belly feathers, droppings, vocalizations, vegetation wear, or birds attending on the surface of the circular plot.

Each visited observational unit was classified into one of four categories: Present – (‘occupied’) evidence detected, Absent – (‘unoccupied’) no evidence detected, Non-Habitat – (‘unoccupied’) the observational unit did not contain any potential

habitat (i.e., was a snow field, water, grass plain with no crevices, etc.), or X – (‘missing’) inaccessible due to occurring on a cliff face or inaccessible beach

section. Density Survey – At most observational units visited during Stage 3 counts of feathers

and droppings were recorded rather than just presence or absence of evidence. Counts at few high density units were omitted due to time constraints.

Expert Search - The center of each cell was located and the whole cell assessed as either: Non-Habitat - cells completely composed of some combination of water, solid plain

of grass with no rocks or talus offering crevice openings, or tundra / snowfields completely covered and offering no crevice openings, etc., or cells that were both

4

Figure 1. Diagram showing sampling design used for documenting occupancy by auklets at Segula Island, Alaska in 2006. Cells shown here were arranged in subregions - 10x10 arrays shown in Figure 2.

5

(i) composed of < 1% (surface area) of crevice-containing habitat and the rest non-

habitat and (ii) that were discontiguous or isolated from the main colony; or Potential Habitat – cells containing crevice openings and that were not clearly Non-

Habitat. All potential habitat cells were searched for evidence of occupancy by investigating,

initially, the potential habitat closest to the initial sample observational unit. The search continued until evidence was found or 5-10 minutes effort had been expended. Thus the Expert Search led to the following cell classifications:

Non-Habitat – defined above; Present (‘occupied’) – evidence found in cell by directed search; Absent (‘unoccupied’) – no evidence found in cell by directed search.

Sample Frame Refinement. At the end of Stage 3, all visited cells that had been classified as Non-Habitat were removed from the sample frame. Note that such cells either formed a boundary around the colony area or occurred within the colony region itself (patches). If cells containing cliffs had inaccessible initial sampling plots (e.g., on a cliff), these plots were treated as ‘missing data’ in the analysis.

Stage 4. [4 days, two people working independently] Revisit a subset of cells. A simple random sample of 80 Potential Habitat cells was selected from those defined at the end of Stage 3. A simple random sample of four more sample units was selected from each of these cells. Each of the additional plots was investigated for evidence of occupancy. The choice of 80 cells and four sample units was based on field logistics; higher values for either choice would have further improved precision of the estimates.

Patch Occupancy Analyses

The sample design satisfies all the assumptions of the patch occupancy models introduced by MacKenzie et al. (2002, 2003) with the extension that only a subset of cells were visited multiple times (Appendix 1). In contrast to their original development (MacKenzie et al. 2002), in this application the multiple samples per cell vary in space (location within the cell) rather than in time (visitation period at the site). Each visited sample unit resulted in one of three responses: 0 (unoccupied / absent), 1 (occupied / present), or x (missing data). The standard model was simplified in this application since the order of observation of the sample units did not matter.

Three models were fit: the standard patch occupancy model that assumed all occupied cells share a common probability of detection, an extension of that model that allowed for two latent classes (strata) of occupied cells with differing probabilities of detection, and a further extension allowing for three strata (Pledger 2000). All three models were fit using maximum likelihood methods and the best fitting model selected using AIC (Burnham and Anderson 2002). Goodness of fit for the selected model was assessed using the standard Chi-square goodness of fit statistic with the null reference distribution estimated via Monte Carlo simulation (1000 simulated samples) from the fitted model (MacKenzie and Bailey 2004). Standard errors of parameters of interest were estimated from 1000 nonparametric bootstrap resamples (Lunneborg 2000). Confidence intervals were calculated as Estimate +/- 1.96 x Standard Error (Lunneborg 2000).

6

7

Prob(cell results |stratum A) 10 Prob(cell results | stratum B)≥ ×

Classification of revisited cells to stratum Fitting a patch occupancy model that allows for strata differing in their probability of

detection does not directly assign each sampled cell to a specific stratum (e.g., unoccupied, occupied - low density, occupied – high density). However, a classification rule for assigning cells to stratum can be developed from the parameter estimates. The essential step is to decide, before viewing the data or parameter estimates, how much more likely a cell’s observations must be, assuming the cell belongs to stratum A compared to stratum B, in order to assign the cell to stratum A. We required a classification likelihood level of 10, e.g. to assign a cell to stratum A rather than stratum B. Software

The survey grid of cells and observation unit samples were generated using programs written by the second author for the statistical data analysis package and environment R (R Development Core Team 2005), as were the summaries and figures. The code is available upon request. Occupancy models were fit using the freeware program PRESENCE (MacKenzie and Hines 2002, http://www.mbr-pwrc.usgs.gov/software/presence.html). Maps were prepared using ArcGIS version 9.1, and all occupancy and density data are archived in a FileMaker Pro version 7 database.

RESULTS Sample Frame Development

Stage 2 identified a contiguous set of 308 cells for visitation as potential habitat. Of these, the Expert Search of Stage 3 identified 92 cells as Non-Habitat (Figure 2), leaving a revised colony sample frame of 216 cells. Six cells had inaccessible initial sampling plots on seaside cliffs; these plots were treated as ‘missing data’ in the analysis.

Based on available field effort, a random sample of 80 cells from the 216 was selected for revisiting in Stage 4. Four (additional) sample units were randomly selected from each of these 80 cells. Initially, we had planned to stratify the cells by cover type – bare rock, moss, vegetation. This intention was dropped in the field during Stage 3 as there was one predominant cover type for all cells – thick vegetation.

The sampling design provides for three naïve colony area estimates as well as that from the occupancy model estimation (Table 1), where an estimate is naïve if it ignores the bias caused by lack of perfect detection. Figure 2 shows occupied cells as designated by each of the three naïve estimate methods. The occupancy model does not assign status to specific cells, so we cannot generate a map using this method (but see below – Classification of revisited cells to stratum).

Figure 2. Map of Gula Point auklet colony, Segula Island, Alaska in 2006. Cells are color coded by occupancy status as determined by the Expert Search, and status of individual sample plots within cells is shown as well.

Table 1. Survey summaries and associated colony area estimates from different methods for sampling the auklet colony for evidence of occupancy. See text for survey methods. Only the occupancy model accounts for less than perfect detection to give an unbiased colony area estimate; the area estimates from the first three methods are all biased low by different amounts. In the first two methods Area Estimate = Number of Occupied Cells * Cell Area, in the last two methods Area Estimate = Occupancy Estimate * Total Number of Cells * Cell Area. Relative Bias was estimated as (Area Estimate / Occupancy Model Area Estimate – 100%). Non-

Habitat Sample Inaccessible

Unoccupied Occupied Total Cells

Occupancy Estimate

Area Estimate (km2)

Relative Bias

Expert Search of cell

90 75 141 308 141/216 = 0.653

1.55 -21%

Stage 3 (1 sample unit / cell)

36 6 98 76 216 76/(216-6) = 0.362

0.83 -58%

Stage 41 (5 sample units / cell)

28 52 80 52/80 = 0.65

1.54 -21%

Occupancy Model

216 0.828 1.96

Patch Occupancy Model Selection

The data clearly supported the model allowing for two cell strata differing in their probability of detection (Table 2, Delta AIC and Model Likelihood columns). The goodness of fit assessment failed to detect any significant departure of the observations from the selected model (Table 2, Goodness of Fit).

The patch occupancy model, by accounting for the less than perfect detection of evidence in a sample unit, estimated colony area at 1.96 km2 (95% bootstrap CI: 1.47, 2.50 km2) (Table 3). The identified strata greatly differed in their probability of detection, 0.20 versus 0.78 (Table 3), with the majority of the occupied colony being attributed to the lower density strata (πLow = 64% of occupied colony, Table 3; Figure 1). Collecting five samples from each of the 80 cells provided fairly precise estimates of detection probabilities; multiple samples would have had to be collected from more than 80 cells to improve precision of the probability of occupancy estimate (Table 3).

1 Treats all visited cells whose five sample units failed to detect any evidence as absent rather than simply undetected.

Table 2. Model selection results from fitting models differing in the number of latent cell strata to data from all potential habitat cells (216 cells) and treating inaccessible plots as missing data. The model with the smallest AIC is considered the best from among those under consideration (Burnham and Anderson 2002). Delta AIC = AIC – smallest AIC across models and provides a relative measure of model performance among the models under consideration. Model Likelihood is a measure of the observations’ weight of evidence for a specific model relative to the others. The Goodness of Fit P values are from a Monte Carlo test of each model’s adequacy as a description of the underlying data. The data clearly are best described by the model allowing for two strata differing in their detection probabilities. Model: Number of Detection Strata AIC Delta

AIC Model

LikelihoodGoodness

of Fit 1 624.03 23.2 0.00 0.00 2 600.83 0.0 0.88 0.22 3 604.83 4.0 0.12 0.03

Table 3. Parameter and Colony Area estimates from the occupancy model with two detection strata (‘Model 2’ Table 2). Bootstrap standard errors (‘SE’) and confidence intervals (‘CI’) are described in the text. Prob(Occupied) is the probability that a randomly selected cell in the sample frame is occupied . πLow is the probability a randomly selected occupied cell has low density of evidence (formally, has low detection probability). Prob(Occupied) (SE)

Area km2 (95 CI%)

πLow (SE)

Prob(Detect | Low Density) (SE)

Prob(Detect| High Density)(SE)

0.83 (0.110)

1.96 (1.47, 2.50)

0.64(0.13)

0.20(0.08)

0.78 (0.8)

Classification of revisited cells to stratum Requiring a classification likelihood level of 10 to assign a sampled cell to a particular stratum based on the parameter estimates from Model 2 (Table 3) assigned cells exhibiting 4 or 5 sample units with evidence to the high density stratum and those with 1 sample unit with evidence to the low density stratum (Table 4). A specific cell exhibiting no evidence on any of 5 sample units could not be specifically assigned to the Unoccupied versus Occupied-but-low-density stratum. Table 4. Probability distributions for number of sample plots with evidence for each possible cell stratum, based on parameter estimates from Model 2 (Table 3). The final column assigns a cell to a stratum based on the requirement of a likelihood ratio ≥ 10. Samples with Evidence (out of 5)

Probability | Unoccupied

Probability | Low Density

Probability | High Density

Detection Strata Classification

0 1 0.3277 0.0005 Unoccupied or Low 1 0 0.4096 0.0091 Low 2 0 0.2048 0.0648 Occupied 3 0 0.0512 0.2297 Occupied 4 0 0.0064 0.4072 High 5 0 0.0003 0.2887 High

10

DISCUSSION Colony area estimates The Segula auklet colony is currently estimated to have an area of 1.96 km2 (Table 3) after accounting for less than perfect detection due to spatial subsampling of each cell. Treating this unbiased estimate as the true colony area resulted in estimated relative biases of -20% to – 50% for the methods that did not account for the lack of perfect detection (Table 1). It is important to note that compared to the occupancy model even the Expert Search underestimated colony area by 21% (Table 1) as some areas identified as ‘Absent’ during the Expert Search were likely occupied. Comparisons with earlier area estimates In the late 1970s the Segula colony area was estimated to be 0.78 km2 (Early et al. 1980), with 60% of that being unused habitat. This estimate was based on walking the perimeter in one day and using a polar planimeter to measure the mapped area. In the early 1990s the colony area was estimated to be 1.28 km2 (Thomson 1995a). This estimate was based on more extensive and intensive surveys but exact detection methods were not described. Both estimates are much smaller than even the lower bound of the 95% confidence interval estimate for our current colony area estimate (1.47 – 2.50 km2, Table 3).

Unfortunately, the differences in colony area estimates cannot be clearly attributed to temporal changes in actual colony area, because they are confounded with changes in survey methods across the years. Due to issues described in Renner et al. (2006), none of the earlier surveys can be reproduced nor their uncertainties or biases estimated. Even so, it is apparent from the current study that ignoring the lack perfect detection severely biases the colony area estimates (Table 1). Relative to former surveys, the current effort was quite rigorous, intensive, standardized, and reproducible, yet ignoring the problem of imperfect detection was seen to cause colony area underestimates by as much as 50% (Table 1). The biases associated with the earlier surveys are plausibly much higher. Thus differences in survey methods are undoubtedly the dominant factor underlying the apparent increase in colony area through time, such as differences in the probability of detecting evidence of occupancy, differences in the underlying definitions of occupancy, and differences in the treatment of enclosed but uninhabited habitat. Additionally, colony area estimates depend on the measurement scale or counting unit (Renner et al. 2006): the current survey used units approximately 100 m x 100 m, the earlier surveys have no associated scale. The differences in scale further undermine the ability to compare area estimates.

Actually, non-quantitative field evidence leads us to strongly believe the colony area has decreased through time. Specific evidence includes: 1) dense overgrowth of the colony and occurrence of large unoccupied areas – though described by Early et al. (1980), the trend in vegetation colonization and success is only expected to have continued since their survey, not reversed; 2) except for a few subsections, the majority of currently occupied locations were restricted to the steeper side-slopes of the ridges and seldom found in the now heavily vegetated gully and valley bottoms; 3) low auklet attendance relative to the descriptions from earlier visits, though this comparison is confounded by observed low attendance at the Buldir colony during our 2006 survey period, presumably due to a late spring; 4) at least one area used for attendance counts in 1995 was completely unoccupied in 2006; and 5) consideration of the potential relative biases in earlier survey methods as described above.

11

The absence of a standardized survey method and, in earlier surveys, of adjustment for lack of perfect detection, prevents any quantification of changes in colony area through time. This is not to cast aspersions on the earlier efforts but rather highlights the importance of employing a standardized survey method as argued by Renner et al. (2006). In particular, the current results clearly demonstrate the importance of using a standardized method that can account for lack of perfect detection of occupancy and provide for associated estimates of uncertainty. Methods assessment

Especially in low density colonies, the methods described in Renner et al. (2006) result in an underestimate of the true number of occupied cells. In this project, we tried to achieve a balance between reproducibility and accuracy. For example, the “Expert Search” technique may be considered the most accurate colony survey method as it incorporates the biological and field experience of the auklet researcher in identifying occupied cells. However, it suffers from three deficiencies: 1) it is least reproducible as there is no standardized method for identifying good habitat and searching a known area, 2) it does not support estimates of uncertainty such as standard errors or confidence intervals, and 3) it is still subject to potentially severe bias in low density colonies (Table 1). Expert Searches are thus probably the least useful method for monitoring trends through time.

The patch occupancy model and associated survey method produced the least biased colony area estimate as it accounts for lack of detection due to spatial subsampling (but not temporal subsampling). However, the occupancy model does not directly classify each cell’s occupancy. While the Expert Search misses some occupied cells, it is able to define the occupancy of some cells that the patch occupancy model is unable to differentiate between unoccupied and occupied-but-low-density (Figure 1). A combination of both approaches, as implemented here, appears to provide the most useful information in terms of both unbiased colony area estimates and standard errors as well as explicit georeferenced maps of known occupied cells.

Monitoring changes

Colony area change could not be quantitatively assessed due to the differences among survey methods described above, especially the lack, in earlier surveys, of a sampling method providing unbiased estimates of colony area by accounting for lack of perfect detection. If the current survey methods had been applied in the late 1970s and early 1990s, two assessments of change in colony area could be applied.

Trends in total colony area could be assessed using weighted regression, where each estimate would have an associated weight equal to 1/(SE2). Either linear or nonlinear models could be used as warranted by the observed trends.

Changes in spatial pattern of occupancy, i.e., changes in occupancy status of georeferenced cells, might be investigated assuming each survey used a similar method of classifying cell occupancy – perhaps through a combination of the patch occupancy survey and the Expert Search survey described above. The most informative summary would be visual through mapping of change in cell occupancy status. Changes between any two surveys could be summarized quantitatively via a sign test (Conover 1999). Changes from high to low density occupancy would be interpretable. However, the assessments are limited by the inability, during each survey, to distinguish between unoccupied and occupied but low density cells.

12

Sample unit choice

The requirement for visiting multiple sample units randomly selected from a grid cell necessitated a choice between hexagonal or square sample units. We chose to use hexagonal sample units to maximize the spatial correspondence between sample unit (hexagon) and observational unit (circle), and thus minimize this component of spatial subsampling in the probability of detection. Given that use of a patch occupancy model simply incorporates this component into the overall probability of detection, we recommend that future surveys use square sample units to simplify the survey design and its description. ACKNOWLEDGMENTS We thank Greg Thomson for providing accurate and useful information regarding our campsite and the colony. Art Sowls kindly loaned us much of the camp equipment used for this project. Danielle Jerry generously supported Joel’s intensive involvement in the field work which resulted in clarifying a number of important biometrical issues. Martin Renner provided invaluable help and ideas at all stages, from study design (love those hexagons!) to packing to telephone field support to creating maps and figures. Conversations with Ian Jones helped clarify a number of points. We especially thank the captain and crew of the M/V Tiglax for safe and comfortable transport to and from the island (and the hot showers). We sincerely appreciate the support of our spouses, especially Carol Ann’s understanding with regard to rescheduled anniversary celebrations.

13

LITERATURE CITED Burnham, K. P. and D. R, Anderson. 2002. Model selection and multimodel inference: a

practical information-theoretic approach (2nd ed). Springer, New York. Conover, W. J. 1999. Practical Nonparametric Statistics (3rd ed). Wiley, New York. Early. T., J. Beall, W. Henry and A. Taber. 1980. Results of bird and mammal surveys of the

western Aleutian Islands, summer 1979. U.S. Fish and Wildl. Serv. Rep., Adak, AK. 140 pp.

Lunneborg, C. E. 2000. Data Analysis by Resampling: Concepts and Applications. Duxbury

Press, Pacific Grove, CA. MacKenzie, D. I. and J. E. Hines. 2002. Presence 2.0 http://www.mbr-

pwrc.usgs.gov/software/presence.html MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, J. A. Royle and C. A. Langtimm.

2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83: 2248-2255.

MacKenzie, D. I., J. D. Nichols, J. E. Hines, M. G. Knutson and A. B. Franklin. 2003.

Estimating site occupancy, colonization and local extinction when a species is detected imperfectly. Ecology 84: 2200-2207.

MacKenzie, D. I. and L. L. Bailey. 2004. Assessing the fit of site-occupancy models. Journal of

Agricultural, Biological and Environmental Statistics 9(3): 300-318. Nelson, W.H. 1959. Geology of Segula, Davidof, and Khvostof Islands Alaska. Geological

Survey Bulletin 1028-K. U.S. Government Printing Office, Washington. D.C. 15pp. Pledger, S. 2000. Unified maximum likelihood estimates for closed capture-recapture models

using mixtures. Biometrics 52(2): 639-649. R Development Core Team. 2005. R: a language and environment for statistical computing. R

Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org.

Renner, H.M., M. Renner, J. H. Reynolds, A. M. A. Harding, I. L. Jones, D. B. Irons and G. V.

Byrd. 2006. Colony mapping: a new technique for monitoring crevice-nesting seabirds. Condor 108: 423-434.

Thomson, G. 1995a. Counts of least and crested auklets following fox removal at Segula Island,

Alaska, in 1995. U.S. Fish and Wildl. Serv. Rep. AMNWR 95/05, Homer, AK. 29 pp.

14

Thomson, G. 1995b. Eradication of arctic foxes from Segula Island, Alaska, in 1995. U.S. Fish and Wildl. Serv. Rep. AMNWR 95/06, Homer, AK. 35 pp.

15

Appendix A. Annotated list of birds observed at the north end of Segula Island, 23 May – 9 June 2006. Cackling Goose Branta hutchinsii leucopareia – Aleutian Cackling geese were observed in small

groups (2-10 birds) most days throughout the duration of our stay. Typically they were seen or heard on the ridge above camp, although numerous droppings were present in the auklet colony.

Green-winged Teal Anas crecca nimia – 2 males were frequently observed in the cove near

camp. Harlequin Duck Histrionicus histrionicus – Groups of up to 15 were frequently observed around

the mouth of Camp Cove. White-winged Scoter Melanitta fusca – One was observed in Camp Cove on 24 and 26 May. Rock Ptarmigan Lagopus muta – Very abundant on higher parts of the island, with numerous

dropping piles in upper areas of the auklet colony. Fork-tailed Storm-Petrel Oceanodroma furcata – Vocalizations were heard nightly in Camp

Cove, indicating a few hundred birds in the cove. A few nested in the auklet colony as well but we did not spend time there during the night. No Leach’s storm-petrels (Oceanodroma leucorhoa) were detected, in contrast to Thomson’s (1995b) findings.

Red-faced Cormorant Phalacrocorax urile – Two or three individuals were occasionally

observed flying in groups of pelagic cormorants near Gula Point, and near the mouth of Camp Cove.

Pelagic Cormorant Phalacrocorax pelagicus – Small groups (up to 10 birds) were regularly

observed flying near the auklet colony cliffs, where nests were documented in 1979 and 1995 (Thomson 1995b). Because we didn’t have a boat, we were unable to see the cliffs to look for nests.

Bald Eagle Haliaeetus leucocephalus – Several pairs probably nested on Segula in 2006. At least

one nest was located in the auklet colony at the southeast edge of Gula Point – it contained two chicks on 3 June. Up to 5 birds (3 adults, 2 immature) were observed flying simultaneously near Camp Cove in late May.

Peregrine Falcon Falco peregrinus – At least three nests were present on the north end of the

island (one above camp, one atop a sea cliff in the auklet colony, and another high up in the auklet colony.

Black Oystercatcher Haematopus bachmani – Two pairs nested in Camp Cove. A nest with

three eggs was observed on 28 May. Wandering Tattler Tringa incana – A single bird was observed in Camp Cove on 28 May.

16

Rock Sandpiper Calidris ptilocnemis – Several birds were routinely observed on the ridge east of

Camp Cove. No nests were found, but birds were both displaying and chasing each other.

Glaucous-winged Gull Larus glaucescens – Up to 20 were observed in Camp Cove, with only

one abandoned nest found. Gulls were also observed in small numbers in the auklet colony, although apparently in much lower numbers than described by Thomson (1995b).

Pigeon Guillemot Cepphus columba – A few birds nested in camp cove and were seen / heard

daily. Parakeet Auklet Aethia psittacula – Nested in and around Camp Cove and were seen / heard

regularly nearshore in all areas we visited. Least Auklet Aethia pusilla – On this short trip we were unable to make it to Chugul Point,

where Thomson (1995b) recorded a small number of nesting auklets. We only observed auklets in the Gula Point colony area described in this report.

Crested Auklet Aethia cristatella – We estimated that crested auklets represent 20% of the birds

nesting in the Gula Point colony but did not conduct quantitative attendance counts. Horned Puffin Fratercula corniculata – Observed daily in Camp Cove throughout our stay. Up

to 10 birds were often observed flying to cliffs on the west side of the cove, indicating probable nesting.

Tufted Puffin Fratercula cirrhata – Small numbers (<5) were observed offshore on most days. Common Raven Corvus corax – Two were observed in the auklet colony on 26 May and again

on 3 June. Winter Wren Troglodytes troglodytes – Observed daily both in and around camp, and in the

auklet colony. Song Sparrow Melospiza melodia – Observed daily both in and around camp, and in the auklet

colony. One nest was located in camp, with an adult observed carrying food on 8 June. Lapland Longspur Calcarius lapponicus – Observed daily both in and around camp, and in the

auklet colony. Snow Bunting Plectrophenax nivalis – Observed daily at higher elevations, both above camp and

in the auklet colony. Gray-crowned Rosy-Finch Leucosticte tephrocotis - Observed daily both in and around camp,

and in the auklet colony.

17

Appendix B. Coordinates of observational plot centers visited on Segula Island in 2006. All points are UTM zone 60N, datum WGS 84. Status indicates presence/absence of auklet evidence on the observational plot. E.S. indicates status of the entire cell based on the expert search. Plot names ending in b-e indicate multiple samples from the same cell (Stage 3), and do not have an expert search designation. Plot name Easting Northing Status E.S.d04-081a 578045 5765202 NH A d04-082a 578181.8 5765119 A A d04-083a 578283.4 5765128 A A d04-084a 578443.7 5765152 NH A d04-087a 578670.4 5765175 NH NH d04-088a 578822.8 5765137 A A d04-089a 578940.1 5765150 A A d04-090a 579018.3 5765141 P P d04-091a 578119.2 5765317 NH A d04-092a 578150.5 5765272 NH NH d04-093a 578248.2 5765252 NH A d04-094a 578439.8 5765231 NH A d04-095a 578486.7 5765268 A P d04-096a 578623.5 5765301 P P d04-097a 578740.7 5765247 A A d04-097b 578666.5 5765268 A d04-097c 578670.4 5765256 A d04-097d 578717.3 5765256 A d04-097e 578748.6 5765279 A d04-098a 578783.7 5765272 A A d04-099a 578873.6 5765288 A P d04-099b 578858 5765243 A d04-099c 578889.3 5765279 A d04-099d 578869.7 5765227 A d04-099e 578912.7 5765306 A d04-100a 578963.5 5765277 NH P d05-081a 579131.6 5765171 A A d05-082a 579170.7 5765216 NH A d05-083a 579272.3 5765148 A A d05-084a 579440.4 5765200 NH A d05-085a 579526.4 5765168 NH NH d05-086a 579620.2 5765159 NH NH d05-087a 579737.5 5765200 NH NH d05-088a 579796.1 5765180 NH NH d05-089a 579913.4 5765125 NH NH d05-090a 580050.2 5765182 NH NH d05-091a 579116 5765315 P P d05-091b 579131.6 5765329 A d05-091c 579100.4 5765324 A

Plot name Easting Northing Status E.S.d05-091d 579100.4 5765238 A d05-091e 579104.3 5765240 A d05-092a 579174.6 5765268 NH P d05-092b 579213.7 5765304 A d05-092c 579237.2 5765295 A d05-092d 579182.4 5765263 A d05-092e 579245 5765231 P d05-093a 579319.2 5765319 P P d05-093b 579295.8 5765301 A d05-093c 579284.1 5765245 P d05-093d 579354.4 5765259 A d05-093e 579346.6 5765240 A d05-094a 579385.7 5765236 NH A d05-095a 579467.8 5765310 A A d05-096a 579577.2 5765288 A P d05-097a 579717.9 5765283 NH NH d05-098a 579827.4 5765297 NH NH d05-099a 579944.7 5765252 NH NH d05-100a 580065.8 5765254 NH NH e03-009a 577845.6 5765389 A A e03-010a 578029.3 5765414 A A e03-010b 577966.8 5765369 A e03-010c 578017.6 5765394 A e03-010d 577982.4 5765423 NH e03-010e 578037.2 5765410 NH e03-019a 577857.3 5765545 NH A e03-020a 577943.3 5765491 NH A e03-029a 577892.5 5765588 NH NH e03-030a 577978.5 5765642 NH A e03-039a 577869.1 5765696 NH A e03-040a 578029.3 5765708 P P e03-048a 577814.4 5765823 NH NH e03-049a 577900.3 5765854 P P e03-050a 578002 5765800 NH A e03-059a 577876.9 5765890 A A e03-059b 577904.3 5765965 NH e03-059c 577849.5 5765893 NH e03-059d 577861.3 5765904 NH e03-059e 577908.2 5765922 NH

18

Plot name Easting Northing Status E.S.e03-060a 578005.9 5765969 A P e03-060b 578029.3 5765956 NH e03-060c 577947.3 5765954 NH e03-060d 578037.2 5765960 NH e03-060e 577986.3 5765922 A e03-069a 577880.8 5766015 A A e03-069b 577919.9 5766042 NH e03-069c 577837.8 5766021 NH e03-069d 577884.7 5766026 NH e03-069e 577869.1 5765981 NH e03-070a 578009.8 5766035 A A e03-070b 577966.8 5766060 NH e03-070c 577959 5766024 NH e03-070d 578029.3 5766033 A e03-070e 577978.5 5765985 NH e03-079a 577912.1 5766100 NH A e03-080a 578029.3 5766191 A A e03-089a 577888.6 5766267 NH NH e03-090a 577994.2 5766233 NH NH e03-099a 577837.8 5766360 NH NH e03-100a 577986.3 5766378 NH NH e04-001a 578127.1 5765430 P P e04-002a 578174 5765358 A P e04-002b 578150.5 5765380 A e04-002c 578185.7 5765383 P e04-002d 578228.7 5765380 A e04-002e 578162.2 5765396 A e04-003a 578267.8 5765385 NH P e04-004a 578365.5 5765401 A P e04-004b 578349.9 5765410 P e04-004c 578428 5765369 A e04-004d 578369.4 5765340 A e04-004e 578435.9 5765360 A e04-005a 578471 5765380 A A e04-005b 578514 5765401 A e04-005c 578537.5 5765387 A e04-005d 578514 5765342 A e04-005e 578482.8 5765338 NH e04-006a 578615.7 5765347 NH P e04-007a 578744.6 5765394 P P e04-008a 578779.8 5765369 A A e04-008b 578807.2 5765367 A e04-008c 578818.9 5765378 A e04-008d 578764.2 5765423 A e04-008e 578779.8 5765396 P e04-009a 578912.7 5765410 P P e04-010a 578994.8 5765349 P P

Plot name Easting Northing Status E.S.e04-011a 578076.2 5765441 P P e04-011b 578131 5765505 NH e04-011c 578048.9 5765507 NH e04-011d 578064.5 5765489 NH e04-011e 578076.2 5765514 NH e04-012a 578166.1 5765534 P P e04-013a 578263.9 5765455 P P e04-014a 578412.4 5765482 P P e04-014b 578404.6 5765446 P e04-014c 578412.4 5765518 P e04-014d 578424.1 5765543 P e04-014e 578412.4 5765545 A e04-015a 578467.1 5765509 P P e04-016a 578603.9 5765466 A A e04-017a 578686 5765505 A P e04-018a 578803.3 5765482 P P e04-019a 578936.2 5765473 A P e04-019b 578940.1 5765507 P e04-019c 578901 5765516 P e04-019d 578869.7 5765466 P e04-019e 578936.2 5765496 A e04-020a 579010.4 5765511 A P e04-021a 578107.5 5765613 P P e04-022a 578228.7 5765647 P P e04-023a 578244.3 5765556 A P e04-024a 578412.4 5765581 A P e04-025a 578537.5 5765604 P P e04-026a 578572.7 5765615 P P e04-027a 578670.4 5765649 P P e04-027b 578693.8 5765586 P e04-027c 578682.1 5765611 P e04-027d 578697.7 5765574 P e04-027e 578662.6 5765613 P e04-028a 578815 5765620 P P e04-028b 578811.1 5765635 P e04-028c 578822.8 5765552 A e04-028d 578752.5 5765647 A e04-028e 578783.7 5765579 P e04-029a 578885.4 5765633 P P e04-030a 579010.4 5765624 A P e04-031a 578119.2 5765656 A P e04-032a 578197.4 5765755 NH P e04-033a 578338.1 5765746 A P e04-034a 578377.2 5765723 P P e04-035a 578525.8 5765705 P P e04-035b 578478.8 5765656 P e04-035c 578486.7 5765705 P

19

Plot name Easting Northing Status E.S.e04-035d 578486.7 5765669 A e04-035e 578447.6 5765728 P e04-036a 578549.2 5765751 A P e04-037a 578693.8 5765735 P P e04-038a 578752.5 5765710 P P e04-038b 578850.2 5765762 P e04-038c 578815 5765687 P e04-038d 578787.6 5765712 P e04-038e 578838.5 5765760 P e04-039a 578947.9 5765674 P P e04-039b 578908.8 5765665 P e04-039c 578897.1 5765744 P e04-039d 578936.2 5765726 P e04-039e 578951.8 5765762 P e04-040a 578998.7 5765730 P P e04-041a 578099.7 5765866 NH A e04-041b 578123.1 5765843 NH e04-041c 578107.5 5765771 NH e04-041d 578119.2 5765814 NH e04-041e 578091.9 5765870 NH e04-042a 578224.8 5765820 A P e04-043a 578303 5765843 P P e04-044a 578443.7 5765784 P P e04-044b 578357.7 5765830 A e04-044c 578420.2 5765793 A e04-044d 578400.7 5765818 A e04-044e 578443.7 5765825 P e04-045a 578533.6 5765854 P P e04-045b 578482.8 5765870 P e04-045c 578455.4 5765787 P e04-045d 578478.8 5765778 A e04-045e 578521.8 5765789 P e04-046a 578564.8 5765809 P P e04-047a 578713.4 5765832 P P e04-048a 578775.9 5765841 P P e04-048b 578779.8 5765802 P e04-048c 578807.2 5765787 P e04-048d 578756.4 5765793 P e04-048e 578787.6 5765789 P e04-049a 578873.6 5765830 A P e04-050a 578990.9 5765784 P P e04-050b 579018.3 5765827 P e04-050c 578994.8 5765841 P e04-050d 579022.2 5765802 P e04-050e 579022.2 5765798 A e04-051a 578127.1 5765949 A A e04-052a 578224.8 5765875 P P

Plot name Easting Northing Status E.S.e04-053a 578338.1 5765967 NH P e04-054a 578353.8 5765945 A P e04-055a 578510.1 5765949 P P e04-056a 578643 5765913 P P e04-057a 578744.6 5765918 P P e04-057b 578678.2 5765924 NH e04-057c 578709.5 5765875 NH e04-057d 578717.3 5765947 A e04-057e 578732.9 5765969 A e04-058a 578834.6 5765947 P P e04-059a 578904.9 5765920 P P e04-060a 579018.3 5765918 P P e04-061a 578119.2 5766030 NH A e04-061b 578138.8 5765992 NH e04-061c 578127.1 5766012 NH e04-061d 578119.2 5766008 A e04-061e 578134.9 5766035 NH e04-062a 578224.8 5766019 A P e04-062b 578142.7 5766030 NH e04-062c 578205.2 5765994 NH e04-062d 578146.6 5766046 NH e04-062e 578213.1 5766062 NH e04-063a 578314.7 5765990 NH A e04-063b 578260 5765990 NH e04-063c 578303 5766055 A e04-063d 578256 5766010 NH e04-063e 578318.6 5766015 NH e04-064a 578357.7 5766087 P P e04-064b 578365.5 5766060 A e04-064c 578431.9 5766053 NH e04-064d 578431.9 5766080 NH e04-064e 578357.7 5765983 A e04-065a 578447.6 5766080 A P e04-065b 578541.4 5766003 A e04-065c 578514 5765997 NH e04-065d 578478.8 5766075 A e04-065e 578498.4 5766073 P e04-066a 578603.9 5766012 A P e04-066b 578619.6 5766003 A e04-066c 578560.9 5766037 NH e04-066d 578627.4 5766066 P e04-066e 578557 5765999 NH e04-067a 578697.7 5766053 P P e04-068a 578815 5765981 P P e04-068b 578811.1 5765987 P e04-068c 578799.4 5766035 P e04-068d 578772 5766037 A

20

Plot name Easting Northing Status E.S.e04-068e 578775.9 5765999 NH e04-069a 578897.1 5766078 A P e04-069b 578908.8 5766035 P e04-069c 578893.2 5765994 A e04-069d 578893.2 5766026 NH e04-069e 578865.8 5765987 A e04-070a 578975.3 5765987 P P e04-070b 578967.5 5766033 A e04-070c 578979.2 5766021 A e04-070d 579033.9 5766057 P e04-070e 578990.9 5766042 A e04-071a 578068.4 5766150 NH A e04-072a 578174 5766107 NH A e04-072b 578166.1 5766148 NH e04-072c 578177.9 5766096 NH e04-072d 578166.1 5766188 A e04-072e 578205.2 5766143 NH e04-073a 578318.6 5766182 A A e04-074a 578369.4 5766166 NH A e04-074b 578388.9 5766150 NH e04-074c 578392.9 5766139 NH e04-074d 578408.5 5766116 NH e04-074e 578353.8 5766103 A e05-075a 579479.5 5766112 A A e04-075b 578447.6 5766148 A e04-075c 578529.7 5766168 NH e04-075d 578517.9 5766112 NH e04-075e 578533.6 5766134 A e04-076a 578635.2 5766125 A P e04-076b 578564.8 5766157 NH e04-076c 578627.4 5766125 P e04-076d 578623.5 5766114 A e04-076e 578643 5766179 NH e04-077a 578662.6 5766195 A P e04-077b 578693.8 5766114 A e04-077c 578713.4 5766134 NH e04-077d 578701.7 5766127 P e04-077e 578732.9 5766191 A e04-078a 578807.2 5766139 A P e04-078b 578815 5766103 NH e04-078c 578830.6 5766103 A e04-078d 578779.8 5766191 A e04-078e 578850.2 5766100 P e04-079a 578944 5766123 P P e04-079b 578932.3 5766098 NH e04-079c 578858 5766168 NH e04-079d 578936.2 5766173 NH

Plot name Easting Northing Status E.S.e04-079e 578885.4 5766143 A e04-080a 578975.3 5766182 P P e04-080b 579018.3 5766112 P e04-080c 578963.5 5766116 NH e04-080d 579022.2 5766118 A e04-080e 579018.3 5766166 P e04-081a 578056.7 5766202 NH NH e04-082a 578224.8 5766263 NH NH e04-083a 578256 5766236 NH A e04-083b 578271.7 5766299 NH e04-083c 578306.9 5766279 NH e04-083d 578342 5766222 A e04-083e 578299 5766261 A e04-084a 578428 5766272 A A e04-085a 578455.4 5766242 A A e04-085b 578447.6 5766238 NH e04-085c 578459.3 5766258 NH e04-085d 578498.4 5766294 NH e04-085e 578486.7 5766265 A e04-086a 578611.7 5766288 A A e04-086b 578596.1 5766261 NH e04-086c 578588.3 5766256 A e04-086d 578592.2 5766231 NH e04-086e 578553.1 5766236 NH e04-087a 578729 5766270 A A e04-088a 578846.3 5766238 A A e04-089a 578893.2 5766229 A A e04-089b 578889.3 5766240 NH e04-089c 578885.4 5766288 A e04-089d 578924.5 5766215 A e04-089e 578893.2 5766202 NH e04-090a 579006.5 5766240 A P e04-090b 579002.6 5766202 A e04-090c 578975.3 5766218 NH e04-090d 579030 5766236 NH e04-090e 578990.9 5766267 P e04-091a 578131 5766308 NH NH e04-092a 578166.1 5766310 NH NH e04-093a 578275.6 5766369 NH A e04-093b 578252.1 5766400 A e04-093c 578263.9 5766321 NH e04-093d 578318.6 5766335 NH e04-093e 578295.1 5766371 NH e04-094a 578381.1 5766308 A A e04-095a 578451.5 5766367 NH A e04-095b 578459.3 5766317 NH e04-095c 578447.6 5766319 NH

21

Plot name Easting Northing Status E.S.e04-095d 578521.8 5766376 NH e04-095e 578482.8 5766376 A e04-096a 578564.8 5766369 A A e04-096b 578639.1 5766340 A e04-096c 578607.8 5766380 A e04-096d 578557 5766378 NH e04-096e 578611.7 5766409 A e04-097a 578697.7 5766378 A A e04-097b 578654.7 5766407 A e04-097c 578689.9 5766396 A e04-097d 578709.5 5766367 NH e04-097e 578686 5766344 NH e04-098a 578822.8 5766382 A A e04-099a 578897.1 5766312 A A e04-099b 578877.5 5766364 NH e04-099c 578897.1 5766407 A e04-099d 578924.5 5766328 A e04-099e 578904.9 5766358 NH e04-100a 578994.8 5766400 A X e05-001a 579088.6 5765362 P P e05-002a 579248.9 5765419 A P e05-002b 579170.7 5765374 A e05-002c 579233.3 5765405 A e05-002d 579166.8 5765349 A e05-002e 579198.1 5765376 A e05-003a 579264.5 5765347 P P e05-003b 579323.2 5765426 A e05-003c 579334.9 5765432 A e05-003d 579264.5 5765414 A e05-003e 579264.5 5765432 P e05-004a 579444.3 5765392 P P e05-005a 579471.7 5765344 P P e05-006a 579604.6 5765426 A P e05-006b 579643.7 5765376 A e05-006c 579565.5 5765412 A e05-006d 579628 5765389 A e05-006e 579577.2 5765342 A e05-007a 579694.5 5765369 NH NH e05-008a 579854.8 5765389 NH NH e05-009a 579870.4 5765371 NH NH e05-010a 580026.7 5765353 NH NH e05-011a 579076.9 5765527 P P e05-011b 579080.8 5765484 A e05-011c 579088.6 5765480 P e05-011d 579119.9 5765529 P e05-011e 579147.3 5765523 P e05-012a 579233.3 5765450 A P

Plot name Easting Northing Status E.S.e05-013a 579291.9 5765520 A P e05-014a 579385.7 5765448 A P e05-014b 579409.1 5765484 A e05-014c 579440.4 5765480 A e05-014d 579362.2 5765511 A e05-014e 579381.8 5765459 A e05-015a 579491.2 5765464 A P e05-016a 579655.4 5765491 NH NH e05-017a 579671 5765527 NH NH e05-018a 579819.6 5765450 NH NH e05-019a 579925.1 5765498 NH NH e05-090a 579991.6 5766236 NH NH e05-021a 579100.4 5765595 P P e05-022a 579186.3 5765563 P P e05-023a 579280.2 5765554 A P e05-023b 579331 5765588 A e05-023c 579350.5 5765631 A e05-023d 579346.6 5765574 A e05-023e 579334.9 5765613 P e05-024a 579362.2 5765556 P P e05-024b 579397.4 5765563 A e05-024c 579397.4 5765568 A e05-024d 579389.6 5765617 A e05-024e 579424.8 5765633 P e05-025a 579534.2 5765624 A P e05-026a 579620.2 5765570 A P e05-027a 579702.3 5765613 NH NH e05-028a 579846.9 5765584 NH NH e05-029a 579952.5 5765599 NH NH e05-030a 580026.7 5765579 NH NH e05-031a 579057.4 5765741 P P e05-031b 579057.4 5765678 P e05-031c 579147.3 5765753 P e05-031d 579123.8 5765730 P e05-031e 579135.5 5765719 P e05-032a 579217.6 5765703 A P e05-032b 579256.7 5765712 P e05-032c 579159 5765687 A e05-032d 579170.7 5765730 P e05-032e 579186.3 5765753 P e05-033a 579338.8 5765746 A P e05-034a 579377.9 5765678 A P e05-035a 579526.4 5765751 A P e05-036a 579604.6 5765692 P P e05-036b 579573.3 5765723 A e05-036c 579600.7 5765667 A e05-036d 579569.4 5765717 A

22

Plot name Easting Northing Status E.S.e05-036e 579651.5 5765751 P e05-037a 579686.7 5765735 NH NH e05-038a 579800 5765728 NH NH e05-039a 579874.3 5765676 NH NH e05-040a 580061.9 5765703 NH X e05-041a 579116 5765830 P P e05-042a 579186.3 5765780 P P e05-042b 579209.8 5765834 A e05-042c 579237.2 5765841 P e05-042d 579252.8 5765782 P e05-042e 579209.8 5765852 P e05-043a 579358.3 5765866 A P e05-044a 579405.2 5765793 A P e05-045a 579479.5 5765850 P P e05-045b 579463.9 5765818 A e05-045c 579546 5765793 A e05-045d 579526.4 5765814 A e05-045e 579499.1 5765852 A e05-046a 579589 5765787 P P e05-046b 579573.3 5765787 A e05-046c 579592.9 5765830 A e05-046d 579643.7 5765791 P e05-046e 579612.4 5765800 A e05-047a 579733.6 5765830 NH NH e05-048a 579803.9 5765802 NH NH e05-049a 579940.7 5765805 NH P e05-050a 580046.3 5765771 NH NH e05-051a 579131.6 5765893 P P e05-051b 579155.1 5765951 P e05-051c 579123.8 5765911 P e05-051d 579088.6 5765890 P e05-051e 579127.7 5765940 P e05-052a 579217.6 5765956 P P e05-052b 579209.8 5765969 P e05-052c 579162.9 5765929 P e05-052d 579186.3 5765902 P e05-052e 579162.9 5765974 P e05-053a 579276.2 5765976 P P e05-053b 579319.2 5765897 A e05-053c 579288 5765933 P e05-053d 579264.5 5765879 P e05-053e 579303.6 5765902 A e05-054a 579377.9 5765954 P P e05-055a 579514.7 5765965 A P e05-056a 579569.4 5765956 A P e05-057a 579741.4 5765915 A P e05-057b 579760.9 5765899 A

Plot name Easting Northing Status E.S.e05-057c 579757 5765978 A e05-057d 579686.7 5765911 P e05-057e 579710.1 5765951 P e05-058a 579768.8 5765949 A P e05-059a 579948.6 5765945 NH NH e05-060a 580065.8 5765958 NH NH e05-061a 579116 5766087 P P e05-062a 579202 5766033 A P e05-062b 579202 5766006 P e05-062c 579194.2 5766028 P e05-062d 579194.2 5766033 P e05-062e 579225.4 5766051 A e05-063a 579260.6 5766066 P P e05-063b 579323.2 5765981 P e05-063c 579338.8 5766044 A e05-063d 579311.4 5766055 P e05-063e 579338.8 5766071 P e05-064a 579436.5 5766060 A P e05-065a 579463.9 5766085 A P e05-066a 579573.3 5766066 A P e05-067a 579702.3 5766069 X X e05-068a 579792.2 5766062 NH X e05-069a 579909.5 5765985 NH NH e05-070a 580034.6 5766075 NH NH e05-071a 579147.3 5766127 P P e05-071b 579065.2 5766116 P e05-071c 579065.2 5766188 A e05-071d 579073 5766170 P e05-071e 579069.1 5766164 P e05-072a 579190.3 5766094 P P e05-072b 579233.3 5766173 A e05-072c 579248.9 5766173 P e05-072d 579245 5766139 A e05-072e 579170.7 5766109 A e05-073a 579303.6 5766154 A P e05-074a 579417 5766103 A P e05-075a 579479.5 5766112 P P e05-075b 579475.6 5766123 A e05-075c 579487.3 5766139 P e05-075d 579499.1 5766177 A e05-075e 579483.4 5766164 A e05-076a 579573.3 5766161 NH NH e05-077a 579757 5766105 P P e05-078a 579811.8 5766091 NH NH e05-079a 579905.6 5766114 NH NH e05-080a 580069.7 5766173 NH NH e05-081a 579100.4 5766263 A P

23

Plot name Easting Northing Status E.S.e05-081b 579143.3 5766292 X e05-081c 579065.2 5766283 P e05-081d 579065.2 5766211 P e05-081e 579104.3 5766238 P e05-082a 579159 5766202 P P e05-083a 579338.8 5766233 X P e05-084a 579377.9 5766265 NH NH e05-085a 579463.9 5766274 P P e05-086a 579573.3 5766265 NH NH e05-087a 579674.9 5766256 NH X e05-088a 579866.5 5766218 NH NH e05-089a 579901.7 5766270 NH NH e05-090a 579991.6 5766236 NH NH e05-091a 579131.6 5766403 NH X e05-092a 579198.1 5766346 NH NH e05-093a 579311.4 5766367 NH NH e05-094a 579370.1 5766337 X X e05-095a 579553.8 5766412 NH NH e05-096a 579592.9 5766380 NH NH e05-097a 579725.8 5766331 NH NH e05-098a 579827.4 5766398 NH NH e05-099a 579889.9 5766394 NH NH e05-100a 579975.9 5766398 NH NH f04-001a 578041.1 5766495 NH NH f04-002a 578217 5766484 NH NH f04-003a 578299 5766441 NH NH f04-004a 578388.9 5766439 NH NH f04-005a 578498.4 5766520 A A f04-006a 578607.8 5766425 A A f04-007a 578682.1 5766509 A A f04-008a 578834.6 5766421 A A f04-008b 578807.2 5766423 NH f04-008c 578815 5766486 NH f04-008d 578768.1 5766509 A f04-008e 578791.6 5766509 NH f04-009a 578885.4 5766495 NH NH f04-010a 579022.2 5766475 NH NH f04-011a 578099.7 5766615 NH NH f04-012a 578209.1 5766606 NH NH f04-013a 578326.4 5766570 A A f04-015a 578510.1 5766567 A A f04-016a 578631.3 5766615 NH P f04-017a 578717.3 5766615 A P f04-017b 578654.7 5766552 NH f04-017c 578705.6 5766545 P f04-017d 578732.9 5766547 X f04-017e 578713.4 5766576 X

Plot name Easting Northing Status E.S.f04-018a 578764.2 5766538 A P f04-019a 578881.5 5766583 NH NH f04-020a 578994.8 5766567 NH NH f04-021a 578107.5 5766692 NH NH f04-024a 578377.2 5766725 A A f04-025a 578455.4 5766694 A A f04-026a 578557 5766662 A A f04-027a 578709.5 5766646 NH X f04-028a 578830.6 5766712 NH NH f04-029a 578928.4 5766669 NH NH f04-030a 579022.2 5766710 NH NH f04-031a 578080.2 5766753 NH NH f04-032a 578224.8 5766791 NH NH f04-033a 578267.8 5766802 NH NH f04-034a 578435.9 5766809 NH NH f04-035a 578514 5766746 NH NH f04-036a 578643 5766784 NH NH f04-037a 578686 5766791 NH NH f04-038a 578850.2 5766768 NH NH f04-039a 578901 5766838 NH NH f04-040a 579002.6 5766798 NH NH f04-041a 578045 5766949 NH NH f04-042a 578224.8 5766917 NH NH f04-043a 578271.7 5766872 NH NH f04-044a 578373.3 5766944 A A f04-045a 578521.8 5766953 NH P f04-046a 578631.3 5766931 NH X f04-047a 578740.7 5766935 NH NH f04-056a 578615.7 5766976 X X f04-057a 578650.8 5766960 X X f04-066a 578603.9 5767105 P P f04-067a 578670.4 5767161 X P f04-077a 578662.6 5767256 A X

24

Appendix C. Simple Random Sampling Patch Occupancy Models

The standard patch occupancy model assumes that an identical number of sample units are randomly selected from each cell and visited (MacKenzie et al. 2002, 2003). That model is presented below, then extended to account for cells only visited once, then further extended to allow for occupied cells to be partitioned into a low detection and a high detection strata.

The presentation below is cast in terms of multiple independently selected sample units within each cell. Each visited sample unit results in a response of either 0 (absent), 1 (present), or x (missing data) response. In this application, the order of the sample units doesn’t matter, thus simplifying the standard model. Basic model Assumptions: 1. A cell’s occupancy status doesn’t change during the period of sampling. 2. Detection probability is constant across all cells (though see below for extensions). 3. Detection of evidence at a cell is independent of detection at any other cell. 4. Detection of evidence in a sample unit is independent of detection at any other sample unit within the cell. 5. Probability of occupancy is constant across the cells. 6. T randomly selected sample units are observed in each cell. Notation ψ Prob(species present at a randomly chosen cell); p. Prob(detect evidence in a sample unit | given the cell is occupied); N total number of surveyed cells; T maximum number of independently randomly selected sample units observed from each

cell; ni total number of sample units in cell i in which evidence was detected (ni = 0, 1, …, T) n. total number of cells at which evidence was detected at least once.

For sites where the species was never detected, the probability of detection is a mixture of the probability the species is present but never detected and the probability the species is absent (site unoccupied). For example, if T = 5 and no evidence was detected in five sample units: (1 )(1 )(1 )(1 )(1 ) (1 ) (1 ) (1 )Tp p p p p pψ ψ ψ ψ− − − − − + − = − + − . Since sites are sampled independently, the full model likelihood is:

( ).sites i with detects sites with nodetects

( , |{ }) (1 ) (1 ) (1 )i in T n Ti

ji

TL p n p p p

nψ ψ ψ−⎡ ⎤ ⎡⎛ ⎞

= − × −⎢ ⎥ ⎢⎜ ⎟⎝ ⎠⎣ ⎦ ⎣

∏ ∏ ψ⎤

+ − ⎥⎦

)ψ

( .. .

sites i with detects

(1 ) (1 ) (1 )i iN nn n T n T

i

Tp p p

nψ ψ

−−⎡ ⎤⎛ ⎞= − × − +⎢ ⎥⎜ ⎟

⎝ ⎠⎣ ⎦∏ − . (Eq A1-1)

Note that this model requires T sample units per site.

25

Basic model + cells with only one sample unit observed

Since multiple sample units were observed on only a subset of cells, the basic model likelihood has to be extended to account for the cells in which only a single sample unit was observed.

( ) .. . .

.

sitesi with detects .

( , |{ }) ( (1 ) 1 )

(1 ) (1 ) (1 )

p p A

i i

S S Si

N nn n T n T

i

L p n p p

Tp p p

n

ψ ψ ψ ψ

ψ ψ ψ−−

= − + − ×

⎡ ⎤⎛ ⎞− × − + −⎢ ⎥⎜ ⎟

⎝ ⎠⎣ ⎦∏

( ) .. . .

sitesi with detects .

(1 ) (1 ) (1 ) (1 )p p i iAN nS S n n T nS T

i

Tp p p p p

nψ ψ ψ ψ ψ

−−⎡ ⎤⎛ ⎞= − × − × − + −⎢ ⎥⎜ ⎟

⎝ ⎠⎣ ⎦∏

(Eq A1-2) where, of the Sp + SA cells from which only a single sample unit was visited, evidence was observed in Sp and no evidence was detected in the other SA cells. Basic model extended to two detection strata (unknown cell status) The basic model can be extended to allow for two (or more) strata in detection rates following Pledger (2000):

( )

( )( )

. . . .

.

1 2 1 . 1 1 1 1 2 2sites i with detects .

1 1 1 2

( , , , |{ }) (1 ) (1 ) (1 )

(1 ) (1 )(1 ) (1 )

i i in n T n ni

i

N nT T

TL p p n p p p p

n

p p

ψ π ψ π π

ψ π π ψ

− −

−

.iT n⎡ ⎤⎛ ⎞= − + − −⎢ ⎥⎜ ⎟

⎝ ⎠⎣ ⎦

× − + − − + −

∏

(Eq A1-3) While Equation 3 assumes T sample units were visited in each cell, it is easily extended to allow for multiple sample units in only a subset of cells (not shown).

26