Factors Associated with Dusky Canada Goose ( Branta Canadensis Occidentalis

Transcription

1 Factors Associated with Dusky Canada Goose (Branta Canadensis Occidentalis) Nesting and Nest Success on Artificial Nest Islands of the Western Copper River Delta Author(s) :Nicole M. Maggiulli and Bruce D. Dugger Source: Waterbirds, 34(3): Published By: The Waterbird Society DOI: / URL: BioOne ( is a a nonprofit, online aggregation of core research in the biological, ecological, and environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne s Terms of Use, available at page/terms_of_use. Usage of BioOne content is strictly limited to personal, educational, and noncommercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder. BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research.

2 Factors Associated with Dusky Canada Goose (Branta canadensis occidentalis ) Nesting and Nest Success on Artificial Nest Islands of the Western Copper River Delta NICOLE M. MAGGIULLI * AND BRUCE D. DUGGER Department of Fisheries and Wildlife, Oregon State University, Corvallis, OR, 97331, USA * Corresponding author; nicole.maggiulli@gmail.com Abstract. Decline of the Dusky Canada Goose (Branta canadensis occidentalis; hereafter, Dusky Goose) population on the western Copper River Delta (CRD) prompted the establishment of an artificial nest island (island) program in A retrospective analysis of the program was conducted to examine general trends in island use and nest success from A series of candidate models was generated to determine how habitat, island and biological variables were associated with island use and nest success from Use of islands by Dusky Geese increased between 1987 and 2005 from 10% to 44%; apparent nest success averaged 64 ± 4% and showed no trend with year. Island use was consistently and strongly associated with the previous year s island status. The odds of nesting on an island that contained a successful nest the previous year were four times greater than for islands not used the previous year. Likelihood of island use was highest at moderate shrub cover and increased with shrub height. Likelihood of nest success increased on islands further from shore. The influence of year suggests the presence of alternate prey and predator abundance is more important to nest success than island features. The increasing use of islands while the CRD Dusky Goose population has been declining indicates that islands may be increasingly important to population productivity. However, quantifying the contribution of the island program requires a better understanding of other population metrics, such as gosling mortality. Received 13 July 2010, accepted 10 June Key words. Alaska, artificial island, Branta canadensis occidentalis, Copper River, wildlife management. Waterbirds 34(3): , 2011 The Dusky Canada Goose (Branta canadensis occidentalis, hereafter, Dusky Goose) breeds on the Copper River Delta (CRD) in south-central Alaska (Pacific Flyway Council 2008 unpublished report). Prior to 1964, Dusky Geese primarily nested on the outer portion of the western CRD, above the mean high tide level in a vegetation community dominated by sedges, grasses, forbs and low shrubs (Trainer 1959). Dusky Goose nest success was high (~89%) and nest losses were mainly attributed to flooding and avian predators, primarily gulls (Trainer 1959). On 27 March 1964, an earthquake measuring 8.5 on the Richter scale elevated the CRD by as much as two meters (Thilenius 1990). The change in elevation dried the CRD, which initiated succession of the plant community (Thilenius 1990; Bromley and Rothe 2003). Intertidal areas dominated by grasses and sedges and used by nesting Dusky Geese began a transition to habitat dominated by shrubs and trees including willow (Salix spp.), sweetgale (Myrica gale), alder (Alnus spp.) and Sitka Spruce (Picea sitchensis) (Crow 1971; Thilenius 1990; Boggs and Shephard 1999). Despite these habitat changes, Dusky Geese continued to use the outer CRD as their core nesting area (Bromley 1976; Campbell 1990). Changes in vegetation on the CRD following the earthquake impacted Dusky Goose reproductive success and subsequent population size. After an initial increase in Dusky Goose abundance following the earthquake (high of approximately 25,000 birds in 1969; Pacific Flyway Council 1973 unpublished report), the Dusky Goose population declined to fewer than 9,000 birds in 1987 (Fig. 2 in Pacific Flyway Council 2008 unpublished report), the lowest population estimate since the 1950s, with the decline primarily attributed to low recruitment caused by an increase in predator abundance that was related to post-earthquake habitat changes. In response to the declining Dusky Goose population, an artificial nest island program was initiated in 1983 to improve recruitment on the western CRD (Babler et al. 1998, unpublished report). Canada Geese in other regions of North America use artificial nest islands (hereafter called islands; Craighead 269

3 270 WATERBIRDS and Stockstad 1961; Brakhage 1965; Will and Crawford 1970; Giroux et al. 1983; Rienecker 1971) and nest success is generally higher than mainland nesting geese (Craighead and Stockstad 1961; Brakhage 1965; Will and Crawford 1970). Artificial nest islands have also been used for other waterbird species, including Common Loons (Gavia immer) and Black Terns (Clidonias niger), with similarly positive results (Piper et al. 2002; Shealer et al. 2006). Monitoring of islands on the CRD indicates that nest success is higher than the surrounding landscape (Miller et al. 2007), but there is considerable variation, and use of islands by Dusky Geese is relatively low compared to artificial nest island programs in other portions of North America (Brakhage 1965; Will and Crawford 1970; Rienecker 1971; Giroux et al. 1983). Identifying the factors that influence use of islands and nest success on islands may improve the management of this program and help improve goose recruitment. Several studies have investigated how habitat features can influence use and nest success on artificial nest islands (Vermeer 1970; Kaminski and Prince 1977; Giroux 1981; Reese et al. 1987; Cline 2004). Most of those studies were conducted at temperate latitudes in man made lakes. The extent to which these factors influence island use and nest success in a largely intact ecosystem is unclear. More recently, research on Dusky Geese nesting on the mainland of the western CRD found a relationship between vegetation characteristics around the nest and nest success (Miller et al. 2007). However, it is not clear those relationships apply to Dusky Geese nesting on islands. Here, we conduct a retrospective analysis of nest island data for Dusky Geese on the CRD to test predictions deduced from hypotheses about factors that influence use and success on islands. Study Area Our study was conducted on the Copper River Delta, the largest Pacific coastal wetland in North America, encompassing 283,380 ha, including 150,000 ha of wetlands with extensive freshwater influences from glacial runoff and precipitation and a bar-built estuarine structure (Thilenius 1990). The nest island project area encompassed 10,800 ha of uplifted marsh near the outer coastal region of the western CRD (Fig. 1). This region is characterized by small freshwater ponds, many engineered by beavers (Castor canadensis), and dissected by a network of small and large sloughs (Babler et al. 1998, unpublished report). During the years of our analysis there were five types of islands present (donut islands, fiberglass floating islands, sandbag islands, barrel-platform islands and rebar-platform islands). We chose to include only the three most abundant island types (donut islands, fiberglass floating islands and sandbag islands) for our analysis, as the presence of the additional island types was too minimal for comparisons (0.4% barrel-platform and 0.2% rebar-platform). Fiberglass floating islands, which were the most abundant island type (58.9% fiberglass floating islands, 33.6% sandbag islands, and 6.9% donut islands), were 2.3 m 2 and created by filling wooden frames with polyurethane foam and covering that with fiberglass. Sandbag islands were created with burlap bags (30 61 cm bags) filled with soil from collapsing slough banks. Sandbags were layered to create a dry surface about 8-20 cm above water level. Sandbag islands ranged in size from 1.4 to 2.7 m 2 and were usually built on high spots in the pond or natural islands that were too wet for nesting. Donut islands ranged in size from 1.7 to 2.1 m 2 and were created with truck tire inner tubes (35 97 cm) filled with 0.28 m 3 of closed-cell polyurethane foam. Fiberglass floating islands and donut islands were both floating islands secured to the pond bottom using a variety of cables and anchors. All islands were covered with erosion mats and topped with sod mats. Islands were constructed with a cm-thick sod mat of natural vegetation cut by hand with shovels. Sod mats were collected in pieces from a marsh near the point of the island installation, and were installed with 15-50% aerial shrub cover for the nesting Dusky Goose (Babler et al unpublished report). METHODS Figure 1. Artificial nesting islands for Dusky Canada Geese were placed in wetlands on the western Copper River Delta (area designated by square), Alaska, USA (adapted from Pacific Flyway Council 2008 unpublished report).

4 ISLAND USE BY DUSKY CANADA GEESE 271 Data Collection United States Forest Service (USFS) personnel have managed the nest islands annually since program inception in Monitoring intensity varied across years; all islands were visited each year from 1984 to 1988 and from 1996 to 2005, and once every two years from 1989 to During years when monitored, each island was visited between 7 May and 29 June to determine use status and nest fate. During the first visit the island was categorized as available or unavailable. Unavailable islands were defined as those that had sunk, been blown to shore, were missing a sod mat, were flooded or were located in a dry wetland. Islands were classified as used or not used based on signs at the nest that included the presence of the incubating adult, young in the nest, egg shell remains and nest characteristics (presence of nest bowl, down and eggs). Dusky Goose nest fate (successful or failed) was determined by signs at the nest. A successful nest contained at least one egg shell membrane in the bottom of the nest. Failed nests contained no membranes and usually some number of wholly or partially destroyed and scattered eggs. Nests were also classified as failed if there were no eggs in an otherwise intact nest bowl or if nests were abandoned. If a nest was active during the first visit, eggs were floated to determine incubation stage (Westerskov 1950) and the nest was visited again after its scheduled hatch date to determine fate. During the first nest visit, USFS staff also recorded information on habitat and island characteristics including: 1) percent aerial shrub cover on the island (AC) comprised of willow (Salix spp.), alder (Alnus spp.) and sweetgale (Myrica gale) and 2) average height of shrubs on the island (SH, m, collected from present). All variables were estimated by USFS biologists and field technicians. New technicians were trained with experienced technicians during the first day of monitoring (Babler et al. 1998, unpublished report). Data Analysis We used results of previous studies, discussions with other goose researchers, our understanding of goose biology and our own observations on the CRD to form hypotheses and develop testable predictions. We hypothesized that island characteristics which would influence detection of and access to islands by predators would have an influence on use and nest success. Consequently, we predicted use of islands and nest success on islands would be positively correlated with the distance between the island and the nearest shore (Giroux 1981; Cline 2004) and the distance between the island and the nearest large slough (Trainer 1959; Bromley 1976; Moffitt et al. 2002; Miller et al. 2006). We predicted use of islands would be highest on islands with moderate shrub cover and height (35-65% aerial cover of shrubs m tall; Bromley 1976). We predicted nest success on islands would be highest on islands with greater cover of tall shrubs (35-65% aerial cover of shrubs m tall; Bromley 1976; Miller et al. 2007). We predicted Dusky Goose nest success on islands would be negatively associated with the distance to the nearest colony or nest of gulls or terns (hereafter larid) as these species actively repel predators (Larson 1960; MacInnes 1962; Vermeer 1968; Vermeer et al. 1992; Nguyen et al. 2006). We also hypothesized that islands more exposed to weather might be used less than those more sheltered and thus predicted use of islands and nest success on islands would be negatively associated with the distance from the island to the nearest shore in the direction of the prevailing winds. Finally, because geese show high fidelity to nest sites, we predicted use of islands would be higher for islands that supported successful nests the previous year (Giroux et al. 1983; Lindberg and Sedinger 1997). Our analysis only included islands defined as available by USFS staff during field visits; island availability ranged from % during any single year. Prior to analysis, we compared field data cards against the digital dataset to estimate data entry quality. We reviewed data for donut and sandbag islands that were available for six years or greater and data for all available fiberglass floating islands. We proofread five variables (AC, SH, species use, Dusky Goose nest presence, and nest fate) for a random sample of 10% of all available island records for each year from The average rate of incorrect entries (±SE) for species use, Dusky Goose nest presence and nest fate was 1.9 ± 0.7%. We proofread all the data for any variable in a year when we found incorrect entries exceeded 5%. We also computed summary statistics for explanatory variables each year to look for outliers. We calculated percent use (defined as the percent of islands used by Dusky Geese) and apparent nest success (the proportion of Dusky Goose nests with 1 egg hatched) for each year and graphed both variables with time ( ) to generally characterize the temporal trend. Estimates of apparent nest success are usually biased high because successful nests are more likely to be detected than failed nests (Mayfield 1961); however, our apparent nest success estimates should be relatively unbiased because all nest locations were known a priori (Mayfield 1961; Jehle et al. 2004). It is possible we failed to detect some nests destroyed very shortly after being initiated (e.g. one egg laid), but this bias should be consistent across all islands so it did not influence our analysis of habitat variables influencing nest fate. We calculated the percent use for each island and the percent of years an island was used when available ( ), for all islands available for at least five years. We compared trends in use and nest success over time using a general linear model (PROC GLM; SAS Institute 2003) and by island type using analysis of covariance (PROC MIXED; SAS Institute 2003) after meeting assumptions of normality and equal variances. We compared use and nest success between islands located in the two major categories of wetlands (sloughs vs. ponds) using logistic regression (PROC GENMOD; SAS Institute 2003). We used odds ratios and 95% confidence intervals to evaluate the strength of evidence for associations among wetland type and island use and nest success. We used ten years of data ( ) to investigate habitat and island features correlated with island use and nine years of data ( ) to investigate habitat and island features correlated with island nest success. We restricted our analysis to a recent block of years in an attempt to control for changes in plant community composition on the CRD following the 1964 earthquake (Thilenius 1990), and to ensure that all islands were monitored yearly and data was gathered for all explanatory variables. In addition to variables collected during field checks of nests, we used a geographical information system to measure Euclidean distance of the nest island to shore (DTS; m); the distance from the island to the nearest large slough including: Tiedeman, Alaganik,

5 272 WATERBIRDS Pete Dahl, Pete Dahl Cutoff, Wahalla and Gus Stevens Sloughs (DTLS; m); pond size (PSIZE, ha); the distance from the island to the nearest mainland shore in the east-southeast (112.5 ; Bishop et al. 2000) direction (FETCH, m); and the distance from the island to the nearest larid colony or nest (DTLAR, m; ESRI 2005). Colonies included Mew Gulls (Larus canus) and Arctic Terns (Sterna paradisaea), and nests included primarily Mew Gulls and Arctic Terns, with some Parasitic Jaegers (Stercorarius parasiticus) and Aleutian Terns (Sterna aleutica). The locations of gull and tern colonies came from surveys conducted from (Miller 2004). We assumed larid colony locations were relatively stable during our study period ( ) based on observations by USGS field technicians and biologists camped near the colonies and information on the nest site fidelity of Mew Gulls and Arctic Terns from the literature (Hatch 2002; Moskoff and Bevier 2002). We categorized the previous year s island status (STATUS) as Not Used, Used and Successful, Used but Failed, or Not Available. We generated univariate statistical hypotheses for associations between island use and nest success with explanatory variables based on our predictions (Table 1). We tested explanatory variables for multicollinearity (Allison 1999) using Pearson s correlation coefficients (PROC CORR; SAS Institute 2003). We tested for relationships between dependent and explanatory variables using logistic regression (PROC GENMOD; SAS Institute 2003). In the analysis of island use, we tested for a linear relationship with five explanatory variables and quadratic relationships for aerial shrub cover (AC 2 ) and average shrub height (SH 2 ). In the analysis of island nest success, we tested for a linear relationship with all seven explanatory variables. For the analysis of nest success we also included Year as a fixed effect because we knew features of the Copper River Delta ecosystem varied among years (e.g. abundance of alternate prey) and that such variability can influence nest success (Miller et al. 2006). We predicted that nest success would vary among years, and that such variability would reflect changes in predator and alternate prey populations, but we lacked the specific data on predator and alternate prey abundance to make a more specific prediction. Canada Geese can exhibit high nest site fidelity (e.g. Craighead and Stockstad 1961; Brakhage 1965; Bromley 1976; Giroux et al. 1983); thus, Dusky Geese that were successful on an island in one year likely returned to that island the next year. Ignoring this fact and treating each nest attempt as independent would have biased test statistics and standard errors (Allison 1999). Consequently, we treated multiple observations per nest island as repeated measures and used the generalized estimating equations (GEE) approach to address the fact that observations on each island were likely correlated among years (Ainley et al. 2009). Because GEE is not likelihood-based, we use quasi-likelihood criteria (QIC) to evaluate competing models (Pan 2001). We calculated QIC (ΔQIC) and QIC weights following equations outlined by Burnham and Anderson (2002). We conducted our modeling as a two-stage process to distinguish the influence of the explanatory variable tied to goose biology (STATUS) from the environmental variables on island use. First, we modeled STATUS vs the null model (intercept-only model) for island use (n = 2 models). We then used the best model from that exercise as the null and base model for all environmental variables in two variable additive models. Models with ΔQIC values less than two were considered competitive; and QIC model weights and variable importance weights (w) for competitive models were used to evaluate and rank models (Burnham and Anderson 2002). We used variable coefficients (β) and likelihood ratio 95% confidence intervals to evaluate the direction and strength of associations between the response variable and the continuous explanatory variables (Burnham and Anderson 2002), and odds ratios to evaluate the direction and strength of evidence for categories of STA- TUS in relation to islands not used the previous year (STATUS, NOT USED). Strength of evidence was considered weak if 95% confidence intervals for variable coefficients included zero or 95% confidence intervals for odds ratios included one. All means are reported ±SE. RESULTS The number of islands monitored on the CRD fluctuated between 37 and 525 from 1984 to 2005 with a total of 897 islands placed on the CRD over time. Of those, 96% were placed in ponds, the remainder in sloughs. Use of islands differed by wetland type. Dusky Geese were greater than six times more likely (95% Confidence Interval (CI): ) to use an island in a pond than a slough; however, average annual apparent nest success was similar between islands in sloughs (71 ± 18%, n = 7) and islands in ponds (64 ± 1%, n = 1,340; odds ratio = 0.7, 95% CI: ). Over half of the islands available for five years or greater were used less than 30% of the time. Combining island types, annual island use averaged 25 ± 2%, but use has increased since 1985 (F 1,20 = 12.19, p = 0.002; Fig. 2). The increase was most evident for fiberglass floating and sandbag islands (F 2,54 = 8.44, p 0.001; Fig. 3). Combining island types and years, apparent nest success averaged 64 ± 4%. Nest success showed no trend with time (F 1,20 = 0.35, p = 0.56) or island type (F 2,51 = 0.02, p = 0.98). Thus, we combined all nest island types for habitat modeling, and because sample size was small, we deleted islands placed in sloughs for all additional analyses. Island Use Annual use of islands averaged 33 ± 1% from (n = 376 islands and 1,028 individual nests). The univariate model with STATUS performed better than the null model (QIC weight = 1.0). Compared to is-

6 ISLAND USE BY DUSKY CANADA GEESE 273 Table 1. Descriptions of univariate model structures and statistical predictions for the logistic regression analyses of artificial nest island use by Dusky Canada Geese and nest success with explanatory variables of island and habitat characteristics on the western Copper River Delta, Alaska, USA. Model a Model Structure Univariate Predictions Associated with Artificial Nest Island Use and Nest Success Predicted Result Island Use FETCH β 0 (FETCH) Increased likelihood of nest island use with a decrease in fetch β 1 PSIZE β 0 (PSIZE) Increased likelihood of nest island use with a decrease in pond area β 1 DTS β 0 (DTS) Increased likelihood of nest island use with an increase in distance to the nearest shore β 1 DTLS β 0 (DTLS) Increased likelihood of nest island use with an increase in distance from the nearest large slough β 1 AC 2 β 0 (AC) + β 2 (AC 2 ) Highest likelihood of nest island use at a moderate percentage of aerial shrub cover β 1, β 2 SH 2 β 0 (SH) + β 2 (SH 2 ) Highest likelihood of nest island use at moderate average shrub height β 1, β 2 STATUS β 0NOT USED + Increased likelihood of nest island use when the island had a successful dusky goose nest the β 1 (STATUS SUCCESSFUL ) previous year relative to islands not used the previous year β 1 Nest Success FETCH β 0 (FETCH) Increased likelihood of nest success with a decrease in fetch β 1 PSIZE β 0 (PSIZE) Increased likelihood of nest success with a decrease in pond area β 1 DTS β 0 (DTS) Increased likelihood of nest success with an increase in the distance to the nearest shore β 1 DTLS β 0 (DTLS) Increased likelihood of nest success with an increase in the distance to the nearest large slough β 1 AC β 0 (AC) Increased likelihood of nest success with an increase in percent aerial shrub cover β 1 SH β 0 (SH) Increased likelihood of nest success with an increase in average shrub height β 1 DTLAR β 0 (DTLAR) Increased likelihood of nest success with a decrease in the distance to the nearest known larid colony or nest β 1 a DTS = distance from the island to the nearest shore; PSIZE = area of the island pond; AC = percent aerial shrub cover of the island, consisting of sweetgale, alder and/or willow; SH = average shrub height on the island (meters); DTLS = distance from the island to the nearest large slough; STATUS SUCCESSFUL = the island had a successful nest the previous year; β 0NOT USED = islands not used the previous year as the reference group; FETCH = distance from the island to shore in the direction of the prevailing wind (112.5? angle); DTLAR = distance from the island to the nearest known larid nest or colony.

7 274 WATERBIRDS Figure 2. Temporal patterns in artificial nest island availability and use by breeding Dusky Canada Geese on the western Copper River Delta, Alaska, USA. lands that were not used in year i, islands with failed nests in year i were 1.8 ( ) times more likely to get used in year i+1and islands that had successful nests were 4.0 ( ) times more likely to be used. Consequently, we used the univariate model with STATUS as the base (and null) model for subsequent modeling of environmental variables. The best two-variable model for island use included STATUS and AC 2 (QIC weight = 1.0; Table 2). An examination of parameter estimates for the other models indicated the confidence interval for the β estimate of DTS did not include zero, but the overall effect was very weak compared to AC 2 as evidence by its model ΔQIC value of Also, an inspection of the parameter estimates suggested that the linear form of SH appeared significant. When we included the linear form of SH along with STATUS and AC 2, a posteriori, that three-variable model performed better than the two-variable model with STA- TUS and AC 2 (QIC weight = 0.86; Table 2). The two-variable model with STATUS and AC 2 was the only model within four ΔQIC values of the best model (Table 2). Likelihood of island use was highest at intermediate aerial shrub coverage and increased with shrub height (Fig. 4). Nest Success Annual apparent nest success averaged 64 ± 2% from (n = 336 islands, 947 individual nests). The best model describing the data for nest success included Year and DTS (QIC weight = 0.98). No other model was within eight QIC values of the best model. Nest success increased as island distance from shore increased (β = 0.015, 95% Confidence limits = ; Fig. 5). Nest success was constant for seven of nine years, but was high in 2002 (β = 0.994, 95% Confidence limits = ) and low in 1998 (β = , 95% Confidence limits = [-0.12]). DISCUSSION Figure 3. Relationship between proportion of islands used by breeding Dusky Canada Geese and year for three types of artificial nest islands used by Dusky Canada Geese breeding on the western Copper River Delta, Alaska, USA (all slopes ± SE). Use of islands by nesting Dusky Geese increased during our study (Fig. 2). There was no clear relationship between use and the number of islands available over time; however, island use increased with time (Fig. 2). If we compare the first three years of the island program ( ) to the last three years in our analysis ( ), mean island use increased from 11% to 39% at the same time the CRD Dusky Goose population declined 35% (Pacific Flyway Council unpublished data). Thus, islands are proportionately more important to population productivity today than 20 years ago.

8 ISLAND USE BY DUSKY CANADA GEESE 275 Table 2. Modeling results investigating factors influencing use of artificial nest islands and nest success on islands for nesting Dusky Canada Geese on the western Copper River Delta, Alaska, USA( ). The top four models are shown for each analysis. Models QIC n k ΔQIC w Nest Island Use STATUS AC 2 SH STATUS AC STATUS SH STATUS SH Nest Success YEAR DTS YEAR DTS SH Figure 5. The probability of Dusky Canada Goose nest success on an artificial nest island versus distance from shore for Dusky Canada Geese breeding on the western Copper River Delta, Alaska, USA. Probabilities are the results of running the data through the best approximating model (nest success: YEAR + DTS) and generating a separate line for each year. Figure 4. The probability of a Dusky Canada Goose using an artificial nest island for nesting versus aerial shrub cover (a) and average shrub height (b) for Dusky Canada Geese breeding on the western Copper River Delta, Alaska, USA. Probabilities are the results of running the data through the best approximating model (island use: STA- TUS + AC 2 + SH) with a separate relationship displayed for each category of nest status in the previous year (island unavailable, island available but not used, island used nest failed, and island used nest successful). Island use rates were similar for fiberglass floating and sandbag islands (Fig. 3). Comparatively lower use rates for donut islands (Fig. 3) could reflect predator avoidance behavior during nest site selection, as observations of damage to donut islands suggests brown bears are attracted to the floating tires (Babler et al unpublished report). Alternately, the donut island nest platform was perched higher above the water, which could have influenced goose use (N. Maggiulli personal observation). Less than half of the islands were used each year for nesting by Dusky Geese (Fig. 2). Use of nest islands may be limited because of failure to provide gander loafing sites (Brakhage 1965; Rienecker 1971; Babler et al unpublished report), beaver and muskrat damage to islands (Babler et al unpublished report), the

9 276 WATERBIRDS limited availability of islands during nest site selection due to snow and ice cover (Bromley 1976; Campbell 1990; Bromley and Rothe 2003) or because nest sites on the western CRD are not limited (Babler et al unpublished report). However, the steady increase in use of islands with time suggests other factors are also likely important. Increased use of islands over time has been reported in other studies of Canada Geese (Craighead and Stockstad 1961; Rienecker 1971; Giroux 1981; Giroux et al. 1983) and could be caused by a variety of mechanisms that explain the recruitment of geese to islands and retention of first time island users including an increase in local population size, nest site fidelity by breeding adults, local imprinting of goslings hatched on islands, and nest prospecting by young breeders (Rienecker 1971; Giroux 1981; Giroux et al. 1983; Eadie and Gauthier 1985; Zicus and Hennes 1989). The Dusky Goose population size has fluctuated since 1985, but the overall trend is declining, so local increases in nesting population size is an unlikely explanation for the trend with year on the CRD. We suspect all the remaining mechanisms may contribute to increased nest use with time. Nest site fidelity, imprinting and prospecting are all supported by our finding that previous year s island status was an important indicator of Dusky Goose island use on the CRD, particularly when islands had a successful Dusky Goose nest in the prior year; but determining the relative importance of each would require tracking individually-marked birds over time, which was beyond the scope of our study. The different mechanisms do suggest potentially different management actions for increasing island use. If site fidelity by adults or gosling imprinting is the primary explanation, then efforts to increase nest success should further increase island use. If nest prospecting by first time breeders is important, and signs of old nests are important cues for prospecting geese (e.g. Zicus and Hennes 1989), then artificially creating successful nests on unused islands may increase use. We also hypothesized that recruitment to islands would be influenced by habitat characteristics of islands or the surrounding habitat. There was strong evidence that moderate shrub cover and tall shrub height were associated with an increased likelihood of island use (Table 2; Figs. 4a, 4b). These results agree with other studies of island-nesting Canada Geese (Craighead and Craighead 1949; Sherwood 1968; Giroux 1981; Reese et al. 1987) and are similar to a previous study of mainland nesting Dusky Geese on the CRD (Bromley 1976). Moderate shrub cover of tall shrubs is likely advantageous for concealing the nest and providing a physical barrier from predators, but also allowing for visibility to facilitate nest vigilance (review in Maggiulli 2008). We did not predict increased island use with taller shrubs; however, this finding suggests that concealing the nest from predators on islands may take precedence over visibility for Dusky Geese. These variables had a strong influence on island use, suggesting that a moderately-covered island nest site with protective tall shrubs is an important aspect of Dusky Goose nest site selection on the CRD. The relationships between island use and the remaining explanatory variables were insignificant. The range of apparent nest success estimates for Dusky Geese on islands of the western CRD in our study (47-82% for ) are similar to another study on the CRD (roughly 50-73%) that included a subset of the years we analyzed ( ; Miller et al. 2007). Miller et al. (2007) estimated daily nest survival rate using estimation procedures that controlled for the biases normally inherent with reporting apparent nest success and the fact that our estimate was similar to their estimates suggests any bias associated with our estimate is minor. Our estimate of nest success is also comparable to estimates of nest success for previous studies of nest islands (e.g. 63%, Craighead and Stockstad 1961; 73%, Brakhage 1965; 85%, Johnson et al. 1994). The majority of previous nest island studies reported apparent nest success, so results are directly comparable. In contrast to our work, most previous

10 ISLAND USE BY DUSKY CANADA GEESE 277 studies were conducted in the lower 48 states of the United States, where the predator community is not as diverse as the CRD. This suggests the nest island program on the CRD is resilient to numerous predator challenges and improves nest success similar to other nest island programs. We found strong evidence that nest success varied by year and the distance of the island from shore (Fig. 5). Recent work attributed 72% of Dusky Goose nest losses on the CRD to Bald Eagles (Haliaeetus leucocephalus; hereafter, eagles ; Anthony et al. 2004). The location and appearance of islands makes them easy for eagles to detect (N. Maggiulli, personal observation), but floating islands are small, unstable, and covered by low-lying shrubs, characteristics that likely discourage eagles from landing on islands (Miller et al. 2007). Additionally, distance from shore would not appear to be a deterrent to eagles. If true, mammalian predation may remain an important factor influencing Dusky Goose nest success on islands of the CRD. Additional research to determine predators frequenting islands and modes of depredation (e.g. Anthony et al. 2004; Anthony et al. 2006) may lead to additional suggestions for how to improve island nest success. We found no evidence that any of our additional habitat variables helped explain patterns of nest success. The higher success of nests on islands compared to the mainland (Grand et al. 2006; Miller et al. 2007) indicates the islands effectively protect Dusky Goose nests from predators. However, Miller et al. (2007) found this relationship was seasonal, with higher nest success rates for island nest sites only early in the nesting season. Miller et al. (2007) also found a seasonal relationship between habitat variables and nest success, with higher nest success on tall shrub and island sites than short shrub sites early in the nesting season, but no differences later in the nesting season (after May 25). Our island monitoring protocol did not allow the determination of nest initiation date. Consequently, if habitat features had seasonal impacts on nest success, the lack of this data may have impacted our ability to detect strong relationships between habitat features and nest success. Nest success was strongly associated with year, which likely reflects annual variability in predator and alternate prey community dynamics. Spawning runs of Pacific Eulachon (Thaleichthys pacificus, hereafter, eulachon ) attract large aggregations of eagles and other piscivorous birds to large sloughs and rivers on the CRD (Miller et al. 2006; N. Maggiulli, personal observation). Eulachon spawning runs often coincide with the Dusky Goose breeding season, and are known to influence Dusky Goose nest survival (Miller et al. 2006). In years with negligible eulachon spawning runs during the Dusky Goose nesting season (e.g. 1998, Moffitt et al. 2002; Cordova Ranger District unpublished report), aggregated predators may resort to predating Dusky Goose nests, regardless of the nest location (island or mainland). In these years habitat variables that can protect Dusky Goose nests on islands (e.g. shrub cover and shrub height) are likely less of a deterrent to predators. However, in years when eulachon are present throughout the nesting season (e.g. 2002, Miller et al. 2006), the majority of predators will be preoccupied with eulachon, and likely avoid predating Dusky Goose nests that are not easy to access (e.g. on islands). In these years, Dusky Goose nests on islands further from shore may have an additional advantage. Associations between habitat features and island use were more consistent than between habitat features and nest success. Lower variability in factors associated with island use is likely because factors associated with nest site selection are controlled by decisions of the nesting goose. Nest site selection by adults can be related to previous year s success (Giroux et al. 1983; Lindberg and Sedinger 1997), predator avoidance (Petersen 1990) and annual snowmelt (Lecomte et al. 2008) and these decision rules likely do not vary appreciably among years. In contrast, nest success is dependent on a more complex interaction of factors that vary within and among years outside the control of a nesting goose (e.g. Ackerman et al. 2004; Miller et al. 2007). Our inability to control

11 278 WATERBIRDS for these complexities may have, in part, confounded our analysis of habitat features important to nest success. Regardless, our analysis does indicate that the distance of the island from shore is strongly associated with nest success, and the influence of year on island nest success suggests the presence of alternate prey and predator abundance are more important to Dusky Goose nest success than island features. Dusky Goose population size on the CRD was relatively constant over the years of our study (Pacific Flyway Council 2008 unpublished report). Assuming gosling survival is independent of nest location (island vs mainland), the increasing use of islands and relatively constant nest success indicates Dusky Goose gosling production attributable to islands is increasing. Additionally, consistently high nest success indicates artificial islands have potential for increasing Dusky Goose productivity on the CRD, but this assumes, in part, that nest success is the reproductive factor most limiting productivity. Quantifying the impact of the nest island program on Dusky Goose population productivity requires a detailed understanding of gosling survival until birds depart for fall migration. ACKNOWLEDGMENTS Funding and resources were provided by the United States Forest Service Pacific Northwest Research Station, Corvallis, Oregon, the Cordova Ranger District in Cordova, Alaska, and the Department of Fisheries and Wildlife of Oregon State University. We thank staff of the Cordova Ranger District for the database they contributed and the housing and resources provided on the Copper River Delta; D. Logan and G. Reeves for their guidance and support; and M. Anthony, T. Fondell, D. Miller and United States Geological Survey for providing data and guidance. We also thank the many people who helped collect data on nest island use and nest success since program inception in LITERATURE CITED Ackerman, J. T., A. L. Blackmer and J. M. Eadie Is predation on waterfowl nests density dependent? Tests at three spatial scales. Oikos 107: Ainley, D. G., K. D. Dugger, R. G. Ford, S. D. Pierce, D. C. Reese, R. D. Brodeur, C. T. Tynan and J. A. Barth Association of predators and prey at frontal features in the California Current: competition, facilitation, and co-occurrence. Marine Ecology Progress Series 389: Allison, P. D Logistic regression using the SAS system: theory and application reprint, SAS Institute Inc., Cary, North Carolina. Anthony, M., J. B. Grand, T. F. Fondell and B. F. J. Manly A quantitative approach to identifying predators from nest remains. Journal of Field Ornithology 75: Anthony, R. M., J. B. Grand, T. F. Fondell and D. A. Miller Techniques for identifying predators of goose nests. Wildlife Biology 12: Bishop, M. A., P. M. Meyers and P. F. McNeley A method to estimate migrant shorebird numbers on the Copper River Delta, Alaska. Journal of Field Ornithology 71: Boggs, K. and M. Shephard Response of marine deltaic surfaces to major earthquake uplifts in southcentral Alaska. Wetlands 19: Brakhage, G. K Biology and behavior of tub-nesting geese. Journal of Wildlife Management 29: Bromley, R. G. H. and T. C. Rothe Conservation assessment for the Dusky Canada Goose (Branta canadensis occidentalis Baird). PNW GTR-591, United States Forest Service, Pacific Northwest Research Station, Portland, Oregon. Bromley, R. G. H Nesting and habitat studies of the Dusky Canada Goose (Branta canadensis occidentalis Baird) on the Copper River Delta, Alaska. Unpublished M.Sc. Thesis, University of Alaska, Fairbanks, Alaska. Burnham, K. P. and D. R. Anderson Model selection and multimodel inference: a practical information-theoretic approach. Second edition. Springer Science+Business Media, Inc., New York, New York. Campbell, B. H Factors affecting the nesting success of Dusky Canada Geese (Branta canadensis occidentalis) on the Copper River Delta, Alaska. Canadian Field-Naturalist 104: Cline, M. L Productivity of Giant Canada Geese (Branta canadensis maxima) in northeastern Illinois. M.Sc. thesis, Southern Illinois University, Carbondale, Illinois. Craighead, F. C., Jr. and J. J. Craighead Nesting Canada Geese on the upper Snake River. Journal of Wildlife Management 13: Craighead, J. J. and D. S. Stockstad Evaluating the use of aerial nesting platforms by Canada Geese. Journal of Wildlife Management 25: Crow, J. H Earthquake-initiated changes in the nesting habitat of the Dusky Canada Goose. Pages in The great Alaska earthquake of 1964: biology. Committee on the Alaska Earthquake of the Division of Earth Sciences, National Research Council, National Academy of Sciences, Washington, D.C. Eadie, J. M. A. and G. Gauthier Prospecting for nest sites by cavity-nesting ducks from the genus Bucephala. Condor 87: ESRI (Environmental Systems Resource Institute) ArcMap [GIS software]. ESRI, Redlands, California. Giroux, J., D. E. Jelinski and R. W. Boychuk Use of rock islands and round straw bales by nesting Canada Geese. Wildlife Society Bulletin 11: Giroux, J. F Use of artificial islands by nesting waterfowl in southeastern Alberta. Journal of Wildlife Management 45:

12 ISLAND USE BY DUSKY CANADA GEESE 279 Grand, J. B., T. F. Fondell, D. A. Miller and R. M. Anthony Nest survival in Dusky Canada Geese (Branta canadensis occidentalis): Use of discrete time models. Auk 123: Hatch, J. J Arctic Tern (Sterna paradisaea). Account 707 in The birds of North America online (A. Poole, Ed.), Cornell Lab of Ornithology, Ithaca, New York. accessed February Jehle, G., A. A. Yackel Adams, J. A. Savidge and S. K. Skagen Nest survival estimation: a review of alternatives to the Mayfield estimator. Condor 106: Johnson, R. R., K. F. Higgins, I. J. Ball and S. C. Kohn Waterfowl productivity and use of nesting structures in the prairie pothole region. Prairie Naturalist 26: Kaminski, R. M. and H. H. Prince Nesting habitat of Canada Geese in southeastern Michigan. Wilson Bulletin 89: Larson, S On the influence of the Arctic Fox (Alopex lagopus) on the distribution of Arctic birds. Oikos 11: Lecomte, N., G. Gauthier and J. Giroux Breeding dispersal in a heterogeneous landscape: the influence of habitat and nesting success in Greater Snow Geese. Oecologica 155: Lindberg, M. S. and J. S. Sedinger Ecological consequences of nest site fidelity in Black Brant. Condor 99: MacInnes, C. D Nesting of small Canada Geese near Eskimo Point, Northwest Territories. Journal of Wildlife Management 26: Maggiulli, N. M Factors associated with Dusky Canada Goose (Branta canadensis occidentalis) nesting and nest success on artificial nest islands of the western Copper River Delta. Unpublished M.Sc. Thesis. Oregon State University, Corvallis, Oregon. Mayfield, H. F Nesting success calculated from exposure. Wilson Bulletin 73: Miller, D. A Factors related to temporal and spatial variation in nest survival for Dusky Canada Geese on the Copper River Delta, Alaska. Unpublished M.Sc. Thesis. Auburn University, Auburn, Alabama. Miller, D. A., J. B. Grand, T. F. Fondell and M. Anthony Predator functional response and prey survival: Direct and indirect interactions affecting a marked prey population. Journal of Animal Ecology 75: Miller, D. A., J. B. Grand, T. F. Fondell and R. M. Anthony Optimizing nest survival and female survival: Consequences of nest site selection for Canada Geese. Condor 109: Moffitt, S., B. Marston and M. Miller Summary of Eulachon research in the Copper River Delta, : report to the Alaska Board of Fisheries. Regional Information Report No. 2A02-34, Alaska Department of Fish and Game, Anchorage, Alaska. Moskoff, W. and L. R. Bevier Mew Gull (Larus canus). Account 687 in The birds of North America online (A. Poole, Ed.), Cornell Lab of Ornithology. edu:80/bna/species/687, accessed February Nguyen, L. P., K. F. Abraham and E. Nol Influence of Arctic Terns on survival of artificial and natural Semipalmated Plover nests. Waterbirds 29: Pan, W Akaike s information criterion in generalized estimating equations. Biometrics 57: Petersen, M Nest site selection by Emperor Geese and Cackling Canada Geese. Wilson Bulletin 102: Piper, W. H., M. W. Meyer, M. Klich, K. B. Tischler and A. Dolsen Floating platforms increase reproductive success of Common Loons. Biological Conservation 104: Reese, K. P., J. A. Kadlec and L. M. Smith Characteristics of islands selected by nesting Canada Geese (Branta canadensis). The Canadian Field-Naturalist 101: Rienecker, W. C Canada Goose nest platforms. California Fish and Game 57: SAS Institute SAS Software, Version SAS Institute, Cary, North Carolina. Shealer, D. A., J. M. Buzzell and J. P. Heiar Effect of floating nest platforms on the breeding performance of Black Terns. Journal of Field Ornithology 77: Sherwood, G. A Factors limiting production and expansion of local populations of Canada Geese. Pages in Canada Goose Management: current continental problems and programs (R. L. Hine and C. Shoenfeld, Eds.), Dembar Educational Research Services, Madison, Wisconsin. Thilenius, J. F Woody plant succession on earthquake-uplifted coastal wetlands of the Copper River Delta, Alaska. Forest Ecology and Management 33/ 34: Trainer, C. E The 1959 Western Canada Goose (Branta canadensis occidentalis) study of the Copper River Delta, Alaska, United States Bureau of Sport Fisheries and Wildlife, Annual Waterfowl Report, Juneau, Alaska. Vermeer, K Ecological aspects of ducks nesting in high densities among larids. Wilson Bulletin 80: Vermeer, K A study of Canada Geese (Branta canadensis) nesting on islands in southeastern Alberta. Canadian Journal of Zoology 48: Vermeer, K., K. H. Morgon and G. E. J. Smith Black Oystercatcher habitat selection, reproductive success, and their relationship with Glaucouswinged Gulls. Colonial Waterbirds 15: Westerskov, K Methods for determining the age of game bird eggs. Journal of Wildlife Management 14: Will, G. C. and G. I. Crawford Elevated and floating nest structures for Canada Geese. Journal of Wildlife Management 34: Zicus, M. C. and S. K. Hennes Nest prospecting by Common Goldeneyes. Condor 91: