In the southern Beaufort Sea of the United States and Canada, prior investigations have linked declines in summer sea ice to reduced physical condition, growth, and survival of polar bears (Ursus maritimus). Combined with projections of population decline due to continued climate warming and the ensuing loss of sea ice habitat, those findings contributed to the 2008 decision to list the species as threatened under the U.S. Endangered Species Act. Here, we used mark–recapture models to investigate the population dynamics of polar bears in the southern Beaufort Sea from 2001 to 2010, years during which the spatial and temporal extent of summer sea ice generally declined. Low survival from 2004 through 2006 led to a 25–50% decline in abundance. We hypothesize that low survival during this period resulted from (1) unfavorable ice conditions that limited access to prey during multiple seasons; and possibly, (2) low prey abundance. For reasons that are not clear, survival of adults and cubs began to improve in 2007 and abundance was comparatively stable from 2008 to 2010, with ~900 bears in 2010 (90% CI 606–1212). However, survival of subadult bears declined throughout the entire period. Reduced spatial and temporal availability of sea ice is expected to increasingly force population dynamics of polar bears as the climate continues to warm. However, in the short term, our findings suggest that factors other than sea ice can influence survival. A refined understanding of the ecological mechanisms underlying polar bear population dynamics is necessary to improve projections of their future status and facilitate development of management strategies.
Introduction
Polar bears are vulnerable to the loss of sea ice, which is projected to induce substantial declines in abundance by mid-century (Amstrup et al. 2008, Hunter et al. 2010) unless global greenhouse gas levels are reduced (Amstrup et al. 2010). Global temperatures will rise as atmospheric greenhouse gas concentrations increase (Pierrehumbert 2011), and the Arctic has warmed at twice the global rate (IPCC 2007), in part due to positive feedback mechanisms referred to as Arctic Amplification (Serreze and Francis 2006, Perovich and Polashenski 2012). Since the advent of satellite observations in 1979, the spatial extent of Arctic sea ice during the autumn ice minimum declined by over 12% per decade through 2010 (Stroeve et al. 2012), a rate of loss greater than predicted by climate models (Stroeve et al. 2007, Overland and Wang 2013). Coastal areas are experiencing longer ice-free periods (Markus et al. 2009), and the remaining ice is increasingly composed of thin, first-year ice (Maslanik et al. 2007, 2011) with greater potential for rapid melt in subsequent years (Stroeve et al. 2012). Projections of global warming and sea ice loss led to the species being listed as threatened under the U.S. Endangered Species Act in 2008, increasing global awareness of its status, and elevating the importance of monitoring by circumpolar nations (Vongraven et al. 2012).Amstrup et al. (2008) introduced a classification of polar bear habitat into ecoregions based on broad seasonal patterns of sea ice dynamics that is useful for understanding regional differences in polar bear ecology (Fig. 1). However, populations sharing an ecoregion may experience localized combinations of environmental and ecological conditions that elicit different responses. For example, in the seasonal ice ecoregion (Fig. 1), where sea ice melts completely in summer and forces all bears ashore until ice re-forms in autumn, reduced access to prey during prolonged ice-free periods is negatively affecting the status of some populations (Stirling et al. 1999, Regehr et al. 2007, Rode et al. 2012), while high prey abundance may be forestalling declines in other populations (Peacock et al. 2013). In the divergent ice ecoregion (Fig. 1), polar bears still have access to some ice all year, although its availability over the continental shelf during summer and autumn is increasingly limited (Markus et al. 2009). Despite extensive sea ice loss throughout the divergent ice ecoregion (Markus et al. 2009, Stammerjohn et al. 2012), an expansive continental shelf and high productivity may have enabled the Chukchi Sea population to maintain condition and recruitment more effectively than the neighboring southern Beaufort Sea (SBS) population (Rode et al. 2014).The SBS population is one of 19 recognized worldwide (Obbard et al. 2010), and it has been studied more intensively than most. The population is thought to have been overharvested prior to the passage of the U.S. Marine Mammal Protection Act in 1972 (Amstrup et al. 1986), and to have generally increased in abundance thereafter through the late 1990s (Amstrup et al. 2001). Estimates from the mid-2000s suggested that abundance had stabilized and possibly declined (Regehr et al. 2006).Recent investigations of the SBS population have revealed early indications of the effects of climate-induced changes in the characteristics and availability of sea ice. Fischbach et al. (2007) documented a shift in the distribution of maternal dens from multiyear pack ice to terrestrial locations, perhaps in response to the reduced availability of ice suitable for denning (Amstrup and Gardner 1994). The summer retreat of sea ice from continental-shelf waters now forces polar bears to either remain with the remnant ice in the central Polar Basin or move to land; both options are hypothesized to reduce fitness compared to historical patterns of habitat availability and use. Although most SBS polar bears currently remain with the sea ice, a growing proportion of the population is utilizing terrestrial habitat (Schliebe et al. 2008; USGS, unpublished data) and accessing remains of subsistence-harvested bowhead whale (Balaena mysticetus) carcasses (Herreman and Peacock 2013). The increasing distance between shore and the summer pack ice increases the potential for long-distance swimming (Pagano et al. 2012), which elevates susceptibility to adverse weather (Monnett and Gleason 2006) and is energetically expensive (Durner et al. 2011). Nutritional stress also appears to be increasing (Cherry et al. 2009, Rode et al. 2014), and may be responsible for observations of reduced body size, growth, and survival of young (Rode et al. 2010, 2014). Regehr et al. (2010) associated reduced ice over the continental shelf in 2004 and 2005 with reductions in survival (Amstrup and Durner 1995, Amstrup et al. 2001).We investigated the population dynamics of SBS polar bears from 2001 to 2010 using Cormack-Jolly-Seber (CJS) mark–recapture models (Lebreton et al. 1992, Amstrup et al. 2005) to (1) determine whether low survival rates reported for 2004 and 2005 (Regehr et al. 2010) persisted into subsequent years, (2) assess the recent trend in abundance, and (3) refine our understanding of the relationship between sea ice and polar bear survival. The spatial and temporal extent of sea ice over the continental shelf generally declined during this period, and we evaluated the utility of measures of ice availability to explain temporal patterns in survival. Our findings provide new information on population status, as well as insights into the ecological mechanisms underlying population dynamics of polar bears.
The polar bear (Ursus maritimus) is a universally recognized symbol of the Arctic. Polar bears prefer sea ice concentrations exceeding 50% in the shallow, productive waters over the continental shelf (Sakshaug 2004, Durner et al. 2009), where ice provides a platform from which polar bears can efficiently hunt marine mammals (see Plate 1). Their primary prey are ringed (Pusa hispida) and bearded (Erignathus barbatus) seals, although diet varies regionally with prey availability (Thiemann et al. 2008, Cherry et al. 2011).
Methods
Study areaThe Beaufort Sea (Fig. 1), unlike most marginal seas in the Polar Basin, has a narrow continental shelf with a steep shelf-break that plummets to some of the deepest waters of the Arctic Ocean (Jakobsson et al. 2008). Pacific waters enter the Arctic Ocean via the Bering Strait and the remnant of the Alaska Coastal Current flows eastward along the shelf (Schulze and Pickart 2012). Nearshore waters carry substantial freshwater inputs, including terrestrial carbon and nitrogen, from the Mackenzie River and numerous smaller river systems (Dunton et al. 2006). Offshore, the anti-cyclonic Beaufort Gyre (Proshutinsky et al. 2002, Giles et al. 2012) and the Transpolar Drift Stream (Serreze et al. 1989) govern sea ice motion basin-wide.During spring, primary production in the SBS is dominated by ice algae (Horner and Schrader 1982). Wind-induced current reversals and storm events pump nutrient-rich basin waters onto the continental shelf, supporting production throughout the year and seeding the algal bloom the following year (e.g., Sigler et al. 2011, Tremblay et al. 2011, Schulze and Pickart 2012, Pickart et al. 2013a, b). Climate warming is increasing primary productivity (Tremblay et al. 2011, Nicolaus et al. 2012, Pickart et al. 2013a) and altering its composition (Lasternas and Agustí 2010). Open water at the interface of land-fast ice and pack ice is an additional source of primary production (Palmer et al. 2011), and these areas are important for numerous Arctic species (Stirling 1997). Zooplankton are thought to underutilize primary production in Arctic ecosystems, thereby favoring a rich benthos (Grebmeier et al. 2006, Logerwell et al. 2011). Arctic cod (Boreogadus saida) are the most abundant fish in pelagic waters (Jarvela and Thorsteinson 1999, Parker-Stetter et al. 2011). Ringed and bearded seals are resident year-round (Stirling et al. 1982, Frost et al. 2004), while beluga (Delphinapterus leucas) and bowhead whales migrate into the SBS during summer (Luque and Ferguson 2009, Ashjian et al. 2010). Polar bears are an apex predator of this food web, which may be sensitive to perturbation due to its simple structure and strong interspecific dependencies (Banašek-Richter et al. 2009).
Data sourcesU.S. Geological Survey (USGS) researchers captured polar bears in the U.S. portion (Alaska) of the SBS (Fig. 1) from approximately late March to early May annually from 2001 to 2010. The communities of Barrow, Deadhorse, and Barter Island (Kaktovik) were used as operational bases each year, excluding Barrow in 2001 and Barter Island in 2006. Helicopters were used to search the sea ice for polar bears, ranging as far as ~160 km from the coast. Bears were immobilized with Telazol (Pfizer Animal Health, New York, New York, USA) administered with projectile syringes fired from a helicopter (Stirling et al. 1989) and given lip tattoos and ear tags with unique identification numbers. Satellite radio collars were affixed to a subset of adult females captured, except in 2010. Captures were generally nonselective with respect to sex and age class, although females wearing radio collars were often targeted to facilitate collar retrieval. Field procedures were approved by the independent USGS Alaska Science Center Animal Care and Use Committee.Researchers from Environment Canada (EC) and the University of Alberta (UA) conducted mark–recapture activities in the Canadian portion of the population range (Fig. 1) using similar methods in April and May from 2003 to 2006. The combined efforts of USGS, EC, and UA resulted in the distribution of capture effort throughout the majority of the SBS population range in those years. UA researchers continued to capture polar bears in Canada from 2007 to 2010, although subadults and females were preferentially targeted, and capture effort did not extend into the easternmost portion of the study area. Animal welfare committees of EC and UA approved bear capture and handling protocols in Canada.We modeled the population dynamics of SBS polar bears using two combinations of data. The first data set (USGS) was compiled from USGS captures in the United States portion of the study area. Capture methods were consistent throughout the study period, and these data were therefore expected to provide the most reliable assessment of trends in survival and abundance, although estimates were applicable only to the portion of the population available for capture within the United States. The second data set (USA and Canada [USCA]), was compiled from the data collected by all three entities. The USCA data had greater spatial coverage and the potential to produce estimates germane to the entire SBS population. However, the geographic and temporal discontinuities in capture effort and differential selectivity for age and sex classes in some years presented modeling challenges, especially with respect to estimation of recapture probabilities and abundance.
Mark–recapture modelingWe estimated the survival and abundance of SBS polar bears using open-population Cormack-Jolly-Seber (CJS) models (Lebreton et al. 1992, Amstrup et al. 2005), similar to several previous mark–recapture investigations of polar bear populations (e.g., Amstrup et al. 2001, Regehr et al. 2007, 2010, Taylor et al. 2008, Stirling et al. 2011). Multiple observations of an individual within a calendar year were amalgamated into a single capture record for that year and the history of annual capture indicators was constructed for each animal. We used information on harvests of marked bears to terminate subsequent modeling of their capture histories.CJS models are composed of sub-models for survival and recapture probabilities, which are typically expressed as linear functions of explanatory variables (covariates) via a logistic link function (Lebreton et al. 1992). Covariates were either single variables or groups of related variables that were employed simultaneously, so we used a single term to reference either case (Table 1; Appendix A). We utilized the regression parameterization of CJS models (McDonald and Amstrup 2001, Amstrup et al. 2005) because of the flexibility with which it incorporates covariates. Parameters of the logistic functions were estimated using maximum likelihood. Survival and recapture probabilities were estimated from the parameters of the logistic functions, and abundance estimates were derived from the estimated recapture probabilities using the Horvitz-Thompson estimator (Horvitz and Thompson 1952, McDonald and Amstrup 2001).
Survival probability modelsWe constructed survival models from combinations of covariates representing age and sex class effects and forms of temporal structure, and modeled survival probabilities separately for each age class (Table 2). Four age classes were defined: cub (Age0), yearling (Age1), subadult (2 to 4 years old, Age2), and adult (>4 years old, Age3), categories similar to those used in other investigations (e.g., Regehr et al. 2010, Stirling et al. 2011). Survival models also incorporated one of four forms of temporal variation: temporal stratification (TS-sur), a cubic function of year (Time-cubic), and two measures of sea ice availability. Data were too sparse to independently estimate an annual survival probability for each age class, but the covariates TS-sur and Time-cubic provided flexibility to model temporal variation using fewer parameters (e.g., Stoklosa and Huggins 2012, Peacock et al. 2013, Thorson et al. 2013). We constructed the ice covariate Summer-habitat by summing monthly indices of the area (km2) of optimal polar bear habitat over the SBS continental shelf (Durner et al. 2009) for July through October of each year (Fig. 2). We expected this covariate to be informative because it was based on the availability of habitat preferred by radio-collared polar bears. The ice covariate Melt-season measured the time between the melt and freeze onset in the Beaufort Sea each summer (Inner Melt Length; Fig. 2; Markus et al. 2009, Stroeve et al. 2014), obtained courtesy of Dr. Julienne Stroeve (National Snow and Ice Data Center, University of Colorado, personal communication). We included Melt-season because it covered a broader geographic region than Summer-habitat and the two covariates were not highly correlated (r = −0.68). The means and standard deviations of the ice covariates from 2001 to 2010 were used to normalize their values before analysis.
Recapture probability modelsRecapture models incorporated covariates for sex and the four age classes. We constrained recapture probabilities of cubs and yearlings to equal those of adult females because family groups were captured simultaneously (Age01.3Fem). We constructed a group of time-varying indicator covariates (UA) to model potential discontinuities in Canadian recapture probabilities after 2006, because UA researchers targeted females and subadults and expended less capture effort from 2007 to 2010.Recapture models incorporated two nonparametric forms of temporal structure, either constant through time or a separate probability for each year (Time). In addition, we constructed covariates to model temporal variation arising from heterogeneity in capture effort. The hours spent searching for polar bears in the United States each year was recorded by USGS (Search-US). The total hours spent flying in Canada was recorded, and we assumed 60% of flight hours were spent actively searching for polar bears, a percentage derived from USGS flight records. We added the resulting approximation of Canadian search hours to USGS search hours to construct a measure of effort for the entire study area (Search). In addition, we stratified annual search hours in the United States and Canada into low, medium, and high effort categories (Eff-US, Eff-CA), which Stirling et al. (2011) found useful.We incorporated geographic structure into some models because recapture probabilities vary within the study area (Amstrup et al. 2001). We assigned individuals to a home stratum based on the proximity of their mean capture longitude to the four operational bases (Fig. 1), and used stratum assignments to construct four indicator covariates: Barrow, Deadhorse, Barter Island, and Canada (Home). These covariates reflected coarse variation in recapture probabilities among regions, rather than distinct locations as would be used in multistate models (White et al. 2006). We collapsed the Home covariates into two covariates corresponding to country (US, CA) for some models. In addition, we used an indicator covariate (BI2006) to accommodate the lack of effort from the Barter Island base in 2006.Two covariates improved model fit so greatly that we included them in all recapture models. The first (Radio) indicated whether a bear was wearing an active radio during the sampling period each year. Instrumented bears were often targeted for sampling to evaluate their condition and retrieve radios, and could therefore have elevated recapture probabilities; similar covariates have been used by other researchers (e.g., Amstrup et al. 2001, Regehr et al. 2010, Stirling et al. 2011). The second covariate (Cap-procliv) modeled potential recapture heterogeneity unexplained by other covariates, using an individual’s prior capture history as a measure of its current tendency to be recaptured (Appendix F), indirectly accounting for individual characteristics, such as behavior, that are difficult to quantify.
We constructed models of recapture probability from combinations of four age and sex class effects and eight covariate combinations representing temporal and geographic (abiotic) structure (Tables 3 and 4).
Modeling strategiesWe formed CJS models from combinations of survival and recapture sub-models (Tables 2–4). We utilized the plausible combinations (PC) strategy (Bromaghin et al. 2013), with an Akaike’s information criterion (AICc) model weight (Burnham and Anderson 2002) of 2.5% as an inclusion threshold, to objectively base inference on a reduced model space. Simulation results (Bromaghin et al. 2013) suggest its performance is similar to that of the “all combinations” strategy recommended by Doherty et al. (2012).We utilized R version 3.0.3 (R Development Core Team 2013) for data manipulation and version 2.14 of the R package mra (available online)9 to estimate model parameters (function F.cjs.estim), using the Horvitz-Thompson estimator of abundance (Horvitz and Thompson 1952, McDonald and Amstrup 2001), and compute model-averaged estimates based on AICc model weights (function F.cr.model.avg).We evaluated estimation precision using bootstrap resampling (Chernick 1999), drawing bootstrap samples of capture histories and individual covariates at random with replacement. For each sample, we implemented the PC strategy to select models, estimate the parameters of all selected models, and compute model-averaged estimates (Burnham and Anderson 2002) of recapture, survival, and abundance. Because of the computational burden of this approach, we drew 100 bootstrap samples. We bias-corrected (Chernick 1999) the model-averaged estimates, so their means equaled the original model-averaged estimates, and constructed nonparametric 90% confidence intervals as the 5th and 95th quantiles of the corrected bootstrap estimates.AICc is a relative measure of model suitability, so confidence in the interpretation of AICc statistics is enhanced if at least one model provides a reasonable approximation to the data. For each data set, we selected one of the most flexible models and assessed goodness-of-fit using tests implemented in the R package mra (function F.cjs.gof), which are derived from procedures common in logistic regression (Hosmer and Lemeshow 2000, Sakar and Midi 2010).
We assessed the importance of covariates using summed AICc model weights (Burnham and Anderson 2002). However, because uninformative covariates accumulate weight when used in combination with informative covariates (Doherty et al. 2012, Bromaghin et al. 2013), we summed weights on the basis of covariate combinations (Tables 2–4), rather than individual covariates (Burnham and Anderson 2002).
Complementary analyses to investigate potential biasAspects of animal movement and their availability for capture are important considerations in mark–recapture modeling. A pulse of permanent emigration, or temporary emigration near the end of a study, can produce the type of decline in apparent survival reported by Regehr et al. (2010). Although the population level effect of permanent emigration is equivalent to death (e.g., Nichols 2005), understanding population dynamics requires differentiating between them. Similarly, because only the nearshore regions of SBS polar bear habitat can be searched by helicopter (Fig. 1), heterogeneity in habitat preference, temporary emigration, and the nonrandom distribution of individuals could bias parameter estimates. For these reasons, we analyzed available location data from radio-collared female polar bears for signs of pulsed emigration or nonrandom movement that might aid interpretation of parameter estimates.Abundance estimates in CJS models are derived from recapture probabilities using the Horvitz-Thompson estimator (Horvitz and Thompson 1952, McDonald and Amstrup 2001). Because un-modeled heterogeneity in recapture probabilities can bias abundance estimates, we investigated the reliability of estimated abundance trends by comparing them to simulated abundance projections based on estimated survival probabilities, which are comparatively robust to recapture heterogeneity (Carothers 1979, Abadi et al. 2013), but also see Peñaloza et al. (2014). The projections incorporated a measure of cub production similar to litter production rate (Ramsay and Stirling 1988) and breeding success (Wiig 1998).
Results
Data summaryThe USGS data contained information on 715 individual bears that were captured a total of 1049 times (Appendix B). Information on the harvests of 21 marked bears was used to censor their capture histories. The USCA data contained information on 1086 individuals captured a total of 1590 times, and capture histories of 38 bears were censored following their harvests. The number of bears captured annually ranged from 75 to 148 individuals in the USGS data, and from 97 to 291 individuals in the USCA data set (Table 5). The proportion of captured bears that were marked increased during the initial years and stabilized approximately midway through the investigation (Table 5). Additional capture statistics are provided in Appendix B.
Model selection and goodness-of-fitThe PC model selection strategy (Bromaghin et al. 2013) with the USGS data resulted in the retention of 60 survival probability models and 20 recapture probability models. The estimation algorithm failed to converge for one combination of survival and recapture models (for survival, Int + Age3 × Melt-season + Age3Fem + Age2 × Melt-season + Age2Fem + Age1 × Melt-season + Age0 × TS-sur; and for recapture, Time + Age01.23Fem + Radio + Cap-procliv). The model selected to evaluate goodness-of-fit contained 34 parameters (covariate structures listed in boldface type in Tables 2 and 3). The results of goodness-of-fit tests did not raise concerns regarding inadequate model fit. The Osius-Rojek test could not be evaluated (it returned a “not-a-number” code), but the Hosmer-Lemeshow test was not significant (P = 0.690), and the Receiver Operating Characteristic (ROC) curve displayed acceptable discrimination (0.736).For the USCA data, the PC model selection strategy resulted in 47 survival probability models and 4 recapture probability models being selected for further consideration. The model selected to assess goodness-of-fit contained 39 parameters (covariate structures listed in boldface type in Tables 2 and 4). Based on that model, the Osius-Rojek test was moderately significant (P = 0.036), but the Hosmer-Lemeshow test was not (P = 0.276) and the Receiver Operating Characteristic (ROC) curve displayed acceptable discrimination (0.727), providing no compelling evidence of inadequate model fit.
AbundanceAnnual abundance estimates based on the USGS data, applicable to the Alaskan portion of the study area, ranged from 376 bears in 2009 to 1158 in 2004 (Fig. 5a). We suspect estimates in the first two years, particularly 2002, were negatively biased by the absence of capture effort from Barrow in 2001, which may have caused an overestimate of recapture probability in 2002 (Appendix C: Figs. C1–C3). In addition, the mixing of marked and unmarked individuals may have been incomplete during the initial years of the study (Appendix D: Fig. D2). The unusually large number of bears captured in 2004 (Table 5) produced a seemingly large estimate with a wide confidence interval. Even though there is uncertainty regarding abundance levels in these years, the broader pattern of a decline in abundance during the middle of the study followed by relative stability at the end of the study was consistent with patterns in survival.Abundance estimates based on the USCA data ranged from 464 bears in 2002 to 1607 in 2004 (Fig. 5b). As with the USGS data, we suspect estimates for the initial years of the investigation were less reliable than those in the latter years, particularly because no capture effort occurred in Canada before 2003 and our models were not robust to this deficiency. Considering this uncertainty in the earliest estimates, the temporal pattern in abundance resembled that of the USGS estimates and was consistent with patterns in survival. The correlation between USGS and USCA abundance estimates was 0.84 across all years and 0.86 excluding 2002.
Complementary analyses to investigate potential biasAnalyses of location data from radio-collared females produced no strong evidence that temporary emigration or non-random movement negatively biased estimates of survival (Appendix D). An exact test of the hypothesis that equal proportions of collared bears were available for capture each year was not significant (P = 0.990), implying that a pulse of permanent emigration did not occur in the middle of the investigation. We did, however, find indications of nonrandom movement between nearshore and offshore habitats among consecutive years. Although such nonrandom movement can bias estimates of recapture and survival probabilities, our data are consistent with conditions in which bias in survival probabilities is small (modest state transition probabilities; Kendall et al. 1997, Schaub et al. 2004).The results of population projections were largely consistent with the estimated trends in abundance (Appendix E). Projected abundance increased during the initial years of the study, though more slowly than suggested by the abundance estimates, supporting our conjecture that abundance estimates in the earliest years were negatively biased. Projected abundance declined through the middle of the study, and then stabilized. The correlation between abundance estimates and the projections based on the USGS data set was 0.85 across all years and 0.90 excluding 2002; the corresponding correlations for the USCA data set were 0.89 and 0.95.
Discussion
Methodological effectivenessWe constructed covariates to account for multiple sources of heterogeneity in recapture probabilities that could otherwise bias survival and abundance estimates (Abadi et al. 2013). Recapture probabilities were allowed to vary by sex and age class. Covariates indexed both capture effort and nonparametric temporal structure, and geographic covariates accounted for previously discovered regional differences in recapture probabilities (Amstrup et al. 2001). The covariate Cap-procliv (Appendix F) indirectly accounted for potential sources of heterogeneity we could not measure. Collectively, these covariates accounted for several known or potential sources of heterogeneity, and any bias originating from un-modeled heterogeneity is likely to be small compared to the temporal variation in survival and abundance estimates.We found no compelling evidence that emigration or nonrandom movement meaningfully biased estimates. Neither analysis of movement data (Appendix D), nor patterns in recapture rates (Appendix B) revealed a pulse of permanent emigration that could explain the mid-study decline in survival. Similarly, a pulse of immigration followed shortly by permanent emigration, which could explain the high USGS abundance estimate in 2004 and the mid-study decline in survival, is unlikely to have occurred given joint consideration of movement data and recapture rates. The most likely sources of immigrants are the neighboring northern Beaufort and Chukchi Seas (Fig. 1), but large-scale immigration from these regions seems unlikely because their ecological conditions are thought to be more favorable than in the SBS (e.g., Stirling et al. 2011, Rode et al. 2014). In addition, our estimates did not exhibit the decline in apparent survival near the end of the investigation that is indicative of temporary emigration (e.g., Langtimm 2009, Kendall et al. 2013). Analyses of movement data suggested that Markovian dependency in the probability of being available for capture between consecutive years remains a potential source of bias (Schaub et al. 2004). However, we view these results with some caution because of the small sample sizes and prior evidence that bears prefer ice in waters over the narrow continental shelf (Durner et al. 2009). Further, there is no reason to suspect that behavior leading to nonrandom movement during the spring capture season changed during the investigation, implying that trends in survival and abundance would not be appreciably affected even if some bias is present. In summary, although we are aware of the potential influence of temporary emigration and nonrandom movement, we believe any bias from these sources is likely to be small compared to the magnitude of temporal variation and trends in survival and abundance estimates.We believe that the assumptions of the Horvitz-Thompson abundance estimator (Horvitz and Thompson 1952, McDonald and Amstrup 2001) were satisfied. Covariates were utilized to model several known and potential sources of heterogeneity in recapture probabilities, so there is no reason to suspect they are systematically biased. The CJS model conditions on initial captures and estimates subsequent recapture probabilities, while the Horvitz-Thompson estimator applies estimated recapture probabilities to both initial captures and recaptures (McDonald and Amstrup 2001). A behavioral response to initial capture that lowers recapture probabilities therefore has the potential to positively bias abundance estimates. However, our capture methods largely preclude such a behavioral response. Bears are often located by following fresh tracks in snow. There are patches of rough ice in which bears could attempt to hide, if they were not being tracked, but we have not observed hiding behavior. Most bears move away as the helicopter approaches, though some display curiosity or aggression toward it. In addition, we are unaware of any information that suggests our abundance estimates are unrealistically high.
Insights into ecological mechanismsThe previously reported low survival of SBS polar bears in 2004 and 2005 (Regehr et al. 2010) continued into 2007, at levels substantially less than earlier estimates for this population (Amstrup and Durner 1995, Amstrup et al. 2001). Annual survival of cubs during that period was estimated to be <0.2 based on the USCA data and near zero with the USGS data. Indeed, only 2 of 80 cubs in the USGS data observed from 2003 to 2007 are known to have survived to an older age class (Appendix B: Fig. B1). Such poor recruitment, combined with reduced survival of other age classes, must have substantially impacted abundance, and our abundance estimates declined accordingly during those years. Estimated survival of adults and cubs from both data sets began to improve in 2007, with estimates based on the USCA data suggesting a somewhat stronger recovery, although confidence intervals overlap broadly. This potential divergence may be an early indication of regional differences in ecosystem function and polar bear response that may become more apparent as the Arctic continues to warm.Several independent sources of information are consistent with reduced survival during the middle of the study period. Cherry et al. (2009) found physiological indications of nutritional stress to be two to three times greater in 2005 and 2006 than in the 1980s. Regehr et al. (2006) reported observations of starvation during this period, and unusual incidents of polar bears stalking and killing other bears occurred in 2004 (Amstrup et al. 2006). Stirling et al. (2008) reported a case of cannibalism and several instances of polar bears penetrating unusually thick ice barriers to reach ringed seal lairs between 2003 and 2007, predatory behavior that is energetically inefficient and a likely indication of nutritional stress. A similar instance of ice penetration was observed in the SBS in 1975 (Stirling et al. 2008), during a period of low seal production and reduced polar bear cub survival throughout the Canadian portion of the Beaufort Sea (Stirling 2002). Similar low productivity of ringed seals and reduced survival of polar bear cubs were documented in the Canadian portion of the Beaufort Sea again in the mid-1980s (Smith 1987, Stirling 2002).Factors leading to improved survival beginning in 2007 are difficult to identify, but there are indications of a transition at this time. Pilfold et al. (2014) reported a shift in the distribution of seal kill sites among land-fast and pack ice between the periods 2003–2006 and 2007–2011, and more kill sites were located in the latter period. The pattern and occurrence of open-water lead systems and ice deformation within the central Beaufort Sea changed in 2006 (Mahoney et al. 2012). Similarly, Melling et al. (2012) reported unusual atmospheric and oceanographic conditions in the Beaufort Sea during 2007. Ringed seal productivity within Amundsen Gulf in the eastern SBS was low in 2005 and 2006, but improved in 2007 (Harwood et al. 2012), although this was attributed to localized ice conditions and is not thought to be indicative of the low ringed seal production broadly observed in the 1970s and 1980s (Stirling 2002). Potential linkages between these indicators of a transition in the 2006–2007 time frame and polar bear survival may become apparent as our understanding of Arctic ecology improves.Improved survival after 2007 might be partially attributable to either density-dependent mechanisms or the altered characteristics or behavior of surviving individuals. Reduced competition for limited resources resulting from lower abundance may have increased survival at the end of the study period. Similarly, consecutive years of unfavorable conditions may have eliminated the less fit individuals from the population, with the survivors collectively displaying seemingly enhanced survival. Finally, some individuals may have adopted behavior that increased their survival. For example, a growing proportion of the population utilizes terrestrial habitat after ice retreats from the continental shelf in summer (USGS, unpublished data). These bears have access to subsistence-harvested bowhead whale carcasses at discrete locations along the Alaskan coast (Herreman and Peacock 2013), an energy-rich alternative food source (Bentzen et al. 2007) that may enhance survival. Bears utilizing terrestrial habitat in summer may also have earlier access to seals in autumn. Land-fast ice begins to form over coastal waters earlier than the expanding pack ice provides bears summering on perennial ice deep in the Polar Basin with access to continental shelf waters. Such behavioral tendencies are likely to be adopted by dependent young and may therefore become increasingly common.
Subadults, unlike adults and cubs, did not display increased survival during the latter years of our investigation. Newly independent and inexperienced individuals are likely to be less efficient hunters, and less capable of competing for limited resources or retaining control of kills, than adults (Stirling 1974), and may be more susceptible to unfavorable conditions. Subadults are also less likely to utilize crowded feeding sites such as bowhead whale remains. Sea ice covariates were more strongly associated with the survival of subadults than adults, which is consistent with findings that the body condition of growing bears is more closely linked to ice conditions than that of older bears (Rode et al. 2010). The lack of an improvement in subadult survival may be, in part, a residual effect of poor conditions in prior years. Yearlings that survived to subadult status during years of low survival may have been in inadequate condition to survive independently. Regardless of cause, the status of subadults in the SBS population merits monitoring, because their continued low survival could ultimately lead to further declines in abundance.
Polar bears depend on sea ice for several aspects of their life history, and multiple characteristics of sea ice can be expected to influence their vital rates, potentially via mechanisms that are complex and nonlinear (e.g., Ellis and Post 2004, Tyler et al. 2008, Derocher et al. 2013). The duration of ice-absence from the continental shelf is thought to directly affect polar bear condition and vital rates through reduced access to prey (Regehr et al. 2010, Rode et al. 2010, 2012). Observations of polar bears and seal kills in low-concentration, unconsolidated ice (Chukchi Sea in 2009, SBS in 2010; S. C. Amstrup, G. M. Durner, and E. V. Regehr, unpublished field observations) testify to the high value of sea ice in biologically productive shelf waters. However, other aspects of sea ice are undoubtedly important. Extensive ice rubble and rafted floes during winter and spring are thought to have led to past declines in polar bear productivity in the SBS (Stirling et al. 1976, Amstrup et al. 1986, Stirling 2002), as well as during our investigation (Stirling et al. 2008). The increased frequency and severity of storms (Sepp and Jaagus 2011, Thomson and Rogers 2014) combined with thinner and more mobile pack ice (Spreen et al. 2011), both consequences of climate warming, are likely to result in a greater prevalence of deformed ice in winter and spring that may result in lower quality ringed seal birth lair habitat and subsequent reductions in reproductive success. Similarly, continued warming may increase the frequency of unsuitable snow and ice conditions for maintenance of ringed seal lairs (Stirling and Smith 2004, Hezel et al. 2012). The increased vulnerability of ringed seal pups to predation could temporarily enhance polar bear survival, though it would likely also lead to subsequent reductions in prey abundance. Finally, access to terrestrial denning sites can be limited by ice conditions (Derocher et al. 2011).
Despite the known importance of sea ice, measures of ice availability did not fully explain short-term demographic patterns in our data, suggesting that other aspects of the ecosystem contribute importantly to the regulation of population dynamics. Prey abundance can obviously affect bear condition and survival independently, to an extent, of sea ice conditions. Numerous factors such as warming-induced increases in primary productivity (Zhang et al. 2010), phenology-based trophic mismatches (Post et al. 2013), changing disease vectors (Jensen et al. 2010), increased contaminant transport (Sonne 2010), and expanding human activity in the Arctic (Smith and Stephenson 2013) may interact with the primary effects of sea ice conditions and prey accessibility. As the Arctic continues to warm, ecosystem structure and function can be expected to respond (e.g., Lasternas and Agustí 2010, Hezel et al. 2012, Carroll et al. 2013, Hinzman et al. 2013, Ji et al. 2013, Iverson et al. 2014, Nahrgang et al. 2014), perhaps in ways that are difficult to foresee. Although sea ice availability is expected to be the dominant driver of population dynamics over the long term (Amstrup et al. 2008, Stirling and Derocher 2012), other aspects of the ecosystem can be expected to either mitigate or exacerbate the effects of sea ice loss in the short term. The changing Arctic ecosystem may induce increased short-term variation in vital rates and elevate the risk of abrupt population decline (Derocher et al. 2013).
ConclusionsOur results collectively suggest that polar bears in the SBS experienced a significant reduction in survival and abundance from 2004 through 2007. However, suspected biases in the abundance estimates for the earliest years of the investigation, potential biases in the USCA estimates in the latter years, and statistical variation associated with estimates necessitate cautious expression of the magnitude of population decline. Conservatively, the decline seems unlikely to have been <25%, but may have approached 50%. Improved survival and stability in abundance at the end of the investigation are cause for cautious optimism. However, given projections for continued climate warming (IPCC 2013), polar bear ecology in the SBS and elsewhere in the Arctic is expected to become increasingly forced by sea ice loss (Amstrup et al. 2008, Stirling and Derocher 2012). Behavioral plasticity and ecosystem productivity may enable some populations to temporarily maintain recruitment and abundance, despite deteriorating habitat conditions (e.g., Rode et al. 2014), but seem unlikely to counterbalance the extensive habitat degradation projected to occur over the long term (IPCC 2013). Continued research into the ecological mechanisms underlying polar bear population dynamics is necessary to refine projections of their future status and develop appropriate strategies for their management.
Acknowledgements
We thank the many members of the U.S. Geological Survey, Environment Canada, and the University of Alberta field crews and the pilots transporting them for collection of the raw data. We also thank Julienne Stroeve of the University of Colorado, National Snow and Ice Data Center for providing the melt-season covariate, and Tom Evans, U.S. Fish and Wildlife Service, Marine Mammals Management, Anchorage, Alaska, and Marsha Branigan, Government of the Northwest Territories, Environment and Natural Resources, Inuvik, Northwest Territories (NT), Canada, for providing access to harvest records of marked bears. William Kendall, Karen Oakley, Karyn Rode, and Michael Runge of the U.S. Geological Survey, and five anonymous referees provided helpful comments on earlier drafts. The Subject Matter Editor’s management of peer review was extremely helpful. Support for long-term research on SBS polar bears conducted in the United States was provided by U.S. Geological Survey, Ecosystems and Climate and Land Use Change Mission Areas, and the U.S. Bureau of Land Management. Support for research conducted in Canada was provided by the Canadian Wildlife Federation; Environment Canada; Circumpolar/Boreal Alberta Research; Inuvialuit Game Council; National Fish and Wildlife Federation; Natural Sciences and Engineering Research Council of Canada; Northern Scientific Training Program; Northwest Territories Department of Environment and Natural Resources; Parks Canada; Polar Bears International; Polar Continental Shelf Project; Quark Expeditions; U.S. Bureau of Ocean Energy Management; U.S. Geological Survey, Ecosystems Mission Area, Science Support Partnership Program (2003–2005); University of Alberta; and World Wildlife Fund (Canada). The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the U.S. Fish and Wildlife Service (E. V. Regehr). Any mention of trade names is for descriptive purposes only and does not constitute endorsement by the U.S. federal government.
Literature Cited
Abadi, F., A. Botha, and R. Altwegg. 2013. Revisiting the effect of capture heterogeneity on survival estimates in capture-mark-recapture studies: Does it matter? PLoS ONE 8:e62636.[CrossRef]
Amstrup, S. C., E. T. DeWeaver, D. C. Douglas, B. G. Marcot, G. M. Durner, C. M. Bitz, and D. A. Bailey. 2010. Greenhouse gas mitigation can reduce sea-ice loss and increase polar bear persistence. Nature 468:955–958.[CrossRef]
Amstrup, S. C., and G. M. Durner. 1995. Survival rates of radio-collared female polar bears and their dependent young. Canadian Journal of Zoology 73:1312–1322.[CrossRef]
Amstrup, S. C., and C. Gardner. 1994. Polar bear maternity denning in the Beaufort Sea. Journal of Wildlife Management 58:1–10.[CrossRef]
Amstrup, S. C., B. G. Marcot, and D. C. Douglas. 2008. A Bayesian network modeling approach to forecasting the 21st century worldwide status of polar bears. Pages 213–268 in E. T. DeWeaver, C. M. Bitz, and L.-B. Tremblay, editors. Arctic sea ice decline: Observations, projections, mechanisms, and Implications. Geophysical Monograph Series 180. American Geophysical Union, Washington, D.C., USA.
Amstrup, S. C., T. L. McDonald, and B. F. J. Manly. 2005. Handbook of capture-recapture analysis. Princeton University Press, Princeton, New Jersey, USA.
Amstrup, S. C., T. L. McDonald, and I. Stirling. 2001. Polar bears in the Beaufort Sea: A 30-year mark-recapture case history. Journal of Agricultural, Biological, and Environmental Statistics 6:221–234.[CrossRef]
Amstrup, S. C., I. Stirling, and J. W. Lentfer. 1986. Past and present status of polar bears in Alaska. Wildlife Society Bulletin 14:241–254.
Amstrup, S. C., I. Stirling, T. S. Smith, C. Perham, and G. W. Thiemann. 2006. Recent observations of intraspecific predation and cannibalism among polar bears in the southern Beaufort Sea. Polar Biology 29:997–1002.[CrossRef]
Ashjian, C. J., et al. 2010. Climate variability, oceanography, bowhead whale distribution, and Iñupiat subsistence whaling near Barrow, Alaska. Arctic 63:179–194.[CrossRef]
Ashjian, C. J., et al. 2010. Climate variability, oceanography, bowhead whale distribution, and Iñupiat subsistence whaling near Barrow, Alaska. Arctic 63:179–194.[CrossRef]
Banašek-Richter, C., et al. 2009. Complexity in quantitative food webs. Ecology 90:1470–1477. [Abstract]
Bentzen, T. W., E. H. Follmann, S. C. Amstrup, G. S. York, M. J. Wooller, and T. M. O’Hara. 2007. Variation in winter diet of southern Beaufort Sea polar bears inferred from stable isotope analysis. Canadian Journal of Zoology 85:596–608.[CrossRef]
Bromaghin, J. F., T. L. McDonald, and S. C. Amstrup. 2013. Plausible combinations: An improved method to evaluate the covariate structure of Cormack-Jolly-Seber mark-recapture models. Open Journal of Ecology 3:11–22.[CrossRef]
Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach. Springer, New York, New York, USA.
Carothers, A. D. 1979. Quantifying unequal catchability and its effect on survival estimates in an actual population. Journal of Animal Ecology 48:863–869.[CrossRef]
Carroll, S. S., L. Horstmann-Dehn, and B. L. Norcross. 2013. Diet history of ice seals using stable isotope ratios in claw growth bands. Canadian Journal of Zoology 91:191–202.[CrossRef]
Chernick, M. R. 1999. Bootstrap methods: A practitioner’s guide. John Wiley and Sons, New York, New York, USA.
Cherry, S. G., A. E. Derocher, K. A. Hobson, I. Stirling, and G. W. Thiemann. 2011. Quantifying dietary pathways of proteins and lipids to tissues of a marine predator. Journal of Applied Ecology 48:373–381.[CrossRef]
Cherry, S. G., A. E. Derocher, I. Stirling, and E. S. Richardson. 2009. Fasting physiology of polar bears in relation to environmental change and breeding behavior in the Beaufort Sea. Polar Biology 32:383–391.[CrossRef]
Derocher, A. E., et al. 2013. Rapid ecosystem change and polar bear conservation. Conservation Letters 6:368–375.
Derocher, A. E., M. Andersen, Ø. Wiig, A. Aars, E. Hansen, and M. Biuw. 2011. Sea ice and polar bear den ecology at Hopen Island, Svalbard. Marine Ecology Progress Series 441:273–279.[CrossRef]
Doherty, P., G. White, and K. Burnham. 2012. Comparison of model building and selection strategies. Journal of Ornithology 152:317–323.[CrossRef]
Dunton, K. H., T. Weingartner, and E. C. Carmack. 2006. The nearshore western Beaufort Sea ecosystem: Circulation and importance of terrestrial carbon in arctic coastal food webs. Progress in Oceanography 71:362–378.[CrossRef]
Durner, G. M., et al. 2009. Predicting 21st-century polar bear habitat distribution from global climate models. Ecological Monographs 79:25–58. [Abstract]
Durner, G. M., J. P. Whiteman, H. J. Harlow, S. C. Amstrup, E. V. Regehr, and M. Ben-David. 2011. Consequences of long-distance swimming and travel over deep-water pack ice for a female polar bear during a year of extreme sea ice retreat. Polar Biology 34:975–984.[CrossRef]
Ellis, A. M., and E. Post. 2004. Population response to climate change: Linear vs. non-linear modeling approaches. BMC Ecology 4:2.[CrossRef]
Fischbach, A. S., S. C. Amstrup, and D. C. Douglas. 2007. Landward and eastward shift of Alaskan polar bear denning associated with recent sea ice changes. Polar Biology 30:1395–1405.[CrossRef]
Frost, K. J., L. F. Lowry, G. Pendleton, and H. R. Nute. 2004. Factors affecting the observed densities of ringed seals, Phoca hispida, in the Alaskan Beaufort Sea, 1996–99. Arctic 57:115–128.[CrossRef]
Giles, K. A., S. W. Laxon, A. L. Ridout, D. J. Wingham, and S. Bacon. 2012. Western Arctic Ocean freshwater storage increased by wind-driven spin-up of the Beaufort Gyre. Nature Geoscience 5:194–197.[CrossRef]
Grebmeier, J. M., L. W. Cooper, H. M. Feder, and B. I. Sirenko. 2006. Ecosystem dynamics of the pacific-influenced northern Bering and Chukchi seas in the Amerasian Arctic. Progress in Oceanography 71:331–361.[CrossRef]
Harwood, L. A., T. G. Smith, H. Melling, J. Alikamik, and M. C. S. Kingsley. 2012. Ringed seals and sea ice in Canada’s Western Arctic: harvest-based monitoring 1992–2011. Arctic 65:377–390.
Herreman, J., and E. Peacock. 2013. Polar bear use of a persistent food subsidy: Insights from non-invasive genetic sampling in Alaska. Ursus 24:148–163.[CrossRef]
Hezel, P. J., X. Zhang, C. M. Bitz, B. P. Kelly, and F. Massonnet. 2012. Projected decline in spring snow depth on Arctic sea ice caused by progressively later autumn open ocean freeze-up this century. Geophysical Research Letters 39:17505.[CrossRef]
Hinzman, L. D., C. J. Deal, A. D. McGuire, S. H. Mernild, I. V. Polyakov, and J. E. Walsh. 2013. Trajectory of the Arctic as an integrated system. Ecological Applications 23:1837–1868. [Abstract]
Horner, R., and G. C. Schrader. 1982. Relative contributions of ice algae, phytoplankton, and benthic microalgae to primary production in nearshore regions of the Beaufort Sea. Arctic 35:485–503.[CrossRef]
Horvitz, D. G., and D. J. Thompson. 1952. A generalization of sampling without replacement from a finite universe. Journal of the American Statistical Association 47:663–685.[CrossRef]
Hosmer, D. W., and S. Lemeshow. 2000. Applied logistic regression. Second edition. John Wiley and Sons, New York, New York, USA.
Hunter, C. M., H. Caswell, M. C. Runge, E. V. Regehr, S. C. Amstrup, and I. Stirling. 2010. Climate change threatens polar bear populations: A stochastic demographic analysis. Ecology 91:2883–2897. [Abstract]
IPCC [Intergovernmental Panel on Climate Change]. 2007. Climate change 2007: Synthesis report. R. K. Pachauri and A. Reisinger, editors. IPCC, Geneva, Switzerland.
IPCC [Intergovernmental Panel on Climate Change]. 2013. Climate change 2013: The physical science basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. T. F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, and P. M. Midgley, editors. Cambridge University Press, Cambridge, UK.
Iverson, S. A., H. G. Gilchrist, P. A. Smith, A. J. Gaston, and M. R. Forbes. 2014. Longer ice-free seasons increase the risk of nest depredation by polar bears for colonial breeding birds in the Canadian Arctic. Proceedings of the Royal Society B 281:20133128.[CrossRef]
Jakobsson, M., R. Macnab, L. Mayer, R. Anderson, M. Edwards, J. Hatzky, H. W. Schenke, and P. Johnson. 2008. An improved bathymetric portrayal of the Arctic Ocean: Implications for ocean modeling and geological, geophysical and oceanographic analyses. Geophysical Research Letters 35:L07602.[CrossRef]
Jarvela, L. E., and L. K. Thorsteinson. 1999. The epipelagic fish community of Beaufort Sea coastal waters, Alaska. Arctic 52:80–94.[CrossRef]
Jensen, S. K., J. Aars, C. Lydersen, K. M. Kovacs, and K. Åsbakk. 2010. The prevalence of Toxoplasma gondii in polar bears and their marine mammal prey: Evidence for a marine transmission pathway? Polar Biology 33:599–606.[CrossRef]
Ji, R., M. Jin, and Ø. Varpe. 2013. Sea ice phenology and timing of primary production pulses in the Arctic Ocean. Global Change Biology 19:734–741.[CrossRef]
Kendall, W. L., R. J. Barker, G. C. White, M. S. Lindberg, C. A. Langtimm, and C. L. Peñaloza. 2013. Combining dead recovery, auxiliary observations and robust design data to estimate demographic parameters from marked individuals. Methods in Ecology and Evolution 4:828–835.[CrossRef]
Kendall, W. L., J. D. Nichols, and J. E. Hines. 1997. Estimating temporary emigration using capture-recapture data with Pollock’s robust design. Ecology 78:563–578. [Abstract]
Langtimm, C. A. 2009. Non-random temporary emigration and the robust design: Conditions for bias at the end of a time series. Pages 745–761 in D. L. Thomson, E. G. Cooch, and M. J. Conroy, editors. Modeling demographic processes in marked populations. Springer, New York, New York, USA.
Lasternas, S., and S. Agustí. 2010. Phytoplankton community structure during the record Arctic ice-melting of summer 2007. Polar Biology 33:1709–1717.[CrossRef]
Lebreton, J. D., K. P. Burnham, J. Clobert, and D. R. Anderson. 1992. Modeling survival and testing biological hypotheses using marked animals: a unified approach with case studies. Ecological Monographs 62:67–118. [Abstract]
Logerwell, E., K. Rand, and T. Weingartner. 2011. Oceanographic characteristics of the habitat of benthic fish and invertebrates in the Beaufort Sea. Polar Biology 34:1783–1796.[CrossRef]
Luque, S. P., and S. H. Ferguson. 2009. Ecosystem regime shifts have not affected growth and survivorship of eastern Beaufort Sea belugas. Oecologia 160:367–378.[CrossRef]
Mahoney, A. R., H. Eicken, L. H. Shapiro, R. Gens, T. Heinrichs, F. J. Meyer, and A. Graves Gaylord. 2012. Mapping and characterization of recurring spring leads and landfast ice in the Beaufort and Chukchi Seas. OCS Study BOEM 2012-067. U.S. Department of the Interior, Bureau of Ocean Energy Management, Alaska Region, Anchorage, Alaska, USA. http://boemre-new.gina.alaska.edu/reports/BOEM_2012-067_Final.pdf
Markus, T., J. C. Stroeve, and J. Miller. 2009. Recent changes in arctic sea ice melt onset, freezeup, and melt season length. Journal of Geophysical Research 114:12024.[CrossRef]
Maslanik, J. A., C. Fowler, J. Stroeve, S. Drobot, J. Zwally, D. Yi, and W. Emery. 2007. A younger, thinner Arctic ice cover: Increased potential for rapid, extensive sea-ice loss. Geophysical Research Letters 34:L24501.[CrossRef]
Maslanik, J., J. Stroeve, C. Fowler, and W. Emery. 2011. Distribution and trends in arctic sea ice age through spring 2011. Geophysical Research Letters 38:L13502.[CrossRef]
McDonald, T. L., and S. C. Amstrup. 2001. Estimation of population size using open capture–recapture models. Journal of Agricultural, Biological, and Environmental Statistics 6:206–220.[CrossRef]
Melling, H., R. Francois, P. G. Myers, W. Perrie, A. Rochon, and R. L. Taylor. 2012. The Arctic Ocean: a Canadian perspective from IPY. Climatic Change 115:89–113.[CrossRef]
Monnett, C., and J. S. Gleason. 2006. Observations of mortality associated with extended open-water swimming by polar bears in the Alaskan Beaufort Sea. Polar Biology 29:681–687.[CrossRef]
Nahrgang, J., Ã. Varpe, E. Korshunova, S. Murzina, I. G. Hallanger, I. Vieweg, and J. Berge. 2014. Gender specific reproductive strategies of an Arctic key species (Boreogadus saida) and implications of climate change. PLoS ONE 9:e98452.[CrossRef]
Nichols, J. D. 2005. Modern open-population capture-recapture models. Pages 88–123 in S. C. Amstrup, T. L. McDonald, and B. F. J. Manly, editors. Handbook of capture-recapture analysis. Princeton University Press, Princeton, New Jersey, USA.
Nicolaus, M., C. Katlein, J. Maslanik, and S. Hendricks. 2012. Changes in Arctic sea ice result in increasing light transmittance and absorption. Geophysical Research Letters 39:L24501.[CrossRef]
Obbard, M. E., G. W. Thiemann, E. Peacock, and T. D. BeBruyn, editors. 2010. Polar bears: Proceedings of the 15th working meeting of the IUCN/SSC Polar Bear Specialist Group, Copenhagen, Denmark, 29 June–3 July 2009. IUCN, Gland, Switzerland and Cambridge, UK.
Overland, J. E., and M. Wang. 2013. When will the summer Arctic be nearly ice free? Geophysical Research Letters 40:2097–2101.[CrossRef]
Pagano, A. M., G. M. Durner, S. C. Amstrup, K. S. Simac, and G. S. York. 2012. Long-distance swimming by polar bears (Ursus maritimus) of the southern Beaufort Sea during years of extensive open water. Canadian Journal of Zoology 90:663–676.[CrossRef]
Palmer, M., K. Arrigo, C. J. Mundy, J. Ehn, M. Gosselin, D. Barber, J. Martin, E. Alou, S. Roy, and J. Tremblay. 2011. Spatial and temporal variation of photosynthetic parameters in natural phytoplankton assemblages in the Beaufort Sea, Canadian Arctic. Polar Biology 34:1915–1928.[CrossRef]
Parker-Stetter, S., J. Horne, and T. Weingartner. 2011. Distribution of polar cod and age-0 fish in the U.S. Beaufort Sea. Polar Biology 34:1543–1557.[CrossRef]
Peacock, E., M. K. Taylor, J. Laake, and I. Stirling. 2013. Population ecology of polar bears in Davis Strait, Canada and Greenland. Journal of Wildlife Management 77:463–476.[CrossRef]
Peñaloza, C. L., W. L. Kendall, and C. A. Langtimm. 2014. Reducing bias in survival under nonrandom temporary emigration. Ecological Applications 24:1155–1166. [Abstract]
Perovich, D. K., and C. Polashenski. 2012. Albedo evolution of seasonal Arctic sea ice. Geophysical Research Letters 39:L08501.[CrossRef]
Pickart, R. S., L. M. Schulze, G. W. K. Moore, M. A. Charette, K. R. Arrigo, G. van Dijken, and S. L. Danielson. 2013a. Long-term trends of upwelling and impacts on primary productivity in the Alaskan Beaufort Sea. Deep Sea Research Part I: Oceanographic Research Papers 79:106–121.[CrossRef]
Pickart, R. S., M. A. Spall, and J. T. Mathis. 2013b. Dynamics of upwelling in the Alaskan Beaufort Sea and associated shelf-basin fluxes. Deep Sea Research Part I: Oceanographic Research Papers 76:35–51.[CrossRef]
Pierrehumbert, R. T. 2011. Infrared radiation and planetary temperature. Physics Today 64:33–38.[CrossRef]
Pilfold, N. W., A. E. Derocher, I. Stirling, and E. Richardson. 2014. Polar bear predatory behaviour reveals seascape distribution of ringed seal lairs. Population Ecology 56:129–138.[CrossRef]
Post, E., U. S. Bhatt, C. M. Bitz, J. F. Brodie, T. L. Fulton, M. Hebblewhite, J. Kerby, S. J. Kutz, I. Stirling, and D. A. Walker. 2013. Ecological consequences of sea-ice decline. Science 341:519–524.[CrossRef]
Proshutinsky, A., R. H. Bourke, and F. A. McLaughlin. 2002. The role of the Beaufort Gyre in Arctic climate variability: Seasonal to decadal climate scales. Geophysical Research Letters 29:2100.[CrossRef]
R Development Core Team. 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/
Ramsay, M. A., and I. Stirling. 1988. Reproductive biology and ecology of female polar bears (Ursus maritimus). Journal of Zoology 214:601–634.[CrossRef]
Regehr, E. V., S. C. Amstrup, and I. Stirling. 2006. Polar bear population status in the southern Beaufort Sea. Open-File Report 2006-1337. U.S. Geological Survey, Reston, Virginia, USA.
Regehr, E. V., C. M. Hunter, H. Caswell, S. C. Amstrup, and I. Stirling. 2010. Survival and breeding of polar bears in the southern Beaufort Sea in relation to sea ice. Journal of Animal Ecology 79:117–127.[CrossRef]
Regehr, E. V., N. J. Lunn, S. C. Amstrup, and I. Stirling. 2007. Effects of earlier sea ice breakup on survival and population size of polar bears in western Hudson Bay. Journal of Wildlife Management 71:2673–2683.[CrossRef]
Rode, K. D., S. C. Amstrup, and E. V. Regehr. 2010. Reduced body size and cub recruitment in polar bears associated with sea ice decline. Ecological Applications 20:768–782. [Abstract]
Rode, K., E. Peacock, M. Taylor, I. Stirling, E. Born, K. Laidre, and Ø. Wiig. 2012. A tale of two polar bear populations: Ice habitat, harvest, and body condition. Population Ecology 54:3–18.[CrossRef]
Rode, K. D., E. V. Regehr, D. C. Douglas, G. Durner, A. E. Derocher, G. W. Thiemann, and S. M. Budge. 2014. Variation in the response of an arctic top predator experiencing habitat loss: Feeding and reproductive ecology of two polar bear populations. Global Change Biology 20:76–88.[CrossRef]
Sakar, S. K., and H. Midi. 2010. Importance of assessing the model adequacy of binary logistic regression. Journal of Applied Sciences 10:479–486.[CrossRef]
Sakshaug, E. 2004. Primary and secondary production in the Arctic seas. Pages 57–81 in R. Stein and R. W. Macdonald, editors. Organic carbon cycle in the Arctic Ocean. Springer, New York, New York, USA.
Schaub, M., O. Gimenez, B. R. Schmidt, and R. Pradel. 2004. Estimating survival and temporary emigration in the multistate capture-recapture framework. Ecology 85:2107–2113. [Abstract]
Schliebe, S., K. D. Rode, J. S. Gleason, J. Wilder, K. Proffitt, T. J. Evans, and S. Miller. 2008. Effects of sea ice extent and food availability on spatial and temporal distribution of polar bears during the fall open-water period in the Southern Beaufort Sea. Polar Biology 31:999–1010.[CrossRef]
Schulze, L. M., and R. S. Pickart. 2012. Seasonal variation of upwelling in the Alaskan Beaufort Sea: Impact of sea ice cover. Journal of Geophysical Research: Oceans 117:C06022.[CrossRef]
Sepp, M., and J. Jaagus. 2011. Changes in the activity and tracks of Arctic cyclones. Climatic Change 105:577–595.[CrossRef]
Serreze, M., and J. Francis. 2006. The arctic amplification debate. Climatic Change 76:241–264.[CrossRef]
Serreze, M. C., A. S. McLaren, and R. G. Barry. 1989. Seasonal variations of sea ice motion in the Transpolar Drift Stream. Geophysical Research Letters 16:811–814.[CrossRef]
Sigler, M. F., M. Renner, S. L. Danielson, L. B. Eisner, R. R. Lauth, K. J. Kuletz, E. A. Logerwell, and Hunt, G. L. Jr. 2011. Fluxes, fins, and feathers: Relationships among the Bering, Chukchi, and Beaufort seas in a time of climate change. Oceanography 24:250–265.[CrossRef]
Sonne, C. 2010. Health effects from long-range transported contaminants in Arctic top predators: An integrated review based on studies of polar bears and relevant model species. Environment International 36:461–491.[CrossRef]
Smith, L. C., and S. R. Stephenson. 2013. New trans-arctic shipping routes navigable by midcentury. Proceedings of the National Academy of Sciences USA 110:E1191–E1195.[CrossRef]
Smith, T. G. 1987. The ringed seal, Phoca hispida, of the Canadian western Arctic. Canadian Bulletin of Fisheries and Aquatic Sciences 216. Fisheries and Oceans Canada, Ottawa, Ontario, Canada.
Spreen, G., R. Kwok, and D. Menemenlis. 2011. Trends in Arctic sea ice drift and role of wind forcing: 1992–2009. Geophysical Research Letters 38:L19501.[CrossRef]
Stammerjohn, S., R. Massom, D. Rind, and D. Martinson. 2012. Regions of rapid sea ice change: An inter-hemispheric seasonal comparison. Geophysical Research Letters 39:L06501.[CrossRef]
Stirling, I. 1974. Midsummer observations on the behavior of wild polar bears (Ursus maritimus). Canadian Journal of Zoology 52:1191–1198.[CrossRef]
Stirling, I. 1997. The importance of polynyas, ice edges, and leads to marine mammals and birds. Journal of Marine Systems 10:9–21.[CrossRef]
Stirling, I. 2002. Polar bears and seals in the eastern Beaufort Sea and Amundsen Gulf: A synthesis of population trends and ecological relationships over three decades. Arctic 55(Supplement 1):59–76.
Stirling, I., and A. E. Derocher. 2012. Effects of climate warming on polar bears: A review of the evidence. Global Change Biology 18:2694–2706.[CrossRef]
Stirling, I., M. C. S. Kingsley, and W. Calvert. 1982. The distribution and abundance of seals in the eastern Beaufort Sea, 1974–79. Occasional Paper 47. Canadian Wildlife Service, Ottawa, Ontario, Canada.
Stirling, I., N. J. Lunn, and J. Iacozza. 1999. Long-term trends in the population ecology of polar bears in western Hudson Bay in relation to climatic change. Arctic 52:294–306.[CrossRef]
Stirling, I., T. L. McDonald, E. S. Richardson, E. V. Regehr, and S. C. Amstrup. 2011. Polar bear population status in the northern Beaufort Sea, Canada, 1971–2006. Ecological Applications 21:859–876. [Abstract]
Stirling, I., A. M. Pearson, and F. L. Bunnell. 1976. Population ecology studies of polar and grizzly bears in northern Canada. Transactions of the 41st North American Wildlife Conference 41:421–430.
Stirling, I., E. Richardson, G. W. Thiemann, and A. E. Derocher. 2008. Unusual predation attempts of polar bears on ringed seals in the southern Beaufort Sea: Possible significance of changing spring ice conditions. Arctic 61:14–22.
Stirling, I., and T. G. Smith. 2004. Implications of warm temperatures and an unusual rain event for the survival of ringed seals on the coast of southeastern Baffin Island. Arctic 57:59–67.
Stirling, I., C. Spencer, and D. Andriashek. 1989. Immobilization of polar bears (Ursus maritimus) with Telazol® in the Canadian Arctic. Journal of Wildlife Diseases 25:159–168.[CrossRef]
Stoklosa, J., and R. M. Huggins. 2012. A robust P-spline approach to closed population capture–recapture models with time dependence and heterogeneity. Computational Statistics and Data Analysis 56:408–417.[CrossRef]
Stroeve, J., M. M. Holland, W. Meier, T. Scambos, and M. Serreze. 2007. Arctic sea ice decline: Faster than forecast. Geophysical Research Letters 34:L09501.[CrossRef]
Stroeve, J. C., T. Markus, L. Boisvery, J. Miller, and A. Barrett. 2014. Changes in Arctic melt season and implications for sea ice loss. Geophysical Research Letters 41:1216–1225.[CrossRef]
Stroeve, J., M. Serreze, M. Holland, J. Kay, J. Malanik, and A. Barrett. 2012. The Arctic’s rapidly shrinking sea ice cover: A research synthesis. Climatic Change 110:1005–1027.[CrossRef]
Taylor, M. K., J. Laake, P. D. McLoughlin, H. D. Cluff, and F. Messier. 2008. Mark-recapture and stochastic population models for polar bears of the high Arctic. Arctic 61:143–152.
Thiemann, G. W., S. J. Iverson, and I. Stirling. 2008. Polar bear diets and arctic marine food webs: insights from fatty acid analysis. Ecological Monographs 78:591–613. [Abstract]
Thomson, J., and W. E. Rogers. 2014. Swell and sea in the emerging Arctic Ocean. Geophysical Research Letters 41:3136–3140.[CrossRef]
Thorson, J. T., S. Zhou, A. E. Punt, and A. D. M. Smith. 2013. A stepwise-selected spline approximation to time-varying parameters, with application to occupancy modelling. Methods in Ecology and Evolution 4:123–132.[CrossRef]
Tremblay, J.-É., et al. 2011. Climate forcing multiplies biological productivity in the coastal Arctic Ocean. Geophysical Research Letters 38:L18604.[CrossRef]
Tyler, N. J. C., M. C. Forchhammer, and N. A. Øritsland. 2008. Nonlinear effects of climate and density in the dynamics of a fluctuating population of reindeer. Ecology 89:1675–1686. [Abstract]
Vongraven, D., et al. 2012. A circumpolar monitoring framework for polar bears. Ursus 23(Supplement 2):1–16.[CrossRef]
White, G. C., W. L. Kendall, and R. J. Barker. 2006. Multistate survival models and their extensions in program MARK. Journal of Wildlife Management 70:1521–1529.[CrossRef]
Wiig, Ø. 1998. Survival and reproductive rates for polar bears at Svalbard. Ursus 10:25–32.
Zhang, J., Y. H. Spitz, M. Steele, C. Ashjian, R. Campbell, L. Berline, and P. Matrai. 2010. Modeling the impact of declining sea ice on the Arctic marine planktonic ecosystem. Journal of Geophysical Research 115:C10015.[CrossRef]
