Abundance estimation
In order for us to evaluate status of the SH subpopulation, it was important for us to be able to assess trend in abundance. To ensure that results would be directly comparable with the 2011/2012 survey, we replicated its design and effort with the exception of inland transects along the Québec coast, a small area adjacent to the Belcher Islands, and a small number of nearshore islands in James Bay that were added to the study design based on input from traditional knowledge holders. Our study design incorporated coastal contour transects, which recognized that the clumped distribution of polar bears along the coast of Hudson Bay in Ontario and Manitoba during the ice-free season (
Prevett and Kolenosky 1982;
Derocher and Stirling 1990a,
1990b;
Obbard and Walton 2004;
Towns et al. 2010) could bias results. Further, our study design also incorporated transects perpendicular to the coastline and following the density of bears, which decreases with distance away from the coast. Arranging the overland transects along this density gradient minimized estimate bias and improved precision (
Buckland et al. 2001). Both features of our study design account for the known distribution of bears during the ice-free season and are important to minimize bias and thereby improve accuracy and precision.
Modest analytical and design differences between the two studies did not significantly influence the results or our ability to evaluate population trend and status. In the previous survey,
Obbard et al. (2015) used mark–recapture distance sampling (MRDS;
Laake and Borchers 2004) rather than conventional or multiple covariate distance sampling (
Marques and Buckland 2003) analyses because preliminary analyses indicated that detection at distance 0 was significantly <1, thereby violating a fundamental assumption of distance sampling (
Buckland et al. 2001). MRDS integrates distance sampling and double-observer analytical methods, allowing for estimation of detection at distance 0 and subsequent correction of the abundance estimate. In this study, our detection of bears on and near the transect line approximated unity, meaning that we were able to generate an accurate estimate of abundance with simpler multiple covariate distance sampling models. Although right-truncation distance differed between the two studies (
Obbard et al. 2015: ∼2%; here: ∼5%), there was a negligible impact on results. The shoulders of the detection function (i.e., the sightings near the transect line) are most important for generating estimates of density and abundance (
Buckland et al. 2001). Truncating the farthest sightings improves model fit by eliminating the need to estimate spurious bumps in this sightings “tail”, with only modest impacts on estimate precision and negligible impacts on the point estimate (
Buckland et al. 2001). Small differences elsewhere, including more widespread sampling in Quebec in 2016 and the incorporation of techniques to estimate abundance for the James Bay islands in 2016 (rather than relying on raw counts) similarly had trivial impacts on the results. Both the previous and current surveys met the core assumptions of distance sampling including detection on the transect line; therefore, despite the small variations in the sampling and analytical methodology, the results from the two surveys are comparable thereby enabling us to assess trend and status.
Delineation of the study area to 60 km inland along the Ontario coast of Hudson Bay and James Bay as in the previous survey was based on available scientific and traditional knowledge of the distribution of bears during the ice-free season and of denning habitat. Although we did not survey the entire inland area of SH in Ontario, extending transects farther inland would have been very expensive for minimal returns. Truncating our transects at 60 km may have resulted in some slight negative bias in the abundance estimate, but all the evidence suggests that few bears are found that far inland in September, and if any were there, they would not comprise a large proportion of the subpopulation.
Traditional knowledge holders indicated that bears were occasionally observed inland in the Québec portion of the study area north of Pointe Louis XIV/Long Island at the junction of James Bay and Hudson Bay (e.g.,
Laforest et al. 2018;
NMRWB 2018). Based on this information, we adapted our study design slightly from the 2012 survey to provide better coverage of this region. However, despite our intensive survey (transects spaced at 6 km intervals), we observed no bears in the area. No doubt, bears occasionally occur here but their density must be very low, and their number would make only a small difference to the total abundance estimate.
Our results suggest that abundance in the SH subpopulation declined by 17% between 2011/2012 and 2016. It has been hypothesized that distribution of bears during the ice-free season along the shores of Ontario and Manitoba varies depending on where the last ice of the season persists resulting in an inverse correlation between counts from coastal strip surveys conducted at the same time in Ontario and Manitoba (
Prevett and Kolenosky 1982). This argument suggests that polar bears, especially males, may choose to remain on the ice until late in the season, and as a result, they may simply occupy coastal areas closest to the location of residual ice once the ice melts completely. Location of residual ice in any year would then influence distribution of bears. Hence, one explanation for the decline in abundance that we noted for SH may be that the distribution of bears shifted due to ice conditions in 2016 such that more bears came ashore within the boundary of WH. However, using a longer dataset,
Stirling et al. (2004) found no significant correlations (negative or positive) between annual counts in Ontario and those in Manitoba bringing this explanation into doubt. Furthermore, results of an aerial survey conducted in WH in 2016 at about the same time as our survey provided no evidence for an unusual number of bears in the eastern portion of the range of WH, in fact that survey suggested that subpopulation had also likely declined in abundance (
Dyck et al. 2017). Lastly, the pattern of ice ablation in Hudson Bay is affected by prevailing northwesterly winds and counter-clockwise ocean currents resulting in residual ice moving south and east and gathering north of the Ontario coast (
Hochheim et al. 2011). Bears remain on the ice as it drifts south and east during the melt season and eventually come to shore in Manitoba or Ontario (
Parks et al. 2006;
Middel 2014); there is no evidence of bears moving north as ice melts. Therefore, the decline in abundance in SH suggested by our results is unlikely to be due to a shift in distribution into WH or FB.
There are interesting parallels in trends between WH and SH. In both subpopulations changes in individual-level traits (e.g., declines in body condition; WH:
Stirling et al. 1999, SH:
Obbard et al. 2016) preceded changes in population-level traits (e.g., declines in survival and in abundance; WH:
Regehr et al. 2007;
Lunn et al. 2016, SH:
Obbard et al. 2007; this study). However, these changes in morphometry and demography appear to have started sooner in WH than in SH. Because of the pattern of ice ablation and formation in Hudson Bay, on an annual basis bears from the two subpopulations are faced with a similar pattern of ice duration, but it is out of synchrony by a few weeks. That is because the last sea ice usually remains off the northern Ontario coast in summer (
Etkin 1991;
Saucier et al. 2004); therefore, bears in SH leave the ice later in spring than bears in WH, but return to the ice later in the fall because ice forms off the Ontario coast later (
Hochheim and Barber 2010). On a long-term basis, changes in the duration of ice cover appear to have started sooner in the western portions of Hudson Bay where bears from WH mainly occur during the ice-covered season compared to the eastern portions of Hudson Bay where bears from SH mainly occur (
Gagnon and Gough 2005;
Hochheim et al. 2010;
Hochheim and Barber 2014). Thus, the two subpopulations appear to have followed similar trajectories in response to changes in duration of sea ice, though starting later in SH, and it seems likely that these trends will continue in the future.
Reproduction
The proportion of cubs in all observations was similar in the 2011 and 2016 surveys, but the proportion of yearlings dropped from 12% to 5% (sample size from the 2012 portion of the original survey is too small for comparison). We are not aware of any evidence of reproductive synchrony in polar bear populations that would infer large variation in size of the year class of cubs born. Therefore, the low proportion of yearlings in the 2016 survey suggests low survival of cubs born in 2015. Whether this represents an ongoing trend, or was the result of unusual conditions in spring and summer of 2015 is undetermined. However, breakup in eastern Hudson Bay was about two weeks later in 2015 than in 2010–2014 (
Andrews et al. 2018) suggesting that ice conditions should not have had a negative influence on cub survival in 2015. The proportion of yearlings in the 2016 SH survey was similar to that observed in the WH surveys of 2011 and 2016 (3%). In contrast, subpopulations with robust vital rates have proportions of yearlings in the 10%–12% range, such as FB and SH in 2011 (
Obbard et al. 2015).
Long-lived animals should maximize their reproductive output, but should not do so at a risk to their own survival. Because the proportion of cubs remained high in 2016, many adult females in SH are still producing litters, but they may be less successful in raising cubs to yearling age.
Obbard et al. (2016) suggested that because the rate of decline in body condition was less for cubs of the year in SH than for their mothers, by the late 2000s adult females were allocating a greater proportion of their reserves to lactation than females did in the mid-1980s. If this is true, but cub survival is still dropping, then adult females may be putting themselves at increased risk, especially in the context of ongoing declines in duration of sea ice that can be expected to have further negative effects on body condition. A more adaptive strategy for a long-lived species with delayed implantation such as polar bears (
Wimsatt 1974;
Ferguson et al. 1996) would be for females to forego reproductive bouts entirely when a reproductive bout is unlikely to be successful. Because the number of litters of cubs observed in 2016 in Ontario and Akimiski Island was similar to the number observed in 2011, females may not yet be adopting that life history strategy. Furthermore, the number of litters observed in SH in both 2011 and 2016 was about double the number observed in WH at the same time suggesting that a higher proportion of SH females continue to attempt to raise cubs successfully (given that population abundance in WH is similar to that of SH) (
Lunn et al. 2016;
Dyck et al. 2017). The low number of litters of yearlings observed in 2016 compared to the number of litters of cubs suggests that much cub mortality may be via loss of entire litters. If females lose litters due to high neonatal mortality, in the absence of lactational suppression of estrus (
Knott et al. 2017), they may become receptive again shortly after returning to the sea ice in spring, resulting in them having litters in consecutive years. This may be happening in SH and, in itself, would put additional energetic demands on adult females in this subpopulation for no increase in fitness. If such energetic costs are additive, they could ultimately have negative consequences for survival of adult females.