Sea ice resource selection models for polar bears in the Barents Sea subpopulation

Abstract

The extent, thickness and age of Arctic sea ice has dramatically declined since the late 1990s, and these trends are predicted to continue. Exploring the habitat use of sea‐ice‐dependent species can help us understand which resources they use and how their distribution responds to a changing environment. The goal of this study was to develop predictive models of the habitat use of an Arctic apex predator. Polar bear Ursus maritimus habitat use in the Barents Sea subpopulation was modelled with seasonal resource selection functions (RSFs) using satellite‐linked telemetry data from 294 collars deployed on female polar bears between 1991 and 2015. Polar bears selected habitat in the Marginal Ice Zone, with a preference for intermediate sea ice concentrations (40–80%). They spent most time in areas with relatively short travel distances to 15 or 75% ice concentration, and during spring and autumn they exhibited a preference for sea ice areas over the continental shelf or over shallower bathymetry). Predictions of the distribution of polar bears in the Barents Sea area can be made for specific sea ice scenarios using these models. Two such predictive distribution maps based on the autumn seasonal model were made and validated against two independent polar bear survey datasets collected in August 2004 and August 2015. The distribution of optimal polar bear habitat has shifted strongly northwards in all seasons of the year during the 25 yr study period.

Hierarchical Bayesian model reveals the distributional shifts of Arctic marine mammals

“…cording to the estimated covariate responses and model comparison, the effect of relative density of seals outweighs the effect of ice concentration on the log relative density of polar bears. Even though the effects of the relative density of seals and ice concentration cannot be fully disentangled, as seals are heavily dependent on ice, the results indicate that the relative density of seals has a clear positive effect on the relative density of polar bears.When excluding the seal covariate, polar bears' response to ice concentration becomes hump-shaped peaking around 70% of ice cover (see Appendix S1), which is similar to the responses found byDurner et al (2009) andLone et al (2018). Thus polar bears and seals follow a similar hump-shaped relationship to ice concentration.…”

“…() and Lone et al. (). Thus polar bears and seals follow a similar hump‐shaped relationship to ice concentration.…”

“…This has also been recognized in more diverse marine systems (Reisinger et al, 2018). Assessments of polar bears' distribution could be improved by linking areal estimates of seals' density to a RSF (Resource Selection Function) of polar bears (Durner et al, 2009;Lone et al, 2018;Wilson et al, 2014). Reviewers of the manuscript were interested in the reasons to leave bathymetry out from the models of relative densities of polar bears and seals.…”

“…Despite earlier estimates of the population sizes of polar bears (3,200 ± 1,100 individuals [Matishov et al., ]), ringed seals (90,000–150,000 [Kelly et al., ]) and walruses (<500 [Born, Gjertz, & Reeves, ]) in the Kara Sea, there are no rigorous distribution estimates for these species. Previous studies on polar bear distributions (Durner et al., ; Lone, Merkel, Lydersen, Kovacs, and Aars (); Matishov et al., ; Wilson, Horne, Rode, Regehr, & Durner, ; Wilson, Regehr et al., ), have not provided a spatiotemporally explicit fine scale prediction about the density of polar bears, whereas estimates for densities of seals and walruses are mostly missing in the entire Arctic.…”

Comparing sea ice habitat fragmentation metrics using integrated step selection analysis

“…Each grid cell, or pixel, had a value from 0 to 200 defining ice cover, which was divided by 2, resulting in values from 0–100 to reflect percentage of sea ice coverage. Within each pixel, ice cover ≥60% was optimal habitat and <60% was considered suboptimal habitat based on other studies where preferred polar bear habitat was defined (Cherry, Derocher, & Lunn, 2016; Cherry et al, 2013; Laidre et al, 2015; Lone, Merkel, Lydersen, Kovacs, & Aars, 2018; Mauritzen et al, 2003; Pilfold, Derocher, & Richardson, 2014; Sahanatien & Derocher, 2012). We used ArcGIS version 10.5.1 (Environmental Systems Research Institute, 2017) to classify the daily ice cover as either optimal or suboptimal habitat.…”