Patterns of sea ice drift and polar bear (Ursus maritimus) movement in Hudson Bay

Abstract

Sea ice habitats are highly dynamic, and ice drift may affect the energy expenditure of travelling animals. Several studies in the high Arctic have reported increased ice drift speeds, and consequently, polar bears Ursus maritimus in these areas expended more energy on counter-ice movement for station-keeping. However, little is known about the spatiotemporal dynamics of ice drift in Hudson Bay (HB) and its implications for the declining Western Hudson Bay (WH) polar bear subpopulation. Using sea ice drift data from 1987-2015 and polar bear satellite telemetry location data from 2004-2015, we examined trends in drift speeds in HB, polar bear movement relative to drift, and assessed annual and individual variation. In contrast to other areas of the Arctic, we did not find an increase in ice drift speed over the period examined. However, variability in ice drift speed increased over time, which suggests reduced habitat predictability. Polar bear movement direction was not strongly counter to ice drift in any month, and ice drift speed and direction had little effect on bear movement rates and, thus, energy expenditure. On an annual scale, we found individuals varied in their exposure and response to ice drift, which may contribute to variability in body condition. However, the lack of a long-term increase in ice drift speed suggests this is unlikely to be the main factor affecting the body condition decline observed in the WH subpopulation. Our results contrast findings in other subpopulations and demonstrate the need for subpopulation-specific research and risk evaluation.

Energy‐based step selection analysis: Modelling the energetic drivers of animal movement and habitat use

“…greater seal kill biomass at greater depths; Pilfold et al, 2014). Lastly, we included sea ice drift speed (drift) as a covariate, as it affects the costs of moving any given geographic distance (Durner et al, 2017; Klappstein et al, 2020). We fitted the SSF using the same implementation techniques as the ESF, but generated controls based on the observed movement of the polar bears: we fitted a gamma distribution to step lengths and a wrapped Cauchy distribution to turning angles, and used these to randomly sample 20 control locations for each observed (case) location (as in Forester et al, 2009).…”

“…Telemetry locations arise from a combination of active bear movement and passive displacement caused by ice drift. Therefore, we defined a step as the active bear movement between telemetry locations, corrected for ice drift followingKlappstein et al (2020), using drift data from the National Snow and Ice Data Center (Polar Pathfinder Daily 25 km EASE-Grid Sea Ice Motion Vectors;Tschudi et al, 2019). In the following, we used the GPS locations to evaluate environmental variables, whereas we used the tracks corrected for ice drift to measure movement speed.At each step, a bear can either be swimming or walking on sea ice, which have different energetic costs (e.g.…”

Drivers of polar bear behavior, and the possible effects of prey availability on foraging strategy

“…They exhibit philopatry to their summering grounds and compensate for sea ice motion in their navigation and for station keeping (Mauritzen et al, 2003;Durner et al, 2017;Klappstein et al, 2020). Polar bears rely both on visual and olfactory search to hunt sparsely distributed prey (Stirling, 1974;Smith, 1980), which may be influenced by presence of daylight (Togunov et al, 2017(Togunov et al, , 2018.…”

Characterising menotactic behaviours in movement data using hidden Markov models

“…Gaspar et al 2006;Klappstein et al 2020;Safi et al 2013). However, this type of correction does not account for the error often in the environmental data(Dohan & Maximenko 2010;Togunov et al 2020; Yonehara et al 2016). Our methods overcome some of these limitations by building on the robust framework of HMMs, which are relatively flexible to uncertainty in both the track and environmental data by distinguishing between latent state and state-dependent processes(McClintock et al 2012; Zucchini et al …”

“…et al 2020;Gaspar et al 2006;Klappstein et al 2020;Safi et al 2013). However, this type of correction does not account for the error often in the environmental data(Dohan & Maximenko 2010;Togunov et al 2020; Yonehara et al 2016).…”