Projected Polar Bear Sea Ice Habitat in the Canadian Arctic Archipelago

Global climate simulations contributed to the Coupled Model Intercomparison Project Phase 5 (CMIP5) are too coarse to effectively resolve the narrow channels of the CAA. Although numerically challenging, one solution is to dynamically downscale a global simulation onto a finer grid using a regional climate model. A comparison between over 30 different CMIP5 models led us to select one simulation from the Geophysical Fluid Dynamics Laboratory Coupled Physical Model (GFDL-CM3) driven by radiative scenario RCP8.5 to pilot the regional model (available from www.gfdl.noaa.gov/coupled-physical-model-cm3). This simulation includes a realistic spatial distribution of sea ice extent and thickness and simulates the trend in observed minimum sea ice extent during the observational record (1979–2013). This pilot simulation was dynamically downscaled using the ice-ocean Massachusetts Institute of Technology General Circulation Model (MITgcm) simulation in regional mode over the Arctic at a resolution of 18km (available from http://mitgcm.org). The 3-hourly atmospheric forcing fields from GFDL-CM3 were bias-corrected at the monthly scale using differences for variables (x,y,z) or ratios for variables (u,t,g) between the Japanese 25 year Reysis (JRA25) and GFDL-CM3. These biases were calculated over the 2005–2011 period, arguably too short to compute climatological means, to ooth the transition from the JRA25 driven MITgcm simulation to the GFDL-CM3 driven simulation occurring at the start of 2012. We choose a period of 7 years to calculate the biases between the two forcing datasets because of the transitory nature of the climate in the early 21^st century with large trends in many of the Arctic climate forcing fields. MITgcm parameters were provided by Nguyen and ocean boundary conditions taken from the Estimating the Circulation and Climate Change of the Ocean Phase 2 (ECCO2) experiment . The MITgcm is run with time steps of 2-hours.

Our model projection is based on the RCP8.5 scenario, which estimates the global average radiative forcing at 8.5 W/m² by 2100, and mean global temperature changes of ∼3.5°C in 2071–2100 when compared with the historical period of 1961–1990 , and represents a worst-case scenario. We compared the seasonal changes in the sea ice cycle between past (1992–2005), near future (2040–2050), and future (2080–2090) by comparing average SIC in each period by month (Figure 2). Population size, survival, and reproduction of polar bears have all been associated with the changes in the seasonal ice cycle, in particular with changes to the ice-free period , , . We assume that effects on polar bears within the CAA will be comparable to those observed in other populations.

To study how habitat could change, we classified each pixel within the CAA as multiyear ice, annual ice, or ice-free. The classification was made based on the SIC of the pixel location over a given year . Multiyear ice, which is ice that persists through the height of the melt season (typically March – September), is found when SIC ≥15% year-round. Should SIC dip to <15% before freeze-up begins, but be ≥15% at least once during the year prior, the pixel is classified as annual ice. Ice-free areas are defined as <15% SIC year-round. Polar bears typically avoid or abandon sea ice when concentrations drop below 30–50% although the rate of loss is also important , . The cutoff of 15% we used is conservative because bears will occupy habitat with as little as 15% SIC , but higher concentrations are more closely associated with habitat use and successful predation , .