3. Results

There were 4346 captures from 1985 to 2018 of 2533 individual bears, with a mean of 127.8 bears (SE = 11.3) captured per year (Table S3). There were 1159 adult male, 540 solitary adult female, 807 adult female with offspring, 296 subadult male, 331 subadult female, 393 yearling, and 820 COY captures (Table S4).

3.1 Age/sex class energy patterns

Energy density declined significantly over time for solitary adult females (linear regression, P = 0.015) and declined non-significantly over time for the other classes, except adult males and adult females with offspring (linear regression, P > 0.05; Fig. 2). Storage energy declined significantly over time for solitary adult females and yearlings (linear regression, P = 0.001, 0.041 respectively) and declined non-significantly over time for the other classes, except adult females with offspring (linear regression, P > 0.05; Fig. 3).
Energy density was significantly lower at earlier sea ice breakup dates for adult males and subadult females (multiple linear regression, P = 0.020, 0.034; Fig. S1; Tables S5, S6) and non-significantly lower at earlier breakup dates for the other classes (multiple linear regression, P > 0.05; Fig. S1; Table S6). Storage energy was significantly lower at earlier sea ice breakup dates for adult males, subadult males, subadult females, and COY (multiple linear regression, P = 0.004, 0.014, 0.012, 0.029 respectively; Fig S2; Tables S7, S8) and non-significantly lower at earlier breakup dates for the other classes (multiple linear regression, P > 0.05; Fig. S2; Table S8). A longer lagged open water period was associated with significantly reduced storage energy and non-significantly lower energy density for solitary adult females (multiple linear regression, P = 0.028 and P = 0.074, respectively; Figs. S3, S4; Tables S6, S8).
Energy density was significantly different among classes (Kruskal-Wallis, χ2 = 958.35, df = 6, P < 0.001). Solitary adult females had significantly higher energy density than all other classes (Dunn’s test, P ≤ 0.05; Tables S4, S9). COY and adult females with offspring had significantly lower energy density than all other classes, while adult males, subadult males/females, and yearlings had intermediate energy density. Storage energy was also significantly different among classes (Kruskal-Wallis, χ2 = 3398.2, df = 6, P < 0.001). Adult males had significantly higher storage energy than all other classes, followed by solitary adult females (Dunn’s test, P ≤ 0.05; Tables S4, S10). Subadult males/females and adult females with offspring had intermediate storage energy. Yearlings and COY had significantly lower storage energy than all other classes.

3.2 Temporal dynamics of population energy

From 1985 to 2018, the total population energy density declined by 53% and population storage energy declined by 56% (linear regression, P < 0.001 and P < 0.001 respectively; Fig. 4). There was a significant positive correlation between yearly population abundance estimates and both population energy density and storage energy (Spearman’s correlation, coefficient = 0.69, P < 0.001 and coefficient = 0.68, P < 0.001 respectively).

3.3 Population energy and the environment

Sea ice breakup varied from 17 May to 10 July and occurred significantly earlier from 1985 to 2018, with mean breakup occurring 5.5 days/decade earlier (linear regression, P ≤ 0.05; Fig. S5). Sea ice freeze-up varied from 4 Nov to 7 Dec and occurred significantly later over time, with mean freeze-up occurring 4.3 days per decade later (linear regression, P ≤ 0.001; Fig. S5). The length of the open water period varied from 102 days to 166 days and significantly lengthened over time, with a mean increase of 9.9 days per decade (linear regression, P ≤ 0.001; Fig. S5).
The top ranked models for population energy density and storage energy included sea ice breakup and the lagged open water period, while AOw, NAOw, and their lagged effects were not included in the top models (Table S11). Total population energy density was significantly lower when sea ice breakup occurred earlier and the lagged open water period was longer (multiple linear regression, P < 0.001 and P = 0.001, respectively; Fig. 5, Table 1). The top multiple linear regression model predicted that at the earliest breakup (ordinal date 137) and 180 day lagged open water period, total population energy density would be 8303.0 MJ kg-1 (58% lower than the mean energy density value that was calculated in our study, 19944.8 MJ kg-1).
Similarly, total population storage energy was significantly lower when sea ice breakup occurred earlier and the lagged open water period was longer (multiple linear regression, P < 0.001 and P < 0.001 respectively; Fig. 5, Table 4). At the earliest breakup (ordinal date 137) and 180 day lagged open water period, population storage energy was predicted to be 838781 MJ (63% lower than our mean estimated storage energy, 2270218 MJ).