Current and future habitat suitability of northern fur seals and overlap with the commercial walleye pollock fishery in the eastern Bering Sea

Table 2 Model sample sizes and performance metrics for northern fur seal habitat selection models

Model	n_seal	n_trip	n_year	AUC_CV	AUC_Time	AUC_Space
Continental shelf
Pribilof	331	1,272	21	0.88 ± 0.005	0.62 ± 0.02	0.51 ± 0.04
East	176	868	16	0.91 ± 0.004	0.76 ± 0.001	0.23–0.58
English Bay	12	39	3	0.86 ± 0.03	0.77 ± 0.07	0.44–0.71
Reef Point	65	199	12	0.74 ± 0.02	0.62 ± 0.02	0.51–0.79
North	49	123	7	0.95 ± 0.01	0.87 ± 0.02	0.28–0.97
South	18	43	6	0.96 ± 0.01	0.80 ± 0.07	0.25–0.91
Basin
Pribilof	148	363	16	0.81 ± 0.006	0.67 ± 0.02	0.71 ± 0.02
English Bay	5	10	3	0.74 ± 0.07	-	0.46–0.58
Reef Point	65	167	12	0.66 ± 0.02	0.53 ± 0.04	0.53–0.69
North	36	65	7	0.89 ± 0.01	0.83 ± 0.02	0.53–0.93
South	40	119	6	0.94 ± 0.01	0.86 ± 0.03	0.55–0.91

Models for each habitat type were run using all of the data (Pribilof models) as well as on subsets based on rookery complex. Sample sizes represent the number of fur seals (n_seal), number of foraging trips (n_trip), and number of years (n_year) within each model, averaged across the 10 replicate datasets with the associated standard deviation. Mean performance metrics (replicate average ± SD), evaluated using the area under the receiver operating curve (AUC), are shown for 10-fold cross-validation (AUC_CV), as well as temporal (AUC_Time) or spatial blocking (AUC_Space; Pribilof models only). Values for AUC_Space for complex-specific models represent the mean model performance in predicting to datasets from other complexes, presented as a range across complexes

ISSN: 2051-3933