Search for: All records

This study compares alternative implementations of species distribution models (SDMs) for quantifying static and dynamic patterns in marine habitat use, with a case study focusing on juvenile salmon in the eastern Bering Sea. We compare the performance of two prevalent SDM frameworks—generalized additive models (GAMs) and vector autoregressive spatio-temporal (VAST) models—in predicting juvenile salmon distributions and assessing interannual variation in habitat utilization. The two SDM frameworks produced similar spatial predictions but performed differently in tests of within-sample and out-of-sample predictive power. Our findings indicate that VAST models may provide more precise estimates of distribution compared to GAMs. Maps of predicted juvenile salmon distributions showed highest salmon densities in habitats within the 50 m isobath of the continental shelf, underscoring the importance of these coastal areas, although among-species differences were evident. Model performance results suggested evidence for spatial variation in juvenile salmon species’ distributions through time. Our findings suggest that an SDM approach can be effective for estimating static and dynamic juvenile salmon distributions, and for providing insights that are useful in spatial fisheries management contexts. 

Abstract Coastal ecosystems in Alaska are undergoing rapid change due to warming and glacier recession. We used a natural gradient of glacierized to non-glacierized watersheds (0–60% glacier coverage) in two regions along the Gulf of Alaska—Kachemak Bay and Lynn Canal—to evaluate relationships between local environmental conditions and estuarine fish communities. Multivariate analyses of fish community data collected from five sites per region in 2019 showed that region accounted for the most variation in community composition, suggesting that local effects of watershed type were masked by regional-scale variables. Seasonal shifts in community composition were driven largely by the influx of juvenile Pacific salmon ( Oncorhynchus spp.) in late spring. Spatiotemporal differences among fish communities were partly explained by salinity and temperature, which accounted for 19.5% of the variation in community composition. We used a multi-year dataset from Lynn Canal (2014–2019) to examine patterns of mean length for two dominant species. Generalized additive mixed models indicated that Pacific staghorn sculpin ( Leptocottus armatus ) mean length varied along site-specific seasonal gradients, increasing gradually through the summer in the least glacially influenced sites and increasing rapidly to an asymptote of ~ 150 mm in the most glacially influenced sites. Starry flounder ( Platichthys stellatus ) mean length was more strongly related to environmental conditions, increasing with temperature and turbidity. Together, our findings suggest that community compositions of estuarine fishes show greater variation at the regional scale than the watershed scale, but species-specific variation in size distributions may indicate differences in habitat quality across watershed types within regions.