Seabirds are abundant, conspicuous members of marine ecosystems worldwide. Synthesis of distribution data compiled over time is required to address regional management issues and understand ecosystem change. Major challenges when estimating seabird densities at sea arise from variability in dispersion of the birds, sampling effort over time and space, and differences in bird detection rates associated with survey vessel type.
Using a novel approach for modeling seabirds at sea, we applied joint dynamic species distribution models (JDSDM) with a vector-autoregressive spatiotemporal framework to survey data collected over nearly five decades and archived in the North Pacific Pelagic Seabird Database. We produced monthly gridded density predictions and abundance estimates for 8 species groups (77% of all birds observed) within Cook Inlet, Alaska. JDSDMs included habitat covariates to inform density predictions in unsampled areas and accounted for changes in observed densities due to differing survey methods and decadal-scale variation in ocean conditions.
The best fit model provided a high level of explanatory power (86% of deviance explained). Abundance estimates were reasonably precise, and consistent with limited historical studies. Modeled densities identified seasonal variability in abundance with peak numbers of all species groups in July or August. Seabirds were largely absent from the study region in either fall (e.g., murrelets) or spring (e.g., puffins) months, or both periods (shearwaters).
Our results indicated that pelagic shearwaters (