Spatiotemporal reconstruction of species ranges
We estimated annual species range edges from the trawl survey data using the spatial generalized linear mixed model implemented in the R package VAST (Thorson and Barnett 2017, Thorson 2019). This model was designed to estimate total abundance and spatial variation in density of species using spatially referenced biomass observations. We fit VAST to data that follow either stratified-random or fixed-station designs; in both cases, VAST predicted densities over a fixed spatial domain. This analysis enabled comparison across years even when survey methodologies were revised and across regions with distinct survey protocols. In addition, this approach controlled for differences in catchability, enabling us to combine the two historical West Coast surveys (Thorson et al. 2016a). The model structure is described in detail in the supplementary materials (Appendix 2).
Of the 209 species-region combinations, VAST models converged for 205 of them (72 of 74 in the Northeast, 53 of 54 on the West Coast, and 80 of 81 in the Eastern Bering Sea). This set included five species found in both the West Coast and the Eastern Bering Sea (Atheresthes stomias, Bathyraja interrupta, Clupea pallasii, Glyptocephalus zachirus, and Hippoglossoides elassodon ), and one found in both the Northeast and the West Coast (Alosa sapidissima ), so the total number of unique species was 199.