Climate change, interspecific competition, and poleward vs. depth distribution shifts: Spatial analyses of the eastern Bering Sea snow and Tanner crab (Chionoecetes opilio and C. bairdi)

Abstract

Snow crab (Chionoecetes opilio), an arctic species, and Tanner crab (C. bairdi), a subarctic species, support large-scale industrial fisheries in the eastern Bering Sea and are spatially overlapping congeneric species at the opposite boundaries of their distributional ranges in the eastern North Pacific. Using summer bottom trawl survey data from 1982 to 2017, spatial analyses and geostatistical regressions were conducted for each species for three population categories (immature females, oldshell mature females, and oldshell large males) to explore intra and interspecies differences in habitat and spatial distributions as well as interspecific spatial co-occurrence patterns. Variability in depth distributions rather than poleward distributional shifts may be a Tanner crab response to ocean warming, the apparent converse of snow crab. Both temperature and depth have strong effects on the Tanner crab distributions and 2 °C is an approximate minimum temperature threshold for Tanner crab habitat. 2 °C is an important maximum temperature threshold for immature female snow crab but temperature effects are weaker for adult snow crab. Snow crab ontogenetic movement results in loose coupling between settlement and adult habitat and in significant spatial overlap between older snow crab and Tanner crab. The similarity in spatial distributions and habitat of older snow crab with juvenile Tanner crab habitat and the predatory nature of snow suggest a possible regulatory role of snow crab on the northern extent of spatial distributions of Tanner crab, along with other possible factors such as groundfish predation. Snow crab distributions will most probably continue to shift northwards with warming ocean temperatures but a poleward shift for Tanner crab appears more uncertain. The utilities of comparing spatial and non-spatial models and accounting for spatial confounding in spatial regression modeling are also addressed.