Multi-scale spatial models reveal the interplay of weather and habitat effects on cold-water fish

Abstract

Climate change impacts cold-water species by altering thermal and flow regimes. These impacts may vary spatially, depending on habitat quality. We developed a Bayesian hierarchical model to investigate the effects of seasonal weather patterns on brook trout (Salvelinus fontinalis) populations occupying suitable and marginal habitats separated by the Eastern Continental Divide in North Carolina, USA. We specified weather-related coefficients separately for the Interior and Atlantic slope populations. We also specified multi-scale random effects to account for heterogeneity due to nested stream habitats. Our cohesive framework accommodates depletion sampling data and allows us to infer variation in capture probabilities between two different sampling protocols. As expected, we found the Atlantic slope (i.e., marginal habitats) population sizes were smaller on average than those of the Interior (i.e., suitable habitats), and the dominant drivers differed for either side of the Divide. Specifically, juvenile fish on the Atlantic slope were most adversely affected by high winter flows, and the Interior juveniles were most adversely affected by high spring flows. We also identified higher spatial variation than temporal variation in trout density conditioned on the covariates, where the primary source of spatial heterogeneity was likely meso-habitat characteristics of stream segments. Our method allowed us to partition uncertainty between population dynamics and imperfect detection, and facilitated scale-specific inference useful to develop management strategies for a species of conservation concern. This approach is widely applicable to other lotic species and ecosystems occupying dendritic habitats.