Hierarchical analysis of relationships between brook trout (Salvelinus fontinalis) density and stream habitat features

Abstract

We used hierarchical linear regression to examine relationships between brook trout (Salvelinus fontinalis) density and habitat features nested at three levels: sections within reaches, reaches within streams, and streams within a basin. Brook trout density and environmental variables were quantified at 600 stream sections distributed among 120 reaches and 22 streams in the Cascapedia River basin, Quebec, Canada. Decomposition of variance showed that variation in density among streams was small relative to that among sections or reaches and not statistically significant. Density was influenced by habitat variables at both the section (current velocity, woody debris, cover) and reach (subbasin area, height increment at flood, valley width) levels. A cross-level interaction between current velocity and subbasin area pointed to a "contextual" effect: density showed stronger decline with current velocity in larger subbasins than in smaller subbasins. This result suggests that there was no single "best scale" for examining fish–environment relationships. Accounting for contextual effects by use of hierarchical models can enhance our understanding of how habitat features influence fish densities at multiple spatial scales.