Path analysis of collective properties and habitat relationships of fish assemblages in coastal plain streams

Abstract

Collective properties of fish assemblages at 44 sites in low-gradient streams (orders 1–3) on the upper coastal plain of South Carolina, U.S.A., were highly variable among sites: richness, 2–17 species per site; density, 0.2–9.5 individuals∙m−2; mean individual mass, 0.4–22.2 g per individual; and biomass, 0.1–42.1 g∙m−2. Principal components analysis reduced 15 habitat parameters to four interpretable components that were used as orthogonal predictive variables in path analysis. Path analysis, with a directional model containing intermediate steps and multiple dependent variables, was used as an alternative to multiple regression. Two principal components, associated with velocity (and its covariates) and depth (and its covariates), were the primary correlates of biomass (R2 = 49%), richness (44%), mean size (36%), and density (20%), with cover variables contributing relatively little to any R2value. Velocity effects were negative and depth effects positive for all collective properties. Habitat effects on biomass were mediated by effects on individual fish size rather than numbers. Correlations of collective properties at sites sampled in 2 years were similar to those from the single year, habitat-based path model, implying that the path model was adequate and inclusive, and accurately reflected collective properties of these communities.