Modeling and mapping fish abundance across wadeable streams of Illinois, USA, based on landscape-level environmental variables

Abstract

To effectively conserve and restore stream ecosystems, we need to better understand the distribution and abundance of individual fish species in relation to natural environments and anthropological stressors. In this study, we modeled the abundance of 97 fish species in small wadeable streams of Illinois, USA, based on random forests regression and landscape-level environmental variables. Model R2 values for intermediately common species were higher than for common species, but highly variable among rare ones. Models for 50 species reached R2 of 0.2–0.70 and were tested with a separate set of samples and applied to unsampled wadeable reaches to show the population hotspots of each species across the state. Furthermore, we evaluated the importance of individual environmental variables to a given fish species as well as the directional responses of each species to top 10 key predictors. Climate and land use were the best predictors for most species, followed by topography, geology, and soil permeability. Spatial connection of a stream also was associated with a large number of species. These findings improved our understanding of the relationships between fish species and landscape environments. The distribution maps could guide resource management, restoration, and monitoring of stream fish assemblages.