ABSTRACT:

Accurate estimates of aquatic species distribution and habitat extent are critical to effectively manage ecological objectives. Many types of habitat models exist, which meet different objectives and have disparate outputs. However, no standard methods compare predictive accuracy of different habitat model types. We compared three aquatic habitat models, which predicted native Bonneville Cutthroat Trout distribution in the Bear River Watershed (USA) at a monthly timestep. Models included an existing hydraulic-habitat model, an existing habitat threshold model, and a geospatial model developed for this study. Validation of environmental predictors used in all models reflected satisfactory to poor fit—no observed conditions were well represented by model estimates—a function of either outdated, incorrect, or over-generalized input data. Validation of habitat suitability predictions using Bonneville Cutthroat Trout presence data showed the habitat threshold model accurately classified 100% of fish presence observations in suitable habitat in all modeled months, though model performance was sensitive to performance criteria selection which favored models with greater precision. Overall, habitat predictions from the simple, generalizable habitat threshold model are most useful for incorporating ecological objectives into water management models, though additional information from the more precise geospatial method may be useful for maximizing native fish conservation efforts.