Logistic Regression Models for Resolving Primary Stream-Bearing Length or Area for Coastal Cutthroat and Steelhead Trout at the Landscape and Reach Scale

Abstract

Steelhead and cutthroat are closely related, difficult to differentiate and commonly co-occur in larger coastal watersheds. Provided cutthroat spawning and early rearing occurs in small streams, it is unknown which reaches are primarily used by each species at the landscape level.  To quantify species dominance at the reach level, I completed meta-analyses using parr-sized trout captures and relative catch frequency (counts per 100) of each from 636 streams or reaches.  A further 561 streams located in the USA were used for model validation. Various GIS map layers and physical habitat measurements were assembled with reach-occurrence data to perform a logistic regression predicting species dominance. Variables included stream width, channel width, stream order, watershed area, unit runoff, ecoregion placement, and long-term mean annual discharge. Watershed area stratified by ecoregion or mean annual discharge (mad) was the single best predictor of cutthroat and steelhead occurrence with a classification rate of 98% correct for the entire species range.  Primary watersheds with an ocean terminus and those not routinely documented as steelhead fishery sites were dominated by cutthroat. Cutthroat dominated in reaches where mad≤800 L•s^-1 and steelhead dominated in reaches where mad>800 L•s^-1. The models have practical application for determining stream-bearing length and area primarily used by each species.