Using Aerial Imagery to Characterize Redband Trout Habitat in a Remote Desert Landscape

Abstract

AbstractRemote sensing products, including aerial imagery, can be used to quantify characteristics of watersheds and stream corridors that often predict the distribution and abundance of aquatic species. We conducted a supervised, object‐oriented classification of imagery from the National Agricultural Imagery Program to develop a high‐resolution (1‐m) land cover data set with four cover classes, emphasizing accurate characterization of woody riparian vegetation along stream corridors in northern Nevada and southwestern Idaho. The overall classification accuracy was 76%, and producer's accuracy (reflecting false positives) and user's accuracy (reflecting false negatives) for the woody vegetation class were 84% and 70%, respectively. Using logistic and quantile regression models in a model‐selection framework, we found woody vegetation to be positively associated with the occurrence and density of Redband Trout Oncorhynchus mykiss gairdneri. In addition, occurrence probabilities and densities were highest at mean August stream temperatures (predicted from a stream temperature model) ranging from 13°C to 17°C. When considered together with stream temperature, percent woody vegetation typically predicted Redband Trout occurrence and density better than most field‐measured instream and riparian habitat variables in northern Nevada. Our study highlights how free high‐resolution imagery can be used to characterize woody riparian vegetation and Redband Trout habitat across a large and remote desert landscape that can be difficult to access for field surveys. It also suggests that imagery from the National Agricultural Imagery Program may have wider application in identifying stream habitat restoration opportunities, where land and water uses have negatively impacted woody riparian vegetation in desert regions of the interior western United States.