Using Aerial Imagery to Predict Occurrence and Density of Redband Trout in a Remote, Desert Landscape

Remotely-sensed data used to characterize watersheds and streams are often useful predictors of aquatic species.  We evaluated aerial imagery as a tool for predicting redband trout (Oncorhynchus mykiss gairdneri) in northern Nevada and southwestern Idaho.  We conducted a supervised, object-oriented classification of National Agricultural Imagery Program (NAIP) imagery to develop a 1-m resolution land cover dataset focusing on accurate characterization of woody riparian vegetation.  The land cover classification had a producer’s error (false negatives) of 16% for woody vegetation, whereas user’s error (false positives) was 30%.  We used logistic and quantile regression models to show that percent woody vegetation in a 5-m stream buffer as classified from NAIP imagery, in addition to mean August temperature from a stream temperature model, were better predictors of redband trout occurrence and density than field-measured instream and riparian habitat.  Quantile (90^th) regression models also showed redband trout densities were higher when there was more woody riparian vegetation, but only when mean August stream temperatures were near 15°C.  Our study shows how free, high resolution imagery can be used to characterize redband trout habitat in remote desert streams, as well as identify place-based opportunities to restore woody riparian vegetation.