Improving habitat-selection models for grizzly bears is important to the conservation and management of the species. Remote-sensing data can be used to derive vegetation indices from Landsat TM, a surrogate for habitat quality. This study compared and evaluated Tasseled Cap Analysis (TCA) 'Greenness' and NDVI based on grizzly bear habitat selection, derived from radio-telemetry, in central British Columbia. Four groups of habitat models were developed for the mountain and plateau portions of the study area. AIC was used to rank the models and a k-fold cross validation to evaluate these competing models. One model using TCS greenness and a 21x21-pixel buffer with data from mountainous regions had the best predictive ability. Bears in the mountains are more dependant on vegetation as a food source, which is reflected in their selection for areas of higher greenness compared to plateau bears which have more access to ungulates and have larger home range sizes.
