- Many habitat selection studies have focused on the importance of spatiotemporal scales and sample size, yet often hidden within is a trade-off between using more animal locations versus more predictive covariates. Few have evaluated the outcome of choosing between these two different paths even though the trade-off can have significant impacts on the conclusions drawn from habitat modeling. We evaluated the covariate resolution versus relocation trade-off across multiple spatiotemporal scales by building habitat selection models using a data-mining approach with 22 years of VHF collar data from the Denali Caribou Herd. We asked whether caribou selected winter habitat based on forage resources after accounting for snow depth. Habitat selection models at three temporal scales (decadal, inter-annual and intra-annual) provide correlative evidence of active selection for winter forage. Based on model performance and predicted occupancy, we found that acquiring appropriate covariate layers is critical and likely even more important than using more animal locations. Based on analyses that utilized more covariates, caribou predominantly selected high graminoid cover, their second most important winter food, over areas with high lichen cover, their primary winter food, at all spatial scales implying a need to balance food quality and accessibility. However, habitat selection differed between temporal scales. Two years in a row, caribou switched to flatter areas with higher lichen cover and low shrub cover while intra-annual variation in habitat selection showed caribou congregated in areas of higher lichen cover and lower snow depths as winter progressed. We conclude that large patches of tussock tundra are high-use areas for wintering caribou in Denali at decadal scales while lichen-rich areas were used at inter-annual and intra-annual temporal scales. Lower snow levels in the future may allow heavier use of lichen woodlands but increased wildfire activity and shrub encroachment may counteract increased lichen availability in woodlands.