Second order selection
To understand scale dependence in deer’s response to anthropogenic features during second order selection, we ran resource selection function (RSFs) models with scale optimization. We considered all of SI available for deer and drew nine random locations anywhere on SI for each used location. We then estimated the area of med-hi-dev within buffers (varying from 100m-800m at 100m increments, thus applying a multi-grain assessment (McGarigal et al. 2016)) surrounding each used and available location. We applied logistic regression generalized linear models (GLMs) to assess the dependence of used/available locations on the area of med-hi-dev. We selected the model with the minimum Akaike Information Criterion (AIC) score (Burnham & Anderson 2002) as the most predictive radii for individual deer.