“…When the used relocation occurred outside of the predetermined radius, we assumed the radius of availability was the distance between the previous and current used relocation (Durner et al ). We recorded the following habitat variables at each used and available cell: 1) land cover category (Table ), derived from Landsat™ imagery, aerial photography, and ground‐collected GPS points; 2) percent tree canopy cover, derived from the 2011 United States Forest Service National Land Cover Database (http://www.mrlc.gov/nlcd11_data.php, accessed 8 Jan 2015), with canopy cover fixed at 0% within areas classified as a forage opening; 3) years since prescribed fire, with a value of 70 years outside all burn units reflecting the last major wildfire; 4) aspect, derived from a digital elevation model (DEM) provided by the Missouri Spatial Data Information Service (http://www.msdis.missouri.edu/data/dem/index.html, accessed 13 Dec 2013); 5) slope, derived from the DEM; 6) Euclidean distance to the nearest wooded (forest or woodland)‐open area edge, derived from the land cover category map within the Geospatial Modeling Environment (Geospatial Modeling Environment Version 0.7.2.0, http://www.spatialecology.com/gme, accessed 17 May 2013); 7) interspersion and juxtaposition index (IJI), calculated using FRAGSTATS (FRAGSTATS v. 4, http://www.umass.edu/landeco/research/fragstats/fragstats.html, accessed 20 Nov 2015) and a moving window of radius 1.922 km reflecting the mean radius of available habitat across all individuals; 8) Euclidean distance to paved road, derived from TIGER 2010 road data from the United States Census Bureau, Missouri Department of Transportation road data, and logging road data from MDC; 9) Euclidean distance to public gravel road; 10) Euclidean distance to 2‐track road; and 11) road density, in units of km paved and public gravel roads per 95 ha (a circle with radius 550 m), reflecting the distance at which elk respond to human disturbance from roads (Benkobi et al , Rumble et al ).…”