“…We extracted attributes of eight variables for each field plot as well as 10,000 points generated randomly across the ecoregion. The variables included elevation (30 m spatial resolution), slope (30 m), aspect (cosine transformed, 30 m), and mean minimum January temperature (1950–2005), mean maximum July temperature (1950–2005), mean wet (November–April), and dry season (May–October) precipitation (1950–2005) from climate data with a native resolution of 860 m (ClimSurf; Alvarez, Guo, Klinger, Li, & Doherty, ), and mean peak normalized difference vegetation index (NDVI, 2000–2010, 250 m native resolution; seasons based on Tagestad et al, ). We used the correlation matrix of the eight variables for the PCA and the Kaiser‐Guttman criterion for retention of PCA axes (McGarigal, Cushman, & Stafford, ).…”