Spatial generality of predicted occurrence models of nesting habitat for the greater sage‐grouse

Abstract

We investigated generality of resource‐based models of predicted occurrence as part of an effort in spatial prioritization of habitat for conservation in shrub‐steppe landscapes of western North America. Using GPS telemetry data on the greater sage‐grouse (Centrocercus urophasianus), we investigated whether spatial models developed in particular regions of Wyoming, USA were generalizable to other portions of the state, whether generality versus region‐specificity were mediated by behavior state, and why models were either general or region‐specific (i.e., what factors constrained versus promoted generality). The behavior states we examined were nest‐site selection, and resource selection during nesting off‐bouts. We tested generality by developing models within the region at which location data were collected, applying models across regions (3 different hydrologic basins), using independent location data to validate, and subsequently analyzing a validation parameter (Spearman's rho) in a beta regression model. In a spatial application, we combined information on nest site selection and off‐bout occurrence to develop predictive occurrence maps that account for multiple biological processes influencing occurrence during the nesting period. Models of nest site occurrence showed considerable generality, but models of occurrence during off‐bouts did not. Habitat‐occurrence relationships associated with generality included inter‐region consistency in sagebrush coverage, and importance (selection/avoidance) of linear features (rights‐of‐way) and natural gas wells. Relationships associated with region‐specificity included importance of mesic habitat and bare ground. These findings suggest generality is mediated by behavior state, with certain behavior more generalizable than other behavior. Features of generalizable behavior included a relatively well‐defined niche and historical constraints (in this case, nest‐area fidelity); features of region‐specificity included influences on behavior that were more local/variable in space and time (i.e., functional responses). Implications for conservation are that, in the shrub‐steppe and perhaps other systems, a steadfast need for region‐specific data to parameterize occurrence models is not universal, with some behavior (i.e., nest‐site selection) amenable to spatial prediction across large areas based on relationships identified in the species elsewhere.