Predictive modeling and mapping sage grouse ( Centrocercus urophasianus) nesting habitat using Maximum Entropy and a long-term dataset from Southern Oregon

Abstract

Predictive modeling and mapping based on the quantitative relationships between a species and the biophysical features (predictor variables) of the ecosystem in which it occurs can provide fundamental information for developing sustainable resource management policies for species and ecosystems. To create management strategies with the goal of sustaining a species such as sage grouse ( Centrocercus urophasianus), whose distribution throughout North America has declined by approximately 50%, land management agencies need to know what attributes of the range they now inhabit will keep populations sustainable and which attributes attract disproportionate levels of use within a home range. The objectives of this study were to 1) quantify the relationships between sage grouse nest-site locations and a set of associated biophysical attributes using Maximum Entropy, 2) find the best subset of predictor variables that explain the data adequately, 3) create quantitative sage grouse distribution maps representing the relative likelihood of nest-site habitat based on those relationships, and 3) evaluate the implications of the results for future management of sage grouse. Nest-site location data from 1995 to 2003 were collected as part of a long-term research program on sage grouse reproductive ecology at Hart Mountain National Antelope Refuge. Two types of models were created: 1) with a set of predictor variables derived from digital elevation models, a field-validated vegetation classification, and UTM coordinates and 2) with the same predictors and UTM coordinates excluded. East UTM emerged as the most important predictor variable in the first type of model followed by the vegetation classification which was the most important predictor in the second type of model. The average training gain from ten modeling runs using all presence records and randomized background points was used to select the best subset of predictors. A predictive map of sage grouse nest-site habitat created from the application of the model to the study area showed strong overlap between model predictions and nest-site locations.