Using Zero-Inflated Models to Predict the Relative Distribution and Abundance of Roe Deer Over Very Large Spatial Scales

Abstract

In Norway, recovering populations of large carnivores commonly prey on roe deer (Capreolus capreolus). Understanding predator habitat use and ecology requires fine-scaled information on prey distribution and abundance. However, the massive spatial scales at which large carnivores use the landscape presents many practical and statistical challenges for developing functional prey distribution models. Pellet-count data from > 1000 km of transects gathered across southeastern Norway between 2005 and 2011 were used to derive a map of the relative prey abundance for roe deer. These data were modeled using zero-inflated hurdle models using both environmental and anthropogenic variables. Snow depth and agricultural fields were the most significant variables in explaining both presence and abundance. Internal k-cross validation of the model showed medium accuracy (Spearman's r = 0.35), whereas external evaluation carried out on the basis of independently collected snow-tracking data (Spearman's r = 0.37) and hunting statistics (Spearman's r = 0.88) showed higher accuracy. The map generated can facilitate both the study of broad scale processes linking predators and prey as well as roe deer management in southeastern Norway.