Do Regional Habitat Models Outperform a Single-model Approach for Resource Selection at a Population Scale? A Case Study with Mule Deer in Nevada
Loading...
Authors
Vasquez, Joshua P.
Issue Date
2024
Type
Thesis
Language
Keywords
Great Basin , Habitat Selection , Habitat Suitability , Mule Deer , Odocoileus Hemionus , Random Forest
Alternative Title
Abstract
Effective conservation and management of ungulate species requires characterization of resource availability, selection, and use. Mule deer ( Odocoileus hemionus ) in the Great Basin are experiencing population reductions that are generally thought to be driven by declines in their preferred sagebrush-dominated habitats. In many parts of the Great Basin, sagebrush habitats are being rapidly lost or degraded due to wildfire, energy development, mining, anthropogenic development, climate change, and overgrazing. Robust models of resource selection by mule deer allow wildlife managers to make more informed decisions about habitat protection and permitting for development projects. We used machine learning (random forest) to evaluate patterns of habitat selection at the population level (second order) during summer by GPS-collared mule deer ( n = 630) across northern Nevada. We divided our study area into four ecologically distinct regions. We compared two alternative modeling approaches: a "region-specific" modeling approach, in which we fit separate resource selection models for each region (thereby accommodating distinct patterns of resource selection within each region), and an "all-regions'' modeling approach, in which we fit a single model of mule deer habitat selection for our study region (assuming similar resource patterns across all four regions). The all-regions model outperformed the regional models in cross-validation, indicating that patterns of selection of resources by mule deer were similar across northern Nevada. Our models indicated that mule deer favored summer habitats near perennial water sources, with higher cover of perennial grasses and forbs, less bare ground, and cooler temperatures than expected on the basis of available resources. Our research is important for mule deer conservation by comparing model performance of summer resource selection at the landscape level (second order), which highlights areas of conservation need from future anthropogenic alterations within the Great Basin.