Exploring the use of fine resolution nested ecological niche models to identify greater sage-grouse (Centrocercus urophasianus ) habitat and connectivity potential across a diverse landscape

摘要

Suitable habitat for greater sage-grouse (Centrocercus urophasianus ) has been greatly reduced over a relatively short ecological scale (1800s -- Present). This reduction of habitat has had a negative impact on the current distribution and connectivity of the species. There has been work to map sage-grouse distribution at small ecological extents with fine resolution, and at broad extents and coarse resolutions. There is a current need to identify sage-grouse habitat at a fine ecological scale across a broad extent. This information will help researchers and land managers to better understand spatial patterns and connectivity associated with sage-grouse habitat and the processes that drive them. I focused my dissertation on testing the feasibility of developing broad spatial extent and fine resolution predictive habitat models for sage-grouse nest and brooding habitats. By using fine resolution mapping, I was able to capture more subtle variation in potential habitat; by using a broad extent I was able to apply these findings at a landscape scale. I also proposed a method of using nested ecological models blended together to predict potential habitat. In order to best predict habitat potential, multiple modeling techniques were applied (nonparametric multiplicative regression, maximum entropy distribution, random forest and generalized additive model). These methods were used to create independent sagebrush presence and total vegetation cover models and these were combined to create sage-grouse habitat predictive models. The statistical strength of each model was tested (logbeta;, R2 and AUC) as well as their predictive ability (overall accuracy and RMSE ). The results of this work produced fine resolution (30m) models, predicted across a broad extent (Utah, 21.9 million ha). The overall accuracy for the final sagebrush model was 72%. The RMSE for the vegetation cover MODEL was between 6.6 and 7.6% cover. In addition to model creation, potential research and management applications for these models are discussed. These models will provide baseline habitat estimations that could be used for better understanding past distributions of sage-grouse and improving current and future management planning. Furthermore, these same techniques could be applied to other species across multiple spatial and temporal scales.