The use of remote sensing for characterizing forests in wildlife habitat modeling

Graduation date: 2015Spatially explicit maps of habitat relationships have proven to be valuable tools for conservation and management applications including evaluating how and which species may be impacted by large scale climate change, ongoing fragmentation of habitat, and local land-use practices. Studies have turned to remote sensing datasets as a way to characterize vegetation for the examination of habitat selection and for mapping realized relationships across the landscape. Although the use of remote sensing in wildlife studies has increased in recent years, the use of these datasets is still limited and some data sources and methods are yet to be explored. The overall goal of this dissertation was to look at the state of the wildlife ecology discipline in the use of geospatial data for habitat mapping, and to advance this area through the fusion of remote sensing tools for the mapping of previously difficult to characterize forest metrics for inclusion in avian cavity-nester habitat models. \ud Chapter 2 reviewed over 60 years of selected wildlife literature to examine the wildlife ecology disciple through historic trends and recent advances in the use of remote sensing for habitat characterization focusing on aspects of scale and the use of available technology. We discuss commonly used remote sensing data sources, point out recent advances in the use of geospatial data for characterizing forest wildlife habitat (the use of lidar data and the creation of spatially explicit habitat prediction maps), and provide future suggestions for increased utilization of available datasets (secondary lidar metrics and time series Landsat data). In chapters 3 and 4 we explored the use of remote sensing for characterizing forest components previously difficult to map across landscapes at scales relevant to local wildlife habitat selection. Chapter 3 found promise in the fusion of lidar structure and Landsat time series disturbance products in the modeling and mapping of post-fire snag and shrub distributions at fine scales and at size/cover thresholds relevant for habitat mapping applications for many wildlife species. The study was conducted within the 2003 B&B Fire Complex in central Oregon. Using 164 field calibration plots and remote sensing predictors, we modeled the presence/absence of snag classes (dbh ≥40cm, ≥50cm, and ≥75cm) and woody shrub cover resulting in 10m output predictive grid maps. Remote sensing predictors included various lidar structure and topography variables and Landsat time series products representing the pre-fire forest, disturbance magnitude, and current forest conditions. We were able to model and map all habitat metrics with acceptable predictive performance and low-moderate errors. The utility of these snag and shrub metrics for representing important nesting habitat features for a cavity-nesting species of conservation concern, the Lewis's Woodpecker (Melanerpes lewis), was demonstrated in Chapter 4. We were able to model nesting habitat with good accuracies according to multiple performance measures and then map realized relationships for this species of conservation concern in an identified source habitat type, providing a potential resource for local scale conservation and management efforts and adding to the regional knowledge of habitat selection for the Lewis's Woodpecker. To our knowledge, these chapters represent first attempts to fuse lidar and time series Landsat disturbance metrics in a post-fire landscape and for the mapping of snag and shrub distributions at scales relevant to avian cavity nesting habitat