Doctor of Philosophy (PhD)
Forestry, Wildlife, and Fisheries
Understanding habitat needs of animal populations is critical for their effective management. In recent years, technological advances have increased the range of methods available to examine habitat selection patterns. However, available habitat data are often either limited to small geographic areas or are of coarse resolution, resulting in a gap in data to model habitat selection at landscape scales. I explored a method of processing Landsat data, the at-satellite reflectance tasseled cap, to address this data gap using black bears in south central Louisiana as a case study. As I showed, this case was particularly instructive because these bears occupy two very different habitat matrices. I examined the information content of resource measures derived from tasseled caps and determined that they contain substantially more information than is represented in coarse habitat maps such as available from the USGS GAP program. Additionally, this process could be applied over large areas and time frames, during different times of the year, and across sensors to produce consistent results that avoid the need to categorize land cover/habitats. I used logistic regression and the information theoretic approach to examine: the spatial scale at which habitat measures were derived, model complexity, and the relative value of groups of derived habitat measures. I grouped derived habitat measures to examine the information content in: images captured in two seasons, measures based on mean and standard deviation filters, and combinations of tasseled cap functions. My work suggests that researchers should consider multiple summary statistics derived over a range of scales, use multi-temporal data, and use all three tasseled cap functions to derive habitat measures. I calculated resource selection functions (RSF) for black bears in south central Louisiana and examined model calibration and discrimination. Mahalanobis distance has been proposed as an alternative to RSF because it does not require delineation of available resources, although results from the two approaches have not been compared. In this study, habitat quality predictions from RFS models more accurately depicted bear habitat preference than those of Mahalanobis. I propose an alternative use of Mahalanobis distance to direct model extrapolation beyond the boundaries of modeled populations.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Wagner, Robert Owen, "Developing landscape-scaled habitat selection functions for forest wildlife from Landsat data: judging black bear habitat quality in Louisiana" (2003). LSU Doctoral Dissertations. 4044.
Richard M. Pace, III