PARENT SESSION
Oral Session - Modeling Species Patterns Chair(s): Bossenbroek, Jonathan¹, ¹ University of Notre Dame, Notre Dame, IN
Friday, April 2, 2004 3:00 PM - 5:20 PM Apollo Room 5

Comparing the performance of high and low resolution satellite data. ^*VITALE, ABIGAIL C. ^1,2 and COOPER, ROBERT J. ¹, ¹ University of Georgia, Athens, GA, USA² Virginia Tech, Blacksburg, VA, USA

ABSTRACT- The utility of high-resolution satellite data in wildlife models has yet to be tested. Coarser resolutions of 20-m alone are often not sufficient to accurately predict the presence of a species across a landscape. The purpose of this study was to assess the utility of high-resolution (HR) satellite data. Specifically, whether increased data resolution improves species model predictions and whether such data can replace the use of field measurements and other variables commonly used in habitat models. Six bird species were grouped into either closed canopy (CC) or early successional (ES) habitat categories based on life history strategies and then modeled using presence/absence data in a mixed hierarchical model. Models for each species utilized a combination of field measurements, GIS and landcover variables at various scales. Landcover variables were derived from Ikonos 4-m and Landsat TM 30-m images (landcovers were comparable in time, space, and classification accuracy). Variables for models were first selected based upon species life history. Using stepwise selection, the top five models were chosen using AIC (within each resolution data type). Those models were then validated through an iterative method and compared to one another based on overall predictive mean (OPM). HR models performed better for ES species, while the 30-m and 4-m models were comparable for CC species. Most models containing combinations of all types of variables performed better than those containing only field measurements for CC species. HR models replaced some field measurements for ES models. Consideration of species, resources, and project goals will determine which satellite data type is suitable for your modeling needs.

KEY WORDS: satellite data, high-resolution , bird models, hierarchical models, logistic regression