OOS 10-6 - Detecting southeastern pine decline in Red-cockaded Woodpecker habitat using hyperspectral remote sensing: A case study from Fort Benning, GA

Tuesday, August 4, 2009: 9:50 AM
Mesilla, Albuquerque Convention Center
Maria Santos, Museum of Vertebrate Zoology, University of California, Berkeley, Berkeley, CA and Susan Ustin, Land, Air and Water Resources, UC Davis, Center for Spatial Technologies and Remote Sensing, Davis, CA
Background/Question/Methods: Conifer forests worldwide are affected by disturbance that cause major ecosystem shifts, and the mixed-conifer forests of southeastern United States are not an exception. Southeastern mixed-conifer forests currently face decline due to the cumulative effect of root pathogens, herbivores and pests, even stand age distribution, drought, past and current land use, and site and soil quality. These old growth stands are home of many wildlife species including the ESA listed red-cockaded woodpecker (RCW) Picoides borealis (35 FR 16047, October 13, 1970), whose recovery plan includes promotion of habitat recovery and mitigation. Our project aimed at assessing tree stress using airborne hyperspectral imagery over Ft. Benning, GA; produce predictive distribution models of senescence using generalized linear models; and assess potential consequences of the anticipated tree senescence to RCW populations. Results/Conclusions: Our results show that species spectral properties are significantly different, and senescence detection can be attained with or without a priori species classification. The species classification was reasonable, and with greater discrimination of long-leaf pine (Pinus palustris) than either loblolly pine (Pinus taeda) or short leaf pine (Pinus echinata); senescence detection without a priori tree classification used similar spectral features as those involved in the classification itself. Burnt soil also accounted for spectral signature differences, and when accounted for in the senescence analysis it increased the separability between live and senescent trees. Species level detectability of senescence was possible using a combination of reflectance indexes derived from the hyperspectral imagery, with accuracy between 50-70%. Pro-active monitoring to predict pine decline in southeastern forests, provides a means to improve effective management of land-use in military installations in consortium with conservation of biodiversity.
Copyright © . All rights reserved.
Banner photo by Flickr user greg westfall.