Remote sensing of forest evergreen understory communities in the central Appalachian highlands.

Robert Chastain, University of Maryland Center for Environmental Science Appalachian Laboratory, chastain@al.umces.edu

Rhododendron maximum (R. maximum) and Kalmia latifolia (K. latifolia) are important components of Appalachian forest communities. These species are important because they: (1) have been shown to inhibit canopy tree regeneration when present in dense thickets; (2) have the potential for slowing N cycling, and therefore have water quality implications in the event of disturbance (they have been referred to as a keystone species in this regard); (3) may significantly influence carbon sequestration in these forests; (4) provide forage and refuge for a number of animal species; and (5) enhance the beauty of Appalachian forests. The accurate landscape and regional scale mapping of forest understory communities dominated by R. maximum and K. latifolia is valuable since their spatial extent and temporal dynamics are currently poorly understood, and this limits the ability to test hypotheses about the importance of dense evergreen understories on carbon and nitrogen cycling. Most existing image classifications exhibit high classification errors for deciduous forests with evergreen understories, as these are mapped either as mixed or evergreen forests (coniferous). Additionally, information about temporal dynamics in both leaf area (density of thickets) and spatial extent will facilitate broad scale studies of their role in the inhibition of canopy tree regeneration.

In this study, R. maximum and K. latifolia dominated understory communities were mapped in the Savage River and Green Ridge State Forests in western Maryland and the Buchanan State Forest in southwestern Pennsylvania using leaf-off Landsat TM and ETM+ imagery obtained between 1984 and 2000. Synthetic aperture radar (SAR), hyperspectral, and topographic data were applied to a series of classification methods in an attempt to improve classification accuracy over traditional approaches. In addition, patterns of changes in their spatial extent and leaf area were examined using change vector analysis (CVA). The CVA results were related to gypsy moth defoliation events, topographic variables, and precipitation trends over the 16-year time period of available image data to investigate the roles of disturbance and climatic trends in patterns of the vigor of these evergreen species. Classification techniques and limitations as well as results will be presented along with preliminary change detection findings.