William N. Tardy, Ohio University, Ashley A. Wick, Drake University, John K. Maingi, Miami Univerisity, and Thomas O. Crist, Miami University.
Forest stand data are an essential ingredient in the development of management and conservation plans for forest resources. Estimates of forest stand attributes are obtained in labor-intensive and time-consuming forest inventories. Efforts to make forest inventories more efficient and reduce cost are increasingly incorporating remotely-sensed data coupled with image processing techniques. We determined whether forest stand attributes such as stem density, basal area, species richness, and Shannon’s diversity index were correlated with spectral data from Landsat TM. In addition, the study sought to establish whether forest community types identified through ecological classifications and ordinations were spectrally distinguishable using Landsat TM data. Only the green band (Landsat TM 2) was found to be significantly correlated with species diversity (r=0.559, p=0.002). The best correlations between basal area and spectral data were obtained with the Moisture Vegetation Index (r=0.526, p=0.004), while the best correlation between stand density and spectral data were obtained with the KT Greenness image (r=-0.645, p=0.0001).The Jefferies-Matusita (JM) distances calculated for the five forest community types defined indicated that most community types were spectrally distinct. Community types that could not be separated using Landsat TM data and image enhancements were Beech-Maple/Tulip Poplar, and Beech-Maple/Riparian forest. Although hyperspectral data may be necessary for greater separability of these stand types, image enhancements of Landsat TM predicted several properties that may be useful in rapid assessments of stand structure and tree diversity.
