Abstract
The classification of forest type by aerial photo interpretation is usually based on color and texture of the image, and also from knowledge of the relationship between vegetation and topography. In this study, use of topographic factors for object-based forest type classification with high resolution satellite IKONOS imagery was examined to improve efficiency and accuracy of making maps of forest types. Two methods, one using spectral features only and the other combining spectral features and topographic factors, were applied to classify the forest in the eastern part of Kyoto City. Using topographic factors (distance from ridge line and distance from thalweg) with spectral features improved the accuracy of classification significantly at the 1% significance level measured by the Kappa coefficient. In the region where the difference in elevation is small, the spatial pattern of forest distribution is affected not by the temperature change with elevation but by the difference in soil moisture condition according to topographic position (ridge or valley). Consequently, the use of topographic factors would improve the accuracy of forest type classification.
