Spatial Analysis of Land Cover at Various Spatial Scales for Identification of Optimum Pixel Resolution

要旨

The spatial resolution of satellite dataset is an essential aspect for understanding any analysis of land use/land cover. This paper presents an approach to identify an optimum pixel resolution for forest cover related analysis and highlights the effect of pixel size on representation of landscape heterogeneity. Two techniques have been evaluated here for this purpose i.e. Landscape Metrics and Geostatistical methods. The results show that Geostatistical approach is potentially a more useful method to detect and describe the heterogeneity in patterns of a landscape, while the spatial methods involving landscape metrics may prove biased to the number of classes or to the spatial complexity of the landscape patterns. The optimum pixel resolution thus derived is 5m (PAN + LISS), which as indicated by geostatistical methods exhibit a spatial variability comparable to higher resolution data (IKONOS). The present study also emphasizes the significance of scale in satellite data interpretation for understanding the vegetation cover on ground.