Landscape Ecology and Diversity Patterns in the Seasonal Tropics from Landsat TM Imagery

Abstract

We examine patterns of landscape diversity in the seasonal tropical forests of northeastern Guatemala with analyses of Landsat Thematic Mapper (TM) imagery. Our primary objective is to demonstrate that TM analyses, without extensive field data, provide valuable information on landscape diversity patterns to aid conservation and development plans. A principal components analyses indicates that the major source of variability at the scale of TM resolution (°30 m) is related to TM band 4 (TM4) radiance, and that the second and third most important sources are related to TM bands 5 and 7, respectively. Cluster analysis of TM4, TM5, and TM7 radiance produced six distinct land‐cover types, which follow a topographic gradient. Both TM radiance and the radiance coefficient of variation (CV) are significant in discriminating land‐cover types. The primary source of variability within upland forests is related to TM4, and the primary sources of variability within lowland swamps are related to TM5 and TM7. We infer from these results that green leaf biomass is the most important variable in discriminating between land‐cover types in the uplands, and that canopy closure and degree of senescence are the most important variables in the lowland swamps. We also examined patterns of landscape diversity reflected in three landscape indices: the number of land‐cover types (LCT), the Shannon‐Weaver index of landscape evenness (S‐W), and a topographic index (TI). The strongest correlations were found for TM4 radiance and TI (r = 0.94, n = 30, P < 0.0001) and S‐W (r = ‐0.72, n = 30, P < 0.0001), and between TM4 radiance CV and LCT (rs = 0.67, n = 30, P < 0.0001). Thus, TM4 provides a good measure of landscape diversity. We found a perfect nonparametric correlation (rs = 1, n = 6) between TM4 radiance CV within a land cover type and the type's TI rank, which probably reflects increasing canopy heterogeneity between lowland swamps and upland forests. Finally, we demonstrate the potential of TM‐based land‐cover maps and diversity indices in designing and monitoring nature reserves.