Linking regional and landscape scales for assessing biodiversity: a case study from Western Ghats

Abstract

A broad scale classification of the Western Ghats into different types of landscapes will be a useful first step towards biodiversity assessment. Such a classification, coupled with point sampling in different landscape types could allow for efficient extrapolation of information to arrive at an estimate of the status of biodiversity in the Western Ghats as a whole. This paper attempts such a broad-scale mapping (at a 1 : 1,000,000 scale) of the Western Ghats into different landscape types, using IRS IB imagery. For this purpose, we use the Normalized Difference Vegetation Index (NDVI) which reduces problems of scene-to-scene radiometric variability. This index is believed to be correlated to vegetation biomass, vigour, photosyn-thetic activity and leaf area index - and is potentially useful for classifying different types of vegetation. Our attempt is to classify at the scale of the landscapes, and not at the level of individual landscape elements (LSEs). Each landscape is a mosaic of different LSE types, and the relative distribution of these different types determines the composition of the landscape. The distribution of NDVI values was therefore taken as the basis for classification. For each 50 by 50 pixel unit constituting a 'superpixel', the four moments of distribution - mean, standard deviation, skew and kurtosis - were calculated. These superpixels were then clustered using unsupervised classification. Small patches were merged with the most dominant LSE type in the vicinity, to obtain a final set of 205 patches belonging to eleven types of landscapes. We interpret the distribution of these eleven types in terms of topography, climate, population, agriculture and vegetation cover.