Generation of fraction images from AVHRR data using linear mixing model

The coarse resolution satellite data have been widely used for regional and global studies as they provide high temporal frequency. The information contained in the coarse resolution pixels are mostly mixture of several components. The extraction of information contained in a pixel find its role in Geosphere-Biosphere context. The present study address the utility of constrained least square model applied to coarse spatial resolution data from NOAA-AVHRR for generating fraction images of vegetation, soil and water/shade. The red and near-infrared channels have been used to run the constrained least square model to generate fraction images. The derived fraction images are related to normalised difference vegetation index (NDVI) for model validation. The results suggest that vegetation fraction components are strongly correlated with NDVI values (r²=0.98). The soil fractions (r²=?0.84) and water/shade fractions (4²=?0.97) are negatively correlated with NDVI. The relationship between the fraction images and NDVI show the potential of the model in deriving sub-pixel component information using coarse resolution data.