Accuracy Assessment and Normalisation of Water Spread Area Estimate from Multi-sensor Satellite Data

Abstract

Water spread area is one of the few parameters that can be easily derived using multispectral satellite data. However, due to limitation in land–water boundary pixel delineation and availability of area estimate from simultaneous images of different spatial resolutions, it poses serious challenges for accurate estimation of water spread area. The area estimated from data of different resolutions is not the same owing to many factors like shape and size of the water body and the efficiency of the approaches to delineate land–water boundary pixel. In addition to this, multiple coverage of same water body from sensors of different resolutions will lead to varying water spread area estimates. In order to assess the effect of these factors on water spread area estimate, a study was undertaken to assess the accuracy of water spread area estimate using the field-based measurement and to establish relationship for normalisation of water spread area estimate from different resolution sensors. In this study, water boundary was traced by collecting continuous measurements of location with the help of satellite-based positioning system. The study showed that water spread area estimated through linear spectral unmixing of images from LISS IV, LISS III, AWiFS and MSI sensors resulted in better estimate compared to the index-based method. The error in estimate was less than 5% in all the cases except for AWiFS-based estimate for the smaller water body. The study also concluded that either simple linear regression equation or multiple linear regression equation can be used to normalise the water spread area estimated from different sensors.