Studies on sub pixel classification of satellite images and spectra of landcover components for improved estimation of the capacity of reservoirs

Abstract

Periodical capacity surveys of multi-purpose reservoirs help in assessing the rate of sedimentation and reduction in the storage capacity. Hydrographic surveys are the common, conventional methods wherein, the water-spread area estimated by these surveys is input into the Trapezoidal formula to estimate the storage capacity of a reservoir. Such methods, however, are cumbersome, time consuming and expensive. As an alternative, multi-date satellite remote sensing provides the water-spread area of a reservoir at different water levels in a cost- and time- effective manner, which can then be input into the Trapezoidal formula. Approaches to water-spread area estimation from satellite image data, such as the maximum likelihood and minimum distance to mean classification, adopt the per-pixel based methodology and assign a pixel to a single land cover type. One of the limitations of these approaches is that the pixels representing the periphery of the reservoir, containing a mixture of water, soil and vegetation, are classi ed entirely as water, to result in inaccurate estimates of the water-spread area. This thesis is an attempt to improve the accuracy in estimating the water-spread area using a sub-pixel or linear mixture model (LMM) approach applied to multi-temporal satellite images of multi-purpose reservoirs. The water-spread areas extracted using the per-pixel and sub-pixel approaches from IRS-1C, 1D and IRS-P6 satellite image data were used to quantify the capacity of three reservoirs, namely Nagarjunasagar and Singoor reservoirs located in Andhra Pradesh state, and the Vaigai reservoir situated in Tamilnadu state of India. Spectral end-members, which are an important input for the sub-pixel approach of water-spread area estimation, were selected accurately using the Pixel Purity Index and Scatter Plot methods The role of the spectral response of soil and vegetation present in the peripheral pixels of a multi-purpose reservoir has also been addressed by carrying out spectro-radiometric surveys of the soil,