Estimation of Reservoir Capacity Using Remote Sensing Data – A Soft Classification Approach

Abstract

Sediments carried by the rivers are deposited in the reservoirs and cause several detrimental effects, which include loss of storage capacity, upstream aggradations, effect on water quality and damage or impairment of hydro-equipments. The deposition of sedimentation not only reduces the capacity but also the water-spread of the reservoir. Satellite data has long been in use to estimate the water-spread area at different elevations of a reservoir, which in turn can be used to quantify the capacity of the reservoir. This methodology to estimate the capacity of a reservoir using remote sensing data involves hard or per-pixel based classification to delineate the water-spread area at a particular elevation. One of the limitations of this approach is that the border pixels, representing soil class with moisture, are classified entirely as water pixels, thereby giving inaccurate estimate of the water-spread area. To estimate the water-spread area of Nagarjuna Sagar reservoir accurately, the sub-pixel or linear mixture model (LMM) approach has been adopted in this study. The sub-pixel approach uses a linear mixture model of the spectra to quantify the proportions of water and other classes present in each of the border pixels of the reservoir. IRS-1C and 1D satellite image data (24m) of eight optimal dates ranging from minimum draw down level (MDDL) to full reservoir level (FRL) were used to estimate the water-spread area of the reservoir. The extracted water-spread areas using per-pixel and sub-pixel approaches were in turn used to quantify the capacity of the Nagarjuna Sagar reservoir. The estimated capacity of the Nagarjuna Sagar reservoir using per-pixel and sub-pixel approaches was 8101.63 Mm3 and 8014.49 Mm3 respectively.