RAINFALL-RUNOFF ESTIMATION USING RS/GIS TECHNIQUES AND NEURAL NETWORK

Abstract

Water management has gained widespread importance in modem societies. Rainfall is the most significant of all the sources of water. Better management of runoff water can improve the availability of water. There are multiple methods for predicting potential runoff in a given area. Traditional methods of runoff estimation, such as rational formula require many parameters and cumbersome calculations, making it difficult to conduct runoff estimation. On the other hand innovative approaches like neural network technique can be experimented, which has emerged as powerful tool for a number of applications including prediction and estimation models. Scan basin is an arid area with few water resources, leading it to rely on rainfall for irrigation and other needs. There is great need of improving the existing infrastructure of water storage, so is to try innovative methods of runoff estimation to enhance the predictability in terms of space and time. Study has used applied GIS techniques on rainfall, runoff, land cover and soil data to compare the results of rational formula method with the results of neural network technique. Neural network has predicted the results within MSE of± 6% compared with original values, with fewer parameters and computational complexities. Study has concluded that provided that sufficient data is available, neural network technique can yield better results with fewer parameters, cost and calculation.