DROUGHT FORECASTING USING ARTIFICIAL NEURAL NETWORK (ANN) FOR VIDARBHA REGION

Abstract

A prolonged shortage in water supply, whether precipitation or ground water leads to drought. So, to reduce the famine risk and to ensure economic activity to continue to the possible extent during drought its forecasting is very important. Among the various parameters used, Standardized Precipitation Index (SPI) is most used due to its standardized form and variable timescale. Nowadays drought forecasting using machine learning model has upper hand over traditional methods due to its adaptability with nonlinear characters. In this study, various machine learning models are used to predict the SPI of different timescales like SPI-3, SPI-6, SPI-9 and SPI-12 of ‘Vidarbha’, the drought prone region in Maharashtra State of India, for the period of 1960 to 2020. Along with SPI some other parameters like minimum and maximum temperature, sunshine hours, wind speed are also given as an input which shows significant increase in accuracy. It produced the satisfactory results, and it could be used as a rapid tool for decision making.