Assessment of Drought Using Standardized Precipitation Index and Reconnaissance Drought Index and Forecasting by Artificial Neural Network

Drought is normal recurring feature of climate occurring due to less than average rainfall at a place during a given period of time, which consequently leads to short term water deficit and economic loss. In this study, the known index for drought assessment based on only precipitation data is Standardized Precipitation Index (SPI) and new index based on precipitation as well as potential evapotranspiration data Reconnaissance Drought Index (RDI) were applied for meteorological drought analysis at Banaskantha District. Long term monthly precipitation and potential evapotranspiration data from 1962 to 2001 were used. Analysis were performed on 3, 4, 5, 6, 9, and 12 month long data sets for both indices SPI and RDI. For finding out the indices DrinC software is used. From the result, the worst drought years were 1973-74, 1986-87. The SPI and RDI drought classifications were forecasted by Artificial Neural Network (ANN). The models were tested and checked using the Root Mean Square Error (RMSE), Coefficient correlation (r) and coefficient of determination (R2). According to SPI and RDI model analysis, it can be concluded that, for Banaskantha District, it is better to use SPI9 and RDI9 for forecasting compare to SPI6, SPI12, RDI6 and RDI 12models. The value of SPI9 and RDI9 for forecasting of drought as it gives least error and maximum value. R2 is 0.99 for training and 0.95 for checking of SPI9. And R2 is 0.98 for training and 0.99 for checking of RDI9. It is recommended to use LM algorithm with Feed Forward Back Propagation Network.