Meteorological drought prediction using heuristic approaches based on effective drought index: a case study in Uttarakhand

Monitoring and prediction of drought using standardized metrics of rainfall are of great importance for sustainable planning and management of water resources on regional and global scales. In this research, heuristic approaches including co-active neuro fuzzy inference system (CANFIS), multi-layer perceptron neural network (MLPNN), and multiple linear regression (MLR) were used for prediction of meteorological drought based on Effective Drought Index (EDI) at 13 stations located in Uttarakhand State, India. The EDI was calculated using monthly rainfall time-series data, and the significant input variables (lags) for CANFIS, MLPNN, and MLR models were derived using autocorrelation and partial autocorrelation functions (ACF and PACF) at 5% significance level. The predicted values of EDI obtained by CANFIS, MLPNN, and MLR models were compared with the calculated values based on root mean squared error (RMSE), Nash-Sutcliffe efficiency (NSE), coefficient of correlation (COC), and Willmott index (WI). The visual interpretation was also made using line diagram, scatter plot, and Taylor diagram (TD). The evaluation of results revealed that the CANFIS and MLPNN models outperformed than the MLR models for meteorological drought prediction at study stations. Also, the results of this research can be utilized for the decision-making of remedial schemes to cope with meteorological drought in the study region.