Stochastic model for drought forecasting for Bundelkhand region in Central India

Abstract

In the present study, standardized precipitation index (SPI) series at 3-month, 6-month, 9-month, 12-month and 24-month time scale has been used to assess the vulnerability of meteorological drought in the Bundelkhand region of Central India. SPI values revealed that the droughts in the region over the study period vary from moderately high to extremely high. Suitable linear stochastic model, viz. seasonal and non-seasonal autoregressive integrated moving average (ARIMA) developed to predict drought at different time scale. The best model was selected based on minimum Akaike Information Criterion (AIC) and Schwarz Bayesian Criterion (SBC). Statistical analysis revealed that non-seasonal ARIMA model was appropriate for 3-month SPI series while seasonal ARIMA models have been found promising for SPI series at 6-, 9,-12 and 24-month time scale. Parameter estimation step indicates that the estimated model parameters are significantly different from zero. The predicted data using the best ARIMA model were compared to the observed data for model validation purpose in which the predicted data show reasonably good agreement with the actual data. Hence the models were applied to forecast drought in the Bundelhand region up to 3 months advanced with good accuracy.