Meteorological drought evaluation in Parbhani, Maharashtra, using multi-model approach for SPI forecasting

Drought is one of the most destructive climate phenomena disasters. Hence, Drought prediction plays an important role in mitigating certain adverse effects of drought. Data-driven models are widely used for drought forecasting for example, ARIMA model, artificial neural network (ANN) Model, Wavelet Neural Network (WANN) Model, Support vector regression model, support vector machine model, grey model and so on . In this study, two data driven models (MLP-ANN model and W-ANN model) are used for SPI-1 forecasting for one month lead time basis. The drought model was developed utilizing SPI data from prabhani district of Maharashtra in India for 1980–2022, wherein the first 30 years are employed for model training, while the remaining 12 years are employed for model testing and validation. The SPI values at previous six-month lag were used to forecast current month SPI values and gamma test was performed on 40 different input combinations to choose the best input combination for SPI- 1.The effectiveness of the two data-models is compared by , RMSE, r and CE. For SPI-1 Prediction the best input combination i.e., combination 32 (SpIt-1, SPIt-2, SPIt-3, SPIt-4) based on the lowest gamma statistics, standard error and V-ratio values. It is clear from the performance of models that the W-ANN model has higher r values and lesser RMSE values as compared to MLP-ANN models. Therefore, the W-ANN model was superior to MLP-ANN models at Prabhani district of Maharashtra.