Combination of data-driven models and best subset regression for predicting the standardized precipitation index (SPI) at the Upper Godavari Basin in India

Standardized precipitation index prediction and monitoring are essential to mitigating the effect of drought actions on precision farming, environments, climate-smart agriculture, and the water cycle. In this study, four data-driven models, additive regression, random subspace, M5Pruned (M5P), and bagging tree models, were adopted to predict the standardized precipitation index (SPI) at the Upper Godavari Basin for various periods (3 months, 6 months, and 12 months). The data-driven models' input data was pre-processed with machine learning models to increase quality and the model's performance a priori. These four models predicted the SPI-3, SPI-6, and SPI-12 months based on three metrological station data. Based on the statistical performance metrics such as correlation coefficient (R), root mean square error (RMSE), mean absolute error (MAE), relative absolute error (RAE), and root relative squared error (RRSE), our findings showed that the bagging was the best model for predicting SPI-3 and SPI-6 while the M5P the best for SPI-12 estimation in station 1, while in stations 2 and 3, M5P was superlative in predicting the SPI-3 and SPI-12 months, and the bagging was best in SPI-6. All the best models had acceptable mid-term drought forecasting based on the SPI-3, SPI-6, and SPI-12 months for three stations in the Upper Godavari Basin in India. The machine learning models created in this study produced satisfactory results in short-term and mid-term drought forecasting, and it will be a new strategy for water developers and planners to use for future management and scheduling.