A New Hybrid Weighted Regional Drought Index to Improve Regional Drought Assessment

Unlike other natural disasters, the temporal characteristics of drought are complex. Therefore, accurate drought monitoring and assessment using advanced statistical and machine learning techniques are essential. This research develops a new regional drought indicator – the Seasonally Weighted Regional Standardized Drought Index (SWRSDI). The methodology for SWRSDI introduces a new regional data aggregation scheme called the Seasonal Interdependence and Discrepancy-based Weighting Scheme (SIDWS). The SIDWS scheme divides the data into twelve distinct seasons and assigns weights based on two criteria: 1) seasonal interdependence and 2) seasonal divergence. In application, we compared the performance of SIDWS with the regional aggregation scheme used in the Seasonally Combinative Regional Drought Indicator (SCRDI) and Simple Model Average (SMA) using Root Mean Square Error (RMSE), Normalized Root Mean Square Error (NRMSE), and Mean Absolute Error (MAE) metrics. To estimate SWRSDI values, we standardized the regionally aggregated precipitation data under SIDWS using K-Components Gaussian Mixture Models (K-CGMM). The outcomes of this research demonstrate the superiority of using SIDWS for regional data aggregation. This conclusion is based on the lower values of MAE (1.9735 vs. 2.8983), NRMSE (0.0322 vs. 0.0485), and RMSE (3.0981 vs. 4.4025). Additionally, the study achieves high accuracy in standardization by using a probability-based standardization mixture. The selection criteria advocate the high accuracy of K-CGMM-based estimation of drought indices. Overall, this article provides a more rational and effective procedure for assessing, monitoring, and forecasting drought at the regional level.