A Novel Statistical Framework for Assessing Future Drought Using Multiple Global Climate Model: The Weighted Multimodal Adaptive Standardized Precipitation Index

Abstract

Drought is one of the major consequences of global warming. Being a complex natural hazard, its accurate assessment is challenging. Simulated data of varying climate parameters from Global Climate Models (GCMs) is a crucial source for assessing the future characteristics of climate change. The objective of this article is to improve future drought assessment based on ensemble of multiple GCMs. Consequently, this study proposes a new statistical framework to improve future drought assessment based on a multiple GCM ensemble. The proposed framework introduces a new weighting scheme for Multi-Model Ensembles (MMEs), called the Precipitation Concentration Index-Based Weighting Scheme for Multi-Model Ensembles (PCIWS-MME), and a drought index known as the Weighted Multimodal Adaptive Standardized Precipitation Index (WMASPI). The application of the proposed research is based on 22 GCMs from the Phase 6 Coupled Model Intercomparison Project (CMIP6) and covers 103 grid points in Pakistan. To assess the effectiveness of PCIWS-MME, we compared its performance with the Simple Multimodel Mean (MME) and Mutual Information (MI) using the Root Mean Square Error (RMSE) and Mean Average Error (MAE). Furthermore, we evaluated the quality of WMASPI by fitting the most appropriate models, whether univariate, mixture-based, or derived from nonparametric probability plotting position formulas. The results of probabilistic modeling indicate that mixture probability models are more appropriate than univariate alternatives. For example, on the 3-month time scale under Scenario 1, the Bayesian Information Criterion (BIC) for the best-fitting univariate distribution is \(-\)708.11, while the K-CGMM model achieves a substantially lower BIC of -7001, reflecting a significantly better fit. Similarly, at the 24-month time scale under Scenario 3, the univariate model yields a BIC of \(-\)301.52, whereas the K-CGMM model attains a much lower BIC of \(-\)980.68, further confirming its superior performance. The results associated with the weighting schemes indicate that PCIWS-MME outperformed both the simple mean-based MME and MI-based schemes, since it consistently exhibited lower RMSE and MAE while demonstrating a higher correlation with the observed data. Furthermore, the study used the proposed multimodel ensemble data from PCIWS-MME to calculate standardized drought indices under WMASPI. To assess long-term drought trends, results obtained by trend analysis using the Mann-Kendall (MK) test indicate that, in the short term (3–12 time scales), trends are generally weak and statistically insignificant, except for SSP1\(-\)2.6, which exhibits a slight but significant decreasing trend at certain intervals. In the medium term (24-time scale), all scenarios show decreasing trends, with SSP5\(-\)8.5 displaying the most pronounced decline. Over the long term (48-time scale), all three scenarios demonstrate statistically significant negative trends. In summary, the study demonstrates the use of advanced statistical tools to model and assess drought under global climate change using simulated precipitation data from GCM.