A novel framework for future drought characterization under ranked-based subset selection and weighted aggregative multi-modal ensemble of global climate models

Future drought characterization often relies on Multi-Modal Ensembles (MMEs) of Global Climate Models (GCMs), particularly from the Coupled Model Intercomparison Project Phase 6 (CMIP6). However, the reliability of projections is often hindered by insufficient ranking methodologies for GCMs and inadequate handling of outliers in regional aggregation. This study presents a novel framework to enhance the reliability of drought projections and standardization by introducing innovative ranking, aggregation, and projection methods. The framework is not limited to a specific region but is adaptable to diverse climatic and geographic contexts. The proposed methodology employs Mutual Information (MI) to evaluate the performance of GCM in simulating historical precipitation, followed by comprehensive rating metrics (CRM) to rank models effectively. A novel regional aggregation technique addresses outlier influence, ensuring robust multi-model ensembles. The approach incorporates top-performing GCMs into MMEs using advanced geometric and regression methods, validated using the Kling-Gupta efficiency with knowable moments (KGE km ). A Gaussian-Norm Weighted Drought Index (GNWDI) was also introduced, offering enhanced drought standardization within the Standardized Precipitation Index (SPI) framework. Applying this framework in Punjab, Pakistan, using 22 GCMs, enabled the identification of high-performing models such as MIROC-ES2L, CMCC-CM2-SR5, and IPSL-CM6A-LR. Future drought trends for 2015–2100 were projected under three Shared Socioeconomic Pathways (SSPs). Results revealed a rise in extreme droughts and wet conditions under high emission scenarios (SSP5-8.5), highlighting the intensification of drought severity over extended periods. Specifically, under SSP5-8.5, the average probability of extreme droughts (ED) across all time scales is approximately 0.0221, which remains comparable to lower emission scenarios but shows slightly elevated values at longer time scales, such as 48 months (0.025). Additionally, severe wet (SW) conditions notably increase under SSP5-8.5, with the probability rising from 0.044 at 1 month to 0.051 at 12 and 24 months, and peaking at 0.051 again at 48 months, suggesting more frequent extreme hydrological swings under intensified climate forcing. This study significantly advances drought projection techniques by addressing critical gaps in model ranking, aggregation, and standardization. The framework offers a reliable, regionally adaptable tool for policymakers and researchers, enabling proactive drought management and improved climate resilience under varying emission scenarios. • Droughts model projections changing due to climate forcing, demanding refined prediction methods. • Novel framework ranks model via mutual information and builds robust climate ensembles. • Introduced new aggregation and drought standardization with Gaussian fitting. • Future work may expand to multivariable drought indices and machine learning tools.