Integrating climate model ensembles for reliable regional drought assessment through redundancy control

Abstract Drought is a complex and recurrent natural hazard that poses substantial challenges to sustainable water management and climate adaptation. Therefore, accurate projection of future drought requires reliable and spatially consistent climate assessments. Most evaluations of global climate models (GCMs) focus only on average errors and overlook the shape, timing and persistence of precipitation patterns as they characterize real drought behavior. This study introduces a novel framework to address these limitations for improving drought projection by selecting high‐performing GCMs and combininIg them into a robust ensemble. Using 22 Coupled Model Intercomparison Project Phase 6 (CMIP6) GCMs across 29 grid stations in Balochistan, models are evaluated through redundancy‐adjusted shape‐based similarity measures to capture precipitation dynamics over a historical period (1950–2014). A multicriteria decision‐making approach is proposed and applied to identify the most reliable models which are further combined using several ensemble techniques. Among these ensembles including regression, geometric, and machine‐learning, the constrained least‐squares ensemble (CLSE) achieved the lowest root mean squared error RMSE (3.217) and mean absolute error MAE (2.563), outperforming all counterparts. Future projections (2015–2100) under Shared Socioeconomic Pathway (SSP)1‐2.6, SSP2‐4.5, and SSP5‐8.5 are analyzed using a probabilistic framework to quantify drought severity and uncertainty, leading to the formulation of a new probabilistic drought index. Results showed a steady intensification of droughts, moderate events increasing by 19%–27% under SSP2‐4.5, while severe and prolonged droughts rise by more than 35% under SSP5‐8.5. Transition analysis further reveals a gradual intensification of drought conditions, with moderate and severe events increasing slightly to 0.09 and 0.05, respectively and a reduced stability of normal climate states. Overall, the proposed framework offers a robust, data‐driven approach for enhancing drought predictability and supports climate‐informed water planning and adaptive management across climate‐vulnerable regions like Balochistan.