Climate change is intensifying the frequency and severity of extreme weather events such as floods and droughts, posing significant challenges to global and regional water resource management. This study develops an integrated framework to optimize th...
Climate change is intensifying the frequency and severity of extreme weather events such as floods and droughts, posing significant challenges to global and regional water resource management. This study develops an integrated framework to optimize the use of Coupled Model Intercomparison Project Phase 6 (CMIP6) global climate models (GCMs) for improving climate forecasts and hydrological planning in South Korea. The objectives include identifying high- performing GCMs, applying bias correction techniques, and analyzing Extreme Climate Indices (ECIs) to support reliable long-term forecasting and risk mitigation strategies. A total of 32 CMIP6 GCMs were evaluated using spatial performance metrics—SPAEF, Kling-Gupta Efficiency (KGE), Cramér’s V, and the Fractions Skill Score (FSS)—based on their ability to reproduce historical precipitation patterns during annual and monsoon (JJAS) periods. Four models (FIO-ESM-2- 0, CESM2-WACCM, CESM2, and CMCC-ESM2) demonstrated the best performance, despite a general tendency to overestimate precipitation. The study highlights that bias correction should be applied after model selection, as it can alter GCM rankings and introduce inconsistencies. For bias correction, the CDF Matching (CDFm) method was employed for temporal adjustments and Empirical Quantile Mapping (EQM) for spatial refinement. A Bayesian linear regression framework implemented in PyMC3 was used to quantify uncertainty in precipitation forecasts. Daily and seasonal precipitation projections (2025–2060 and 2065–2100) exhibited relatively low uncertainty, reinforcing the reliability of the corrected GCM outputs for policy and planning purposes. To assess model performance in simulating extreme events, ECIs—including CDD, PRCPTOT, R20mm, R95p, Rx1day, Rx5day, and SDII—were calculated using daily precipitation data. The analysis, supported by statistical tools such as the Modified Mann-Kendall (MMK) test, Theil-Sen Estimator (TSE), Taylor diagrams, and Kernel Density Estimation (KDE), revealed that BCC-CSM2-MR and GFDL-CM4 accurately reproduced indices like PRCPTOT, R20mm, Rx1day, and Rx5day. Additionally, GISS-E2-1-G, ACCESS-CM2, and IITM- ESM showed strong performance for CDD and SDII. Regional discrepancies in ECI spatial distribution also underscored the need for region-specific bias correction. Overall, this multi-scale, model-agnostic approach significantly enhances the accuracy of extreme event projections, supporting adaptive water management strategies. The framework is flexible and can be extended to other climate variables, regions, and time scales. This study provides a robust scientific foundation for sustainable climate adaptation, resource management, and disaster risk reduction in Korea and beyond.