Advancements in Extreme Precipitation Projections for South Asia: A Comparative Evaluation of CMIP5 and CMIP6 Models

Abstract

Global climate models (GCMs) are essential for projecting future climate changes, yet their ability to accurately simulate extreme precipitation, particularly in South Asia, remains a major challenge. This study assessed the performance of GCMs from CMIP5 and CMIP6 in replicating 11 extreme precipitation indices, using ERA5 data from 1975 to 2005. The results revealed substantial variability across individual models, with CMIP6 generally outperforming CMIP5, though certain inconsistencies persisted. Both CMIP5 and CMIP6 multi-model ensemble means (MMEs) exhibited higher root mean square error (RMSE) than the best individual models, highlighting the need for further improvements in model accuracy. On average, CMIP6 models achieved a Kling–Gupta efficiency (KGE) of 0.42, outperforming CMIP5's 0.38, and demonstrated better agreement in Taylor diagrams, with an average r² of 0.65 compared to 0.59 for CMIP5. CMIP6 also showed reduced uncertainty in interannual monthly precipitation variability projections. EC-Earth3 (CMIP6) and EC-Earth (CMIP5) consistently correlated well with various indices, while MIROC-ESM was also a strong performer in both generations. The CMIP6 MME performed better overall, with a KGE of 0.48 and r² of 0.71, surpassing CMIP5 MME's 0.44 and 0.67. Future projections indicate significant changes in precipitation extremes under different emission scenarios for the 2040s and 2080s. While CMIP6 shows clear advancements over CMIP5, continued model refinement is essential to more accurately simulate extreme precipitation events.