Evaluation of CMIP6 GCMs Simulation Performance Using a Hashing Algorithm and Its Application to Future Precipitation Projections

Climate change has led to an increased frequency of extreme hydrological events at the watershed scale. Evaluating the capability of CMIP6 global climate models (GCMs) to simulate precipitation and projecting future precipitation trends are essential for understanding future hydrological characteristics. Based on daily precipitation from 18 GCMs and 9 meteorological stations in the Lushi Basin during 1976—2000, an improved Hashing algorithm (ISHA) incorporating a spatiotemporal correlation mechanism was developed to select optimal GCMs, overcoming the limitation of fixed-grid partitioning in the traditional SHA. Non-parametric quantile mapping (QM) was then applied for bias correction. Based on the selected models, precipitation changes during 2025—2100 were projected under four SSP scenarios. The CNRM-CM6-1-HR, FIO-ESM-2-0, INM-CM5-0, and NorESM2-MM showed the best overall performance. The maximum monthly and long-term mean precipitation bias decreasing from 9.09%-16.04% and 10.41%-23.12% to 0.27%-2.34% and 0.48%-1.89%, respectively. The relative change in future annual precipitation increases with emission scenarios, ranging from -7.71% to 10.50% under low-emission conditions and expanding to -1.89% to 18.02% under high-emission scenarios. The basin may face an increasing risk of enhanced precipitation during the mid-to-late 21st century, thereby intensifying challenges in water resources management and flood control.