Reliability assessment and bias correction of CMIP6 climate models for monthly rainfall projection over Bengaluru Urban

Reliable rainfall projections are essential for water resources planning and climate adaptation in rapidly developing urban areas. This study evaluates 10 CMIP6 climate models for Bengaluru Urban using monthly rainfall data for 2000–2014 and projects rainfall for 2015–2100 under SSP245, SSP370, and SSP585. Model performance was assessed using R2, root-mean-square error, mean absolute error, Nash–Sutcliffe efficiency, Kling–Gupta efficiency, and percent bias, with second-degree polynomial regression used as a supplementary analysis. MPI-ESM1-2-LR showed the highest overall performance, followed by EC-Earth3, IPSL-CM6A-LR, and NorESM2-MM. These four models were selected for bias correction and ensemble development. Among linear scaling, quantile mapping, and empirical quantile mapping (EQM), EQM showed the closest agreement with observations during the historical calibration period, although independent validation was not performed. Annual rainfall is projected to increase by approximately 9.9, 13.1, and 15.7% under SSP245, SSP370, and SSP585, respectively. Sen's slope ranged from 2.17 to 4.07 mm/year, while P95, P99, and monthly return period rainfall estimates also increased. The findings support long-term water resources and climate-adaptation planning, but monthly projections should not be considered design rainfall for urban drainage or flood control systems.