Performance Evaluation of Historical CESM2 Rainfall Before and After Linear Scaling Bias Correction in the Sadar Watershed, Indonesia

Accurate rainfall data is fundamental for hydrological analysis, yet Global Climate Models (GCMs) such as CESM2 exhibit systematic biases compared to ground observations. This study evaluates historical CESM2 monthly rainfall before and after Linear Scaling (LS) bias correction in the Sadar Watershed, East Java, Indonesia (2000–2014). Observed rainfall from seven stations was spatially averaged using Thiessen Polygon method and compared against downscaled CESM2 outputs at 0.25° × 0.25° resolution. Model performance was assessed using Nash–Sutcliffe Efficiency (NSE), RMSE-Observations Standard Deviation Ratio (RSR), and Percent Bias (PBIAS). Prior to correction, CESM2 showed unsatisfactory performance (NSE = 0.410, RSR = 0.768). After LS correction, NSE improved to 0.516 (marginally satisfactory, near 0.50 threshold) and RSR to 0.696 (satisfactory), with PBIAS improving from −10.218% to 1.439% (very good). However, LS corrects only monthly-scale mean bias; daily-scale temporal variability and extreme rainfall remain unimproved. The 15-year period is shorter than the WMO-recommended 30-year climate baseline. Results demonstrate that LS-corrected CESM2 data are suitable for watershed-scale water balance assessment but insufficient for daily hydrological modeling without advanced bias correction methods, highlighting the importance of method selection for sustainable water resource management under climate change (SDG 6 & 13).