Bias Correction Methods Applied to Satellite Rainfall Products over the Western Part of Saudi Arabia

Accurate rainfall data with good spatial–temporal distribution remain a challenge worldwide, particularly in arid regions such as western Saudi Arabia, where variability critically influences water resource management and flood mitigation. This study evaluates five satellite-based rainfall products—GPM, GPCP, CHIRPS, PERSIANN-CDR and PERSIANN—against observed monthly rainfall at 28-gauge stations, using the correlation coefficient (CC), root mean square error (RMSE), relative bias (RB) and mean absolute error (MAE). Among uncorrected products, GPM achieved the highest mean CC (0.52), and lowest RMSE (17.0 mm) and MAE (9.18 mm) compared with CC = 0.39 (RMSE 19.9 mm) for GPCP, CC = 0.20 (RMSE 21.6 mm) for CHIRPS, CC = 0.43 (RMSE 19.2 mm) for PERSIANN-CDR and CC = 0.26 (RMSE 57.3 mm) for PERSIANN. Four bias correction methods—linear scaling, nonlinear adjustment, quantile mapping and artificial neural networks (ANN)—were applied. The ANN reduced GPM’s RMSE by 19% to 13.8 mm, increased CC to 0.59, lowered RB to 2.5% and achieved an MAE of 6.89 mm. These results demonstrate that GPM, particularly when bias-corrected via ANN, provides a dependable rainfall dataset for hydrological modeling and flood risk assessment in arid environments.