Accurate precipitation data is critical for hydrological modeling, flood forecasting, and water resources planning. This study evaluates the performance of satellite-based rainfall estimates from the Integrated Multi-satellite Retrievals for GPM (IMERG) Final Run Version 06 by comparing them with ground-based observations from six stations in the Jatigede Reservoir catchment, West Java, Indonesia. The analysis covers the 2014–2023 period, aligning with the reliable availability of IMERG Final Run products, and examines three temporal resolutions: monthly, daily, and hourly. Statistical evaluation employed Pearson correlation coefficient (r), the ratio of RMSE to observed standard deviation (RSR), and Percent Bias (PBIAS). Results show strong agreement at the monthly scale (r = 0.84, RSR = 0.34, PBIAS ≈ +24%), suggesting suitability for long-term water resource assessments. However, performance declines at shorter timescales. At the daily scale, IMERG underestimates rainfall (PBIAS ≈ -27%) with moderate correlation (r = 0.24). The hourly scale shows the poorest performance (r = 0.10, RSR > 3.0, PBIAS < -50%), indicating limitations in capturing short-duration, high-intensity rainfall typical in tropical regions. These findings underscore the importance of temporal aggregation and bias correction when applying IMERG data for operational hydrology and flood modeling

This study evaluates the performance of Global Precipitation Measurement (GPM) satellite-based rainfall data in comparison to Automatic Rainfall Recorder (ARR) data in forming rainfall distribution patterns and assessing its impact on flood discharge simulation using the HEC-HMS model. Statistical validation was conducted using the Pearson Correlation Coefficient, the ratio of standard deviation of observations to RMSE (RSR), Percent Bias (PBIAS), and Mean Absolute Percentage Error (MAPE). The results show that GPM has a strong correlation with ARR (r = 0.875) and a low RSR value (RSR= 0.256), yet it exhibits a notable negative bias (PBIAS = –24.41%), indicating an underestimation of rainfall values. In contrast, simulations using ARR rainfall patterns produce peak discharges that closely match actual discharge records at the Jatigede Dam outlet, with an average deviation of less than 3% and a MAPE of 1.17%, categorized as very good. The GPM simulation produces peak discharges 13–16% higher than actual observations, with a MAPE of 14.53%, which still falls into the good category. These results suggest that while ARR provides higher accuracy, GPM remains a viable alternative source, especially in data-scarce areas, provided that appropriate calibration methods such as bias correction are applied. This study supports future research in satellite data calibration using machine learning and multivariate statistical approaches.