<![CDATA[Precipitation is a crucial component of weather and climate, playing a fundamental role in the Earth's water cycle. However, in situ, rain gauge networks are still limited in their ability to comprehensively monitor precipitation across all regions, including Semarang regency and city. Satellite – based remote sensing and cloud computing technology, such as Google Earth Engine (GEE), offer a solution for generating rainfall estimates with spatial coverage. This study optimizes CHIRPS rainfall estimates through a calibration process using BMKG rain gauge data over the Semarang region for 2021 – 2023. It evaluates the spatial distribution and performance of CHIRPS before and after calibration. Compared to observational data, the original CHIRPS dataset exhibited significant spatial discrepancies, with a daily RMSE of 44 mm/day, a coefficient of determination (RSQ) of 0.02, and a SMAPE of 99%. The collinearity analysis showed that the relationship between CHIRPS and observational data tends to be scattered and less linear on a daily scale, but after calibration, this relationship becomes stronger. Calibration using the Geographical Differential Analysis (GDA) method successfully improved CHIRPS accuracy, as indicated by a reduction in daily RMSE to 25 mm/day, an increase in daily RSQ to 0.62, and a decrease in daily SMAPE to 70%. These improvements were also observed in monthly and annual rainfall estimates. The calibrated CHIRPS data exhibited enhanced spatial distribution and performance, with a 10% reduction in annual RMSE, a 25% increase in annual RSQ, and a 20% decrease in annual SMAPE compared to the original dataset. Furthermore, sensitivity to rainfall intensity improved, particularly for heavy to extreme rainfall events, as evidenced by a 58% reduction in the FAR, a 73% increase in the POD, and a 48% improvement in the CSI.]]>
Copyrights © 2025
