Comparison of the accuracy of four satellite rainfall estimates during the rainy season, dry season and transition in South Sulawesi which has complex topography

Abstract

The sparse distribution of rainfall gauges in Indonesia can be supplemented with satellite products. However, due to the highly dynamic and random nature of rainfall in the region, it is crucial to evaluate the accuracy of these rainfall estimation data. To assess the accuracy of CHIRPS, GCPC, GsMap, and IMERG gridded rainfall data products, numerical comparisons using root mean square error (RMSE) and dichotomous method indicators such as percent correct (PC), false alarm ratio (FAR), bias score (BIAS), probability of false detection (POD), and critical success index (CSI) will be conducted. Analyzing 40 years of data reveals that the accuracy of satellite products is highest in the southern part of South Sulawesi Province, particularly along the west coast. The southern region shows better accuracy than the northern part, and the western area outperforms the eastern region. The accuracy of satellite rainfall estimates varies by location and time, with RMSE values increasing during periods of high rainfall, such as the peak of the rainy season, and decreasing during the dry season. Dichotomous parameters indicate a higher false alarm error compared to missing detections. Among the products, CHIRPS demonstrates the most consistent accuracy, maintaining performance close to the best values. Overall, CHIRPS outperforms the other products, followed by IMERG, GCPC, and GsMap. However, this ranking is subject to change over time and location, necessitating re-evaluation for each application and time period