Satellite-based rainfall estimation plays a crucial role in regions with limited ground-based observations, such as South Sulawesi. However, the accuracy of these products varies spatially, underscoring the need for localized performance assessment. This study evaluates four widely used satellite rainfall products: Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS), Global Precipitation Climatology Project (GPCP), Global Satellite Mapping of Precipitation (GSMaP), and Integrated Multi-satellitE Retrievals for GPM (IMERG) by comparing them against station observations using seven statistical accuracy indicators. To capture spatial variability in product performance, a K-Means clustering algorithm was applied to group regions with similar accuracy characteristics. This approach enables the classification of the study area into performance-based zones, facilitating region-specific recommendations. The results show that no single product consistently out-performs others across all clusters. IMERG performed best in Cluster Type 1 (North Toraja), with a Probability of Detection (POD) of 0.87 and a Critical Success Index (CSI) of 0.89. CHIRPS was most suitable in Cluster Type 2 (Pinrang, Enrekang, Sidrap, Soppeng, Bone) with a Root Mean Square Error (RMSE) of 12.3 mm. GPCP demonstrated the highest accuracy in Cluster Type 3 (Luwu, East Luwu, Palopo, Tanah Toraja, North Toraja, Makassar, Maros, Gowa, Takalar) with an RMSE of 10.5 mm, Pearson Correlation (PC) of 0.92, and BIAS of 0.05. Higher-elevation generally exhibited larger errors and lower detection capability, suggesting that terrain complexity influences satellite rainfall performance. These findings highlight the value of integrating clustering and terrain context into satellite rainfall validation frameworks.
Giarno et al. (Wed,) studied this question.