How Do Gridded Rainfall Datasets Represent Climate Zones in the Philippines?

Previous validation studies of gridded rainfall over the Philippines have typically relied on established climate classifications, such as the Modified Coronas Climate Classification, which defines four climate zones in the country. In contrast, more recent research using objective clustering methods on rainfall, such as Self-Organizing Maps (SOM), has identified only three climate zones across the country. However, determining whether these three zones are reflected in various gridded rainfall datasets remains unassessed. Thus, this study fills that gap by applying a two-phase SOM method to APHRODITE, PERSIANN-CDR, TRMM, CMORPH, and CHIRPS, which are commonly used gridded rainfall datasets. This study explored the influence of SOM hyperparameters such as grid size, topology, and distance function on the derived climate zones. The results showed that the different rainfall datasets are sensitive to the number of grid sizes, topology, and distance functions, particularly the spatial patterns and number of generated climate zones. Using the climate zones detected in the APHRODITE dataset as the reference, a summary of the recommended hyperparameters for each dataset that well-represents the three climate zones is presented. The findings of this study may serve as a basis for characterizing both the climatological and projected rainfall patterns in higher-resolution datasets.