Assessment of Tropical Cyclone Rainfall from GSMaP and GPM Products and Their Application to Analog Forecasting in the Philippines

Tropical cyclone (TC) rainfall is both a resource and a hazard in the Philippines. Observation of its spatiotemporal distribution is necessary for water and disaster mitigation management. This study evaluated the performance of two high-resolution satellite precipitation datasets-the GSMaP and GPM-IMERG-in estimating accumulated TC rainfall in the Philippines from 2000 to 2021. TC rain is defined as rainfall within 5 degrees of a TC center. Several estimation algorithms were included in the assessment. The uncalibrated near-real-time GSMaP_NRT and early version GPM_ER, the reanalysis GSMaP_RNL, and the gauge-calibrated GSMaP_G and GPM_G. The assessment shows the worst scores for the uncalibrated GSMaP_NRT and GSMaP_RNL, followed by GPM_ER with station correlation coefficient (CC) values of 0.63, 0.67, and 0.73, respectively, compared to GSMaP_G CC of 0.79 and GPM_G CC of 0.77. GSMaP_G and GPM_G also gave the least bias and error, with a consistently high (>0.6) probability of detection (POD) and Pierce skill score (PSS) up to rainfall of 300 mm. In addition to the evaluation, the GSMaP_G and GPM_G were used in the analog forecasting of TC rain. Analog forecasting is based on the principle that past weather conditions can occur again. In TC rain analog forecasting, past TCs with similar intensities and tracks are assumed to bring similar rainfall amounts and distribution as a current TC. Composite mean TC rainfall from historical satellite precipitation estimations was produced to create TC rain forecasts. Results show the analog TC rain forecasts generally captured the spatial distribution of TC rain and performed better than the uncalibrated GSMaP_NRT, with a mean station correlation of 0.62-0.67, POD greater than 0.7, and positive PSS indicating good skills. However, forecasts have a false alarm ratio greater than 0.8 for 150 mm rain and have difficulty producing extreme rainfall (>250 mm).

Automated summary

This study evaluated the performance of two high-resolution satellite precipitation datasets, GSMaP and GPM-IMERG, in estimating accumulated tropical cyclone rainfall in the Philippines. The assessment showed that the calibrated GSMaP_G and GPM_G performed the best, with higher correlation coefficients and smaller biases and errors. These datasets also maintained good detection capabilities and skill scores even at higher rainfall amounts.