Sensitivity of WRF tropical cyclone simulations in the Philippines to different SST data

Improving the accuracy of tropical cyclone (TC) simulations in numerical weather prediction (NWP) models, such as the Weather Research and Forecasting (WRF) model, has been an increasingly important endeavor. Sea-air moisture and energy fluxes are mainly driven by sea surface temperature (SST) and the mixed layer underneath, which have significant effects on the formation and intensification of TCs. We investigated the sensitivity of three Philippine TCs - Typhoons Mangkhut, Goni, and Rai - to different SST datasets and a 1-D ocean mixed layer (OML) model in WRF. We found that the WRF runs with the high-resolution SST data update showed improvements in modeled maximum wind speed and consequently improved the simulated tracks over the archipelagic and/or coastal waters of the Philippines, as it gave better confidence in the (intensity-based) tracking algorithm after TCs made landfall in the country. TC-associated rainfall was also found to be sensitive to SST-updated model runs. Our results show that the use of SST significantly reduces the dry bias of WRF-simulated TC rainfall. The use of the high-resolution GHRSST dataset yielded the best TC simulation results over other SST data by simulating the sensible and latent heat or moisture fluxes over land and sea along coastlines better across the inland archipelagic waters of the Philippines. Disasters due to TCs are often brought about by strong winds and heavy rains over land. Considering that virtually no added computational cost is incurred in including SST update in the WRF model, the use of SST in TC modeling is an efficient method to improve TC hazard simulations. • Different SST dataset and 1-D ocean model using in WRF simulation of 3 tropical cyclones • High-resolution SST dataset reduced the dry bias of simulated TC rainfall • High-resolution SST dataset improved simulated maximum wind speed after TC landfall or over archipelagic waters • Use of SST update in WRF requires little to no additional computational cost