Evaluation of Initial Sea Surface Temperature Uncertainty on HWRF-Ensemble Tropical Cyclone Intensity Forecasts

Errors in Numerical Weather Prediction (NWP) model forecasts can originate from initial condition uncertainty and physics uncertainty. As a consequence, tropical cyclone (TC) ensemble prediction systems (EPS) should account for both sources. Unfortunately, most TC EPS only take into account atmosphere initial condition uncertainty, with limited treatment of physics uncertainty and no uncertainty in the ocean state. The insufficient treatment of all sources of uncertainty can be responsible for the numerous instances where verification falls outside the ensemble distribution.

This study evaluates the impact of initial condition uncertainty in the ocean on tropical cyclones (TCs) on Atlantic and eastern Pacific basin TCs using the 2017 configuration of the Hurricane WRF (HWRF) EPS. Initial-time ocean uncertainty is determined from climatological samples taken from the Global Forecast System (GFS) surface analysis dataset, which is subsequently mixed downward into upper ocean using the Princeton Ocean Model (POM) initialization procedure. Preliminary results show that including SST uncertainty results in larger 72-h ensemble standard deviation in TC intensity compared with atmospheric uncertainty alone. Moreover, the ensemble with oceanic uncertainties is characterized by larger differences in latent heat flux between intense and less intense TCs relative to the atmosphere-only experiment, which in turn yields larger intensity differences.