Sensitivity Analysis of Gravity Wave Characteristics to Physical Parameterization Options in WRF Simulations : A Case Study of Typhoon Soudelor (2015)

Tropical cyclones (TCs) are a source of atmospheric gravity waves, which contribute to mixing in  the upper troposphere and lower stratosphere. Here, we conducted a large ensemble simulation run of the Weather and Forecasting Research (WRF, V4.4.1) model, assessing the impact of 15 combinations of microphysics (MP), planetary boundary layer physics (PBL), and a cumulus scheme (CU) on the model's ability to simulate the physics of Typhoon Soudelor (2015) and this typhoon's generation of gravity waves. The simulation is performed using a moving nested domain at 3 km  horizontal resolution, with a 15 km exterior main domain. We use data from International Best Track Archive for Climate Stewardship to measure bias in track position and intensity of the typhoon, supported by the use of AIRS/Aqua satellite observations as a benchmark. Moving beyond traditional analyses, we also apply a kernel density estimator (KDE) approach to produce more comprehensive results. 

Our results indicate that, while track errors remain below 100 km for the first 42 hours of the run, the simulated storm intensity and speed varied significantly from observations. Notably, simulations incorporating cumulus parameterization generally yield wider track spreads, whereas microphysics produced higher storm intensities and a more accurate representation of deep convective clouds compared to WSM6, despite an overall tendency to overestimate storm strength. We then examined coupling between tropical cyclone dynamics and stratospheric wave generation by comparing simulated Outgoing Longwave Radiation (OLR) and vertical wind speeds against satellite and reanalysis data. KDEs of OLR suggests, that while the Goddard MP effectively captures deep convection, the addition of a Grell-3 CU parameterization tends to produce more extensive mid-to-high-level cloud cover but underestimates the deepest convective cores. In the stratosphere, vertical wind speed profiles indicate that the MYJ and Goddard combinations produce the strongest wave activity, especially during the chosen peak events. Although the simulations slightly overestimate background wind speeds near the tropopause compared to ERA5 reanalysis output, the overall wave morphology remains consistent with observations. These findings reinforce the conclusion that no single physics combination optimally captures all TC attributes, though Goddard MP and specific PBL schemes offer superior performance in representing the convective forcing essential for stratospheric gravity wave excitation.