Model Sensitivity to Perturbations of Environment, Structure, and Model Parameters in Idealized, Ocean-Coupled Tropical Cyclone Simulations

Idealized simulations are important tools to investigate in detail the dynamical evolution of a tropical cyclone for various environmental and/or structural characteristics. We present here a systematic sensitivity analysis using an idealized version of the Hurricane Weather Research and Forecasting (HWRF) model. The tropical cyclone environment is initialized with the tropical moist sounding of Dunion (2011, J. Climate). The westerly 850-200-hPa vertical wind shear is thermally balanced in the meridional direction. The zonal wind field is adjusted to yield a vertically integrated westward mean flow typical of Tropical Atlantic hurricanes. Lateral boundaries are forced with the same initial environmental profiles as in the computational domain interior to minimize imbalances. Coupling with a one-dimensional column ocean model introduces ocean cooling due to surface wind stress and modifies surface fluxes. The ocean column is initialized with prescribed temperature and salinity profiles that exhibit hurricane-season Tropical Atlantic characteristics with a deep, well-mixed upper ocean. The initial vortex is a wavenumber-0 composite of thousands of hurricane reconnaissance (dropwindsonde and tail Doppler radar) observations and historical height-radius cross-sections of steady-state, category-one Tropical Atlantic hurricanes over water. An analysis of the 5-day control simulation obtained in this manner will be presented first.

Model sensitivity to perturbations in parameters that include magnitude of zonal shear, vertically integrated atmospheric mean flow (storm speed), initial SST, environmental low-level and mid-level moisture and temperature, initial intensity, initial radius of maximum wind (RMW), as well as model parameters that control horizontal diffusion, vertical eddy diffusivity, and exchange coefficients of surface momentum and heat flux is then investigated, especially focusing on the quasi-steady-state regime that is observed in the 48-96 hours of the control simulation. Detailed analyses of parameter-model correlations, simulation spread, and response function will be presented for a systematic evaluation of model sensitivity. Suggestions will be made for calibrating the range of parameter values to improve the signal-to-noise ratio for the possibility of multiple, simultaneously perturbed parameters. Implications for ensemble-based data assimilation will be discussed.