The Sensitivity of Tropical Cyclone Simulations in the WRF Model to Surface Layer and Planetary Boundary Layer Parameterization

Abstract

The high wind speeds found in tropical cyclones fundamentally change the physical processes by which heat, moisture and momentum are transferred between the ocean and the lower atmosphere. Despite this fact, surface and boundary layer parameterization schemes in many numerical models that are frequently used for tropical cyclone simulations are based on assumptions made in more tranquil atmospheric conditions. Limited observations in the high wind speed conditions found in strong tropical cyclones suggest that spray and foam can enhance the transfer of heat and moisture from the ocean to the atmosphere, while reducing drag. Inclusion of the effects due to sea spray in a numerical model leads to stronger tropical cyclones (Wang et al. 2001, Perrie et al. 2005).

Based upon the absence of sea spray effects and the values of the exchange coefficients in the WRF model, it was anticipated that simulations using an idealized vortex and ambient environment would not reach the thermodynamically estimated theoretical maximum intensity (MPI) limit of Emanuel (1986). In addition, it was expected that simulations of Hurricane Ivan would not reach the intensity of the observed storm. The sensitivity of the model results to surface layer and PBL parameterization, and model grid spacing was tested, with the hypothesis that the simulated tropical cyclones would remain weaker than MPI theory (for the idealized simulations) or observations (for the Hurricane Ivan studies) regardless of the model physical parameterization choice. Grid spacing was also hypothesized to impact the simulated TC intensity, with the expectation that simulations with smaller grid spacing would produce more intense TCs, based on the results of previous studies.

Simulated TC intensity is found to be highly sensitive to model grid spacing in experiments with Hurricane Ivan or with an idealized initial vortex. Simulations using 4-km grid spacing were able to produce TCs that exceeded the MPI of the idealized environment (determined by minimum sea level pressure), while simulations using coarser (12 or 36-km) grid spacing were not. Simulations of Hurricane Ivan using 4-km grid spacing, initialized with a vortex that was ˜60 hPa weaker than observations reached the maximum intensity of the observed system, and exceeded the observed intensity during the latter stages of the simulation. These results suggest that the WRF model, in its current configuration, overestimates TC intensity, especially with small values of horizontal grid spacing. If the exchange of moisture, heat and momentum were adjusted to more accurately portray the conditions found in high wind speed conditions, the idealized tropical cyclone would likely exceed the theoretical MPI by an even larger amount. Therefore, it is concluded that some other aspect of the model formulation may lead to an overestimation of tropical cyclone intensity, or that there are deficiencies in MPI theory.

In the near future, a version of the WRF model designed for the prediction of hurricanes (HWRF) will be released, and many avenues exist for future research. The HWRF model will require extensive testing before being relied upon by the operational forecasting community. A methodology similar to the one used in this study should be employed, whereby the HWRF model is used in idealized simulations and compared to theoretical intensity limits. The idealized simulations would help to assess the ability of the new model to represent TC intensity in an accurate manner. The surface fluxes and exchange coefficients produced by the model's surface layer parameterizations at high wind speeds should be compared to observational data, to assess the schemes in their current state. Ideally, the schemes in the new model would produce values of fluxes and exchange coefficients that more closely match observations than the schemes in the current version of WRF. If this is not the case, the schemes should be modified to produce conditions that are more consistent with those found in strong TCs.