STEADY-STATE HURRICANE INTENSITY IN THE WRF MODEL: COMPARISON TO MPI THEORY AND SENSITIVITY TO PBL AND SURFACE FLUX PARAMETERIZATION

Despite significant improvements in numerical prediction models and computing technology, numerical model predictions of hurricane intensity have improved much more slowly than track predictions. The Weather Research and Forecast (WRF) modeling system is currently under development by the U.S. weather research and operational modeling communities. In addition, a high resolution, coupled air/sea/land hurricane version of WRF (HWRF) will replace the GFDL hurricane model at the NHC in the near future. Before this transition can occur, it is important to understand limitations with the current model physics in WRF that will need to be addressed. This research is designed to (i) test the ability of different Planetary Boundary Layer schemes in WRF to intensify a tropical system in an idealized testing environment to its theoretical maximum potential intensity (MPI), and (ii) analyze the individual PBL scheme formulations in order to gain an understanding of how the different parameterizations influence the model results under the extreme conditions accompanying tropical cyclones.

Emanuel's MPI theory provides a quantitative estimate of the maximum intensity that a tropical cyclone could reach if certain atmospheric/oceanic conditions were satisfied. These conditions typically are not found in the natural environment, but can be satisfied in an idealized model environment. In order to more effectively judge the accuracy of each PBL scheme, an environment was created that was consistent with the assumptions underlying Emanuel's MPI theory. An initial vortex obtained from gridded analyses prior to the formation of hurricane Ivan (2004) was inserted within an observationally-based, idealized tropical testing environment, and WRF simulations were run for 20 days until a quasi-steady maximum intensity was attained. The sensitivity of the simulations to PBL scheme choice was examined. Preliminary results indicate that both PBL scheme choices produce simulations where the vortex exceeds the theoretical MPI, with the Mellor-Yamada-Janjic (MYJ) PBL scheme generally producing a system with a lower central pressure for simulations featuring 20 km grid spacing and using the Kain-Fritsch cumulus parameterization scheme.

The simulated storm's strength and structure showed marked sensitivity to the PBL scheme choice. In order to understand these differences, values of fluxes and other variables used in surface layer parameterizations are examined. Additional experiments document sensitivity of these results to horizontal and vertical resolution, as well as the inclusion of sea-spray effects.