The impacts of resolution and boundary layer parameterization on the structure of the wind field in high resolution simulations of Hurricane Isabel (2003)

We present detailed analyses of high resolution simulations of Hurricane Isabel (2003) initialized from GFDL model initial conditions at 00Z on Sept. 12th, 2003. The effects of horizontal resolution, vertical resolution, nested grid size, and boundary layer parameterization on the accuracy of the simulation are illustrated. In particular, the vertical structures of the inner-core wind fields are evaluated by direct comparison to the wind fields observed with pseudo-dual-Doppler radar analyses, and the boundary layer winds are evaluated by comparisons with dropsondes and stepped descent observations from CBLAST flights into the storm. Simulations with 4 km resolution and 1.33 km resolution are considered. On the 12th and 13th of September, the simulated boundary layer structures compare remarkably well with those observed, with the YSU PBL scheme using the Donelan et al. correction for the drag coefficient giving the best results. Increasing the resolution in the boundary layer does not give improved results. The azimuthally averaged middle and upper-level wind fields also match quite well. However, comparison with the observed wind fields shows a consistent bias in all the simulations, that the maximum winds at the top of the boundary layer are much larger than observed. In fact, even though the simulated surface winds are lower than those reported in the best track data set, the peak winds aloft are far larger than what was observed by dropsondes. This bias appears to be common to many hurricane simulations. We investigate modifications to the parameterizations that might improve these defects.