The Impact of Horizontal Grid Spacing on the Microphysical and Kinematic Structures of Strong Tropical Cyclones Simulated with the WRF-ARW Model

Using the Advanced Weather Research and Forecasting numerical model, the impact of horizontal grid spacing on the microphysical and kinematic structure of a numerically simulated tropical cyclone (TC), and their relationship to storm intensity was investigated with a set of five numerical simulations using input data for the case of Hurricane Rita ( 2005). The horizontal grid spacing of the parent domain was systematically changed such that the horizontal grid spacing of the inner nest varied from 1 to 5 km by an increment of 1 km, this while keeping geographical dimensions of the domains identical. Within this small range of horizontal grid spacing, the morphology of the simulated storms and the evolution of the kinematic and microphysics field showed noteworthy differences. As grid spacing increased, the model produced a wider, more tilted eyewall, a larger radius of maximum winds, and higher-amplitude, low wavenumber eyewall asymmetries. The coarser-resolution simulations also produced larger volume, areal coverage, and mass flux of updraft speeds >= 5 m s(-1); larger volumes of condensate and ice-phase particles aloft; larger boundary layer kinetic energy; and a stronger secondary circulation. While the contribution of updrafts >= 5 m s(-1) to the total updraft mass flux varied little between the five cases, the contribution of downdrafts <=-2 m s(-1) to the total downdraft mass flux was by far the largest in the finest-resolution simulation. Despite these structural differences, all of the simulations produced storms of similar intensity, as measured by peak 10-m wind speed and minimum surface pressure, suggesting that features in the higher-resolution simulations that tend to weaken TCs (i.e., smaller area of high surface fluxes and weaker total updraft mass flux) compensate for features that favor TC intensity (i.e., smaller-amplitude eyewall asymmetries and larger radial gradients). This raises the possibility that resolution increases in this range may not be as important as other model features (e.g., physical parameterization and initial condition improvements) for improving TC intensity forecasts.