High-Resolution Radar Data Assimilation For Hurricane Isabel (2003) At Landfall

摘要

Major damage caused by hurricanes occurs over land during and after landfall. Accurate predictions of winds and precipitation in and around hurricanes at or near landfall are therefore of vital importance for hurricane preparation and damage mitigation, yet they continue to present a challenge for the hurricane research and numerical weather prediction (NWP) communities. This is, in part, due to rapid changes in hurricane intensity and structure during landfall associated with multiscale dynamical and physical interactions in the hurricane core regions and outer spiral rainbands, and also associated with sudden changes of surface conditions. In this study, we demonstrate the capability of improving predictions of hurricane intensity and structures near landfall by assimilating high-resolution, three-dimensional observations from land-based radars in the landfall regions into a mesoscale NWP model. The landfall of Hurricane Isabel on the east coast of the United States in 2003 is the focus of this study. Observations of Dop-pler radial velocity and reflectivity from five Doppler radars in the landfall region were collected and assimilated into the Navy's Coupled Ocean-Atmosphere Mesoscale Prediction System in a variational data assimilation framework. Four cycles of hourly radar reflectivity data assimilation effectively correct the overprediction of hydrometeor fields by the model, and move the maximum reflectivity regions to the observed locations. Better hurricane structures, including increased maximum wind speed, a tighter inner core, and better organized outer rainbands, are obtained by the radar radial velocity assimilation. Much-improved forecasts of 24-h accumulated precipitation during and after hurricane landfall have also been achieved by the radar data assimilation. The positive results from this study indicate the potential for improving hurricane intensity and structure forecasts by assimilating radar observations into NWP models.