The sensitivity of numerical weather models to initial conditions is well documented. In hurricane modeling, the lack of initial data has meant that the initial hurricane circulation has to be artificially introduced into the model using a so-called ‘bogus' vortex. A hurricane-like vortex of the same size and with the same maximum wind speed as the real storm is inserted in the larger scale model fields. In current operational models bogus vortices are symmetric and based on few observations. Hence they do not accurately represent the real hurricane. This has lead to inaccurate model track forecasts and very little skill in predicting hurricane intensity.

In this study, the sensitivity of model hurricanes to initial conditions will be explored further. Preliminary simulations have already demonstrated a strong sensitivity of storm intensity to the initial moisture field. Hurricane Danny (1997) was chosen due to the availability of an extensive WSR-88D data set covering landfall of the storm. This will allow improved initialization as well as validation of the model forecast. The Penn State/NCAR mesoscale model (MM5) is used. The model initial conditions, obtained from GFDL, include a symmetric bogus vortex. This vortex does not accurately represent hurricane Danny at the time of model initialization and will be removed. The available WSR-88D radar data and aircraft drop sonde data will be used to design a more realistic, but still symmetric, bogus vortex. After that, the possibility of initialization with an asymmetric bogus vortex will be explored. With improved initial conditions and fine horizontal resolution (3 km) a more accurate simulation of Hurricane Danny should be obtained. This will allow careful study of the storm's evolution during landfall and an increased understanding of some of the physical mechanisms at play when a hurricane crosses from water to land.