Numerical Hurricane Prediction Using Assimilation of Remotely-Sensed Rainfall Rates

Abstract

Rainfall rates determined from airborne radar and infrared satellite images are combined to construct a space- and time-dependent heating function for Hurricane Anita (1977). The heating is assimilated into a three-dimensional primitive equation prediction during a 12 h pre-forecast integration, after which the heating rate is computed internally by the model. The specified heating forces initial wind and mass fields toward their observed values, and produces improved 12 and 24 h forecasts of both track and intensity compared to a control integration, for which the heating is computed internally for the entire period. Calculations indicate that model adjustment during the period of heating can be viewed as a slow response of the vorticity field to continuous forcing of the divergence by the heating. The location and pattern of the heating relative to the center appear to be of greater importance than the magnitude of the heating. This may be of significance because remotely-sensed rainfall estimates are more likely to be accurate in the positioning of heavy rainfall than in its intensity. The initialization procedure appears capable of producing useful improvement in short-term hurricane prediction, particularly prior to landfall, when data coverage is best and accuracy is of greatest concern. A number of authors have noted the importance of upper-level inward eddy momentum fluxes for hurricane intensification. Calculations from the simulated storm indicate that such eddy fluxes are present in Hurricane Anita and are associated in part with an anticyclonic outflow eddy over an intense local rainfall area 300 km east of the center.