Dynamic Initialization of a Three-Dimensional Primitive-Equation Model of Hurricane Alma of 1962

Abstract

A technique of dynamic initialization previously studied with two-dimensional models is applied to a forecast of Hurricane Alma (1962) over a 24 h period. In the dynamic-initialization experiment, first-guess horizontal winds and the surface pressure in a four-layer primitive-equation model are driven, or relaxed, toward the observed fields during a 12 h preforecast integration, while the temperature and moisture fields are allowed to adjust freely. The observed data are successfully assimilated into the model during the dynamic initialization. A reasonable 12 h forecast is produced in terms of storm track and maintenance of storm structure. The reduced dynamic imbalance at the beginning of the forecast utilizing dynamic initialization is beneficial. First of all, because the model state is not obscured by nonmeterological oscillations, this portion of the forecast is more useful than that of a forecast without dynamic initialization. Second, the meteorologically important features of the atmospheric circulation are not modified as much by a large mutual adjustment of the mass and momentum fields at the outset of the forecast. Also, the additional integration time in the dynamic-initialization experiment permits the small-scale forcing processes in the model to generate a finer scale precipitation band than present in the initial conditions. The forecast of the storm track was slightly better during the first 6 h but was slightly worse by the end of the forecast, in the experiment with dynamic initialization, however. This latter result, together with the fact that these experiments pertain to only one case, makes the conclusion of the superiority of the dynamic-initialization scheme tentative.