Inversion Methods of Three Cumulus Parameterizations for Diabatic Initialization of a Tropical Cyclone Model

Abstract

To generate convective precipitation consistent with observations at the beginning of a forecast with an atmospheric prediction model, the irrotational circulation and moisture fields must be initialized properly for dynamical balance with the input rotational and mass fields. This article describes a diabatic initialization procedure for a tropical cyclone forecast model, including the adjustment of the input moisture field with the use of ?observed? precipitation data from satellites. The condensation heating rate estimated from ?observed? precipitation data and the radiative heating rate calculated from the input data are used as the diabatic heat source to calculate the balanced irrotational wind field using a diabatic nonlinear normal mode initialization algorithm. The adjustment of the moisture field is performed by the inversion of cumulus parameterization, knowing the input precipitation rate. The inversion methods of three cumulus parameterization schemes are presented. The three cumulus schemes considered here are the Kuo scheme modified by R. Anthes, the relaxed Arakawa?Schubert scheme developed by S. Moorthi and M. Suarez, and a mass-flux scheme formulated by J. Hack. The diabatic and moisture initialization procedures are tested using a regional model for prediction of Typhoons Ed and Flo in the western Pacific during September 1990. The regional prediction model is an augmented version of the Pennsylvania State University?NCAR mesoscale model. The combined diabatic and cumulus initialization can ameliorate the problem of precipitation spinup. However, the adjusted moisture state after cumulus initialization differs considerably, depending on which cumulus parameterization is used. The inversion of cumulus parameterization requires the information of sensitivity of precipitation rate δP to a small variation of moisture δq. The sensitivity δP/δq is a measure for the behavior of different cumulus parameterizations.