VARIATIONAL DATA ASSIMILATION IN THE TROPICS USING PRECIPITATION DATA.1. COLUMN MODEL

This paper describes initial effort in the development of a four-dimen sional variational data assimilation (4D-Var) in the tropics using pre cipitation data derived from remote sensing. The method of 4D-Var usin g precipitation data is formulated, and modifications to the parameter ization schemes of moist processes to remove zeroth-order discontinuit ies are described. Variational data assimilation experiments are carri ed out using a column model to investigate the problems caused by disc ontinuities in parameterization schemes and assess the impact of assim ilating precipitation data in the tropics. It is found that variationa l data assimilation with discontinuous parameterization schemes exhibi ts large fluctuations during the minimization process, slow convergenc e rates, and large analysis errors. The fluctuations become much more serious when precipitation data is assimilated. Precipitation data is very useful to estimate divergence in the tropics, provided that the t emporal resolution of the data is sufficiently high. However, its impa ct on the analysis of temperature and moisture is not clear in the col umn model assimilation experiments, possibly due to the absence of hor izontal advection.