A Variational Method for Doppler Radar Data Assimilation

A variational data assimilation method for meso-scale models is developed. Radar data and other observations are used as the observation constraints. The equations of a meso-scale model are imposed as the weak constraints and the analysis of the lastest assimilation cycle (referred to as 'last analysis' hereafter)and the current analysis are used to determine the time tendencies and other terms in these equations. The model prediction valid at the current time is used as the first guess.

The method is tested on model-generated radar data and last analyses with varying degree of errors. It is shown that when observations, last analysis and constraint equations are accurate, the basic variables (wind, temperature and pressure) can be analyzed quite accurately. The larger the errors in the observations, last analysis and equations are, the bigger are the analysis errors. If the errors in the observations, last analysis and equations are too large, the analyzed variables are unreliable.