Abstract: A variational assimilation technique for the retrieval of wind fields from Doppler weather radar data is introduced. The radial velocity (RV) is the unique signal to be used directly in threedimension variation assimilation system to retrieve wind-field from Doppler radar data. Since wind vector has two or three components, the radial velocity, as one component of wind vector, can not provide enough information for retrieving wind vector. Thus, if three-dimension variation assimilation system only used to assimilate the radial velocity, the indeterminacy would bring some mistakes into the retrieved wind-field. The assimilation technique suggested in this paper, not only assimilates the radial velocity, but also assimilates the information about the movement of radar echo. The key point is to transform the movement of radar echo to a new radar measuring variable -- “seeable velocity”(SV). Because the SV is possessed of the information of wind, more information is added to three-dimension variation assimilation system, and the indeterminacy that would come forth in retrieving wind vector by using RV only was overcome effectively by combining RV and SV. Using CMA GRAPES-3Dvar and CINRAD data, some tests were performed. The results showed that the method of retrieving wind-field was useful and effective for obtaining the construction of the weather system, and the scheme was also useful to obtain initial fields of model. It is an effective technique for meso-scale numerical weather prediction.