It is very effective to improve new generation weather radar application that re liable 3D wind fields are retrieved from real time radar radial velocities. The retrieved wind field can help forecaster to identify the mesoscale structures. T he potential usage of 4DVAR assimulation technique with pure dynamical process t o retrieve wind field in real time is examined by using the Doppler radar data in Guangzhou and Jinan. It is argued whether the we t process needsto be input into cloud model, and what should iteration number, retrieval area, background and initial fields be, etc. In addition, the retrieva l results are analyzed from different aspects including wind field structure, co mputer time, mean square deviation etc. Tests show that there is little difference between retrieved results from th e dry and wet 4DVAR systems. Given a background field, the basic characteristi cs in lowlevel wind fields can be presented from dry model by 1520 iteration s, and model results are high sensitive to initial and background fields. Under the condition of background field, dry results iterated 1520 times can be effecti vely operated, which are beneficial to improve the accuracy of meso , micros cale weather system forecast.