Abstract: One of the Microphysical variables,the rainwater mixing ratio qr,was retrieved from Doppler radar observed reflectivity by a 3D variation data assimilation system.The qr as analysis variable was determined by minimizing a costfunction defined by the difference bet ween radar observed reflectivities and their retrieve from qr,plus the difference between qr and their background field from a mesoscale model prediction.The covariance matrix of background field errors was determined by so called NMC method.The Second-order Auto-regressive (SOAR) method is used to calculate the coefficients of regressive filtertofit in with small spatial scale such as cumulus in the process of spatial transformation.An ideal ex periment demonst rates the correctness of this system and a sensitive experiment proves that the random errors of observed reflectivity have effect on the analyzedre sults.Atlast an experiment with real data from the Doppler radar at Maanshan City in Anhui Province on June 19 of 2002 was performed.The retrieved analysis variable qr in this test shows the detailed structures,which coincide with the distribution of the echopicture observed by the radar.