The atmospheric humidity are estimated with Doppler radar echo intensity. The relative humidity values are equaled to 100% in the areas of the echo intensity greater than 40dBz and relative humidity values are equaled to 90% in the areas of the echo intensity less than 20dBz, the relative humidity values are kept to linear correlation with the echo intensity in the areas of 20dBz-40dBz. The corresponding temperature values are gained by the control experimentation. Then the correspond dew point temperature values are calculated with iteration method from the humidity and temperature. Through these processes, radar echo intensity values are transformed to the variable values of the numerical model MM5 and are use in the MM5. The simulation experimentations are made to validate the method, the example is a heavy rainfall event in Quanjiao area, Anhui Province on the 13 July 2001.The background fields use the T213 data. The initial fields of control experimentation are evolved by the analysis from routine sounding and surface data. The echo intensity data of Hefei Doppler radar observations come into being the initial fields of MM5 sensitive experimentation. In this test the horizontal grids are 103×103,the distance is 10km, There are 33 levels at Perpendicularity, the central position is located at the point (31.00°N,117.66°E). The result shows that predicating precipitation areas are improved obviously and indicates that this method is feasible.