Abstract: Heavy rainfall is mainly generated by the mesoscale and microscale synoptic systems that are short life cycle.Presently,the forecast with this kind of synoptic systems can only rely on the combining real-time monitoring by satellite and radar and extrapolating by the conceptual model of the mesoscale systems.Because the terrain is very complex in the Nort hwest China,the terrain creats radar shadow that have influence on estimating rainf all rate.Therefore,the information of GMS-4 satellite infrared and visual stretched image have been used.These data are converted and reduced.The digitial terrain data of corresponding grid mesh of the satellite data are a available factor for use.The multilevel stepwise discriminatory model is used to estimate rainfall rate grade.At last,a set of software system which can be performed false color image display and process of rainfall rate field at micro computer has been created.This system has been satisfied a great need for now casting.The result has been proved that the estimation of area and shape,of hourly rainf all rate grade that is more than light rain,is generally consistent with the real-time rainfall.