Through the study of six-hour raingauge data from about 120 stations in East China and the corresponding 4400 multi-channel cloud imagines of GMS-5 during the summer of 2001—2002,it is found that the temporal and spatial variation of the TBB and its gradient of the infrared channel,the TBB difference between the infrared and the vapor channel have certain function to eliminate error caused by cirrus,even to solve the absence of visible-channel data. Based on the study above,the rainfall estimate equations between six-hour rainfall and GMS-5 data are obtained by using double-judging & double-MOS regression method,which can be used to estimate six-hour rainfall. In order to further eliminate error caused by deep cirrus or special terrain,a day-by-day rainfall estimate equation by using real-time data is set up. The operational application in Shanghai Meteorological Center shows that its average estimate accuracy reaches about 70%.