Abstract:Typhoon is one of the most important disasters in the southeast region along oce an in China, and it can cause strong rainstorm and heavy rainfall. Based on the opinion of stressing typhoon type rainstorm forecast importance, a new forecast technology SMAT (i.e. synthetic multilevel analogous forecast technology) is use d to study typhoon caused rainstorm in the southeast region of Zhejiang Province. Since it is seldom used as analogous elements in pr evious analogous forecast method, the numerical model integral products are used in the third level forecast in SMAT. In SMAT, similarity degree between samples is described by meteorological elements combination, which makes further progre ss than using single element to describe similarity in earlier analogous forecas t. Day to day sounding data of National Meteorological Center, weather map dat a and surface observation data in study range during 1990－2002 are used， and f o recast model of typhoon rainstorm （SMAT Typhoon） is completed. New forecast s ys tem’s research process，study result and key parameters are introduced, which c an demonstrate strong capability of forecast model. CSI (forecast successful ind ex) is a key index in evaluating rainstorm forecast level, and its mean value is 0.35 in the current field. The value will be declined in heavy rainstorm foreca st. Results show that CSI of rainstorm trial is larger than 0.7, and POD (lo st forecast rate) is 0 in exploiting model phase. In the test phase of forecast mod el, CSI can also be larger than 0.45. Moreover, according to COR (forecast p reci se rate, 0.72) and POD (0.26) of SMAT Typhoon, they are also better than those from other work. According to effect analysis of trail and revision in establish ed model, it shows that the meteorological elements are very important while ana log range is less important to trail effect in describing analog degree. This can di rect the later similar trail: based on the prime evaluation of various elements , physical elements should be paid much more attention instead of study range. W hen selecting the key range, we should pay more attention to its weather type. M oreover, the main causing rainstorm system should be taken more account wether t he main system is deep or not when selecting analogous elements. In the end, some important physical elements are checked out to instruct operational foreca st. All of these show that SMAT model has strong forecast ability on the averag e.