By using SVD method,this paper develops a multivariate linear statistical climatic forecast model to perform a short term climatic sea surface temperature(SST) forecast test in several areas of the tropical Pacific.The forecast correlation skills for the different forecast periods are examined by historical independent sample tests.The results indicate that:the SST forecast correlation skills for each season are very different in different sea areas;there is the highest predictability in the west of equatorial Pacific(Nino4 area);and the forecast performance is stable under most conditions in this areas,with an excellent skill for winter SST forecast made at the end of summer.It is demonstrated that better forecast performance would be obtained when SST and SOI are included in predictors than when single SST or SOI is included.Sometimes,the forecast skills are more excellent when the earlier successive two or three seasons SST and SOI are taken as predictors,although,most of forecast skills are higher when one single seasons SST and SOI is taken as predictors.