Abstract: Rainy season in South China has two periods,Apr to Jun and Jul to Sep.Especially, floods of ten occured in the former rainy season,it caused heavy economic losses in locality.Therefore,it's very important to predictrainfall of former rainy season more accurately.It's well know that the abnormal change of SST can cause abnormality of the atmospheric circulations,and influence precipitation distribution. In this paper,the method of singular value decomposition analysis is used to decompose monthly mean SST field of North Pacific in winter and rainfall field of Guangdong province during the next Apr to Jun.We obtained six space distribution patterns of SST with its close related rainfall field.Therefore by using of SST distribution in winter we can estimate precipitation tendency of next Apr to Jun in Guangdong.