Features and patterns of spatial and temporal distribution of rainfall during Sep and Oct in Guangdong province have been estimated by means of principal components analysis (PCA). The results show that the distribution of the first four eigenvectors represent four main rainfall types

which account for 76% of total variance of rainfall. The dryest(or wettest) years in Guangdong are characterized by the first weighting coefficient of a temporal function with absolute values heigher than 5 0. The results also show that the areas with more rainfall are Leizhou peninsula

Zhujiang Delta and Shantou region

and the areas with less rainfall are Shaoguan

western part of Zhaoqing and Meizhou.