Abstract: In order to overcome the limitation of traditional rainfall patterns in reflecting intraseasonal variation,new anomalous large-scale rainfal regimes are identified in this paper,by introducing the phase space interms of a dynamical point of view,and estimating probability density functions in the phase plane spanned by the first two EOFs,as well as examining local maximum regions.The analysises show that the new rainfall regimes reproduce not only the three kinds of traditional rainfall patterns and the others,but also describe the intraseasonal features of anomalous precipitation distributions,which evidently can use the daily precipitation anomalies involving the very important low-frequency components in short-term climate forecasts.On the other hand,the rainfall regimes from the 10-day and 20-day low-pass filter data,respectively,are approximately accordant,though there exist also some obvious differences between them such as the characters of classification,duration,and intraseasonal distribution,which exhibits the distinct contributions of the two low-frequency scales to rainfall regimes.Furthermore,in the present paper,the climatological distributions of rainfall regimes are in certain agreement with usual monsoon rainbelts in the flooding seasons,and the formation mechanism of rainfal regimes are closely associated with the corresponding summer monsoon anomalies.By investigating the visual and intuitional problems of 2-dimension phase plane,it is verified not only the rationality and feasibility of identifying large-scale anomalous rainf all regimes in the EOF phase space,but also presented the evidences of identifying rainfall regimes in multi-dimensional phase space.