Nonlinear characteristics of the precipitation anomaly over china during the last five decades were investigated by applying a back-propogation neural network-based nonlinear principal component analysis(NLPCA) method to the seasonal rainfall anomaly data.The results show that the NLPCA could describe more general characteristics of the low-dimension structure of precipitation.The precipitation anomaly of each season has a certain extent nonlinear space pattern,but the nonlinearity of autumn and winter precipitation is weaker than spring’s and summer’s.The spatial pattern based on the leading NLPCA mode approximation manifests obvious asymmetry when the nonlinear principal component takes the maximum and minimum value.The leading NLPCA made approximation is more realistic to the original precipitation anomaly data than the leading PCA mode approximation,and changes with season.