Abstract: A method for creating a long range flood prediction model is developed by using the principal components of factor field.First,the principal component of fields of 500 hPa height and Pacific sea surface temperature which have important effect on rainstorm and flood of region are analysed with their principal component as predictor.Then,compulation of correlation and stepwise regression is operated to creat the model.The compulation procedue are quite complete.Applying that model to predict the yearly maximum stream content of Beijiang River flow region shows that the model is effective.