Application of nonlinear stepwise regression in prediction of annual discharge at Yichang station

In order to analyze the influences of introducing nonlinear relationship on the precision of middle-and long-term hydrological prediction,the mean annual discharge at Yichang station on Changjiang River was employed for the hydrological prediction based on the nonlinear stepwise regression model through log-transformation.The predicted values were compared with those by the traditional stepwise regression models.The results show that the annual discharge at Yichang station has a remarkable nonlinear relationship with its discharge in July,August and October in pervious year,and it is sensitive to the activity of Pacific Subtropical High and El Nino Effect.The precision of the fitting values of the nonlinear stepwise regression model and the predicted values is higher than that of the linear models.Introducing the nonlinear relationship may improve the model precision to a certain extent and is applicable.