To build a time sequence prediction model with excluded seasonality out of the sequence with prominent
seasonal features, primary treatment of seasonality exclusion shall firstly be done, to obtain the pass of stationary test; and
then, by integrating autocorrelogram, partial autocorrelogram and AIC codes, ARMA model shall be identified and be
able to pass residual correlation test and the optimum hydrological probability distribution function shall be finally
determined. In accordance with the daily runoff of the Three Gorges from 1950-2009, the predicted daily runoff in 2010
via R software has small error compared with the observed daily runoff, obtaining good prediction performance.