Study On Medium-and-long-range Prediction Of Water Regime In Shanghai

In this paper, a set of medium-and-long-range hydrologic forecast models were established for predicting water regime in Shanghai, including three hydrological series, annual rainfall and rainfall during flood season for one representative station and annual maximum water tide level at Huang-pu park. The prediction models includes two parts .One is quantitative ones which are mainly studied in the paper , including AR(p),GM(l,l) ,modified GM(1,1) and Threshold AR modeling; The other is Markov chain qualitative modeling. In the study of quantitative forecast models, composition analysis for three different time series are firstly investigated by statistic test methods .The conclusions are as follows: there is no evident tendency for above three series, and evident periodicity exist only in precipitation series. After the removal of periodicity component ,the forecasting is made by AR(p )model. In order to analyze whether the prediction accuracy is improved after the removal of periodicity ,the forecasting is made by AR (p) model using original series data .The results indicate that accuracy may not be evidently improved. Secondly, the other forecasting is made by using GM( 1,1 )~the grey model and a modified GM(l,l)-heuristic method of using grey system theory .The prediction results shows that the modified GM(1,1) model is slightly better than GM(1,1) , and both of them are better than AR (p)model. In order to improve further the prediction accuracy of the series studied, a TAR model is also used to make the prediction. Although the calculation of TAR model is complicated, the result of forecasting is better than the others. Finally, a qualitative graded prediction model-Markov chain qualitative modeling, is established for above two precipitation series by using meteorological graded division criteria .The forecasting result is good. The study is valuable to the flood defense decision-making in Shanghai.