Forecast Of The Meiyu Rainfall In Yiwu City Based On BP Neural Network

The amount of Meiyu rainfall could decide how much water to be stored in reservoirs and could help to determine the exact moment of artificial precipitation operation. Therefore, aim at this practical problem, this paper constructed the BP neural network forecast model of Meiyu rainfall with the definitive climate background to analyze and forecast the Meiyu rainfall of the coming year (2010) in Yiwu city. At the same time, it had established the stepwise regression forecast model by the historical samples that consist of the same predictor, which was as a model for comparison to compare the various results. Main research results were as follows:(1) The Meiyu in Yiwu city generally existed relative change characteristics of flood period and dry season over an quasi-periodic 11-year cycle, and there were corresponding relationship of sun's activity waxes and wanes on an 22-year cycle.(2) There were relatively remarkable response relationship of the Meiyu rainfall in "Yiwu city to El nino/la Nina.(3) The 3 items factor of the northern hemisphere 500-hPa geopotential height field and the 7 items factor global SST field,which affected the Meiyu rainfall in Yiwu, respectively concentrated in the northern hemisphere middle and high latitudes and the north-central areas of northern Pacific Ocean near the equator.(4) The 3 items factor of the northern hemisphere 500-hPa geopotential height field,which affected the Meiyu rainfall in Yiwu, mainly concentrated in the northern hemisphere high latitudes, thus validated that the persist stabilization of the Meiyu period of the lower-middle reaches of the Yangtze River correlated closely to the establishing and maintaining of the ridge of high pressure or blocking high above the Urals and the Sea of Okhotsk; The prophase abnormality of SST had a strong correlation with the Meiyu rainfall in Yiwu, and the most correlated zones were as follows:the Western Pacific Warm Pool and the North Atlantic between last December to the following April, Canary cold snap zone of the north-central areas of northern Pacific Ocean near the equator between last December to the following February, and the southern Indian Ocean near the equator on January. But as a omen signal of the previous primaries predictor,the physical process which is how to affect the Meiyu rainfall in Yiwu needed to be further explored.(5) Compared with the traditional stepwise regression forecast model, the BP neural network forecast model for the forecasting of Meiyu rainfall in Yiwu city had certain advantages,which could be used to forecast the Meiyu rainfall of the coming year in Yiwu city. Finally, it had set a standard and process for the classification of the Meiyu rainfall,in order to rate and evaluate the predictive values which were the results from the BP neural network forecast model that had the better result of test forecast. That in turn modified the basic grades of the Meiyu rainfall and obtained the intervals of forecast data by the synthesis of the processing results. In operational practice forecasting, to the grading prediction of Meiyu rainfall, there were cyclical factors and the heating state above the surface between sea and atmosphere in the year, concentrating on whether the El nino/la Nina actually happened, that had to be taken into account.(6) In 2010, the natural precipitation in this region could basically meet the water demands of actual productive and domestic uses in summer drought season. Therefore, it did not need to take artificial precipitation operation in Yiwu city this year.