| In this paper, the Hunchun River measured during 1960-2010 precipitation data and global sea level pressure, sea surface temperature, geopotential height data as the research object, using statistical analysis method combining with the knowledge of hydrology, Hun chun River, the characters of precipitation time series analysis and on global sea level pressure, sea surface temperature and geopotential height three factors and flood season precipitation correlation research, Finally, based on BP neural network Predict the precipitation of the main flood season in Tu men River downstream. The results show that:(1) The precipitation of the main flood season in Hunchun river is about 3～4 years,7 years or so, and the period of 25 years. The precipitation of the main flood season has the characteristic of multi-scale, and the trend of the main flood season is decreasing. The time series of the decomposition of the ingredients are relatively stable fluctuations in the average, the separation effect is better.(2) The relationship between SST, air pressure field and 500hPa height field and the precipitation of the main flood season in Tu men River downstream are mainly based on the inter annual variation, and the inter annual variation is the second. The best SST factors are mainly distributed in the westerly drifting area and the South India ocean, and these seas are negatively related to the precipitation of the main flood season in Tu men River downstream. The best atmospheric pressure factor is mainly concentrated in the Mongolia high pressure zone and South Africa depression area, which is negatively related to the precipitation of the main flood season in Tu men River downstream. And best 500hPa height field factor mainly concentrated distribution in Africa over the height field and main flood season precipitation was negatively correlated and with the lag time prolonged significantly related area showed a trend of decrease.(3)The effect of the prediction of the advanced six months based on the BP neural network is the best(average relative error is -0.012), Within the precipitation forecast permit error (0.2), the model can provide certain reference for the prediction of the main flood season precipitation in Tu men River downstream. |
