Study Of Precipitation Prediction Based On T639L60 Numerical Prediction Products Around The Bohai Bay

Precipitation is one of the important weather phenomena, timely proper precipitation can provide favorable conditions for agricultural production, but otherwise, heavy precipitation will bring disaster. Precipitation forecast is very important in conventional operation, so how to improve the accuracy of precipitation forecast is one of the aims of meteorologists.Numerical weather prediction(NWP) method is currently one of the main methods in weather forecasting, in recent years, with the improvement of numerical model resolution, observation and data assimilation technology, the numerical model can provide a large number of forecast products, become the most important reference basis for making the daily weather forecast. A large number of studies have shown that, reasonable interpretation to numerical forecast products based on statistical methods can greatly promote the forecast accuracy. This paper chooses the bohai rim (including bohai) as the study area, every 3 hours rainfall in the next 72 hours as the forecast object. It analyzed the climate characteristics of precipitation in detailed and summarized the strong precipitation weather system at 500hPa around the bohai bay, then it established precipitation forecast factor library, Finally, the T639 model outputs were interpreted based on both regression estimate of event possibility and BP neural network methods. The main results are as follows:(1)Using high-resolution fusion precipitation products and daily precipitation data, the spatial-temporal distribution of precipitation around the bohai bay were analyzed in detailed, the result shows that:the average annual rainfall in the area is increase from northwest to southeast, the average annual rainfall in eastern Liaoning province is the largest, followed by south of Shandong, Shandong peninsula, and the bohai sea, inland in the west of the religion has less annual rainfall. Since 1982, the average annual total precipitation in the region shows the change trend of "increase-reduce-increase", it has a slightly increasing trend from 1982 to 1988 and presents a decreasing trend from 1988 to 1988 while it shows a larger increasing trend from 2002 to 2012; Precipitation mainly concentrated from April to October, it accounted for 92% of its total annual rainfall; Precipitation characteristics in different areas have obvious difference, overall, moderate rain in the region contribution to the total rainfall is the biggest; At the same time, precipitation also has obvious daily variation characteristics than the morning and evening rainfall is larger, less rainfall around noon.(2) Using the daily precipitation data of 60 meteorological stations around the bohai bay and the NCEP/NCAR data, the 500hPa weather system when regional heavy rain has carried in nearly 30 years were summarized. It can be classified as the upper trough, subtropical high borderline, low vortex type and short wave type four categories. Based on the analysis of the commonness and difference, the diagnosis model of precipitation around the bohai bay was established. Then the paper constructed the precipitation forecast factor library comprehensively by T639 numerical products outputs.(3)The precipitation forecasting model based on the regression estimate of event possibility method was established in this chapter. The moisture elimination can effectively reduce the false alarm rate of precipitation forecasting. Using the T639 numerical product data and precipitation data from 2010 to 2012,the every 3 hour precipitation forecasting equation were established on different lattice of four seasons respectively. Comparison analysis shows that the forecast result of he regression estimate of event possibility can effectively improve the forecast accuracy. Although he forecast effect of different parts is slightly different, but generally speaking, the summer forecast effect is the best, followed by spring and autumn.(4) Established the BP neural network forecasting model based on nonlinear dynamic system for precipitation. The forecast results of two methods and T639 model output products were compared. The results shows that:Whether in the spring, summer or autumn, event probability regression model has better effect than neural network forecasting model in TS score, S score and empty forecast, but in terms of omission of precipitation, the BP neural network model is much better; Both of two kind of forecasting models, the forecast results are the best in summer, the empty rates and omission rates are significantly decreased, the omission rates in spring and autumn are relatively high; Overall, it have better forecast effect significantly in more precipitation seasons (summer) and precipitation areas(eastern Liaoning) both of two kinds of forecasting models.In short, Using the fusion of high space-time resolution precipitation and T639 products, the characteristics of precipitation in bohai rim were analyzed in detailed and objectively in the paper, and the precipitation forecasting models for every 3 hours were established based on the egression estimate of event possibility and BP neural network methods. It is useful not only for interpretation of T639 numerical products, also have certain guiding significance for short-time rainfall forecast while the forecast accuracy has a certain improvement.