| The numerical weather forecast has been the important ways and tools in the prediction of weather.As the description of various physical processes are more realistic and the resolution of numerical prediction model is also rising, the numerical forecasting products constantly enrich and improve. It has been one of the main directions that how to realize the resolution of numerical prediction in the development of numerical prediction model. In this paper,it has examined the EC thin, JAPAN-thin, T639high resolution numerical model and WRF mesoscale numerical model of numerical prediction product application in the Jilin province short-term business forecast effect in2011-2012, according to the China Meteorological Administration issued the Short-term Weather Forecast Quality Inspection Method. The results show that:(1) The precipitation properties are single in winter and summer, each model of small scale precipitation prediction are similar.For heavy precipitation,the EC-thin model is a good reference. But for the transformation season of rain and snow, there are great difference among each model.JMA model forecast performs fine on the prediction of heavy snow in spring, the accuracy rate of the big to blizzard snow is above30%, has good reference value. JMA model forecasts well for the autumn rains.(2)These patterns have good prediction ability for high level cold eddy and typhoon which have obvious precipitation process with strong signal; For obvious precipitation process because of the upper trough, EC-thin model has good forecasting ability. These models have high empty non-response rates for shear type heavy rain process, the forecast result is bad so there is little reference.(3)The prediction error in obvious precipitation process among these numerical models shows that the error for bigger is more than for smaller of high resolution numerical model for most obvious precipitation forecast. For more than25mm scale precipitation,the average absolute deviation of JMA model is much higher than other several models, and the error grows bigger with the aging increasing.(4)The temperature prediction accuracy of less than2results of these numerical models show that:the EC-thin model has a little higher prediction accuracy than other several patterns throughout the year and each season, with high reference value, the prediction accuracy of JMA model is low. Within72hours, the accuracy of highest temperature in the Midwest is significantly higher than in the Southeast, the difference of the minimum temperature accuracy is bigger. JMA model prediction accuracy of the lowest temperature in Yanbian region is higher.(5)The mean absolute error of temperature forecast results show that:the average absolute error of minimum is lower than the maxmum temperature, especially in autumn and winter mean the absolute error is significantly larger than the other two season. The mean absolute error of lowest temperature in summer under2。C. The mean absolute error of the lowest temperature in winter is between3-5。C, especially the JMA model’s mean absolute error is between4。C to5。C, the reference is poorer. With the forecast time increasing, forecast average absolute error of these numerical models grow obviously.(6) For the contrast analysis of average error for forecasting temperature shows that:the minimum temperature forecast of several patterns is higher in autumn an winter, and lower in spring and summer. The highest temperature forecast are all slightly smaller in four seasons, the forecast deviation of EC-thin model is small with a good reference, JMA model is larger.Through the analyzing of several numerical models forecast product, on the one hand,it can provide a reference basis for forecasters to use the numerical model scientific and effectively to make fine forecast. On the other side,it also can improve the using product of numerical mode of, provide methods to improve the forecast accuracy. And it can also provide the basis for local WRF model in numerical sensitivity test. |
PDF Full Text Request