Studies On Forecasting Technology And Operational System For Three Typical Dangerous Weather Affecting Aviation

Aviation meteorological assurances have attracted wide attention from the airlines and meteorological scientists and technology workers around the world since1920’s. This is mainly because that weather is one of the most important factors which affect aviation safety, normal and efficiency. It’s the aviation meteorological service’s responsibility and task to provide timely and accurate weather forecast for airlines and pilots to enhance flight safety. With the rapid development of the aviation industry, a newer and higher requirement for aviation weather service is needed. Thunderstorms are known as the killer of summer flight in many weather factors. Low visibility caused by fog often results in the closure of the airports. The bump caused by vertical wind shear is the most dangerous mereorological factors which affect flight safety. Gale not only affects the aircraft’s normal taking off and landing but also damages airport facilities. Therefore, the in-depth study on the short-term forecast for three kinds of dangerous weather has important practical significance for improving aviation weather forecasting operational capacity. In this paper,the temporal and spatial distribution of three typical aviation dangerous weather for nearly30years were analyzed and their main weather patterns were summarized which can provide effective supports to understand the climatic probability of three typical aviation dangerous weather, pathogenesis and to improve the forecast level;after that analyzed the main physical factors of the three typical aviation dangerous weather and establish the dynamic-statical model using the event probability regression and BP neural network, in addition, the forecasting performancies of two model were tested. at last, the short-term forecasting operational system of three typical aviation dangerous weather which can be used for aviation meteorological services was built basing on software development techniques. The system is being used for quasi-operational running.The main conclusions are as follows:1. The results show that there were different spatial and temporal variations of thunderstorms, fog-haze and gale in the past three decades in mainland China in the context of global extreme weather events’ increasing. Thunderstorms are more frequent in the southeast coast of China, southwest and eastern Tibetan Plateau region. In the past30years the annual thunderstorm days showed two periods of higher value and a lower period. Fog occurs frequently in south China and the coastal areas such as Liaodong Peninsula, Shandong Peninsula, Fujian, Jiangsu and Zhejiang coastal region, Sichuan Basin and the southwest of Yunnan. The number of foggy days in the south of the Yangtze River region is more significantly than the northern part of the Yangtze River, foggy days in the past30years showed a decreasing trend, howerver the number of mist days showed an increasing trend. Spatial distribution of Haze day showed more in east China and less in west. In east China area, there are three more frequent centers as follows:the middle-lower reaches of the Yangtze River, northern and southern China. Haze days showed an increase trend after2001. Gale mainly occurred in the northern China; such as northern Xinjiang, central and western Inner Mongolia, eastern Tibetan Plateau and the eastern coastal region.2. The main physical factors of the different forecasting objects had been selected by using statistical analysis. The top factors in the thunderstorm forecasting equations are water vapor content, dynamic conditions and unstable energy. The top factors in the fog-haze forecasting equations are water vapor content, wind speed and stability index. The top factors in the gale forecasting equations are U, V component of the model output, the middle and lower troposphere wind speed calculated by meridinal wind and zonal wind. In addition, the strength of the trough, temperature and vorticity advection also account for a larger proportion.3. Established the probability regression forecasting models basing on linear regression theory for three typical dangerous weather and the models’s forecasting performances were tested.the TS score is higher than the climate probability value in different regions and different prediction time. The heidike score of thunderstorms forecast was higher in south China, the eastern coast and northeast. The sites which have good prediction effect on haze and mist that horizontal visibility is less than10km haze are mainly distributed in the southeast part of northwest China, the northern part of north China, the eastern part of Huanghuai, and the middle and eastern past of jiang-huai River in spring. In summer, there are two regions, one is southern part of northeast China and north China, another is the middle and western part of south of the Yangtze River. In autumn and winter, the sites having good forecast performance distribute in the southern part of northeast China, north China,Huang-huai area and eastern part of Jiang-huai and better in winter than autumn in the southeast coast and south China. The spatial distribution of fog’s score demonstrates good consistency with the spatial distribution of the number of foggy days.The score is higher in northeast China,huang-huai,eastern part of south of the Yangtze River and southwest region. The gale’s forecast performance shows better overall in north China and the score is higher in the southeast coastal in autumn。4. Established the BP neural network forecasting models basing on nonlinear dynamic system. for three typical dangerous weather.the models’s forecasting performances were tested and compared with the REEP forecasting models.Overall, because of the same input factors of the two models, the spatial distribution of two prediction model scores is basically same. For thunderstorm, With the extension of perdition time the score significantly reduced fast, the score at night is lower than that during the day, and the neural network prediction model is obviously better than regression estimation of event probabilities, which is more obvious in the day time forecast period. BP Neural Network model is better for fog which horizontal visibility less than lkm. The regression estimation of event probabilities showed better performance on gale forecasting. The cases forecast results analysis indicates the regional characteristics of the three types of weather forecasts have more consistent with observations.5. Developed short-term forecasting system and achieved a quasi-operational running. According to characteristics of short-term forecasting task for the security of aviation weather services and forecast demand, Depending on the interpolation and explanation method of numerical forecast products and software development technology, the operational forecasting system was developed. The system is divided into five modules including forecasting model construction, data acquisition and storage, basic and characteristic physical element’s computing, automatic generation of forecast data, geographic information data access and forecast information displaying. The system has many advantages such as modeling approach reasonably, operating easily and having a higher degree of objection and automation etc. It can help weather forecasters to make the forecasts of three types of dangerous weather efficiently. The system has been integrated into the aviation weather hazard forecasting support system for a quasi-operational running.