| In order to establish a classified forecast model of thunderstorm weather over Yinchuan Hedong Airport,analyzed the mesoscale weather background for the small sample data of thunderstorm weather in the severe arid area with an average annual precipitation of only 180 mm.The operational mapping of the data was carried out,and inspected the normal distribution characteristic of the data with coefficients of Kurtosis and Skewness,Chi-square and Q-Q plot.Linear,nonlinear and artificial intelligence are used to build the model,and the test results show that the forecasting ability is better.Because Yinchuan Hedong Airport is located in the city with the lowest annual precipitation in all provincial capitals of China,and the average annual precipitation is only 180 mm,the airport is built late and the data records are few,the thunderstorm weather occurs over Yinchuan Hedong Airport is analyzed by various statistical methods,including annual,monthly,ten-day,pentad,hourly and duration.The results are as follows:(1)the annual thunderstorm days over Yinchuan Hedong Airport shows an increasing trend in fluctuation.(2)The thunderstorm mainly occurs in July,the month of high thunderstorm occurs at the same time as the summer peak of flights over Yinchuan Hedong Airport,the thunderstorm has a great impact on flights.(3)There are more thunderstorms in the later part of July,and more thunderstorms in the sixth pentad of July.(4)15:00-16:00 and 20:00-21:00 are the two most frequent periods of thunderstorms over Yinchuan Hedong Airport.(5)In the cases of thunderstorm over Yinchuan Hedong Airport,56% of the cases lasted no more than 1h,41% of the cases lasted 1?3h,and 3% of the cases lasted more than 3h.The mesoscale system analysis of thunderstorm weather over airport was carried out.The following conclusions are drawn:(1)the thunderstorm with gale,there is a high possibility of a significant airflow convergence zone at 500 hPa.(2)The thunderstorm with rainfall and gale,there are mesoscale systems with north-south and east-west orientation in Mongolia at the northern of Yinchuan,and the systems overlap with each other.There is a 24-hour cooling zone around Yinchuan.(3)The thunderstorm with rainfall,mesoscale systems near Yinchuan are generally north-south orientation,there are also north-south orientation mesoscale systems in Mongolia at the northern of Yinchuan,which overlap with each other,and there are cold centers in the western of Yinchuan at 700 and 500 hPa.In the practice of local aviation meteorological operation,the original value of conventional meteorological elements can't directly reflect the essential needs of aviation meteorological support.In order to objectively and quantitatively forecast the thunderstorm process accompanying different weather phenomena,starting from the reality of aviation meteorological support,under the guidance of meteorological principles,according to the requirements of aviation meteorological support and related physical parameters of Hedong Airport,transforming thunderstorm samples into classified operational mapping,and operational mapping is carried out for different weather phenomena such as strong wind and precipitation associated with airport thunderstorms.The normal distribution characteristic of the data was inspected with coefficients of Kurtosis and Skewness,Chi-square and Q-Q plot.The results show that the samples classified by weather phenomena are tested by normality,and the unclassified samples basically pass the normality test.According to the small sample data,forecast the weather phenomena and intensity accompanied with thunderstorms by establishing various objective forecast models,and select the model with better effect.According to the idea of "Ingredients-based method" for the occurrence of strong convection,selected the parameters closely related to thunderstorm occurrence as the forecast factors.Linear,nonlinear,BP neural network and support vector machine(SVM)forecast models are established for classified and unclassified samples respectively.Factor backdating validation shows that the linear and nonlinear regression models of classified samples are highly approximate to the sample values,while there is a large deviation between the unclassified sample and the actual situation.The main reason is that the operational mapping values of unclassified samples do not strictly obey the standard normal distribution.The forecast model was put into Hedong Airport in 2017.The results showed that for various weather phenomena,the forecasting effect of each model was different:(1)BP artificial neural network and SVM can forecast the weather phenomena of thunderstorm more accurately,reaching 66.7%.It is particularly prominent for the forecast of weak rain,gale and rainfall.(2)The accuracy of classified stepwise regression model in forecasting the intensity of weak rain is higher,as high as 75%.(3)Classified multivariate regression model has a higher forecasting accuracy of the intensity of heavy rain,reaching 66.7%.(4)Classified nonlinear regression model has a better forecasting of the intensity of gale and rain,reaching 60%.Five models have their own advantages and disadvantages.According to the nature of different models,their respective advantages are used to synthetically design operational processes.First,BP artificial neural network and SVM model are used to judge thunderstorm weather phenomena.Then,forecast the intensity of weak rain,heavy rain or gale,gale and rainfall thunderstorm phenomena with classified stepwise,multivariate and nonlinear regression models respectively.Starting from the operational forecasting demand of Hedong Airport and combining with the general idea of short-term forecasting system,developed a thunderstorm operational forecasting system under the guidance of software engineering principle.The function modules of the system include: 1.collecting data,2.processing data,3.forecasting the phenomena of weather,4.forecasting the intensity of weather,5.generating early warning and forecasting documents,6.pushing message.The system realizes the high efficiency and intellectualization of the thunderstorm forecast model over Yinchuan Hedong Airport. |
PDF Full Text Request