Study On Automatic Synoptic Type Discrimination And Methods Of Forecast Of Fog In Eastern China

Fog is the main weather phenomenon leading to the visibility impairment, it causes enormous economic losses to transportation, electrical equipment, industrial and agricultural production, it also threatens the health and life of human beings directly. Transportation is well developed in eastern China, intensive highway network is susceptible to fog. By using observing data from 1981 to 2012 in eastern China, the climatic character of fog’s spatial and temporal distribution in eastern China is analyzed in detail. Based on concluding and summarizing the air circulation status, synoptic type has been identified into three kinds, automatic synoptic type discrimination system has been established quantitatively. On the base of automatic synoptic type discrimination system, diagnostic analyses have been applied to some sensitive quanitities. Using event probability regression and BP neural network, the fog forecasting model in eastern China were established, the forecasting process were built. The main results are as follows:(1) The spatial distribution of the foggy day in eastern China has much difference, there is an obvious increasing trend of the foggy day from coastal region to inland, and the mean annual foggy day is 10d. Foggy days in the past 30 years showed a increasing trend, howerver the number of mist days showed an increasing trend since 1995. Fog has a wide distribution in autumn and winter seasons, fog appears in coastal and inland area. The distribution of the foggy days is bimodal, the peak of wintertime and summertime are December and June. Four seasons obey the same law that fog occours most frequently from 5:00 to 8:00 in a day.(2) The regional foggy day is 85d from 2002 to 2012, there are 7.7 days a year. Based on synoptic map of the 500 hPa height field and surface field, the synoptic situation which is conducive to the formation of fog can be classified into three kinds:behind the trough while before the ridge, straight airflow, trough region.The most common synoptic type of fog is the synoptic situation behind the trough while before the ridge, which account for 50.2%, the synoptic situation with straight airflow accout for 21.2%, the synoptic situation in trough region accout for 15.3%. According to statistical analysis month by month, all regional foggy days appear during October-April, there is no regional foggy day in the other months.(3) For three kinds of synoptic situation which is conducive to the formation of fog, quantitative standard is established to classify the synoptic situation. The synoptic situation of 4000 days from 2002 to 2012 can be judged and classified on the base of automatic synoptic type discrimination system, there are 1671 days in synoptic situation which is conducive to the formation of fog after the classification, about 2300 days are eliminated, the eliminating results are satisfying, only two days are left out from 85 regional foggy days.(4) Some sensitive physical quanitities have been choosen according to the humidity conditions, stratified conditions, wind conditions which is needed to the formation of fog, the threshold value of these sensitive physical quanitities are determined based on historical data. Of 1671 days after the automatic synoptic type discrimination, only 290 days left,1381 days are eliminated, of which the regional foggy day is 83d in reality with no omissions. After synoptic type discrimination and physical diagnosis, the forecasting accuracy of regional fog is 28.6%.(5) Using the event probability regression method, TS score in central Anhui Province is high, with a 65% TS score in some local areas. TS score in coastal regions is lower than inland regions while false negative rate in coastal regions is higher than inland regions. Using BP neural network method, TS score in the borderland between Hebei and Henan Province, between Zhejiang and Jiangxi Province is high, TS score in coastal regions is lower than inland regions, false negative rate in the north of Shangdong Province and in the borderland between Jiangsu and Zhejiang Province is high, false positive rate in the borderland between Shandong and Hebei Province is low. Comparison of two methods shows that TS score of event probability regression method is higher than BP neural network method, the forecasting performance of event probability regression method in Anhui Province is better, the forecasting performance of BP neural network method in the borderland between Zhejiang and Jiangxi Province is better while false negative rate in the north of Shandong Province is high. False positive rate of two methods is not low, false positive rate of event probability regression method is lower than BP neural network method.The results of this paper can provide a useful reference for the synoptic type discrimination and forecast of fog in eastern China.