Investigation of fog characteristics of airports in Turkey

The aviation sector developing in the world and our country faces many problems. At the beginning of these problems technical and meteorological originated procession comes first. Problems due to technology are being solved rapidly with the development of technology, while in the problems of meteorological origin, mankind is in a great struggle with nature. In order to win this struggle, people first try to understand the mechanisms of the occurrence of these events and analyze these complex processes caused by many parametres. As a result of these analyzes, they are looking for solutions. Major meteorological events affecting the aviation industry; such as lightning, fog and wind which cause great material and spiritual loss. It is very important to understand, model, analyze and forecast the formation processes of fog due to the large financial losses that caused in the aviation sector. In this study, METAR and SPECI reports belonging to 37 international, 17 national, 12 military and 1 education airports, were analyzed within the period between 2011 and 2016. The data were obtained from Turkish State Meteorology Service. METAR and SPECI reports are meteorological codes prepared for aviation purposes. Approximately 39 million METAR and SPECI reports of all airports were analyzed during the study. Based on the FG, ie the fog code found in these reports, the basis for working was established. Using the reports included in the study period, 67 airports were analyzed individually for one year, monthly and hourly during the first phase of the study. Besides these, all the fogs categorized according to their types. In addition, fog estimation is performed using BayesNet, Navian Bayes, J48, Ridor and RBFNetwork algorithms in the WEKA program which is a machine learning program. When annual analyzes are investigated, it is observed that none of the airports have a trend. When the monthly analyzes are performed, it is seen that most of the fog incidents occur in January, February and December. In the hourly analyzes, it is observed that fog is mostly crowded with sunrise and sunset. The difference between air temperature and land temperature plays an important role in this. According to the fog types such as FG, MIFG, BCFG, FZFG, PRFG, it was seen that FZFG generally seen mostly the place where low average temperature and continental climate are dominant. In general, FZFG, FG and BCFG fog types are the most common fog types found in the METAR and SPECI reports. Prior to the fog estimation, the JetBrains PyCharm program is used and the WEKA program is ready for use. The WEKA program, a Java based open source machine learning program, was used to estimate the fog for each airport using Bayes Net, Navian Bayes, Ridor, J48 and RBFNetwork algorithms. In selecting these algorithms, 50 machine learning algorithms in the WEKA program were applied to International İstanbul Atatürk Airport, which is the largest airport in terms of passenger and cargo transportation in Turkey, and Bayes Net, Navian Bayes, which has the most successful results in these algorithms, and J48, Ridor and RBFNetwork algorithms have been chosen. The success of algorithms depends primarily on the number of data and the number of fog lines compared to this multiplier. When the estimation results for all the airports are analyzed, it is observed that at the airports located in Batman, Erzurum, Şanlıurfa and Sivas, where the fog hash is observed in large numbers, it can be estimated with about 80% accuracy with the selected Bayes Net and Navian Bayes algorithms for the successful estimation of the fog forecast rate. At the airports such as Dalaman, Bodrum and Çanakkale where the fog has a small number and the fog forecasting success falls to zero. The results of the individual analyzes of airports with extreme characteristics to work according to the results obtained from annual, monthly, hourly analyzes and work done as a result of machine learning are given in detail. As a result of the analyzes and estimations made, the average fog numbers are compared with the altimeters of the airports. Pearson correlation showed that fog and altitude were 0.55 in winter, 0.17 in spring, 0.03 in summer and 0.22 in autumn. When the correlation result is examined, a statistically low result is obtained between fog and altitude. Although this result is not statistically significant, it does not mean that it is not meaningful as a physical structure. All of the obtained average fog values were used and the Turkish fog maps of autumn, winter, spring and summer seasons were drawn in Surfer 13 program. The maps are parallel to the results obtained. When the maps are examined, it is seen that the terrestrial effect is more frequent in autumn and winter months in closed regions and in summer and spring seasons in marine active regions. The same situation shows that the number of fog is high in regions with low temperature values. The study is one of the few studies on this field in Turkey, where an incident such a fog that occurs after predictable power and complex processes. The studies will continue during PhD studies in order to improve the study and achieve more successful results.