Thunderstorm Nowcasting Using Machine Learning Techniques: A Case Study of the Airports in Northern Thailand
Keywords:
thunderstorm, nowcasting, machine learning, classification model, ensemble learning algorithmsAbstract
The objectives of this research were to (1) create thunderstorm nowcasting models and (2) evaluate the efficiency of the nowcasting models for thunderstorm forecasting at northern Thailand’s airports with forecasting times in one hour and two hours using machine learning techniques. The input datasets were obtained from the Meteorological Terminal Air Report (METAR) of nine airports located in northern Thailand. The hourly METAR reports generated by the Thai Meteorological Department during January 2015 and December 2022, covering a total of 364,382 datasets, were analyzed. All input data were divided into 5 groups: the group of all airports, airports in the upper northern, airports in the lower northern, airports surrounded by mountains, 1-2 sides, and 3-4 sides. By using machine learning algorithms, three classification standard algorithms were applied, including Naïve Bayes, Decision Tree, and Neural Networks. Three algorithmic methods with ensemble learning algorithms, including Bagging, AdaBoost, and Random Forest, were also used to create classification models. The Synthetic Minority Oversampling Technique (SMOTE) was used for balancing datasets, and the 10-fold cross-validation method was employed to evaluate predictive models. According to the results of the data group that included all airports for one-hour forecasting time, the Random Forest was the most effective model, with the F-measure value of 76.45% and the Area Under the Curve (AUC) of 0.888. For a two-hour forecasting time, the combination of neural networks and bagging was the most effective model, with an F-measure of 30.62% and an AUC of 0.746.
References
Aeronautical Meteorology Division. (2015). Meteorological messages for aviation and dissemination. Retrieved from http://www2.aeromet.tmd.go.th/KnowledgeNews1.php
Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: Synthetic minority over-sampling technique. Journal of Artificial Intelligence Research, 16(2002), 321-357. https://doi.org/10.1613/jair.953
Chen, H., Li, T., Fan, X., & Luo, C. (2019). Feature selection for imbalanced data based on neighborhood rough sets. Information Sciences, 483(2019), 1-20. https://doi.org/10.1016/j.ins.2019.01.041
Dey, I., & Pratap, V. (2023). A Comparative Study of SMOTE, Borderline-SMOTE, and ADASYN Oversampling Techniques using Different Classifiers. Proceeding of 2023 3rd International Conference on Smart Data Intelligence (pp. 294-302). Trichy, India: ICSMDI.
Environment and Pollution Control Office 2. (2021). Analysis of area potential to support water pollution risks in 17 provinces of the northern region. Retrieved from https://shorturl.asia/yEVaS
Gupta, V., Mishra, V. K., Singhal, P., & Kumar, A. (2022). An overview of supervised machine learning algorithm. Proceeding of International Conference on System Modeling & Advancement in Research Trends (pp. 87-92). Moradabad, India: IEEE.
Jantarakongkul, B., Rasmequan, S., Rimcharoen, S., Kulkasem, P., Chinnasarn, K., Rodtook, A., Vorraboot, P., & Onpans, J. (2014). Optimal methods for classification of highly imbalanced datasets (Research report). Bangkok: National Research Council of Thailand (NRCT). (in Thai)
Khunsuk, T., & Thongkam, J. (2020). Feature selection method for improving customer reviews classification. RMUTI Journal Science and Technology, 13(1), 129-143. (in Thai)
Lamprakhon, B. (2017). A proposed method to compare areas under the roc curves for a single dataset: A case study of credit scoring model (Master’s thesis). Chulalongkorn University. Bangkok. (in Thai)
Manokij, F. (2019). Thailand’s precipitation forecasting using deep learning approach model (Master’s thesis). Chulalongkorn University. Bangkok. (in Thai)
Moon, S. H., Kim, Y. H., Lee, Y. H., & Moon, B. R. (2019). Application of machine learning to an early warning system for very short-term heavy rainfall. Journal of Hydrology, 568(2019), 1042-1054. https://doi.org/10.1016/j.jhydrol.2018.11.060
Nasritha, K., Kerdprasop, K., & Kerdprasop, N. (2018). Comparison of sampling techniques for imbalanced data classification. Journal of Applied Informatics and Technology, 1(2018), 20-37. https://doi.org/10.14456/jait.2018.2 (in Thai)
Northern Met Center. (2023). Today forecast. Retrieved from http://www.cmmet.tmd.go.th/forecast/todayforecast.jpg (in Thai)
Phaeobang, A., & Sinsomboonthong, S. (2019). Adjusting the imbalanced data with 5 classification methods. Thai Journal of Science and Technology, 9(4), 418–435. https://doi.org/10.14456/tjst.2020.46 (in Thai)
Promla, P., & Sanrach, C. (2020). The comparison of efficiency on the analysis of satisfaction on teaching performance using sentiment analysis by ensemble technique. KKU Research Journal (Graduate Studies), 20(4), 140-149. (in Thai)
Senawong, N., Wichitchan, S., & Kumphon, O. (2022). Large-scale data classification based on K-means clustering and deep learning. The Journal of King Mongkut’s University of Technology North Bangkok, 32(4), 957-965. (in Thai)
Sridee, A. (2021). Applying machine learning techniques to detect failure in hard disk drive test process (Master’s thesis). Chulalongkorn University. Bangkok. (in Thai)
Srithagon, S. (2021). Rainfall nowcasting using learning machines based on rain gauge data in Bangkok Metropolis (Master’s thesis). Kasetsart University. Bangkok. (in Thai)
Tanwanon, P., Yingsayree, C., Pongphet, W., & Kangkajit, T. (2017). Predict stock price trends in stock exchange of Thailand using ensemble model. Journal of information Science and Technology, 7(1), 12-21. https://doi.org/10.14456/jist.2017.2 (in Thai)
Thai Meteorological Department. (2023). Meteorological knowledge. Retrieved from https://www.tmd.go.th/info/การพยากรณอากาศ-1
Thummikarat, H. (2020). Parkinson’s disease diagnosis using machine learning (Master’s thesis). Chulalongkorn University. Bangkok. (in Thai)
