Will a tropical cyclone make landfall?

• Published in: Neural computing & applications 2023-03, Vol.35 (8), p.5807-5818
• Main Authors: Kumar, Sandeep, Biswas, Koushik, Pandey, Ashish Kumar
• Format: Article
• Language: English
• Subjects: Accuracy; Artificial Intelligence; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer Science; Cyclones; Data Mining and Knowledge Discovery; Disaster management; Hurricanes; Image Processing and Computer Vision; Ocean basins; Original Article; Probability and Statistics in Computer Science; Questions
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-022-07996-7

In the different development phases of a tropical cyclone, the most exciting and complex phase is its landfall, which is when a tropical cyclone moves over to the land after crossing the ocean’s coast. The location, time, and intensity at landfall of a tropical cyclone determine the extent of the disaster caused by it. In this work, we investigate a fundamental question: will a tropical cyclone make a landfall? Knowing the answer to this question with high accuracy will have huge benefits as the preparedness for a potential landfall involves mobilizing substantial human and economic resources. To answer this fundamental question, we have used high-resolution reanalysis data ERA5 (ECMWF reanalysis 5 th generation) and best track data IBTrACS (International Best Track Archive for Climate Stewardship) to develop a deep learning model that can predict the landfall event in the early phase of a tropical cyclone—in particular, using any 12 hours or 24 hours of data from the first 72 hours of its inception with very high accuracy. We tested the model for six ocean basins of the world and achieved a fivefold accuracy in the range of 97.6 % to 99.2 % across all basins. The model can be trained within 05 to 20 minutes depending on the ocean basin and can predict the above-stated problem within seconds, making it suitable for real-time application.