Scholarlink

<p>Rapid intensification (RI) of Tropical Cyclones (TCs) is a significant threat to coastal regions worldwide. Although progress has been made in predicting TCs, RI events remain difficult to forecast accurately and in a timely manner. This study proposes a Machine Learning (ML)-based classification framework for RI prediction that utilizes Support Vector Machines (SVM) in conjunction with the Synthetic Minority Oversampling Technique (SMOTE) to handle the class imbalance of RI and non-RI cases. The Statistical Hurricane Intensity Prediction Scheme (SHIPS) data for the years 1982 to 2017 for the Atlantic Ocean basin and 1990 to 2010 for the Indian Ocean basin, respectively, are used to train and evaluate the proposed framework. Independent testing is conducted on operational data from 2010 to 2020 for the Atlantic basin and reanalysis data from 2011 to 2017 for the Indian basin. The framework demonstrates high skill, achieving a high Probability of Detection (POD) with a 24-h lead time in both basins, making it a potentially valuable tool for forecasters to issue timely warnings and prepare for the impacts of RI events. Overall, this study highlights the potential of ML-based approaches for improving RI prediction, and emphasizes the importance of ongoing efforts to develop more accurate and reliable forecasting methods for TCs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Machine Learning Approach for Detecting Rapid Intensification in Tropical Cyclones

  • 期刊论文

  • 发表日期：

    2025年1月17日

查看全文
  • Tushar Sharma,
  • Manish Mawatwal,
  • Saurabh Das

摘要

Rapid intensification (RI) of Tropical Cyclones (TCs) is a significant threat to coastal regions worldwide. Although progress has been made in predicting TCs, RI events remain difficult to forecast accurately and in a timely manner. This study proposes a Machine Learning (ML)-based classification framework for RI prediction that utilizes Support Vector Machines (SVM) in conjunction with the Synthetic Minority Oversampling Technique (SMOTE) to handle the class imbalance of RI and non-RI cases. The Statistical Hurricane Intensity Prediction Scheme (SHIPS) data for the years 1982 to 2017 for the Atlantic Ocean basin and 1990 to 2010 for the Indian Ocean basin, respectively, are used to train and evaluate the proposed framework. Independent testing is conducted on operational data from 2010 to 2020 for the Atlantic basin and reanalysis data from 2011 to 2017 for the Indian basin. The framework demonstrates high skill, achieving a high Probability of Detection (POD) with a 24-h lead time in both basins, making it a potentially valuable tool for forecasters to issue timely warnings and prepare for the impacts of RI events. Overall, this study highlights the potential of ML-based approaches for improving RI prediction, and emphasizes the importance of ongoing efforts to develop more accurate and reliable forecasting methods for TCs.