Scholarlink

Accurate typhoon track prediction is vital for disaster mitigation in the Northwestern Pacific. Traditional models like the Weather Research and Forecasting (WRF) model face limitations due to complex atmospheric dynamics. This study introduces advanced deep learning models—Wide and Deep Learning (WDL), Deep Cross Network (DCN), Deep Factorization Machine (DeepFM), and Kolmogorov–Arnold Networks (KAN)—integrated with Convolutional Long Short-Term Memory (ConvLSTM) to refine WRF forecasts. Results show that the Bidirectional Long Short-Term Memory (BiLSTM) + ConvLSTM + WDL model significantly improves prediction accuracy, particularly in long-term forecasts, reducing errors in key metrics like Mean Squared Error (MSE) and Bias2. The findings highlight the potential of deep learning to enhance typhoon track forecasts and inform future research directions.

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

Enhanced Typhoon Track Prediction: A Deep Learning Approach for the Northwestern Pacific

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Chengchen Tao,
  • Zhizu Wang,
  • Changsheng Zuo,
  • Yaoyao Han,
  • Xu Zhang

摘要

Accurate typhoon track prediction is vital for disaster mitigation in the Northwestern Pacific. Traditional models like the Weather Research and Forecasting (WRF) model face limitations due to complex atmospheric dynamics. This study introduces advanced deep learning models—Wide and Deep Learning (WDL), Deep Cross Network (DCN), Deep Factorization Machine (DeepFM), and Kolmogorov–Arnold Networks (KAN)—integrated with Convolutional Long Short-Term Memory (ConvLSTM) to refine WRF forecasts. Results show that the Bidirectional Long Short-Term Memory (BiLSTM) + ConvLSTM + WDL model significantly improves prediction accuracy, particularly in long-term forecasts, reducing errors in key metrics like Mean Squared Error (MSE) and Bias2. The findings highlight the potential of deep learning to enhance typhoon track forecasts and inform future research directions.