Scholarlink

The study investigates the predictability of the Kuroshio Front, a typical mesoscale oceanic phenomenon in the East China Sea, known for its significant influence on the temperature and salinity distribution and acoustic characteristics in the region. Utilizing CTD data from 192 seasonal cruises along the PN section of the East China Sea, the research analyzes the spatial distribution patterns of temperature and salinity, explores the spatiotemporal characteristics of the front using a front extraction method, and employs Mann-Kendall testing to analyze its interannual variability. Furthermore, a CNN-LSTM-Attention machine learning model is utilized to forecast the trend of the front. Results reveal a westward-strong/eastward-weak spatial distribution of the Kuroshio Front, with pronounced occurrences in spring and winter compared to summer and autumn. Notably, a significant interannual intensification trend is evident, with a distinct intensity shift around 1980. The predictive model effectively captures critical features and predicts trend changes, with an average temperature gradient MAE of 0.0221 °C/km (area I) and 0.0031 °C/km (area II). This research enriches the study of oceanography and provides a foundation for research into marine environments and ecosystems.

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

Predictability Study of the Kuroshio Front in the East China Sea Based on In-Situ Observational Data: Analysis and Prediction Using Mann-Kendall Test and CNN-LSTM-Attention

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Lei Zhang,
  • Weishuai Xu,
  • Xiaodong Ma,
  • Maolin Li

摘要

The study investigates the predictability of the Kuroshio Front, a typical mesoscale oceanic phenomenon in the East China Sea, known for its significant influence on the temperature and salinity distribution and acoustic characteristics in the region. Utilizing CTD data from 192 seasonal cruises along the PN section of the East China Sea, the research analyzes the spatial distribution patterns of temperature and salinity, explores the spatiotemporal characteristics of the front using a front extraction method, and employs Mann-Kendall testing to analyze its interannual variability. Furthermore, a CNN-LSTM-Attention machine learning model is utilized to forecast the trend of the front. Results reveal a westward-strong/eastward-weak spatial distribution of the Kuroshio Front, with pronounced occurrences in spring and winter compared to summer and autumn. Notably, a significant interannual intensification trend is evident, with a distinct intensity shift around 1980. The predictive model effectively captures critical features and predicts trend changes, with an average temperature gradient MAE of 0.0221 °C/km (area I) and 0.0031 °C/km (area II). This research enriches the study of oceanography and provides a foundation for research into marine environments and ecosystems.