Scholarlink

<p>In this study, convolutional long short-term memory (ConvLSTM) model is used to predict sea level anomaly (SLA) in the Kuroshio Extension (KE) region, utilizing daily satellite altimetry data (1993–2016). The model captures regional averaged SLA variability, achieving a correlation coefficient of 0.98 for prediction horizon up to 23 d. Propagating features of Rossby waves are also reproduced in the prediction model. While in spatial, discrepancies between predicted SLA and observed SLA are quite large, especially in regions with strong eddy activities. Incorporating equation of motion for the <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(1{1\over{2}}\)</EquationSource> <EquationSource Format="MATHML"><math display="block"> <mn>1</mn> <mrow> <mfrac> <mn>1</mn> <mrow> <mn>2</mn> </mrow> </mfrac> </mrow> </math></EquationSource> </InlineEquation>-layer reduced-gravity model, the performance of the model has a significant improvement spatially and temporally. Challenges persist in high-variability regions, underscoring the need for advanced models. This study highlights ConvLSTM’s potential for SLA forecasting with wind driven physical constraints, offering insights into wind-driven and eddy-influenced processes in the KE region.</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

Sea level prediction in the Kuroshio Extension region using ConvLSTM with wind-driven physical constraints

  • 期刊论文

  • 发表日期：

    2025年12月1日

查看全文
  • Duotian Huang,
  • Xuhua Cheng

摘要

In this study, convolutional long short-term memory (ConvLSTM) model is used to predict sea level anomaly (SLA) in the Kuroshio Extension (KE) region, utilizing daily satellite altimetry data (1993–2016). The model captures regional averaged SLA variability, achieving a correlation coefficient of 0.98 for prediction horizon up to 23 d. Propagating features of Rossby waves are also reproduced in the prediction model. While in spatial, discrepancies between predicted SLA and observed SLA are quite large, especially in regions with strong eddy activities. Incorporating equation of motion for the \(1{1\over{2}}\) 1 1 2 -layer reduced-gravity model, the performance of the model has a significant improvement spatially and temporally. Challenges persist in high-variability regions, underscoring the need for advanced models. This study highlights ConvLSTM’s potential for SLA forecasting with wind driven physical constraints, offering insights into wind-driven and eddy-influenced processes in the KE region.