Scholarlink

<p>Tropical cyclones (TCs) are cyclonic circulations that occur on the sea surface in tropical and subtropical regions. Due to the significant impact of various ocean and atmospheric conditions on their intensity, forecasting the intensity of TCs is challenging. Additionally, the differences in meteorological elements among TC events complicate the extraction of features from cloud maps. To address above problems, we present a novel TC intensity forecasting model consisting of a TC cloud map prediction module and a TC intensity estimation module. The TC cloud map prediction model, namely SE-SimVP, is capable of capturing the important features in cloud maps effectively by adopting SENet. Furthermore, we design a TC estimation model based on residual structure and CNN. This model effectively captures spatial and structural features in satellite cloud maps, enabling accurate estimation of predicted TC cloud maps. Experiments are conducted on the HURDAT2 and HURSAT-B1 datasets from 2000 to 2016. The experimental results demonstrate the superiority of the proposed model over existing deep learning methods for TC intensity forecasting.</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

SE-SimVP: a novel tropical cyclone intensity forecasting model

  • 期刊论文

  • 发表日期：

    2026年8月3日

查看全文
  • Wei Fang,
  • Wenhe Lu,
  • Xiaomei Zheng

摘要

Tropical cyclones (TCs) are cyclonic circulations that occur on the sea surface in tropical and subtropical regions. Due to the significant impact of various ocean and atmospheric conditions on their intensity, forecasting the intensity of TCs is challenging. Additionally, the differences in meteorological elements among TC events complicate the extraction of features from cloud maps. To address above problems, we present a novel TC intensity forecasting model consisting of a TC cloud map prediction module and a TC intensity estimation module. The TC cloud map prediction model, namely SE-SimVP, is capable of capturing the important features in cloud maps effectively by adopting SENet. Furthermore, we design a TC estimation model based on residual structure and CNN. This model effectively captures spatial and structural features in satellite cloud maps, enabling accurate estimation of predicted TC cloud maps. Experiments are conducted on the HURDAT2 and HURSAT-B1 datasets from 2000 to 2016. The experimental results demonstrate the superiority of the proposed model over existing deep learning methods for TC intensity forecasting.