通过深度学习技术加强卫星图像的热带气旋强度估算

IF 2.8 3区 地球科学 Q3 METEOROLOGY & ATMOSPHERIC SCIENCES
Wen Yang, Jianfang Fei, Xiaogang Huang, Juli Ding, Xiaoping Cheng
{"title":"通过深度学习技术加强卫星图像的热带气旋强度估算","authors":"Wen Yang, Jianfang Fei, Xiaogang Huang, Juli Ding, Xiaoping Cheng","doi":"10.1007/s13351-024-3186-y","DOIUrl":null,"url":null,"abstract":"<p>This study first utilizes four well-performing pre-trained convolutional neural networks (CNNs) to gauge the intensity of tropical cyclones (TCs) using geostationary satellite infrared (IR) imagery. The models are trained and tested on TC cases spanning from 2004 to 2022 over the western North Pacific Ocean. To enhance the models performance, various techniques are employed, including fine-tuning the original CNN models, introducing rotation augmentation to the initial dataset, temporal enhancement via sequential imagery, integrating auxiliary physical information, and adjusting hyperparameters. An optimized CNN model, i.e., visual geometry group network (VGGNet), for TC intensity estimation is ultimately obtained. When applied to the test data, the model achieves a relatively low mean absolute error (MAE) of 4.05 m s<sup>−1</sup>. To improve the interpretability of the model, the SmoothGrad combined with the Integrated Gradients approach is employed. The analyses reveal that the VGGNet model places significant emphasis on the distinct inner core region of a TC when estimating its intensity. Additionally, it partly takes into account the configuration of cloud systems as input features for the model, aligning well with meteorological principles. The several improvements made to this model’s performance offer valuable insights for enhancing TC intensity forecasts through deep learning.</p>","PeriodicalId":48796,"journal":{"name":"Journal of Meteorological Research","volume":"47 1","pages":""},"PeriodicalIF":2.8000,"publicationDate":"2024-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Enhancing Tropical Cyclone Intensity Estimation from Satellite Imagery through Deep Learning Techniques\",\"authors\":\"Wen Yang, Jianfang Fei, Xiaogang Huang, Juli Ding, Xiaoping Cheng\",\"doi\":\"10.1007/s13351-024-3186-y\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>This study first utilizes four well-performing pre-trained convolutional neural networks (CNNs) to gauge the intensity of tropical cyclones (TCs) using geostationary satellite infrared (IR) imagery. The models are trained and tested on TC cases spanning from 2004 to 2022 over the western North Pacific Ocean. To enhance the models performance, various techniques are employed, including fine-tuning the original CNN models, introducing rotation augmentation to the initial dataset, temporal enhancement via sequential imagery, integrating auxiliary physical information, and adjusting hyperparameters. An optimized CNN model, i.e., visual geometry group network (VGGNet), for TC intensity estimation is ultimately obtained. When applied to the test data, the model achieves a relatively low mean absolute error (MAE) of 4.05 m s<sup>−1</sup>. To improve the interpretability of the model, the SmoothGrad combined with the Integrated Gradients approach is employed. The analyses reveal that the VGGNet model places significant emphasis on the distinct inner core region of a TC when estimating its intensity. Additionally, it partly takes into account the configuration of cloud systems as input features for the model, aligning well with meteorological principles. The several improvements made to this model’s performance offer valuable insights for enhancing TC intensity forecasts through deep learning.</p>\",\"PeriodicalId\":48796,\"journal\":{\"name\":\"Journal of Meteorological Research\",\"volume\":\"47 1\",\"pages\":\"\"},\"PeriodicalIF\":2.8000,\"publicationDate\":\"2024-09-06\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Meteorological Research\",\"FirstCategoryId\":\"89\",\"ListUrlMain\":\"https://doi.org/10.1007/s13351-024-3186-y\",\"RegionNum\":3,\"RegionCategory\":\"地球科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"METEOROLOGY & ATMOSPHERIC SCIENCES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Meteorological Research","FirstCategoryId":"89","ListUrlMain":"https://doi.org/10.1007/s13351-024-3186-y","RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"METEOROLOGY & ATMOSPHERIC SCIENCES","Score":null,"Total":0}
引用次数: 0

摘要

本研究首先利用四个性能良好的预训练卷积神经网络(CNNs),使用地球静止卫星红外图像测量热带气旋(TCs)的强度。这些模型在北太平洋西部从 2004 年到 2022 年的热带气旋案例中进行了训练和测试。为提高模型性能,采用了多种技术,包括微调原始 CNN 模型、对初始数据集引入旋转增强、通过连续图像进行时间增强、整合辅助物理信息以及调整超参数。最终获得了用于 TC 强度估计的优化 CNN 模型,即视觉几何组网络(VGGNet)。将该模型应用于测试数据时,其平均绝对误差(MAE)相对较低,仅为 4.05 m s-1。为了提高模型的可解释性,采用了 SmoothGrad 与 Integrated Gradients 相结合的方法。分析表明,VGGNet 模式在估计热气旋强度时,非常重视其独特的内核区域。此外,它还部分考虑了云系统的配置,将其作为模型的输入特征,非常符合气象学原理。该模型性能的多项改进为通过深度学习增强热气旋强度预报提供了有价值的见解。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Enhancing Tropical Cyclone Intensity Estimation from Satellite Imagery through Deep Learning Techniques

This study first utilizes four well-performing pre-trained convolutional neural networks (CNNs) to gauge the intensity of tropical cyclones (TCs) using geostationary satellite infrared (IR) imagery. The models are trained and tested on TC cases spanning from 2004 to 2022 over the western North Pacific Ocean. To enhance the models performance, various techniques are employed, including fine-tuning the original CNN models, introducing rotation augmentation to the initial dataset, temporal enhancement via sequential imagery, integrating auxiliary physical information, and adjusting hyperparameters. An optimized CNN model, i.e., visual geometry group network (VGGNet), for TC intensity estimation is ultimately obtained. When applied to the test data, the model achieves a relatively low mean absolute error (MAE) of 4.05 m s−1. To improve the interpretability of the model, the SmoothGrad combined with the Integrated Gradients approach is employed. The analyses reveal that the VGGNet model places significant emphasis on the distinct inner core region of a TC when estimating its intensity. Additionally, it partly takes into account the configuration of cloud systems as input features for the model, aligning well with meteorological principles. The several improvements made to this model’s performance offer valuable insights for enhancing TC intensity forecasts through deep learning.

求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
Journal of Meteorological Research
Journal of Meteorological Research METEOROLOGY & ATMOSPHERIC SCIENCES-
CiteScore
6.20
自引率
6.20%
发文量
54
期刊介绍: Journal of Meteorological Research (previously known as Acta Meteorologica Sinica) publishes the latest achievements and developments in the field of atmospheric sciences. Coverage is broad, including topics such as pure and applied meteorology; climatology and climate change; marine meteorology; atmospheric physics and chemistry; cloud physics and weather modification; numerical weather prediction; data assimilation; atmospheric sounding and remote sensing; atmospheric environment and air pollution; radar and satellite meteorology; agricultural and forest meteorology and more.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信