摘要
构建了一种适用于CALIOP透明云下数据的深度神经网络模型,用于反演海面风速。通过使用2017年1、4、7、10月的CALIOP夜间数据和准同步的AMSR-2风速数据对模型进行训练,然后将该模型应用于2018年1—9月的夜间云下数据,实现海面风速的反演。与无云数据对比,所得反演结果表明精度接近,标准偏差最大为0.89 m/s,最低相关系数为0.94。在引入ERA5有效波高数据后,反演精度进一步提升,标准偏差最大为0.68 m/s,相关系数达到0.96以上。研究结果表明,透明云下数据同样可用于风速反演,深度神经网络能够有效地从CALIOP数据中提取风速信息,并结合有效波高数据进一步提高反演精度。
A deep neural network model was developed to retrieve sea surface wind speed from CALIOP transparent cloud data.The model was trained using CALIOP nighttime data from January,April,July,and October 2017 along with quasi-synchronous AMSR-2 wind speed data.Subsequently,it was applied to nighttime sub-cloud data from January to September 2018 for surface wind speed inversion.Comparative analysis with cloud-free data revealed close accuracy,a maximum standard deviation of 0.89 m/s,and a minimum correlation coefficient of 0.94.Upon integration of ERA5 significant wave height data,the inversion accuracy improved further with a reduced maximum standard deviation of 0.68 m/s and a correlation coefficient exceeding 0.96.These results demonstrate the potential use of transparent cloud data for wind velocity inversion and highlight the effectiveness of deep neural networks in extracting wind velocity information from CALIOP data while enhancing inversion accuracy through integration with significant wave height data.
作者
罗敦艺
吴东
张馨毅
贺岩
Luo Dunyi;Wu Dong;Zhang Xinyi;He Yan(College of Marine Technology,Faculty of Information Science and Engineering,Ocean University of China,Qingdao 266100,China;Laboratory for Regional Oceanography and Numerical Modeling,Laoshan Laboratory,Qingdao 266237,China;Key Laboratory of Space Laser Communication and Detection Technology,Shanghai Institute of Optics and Fine Mechanics,Chinese Academy of Sciences,Shanghai 201800,China)
出处
《中国海洋大学学报（自然科学版）》
CAS
CSCD
北大核心
2024年第12期130-139,共10页
Periodical of Ocean University of China
基金
中国科学院空间激光信息传输与探测技术重点实验室开放基金项目“星载激光海面粗糙度探测及其应用研究”资助。
关键词
遥感
星载激光雷达
海面后向散射
海面风速
有效波高
深度学习
remote sensing
spaceborne lidar
sea surface backscatter
sea surface wind speed
significant wave height
deep learning
