基于多层感知器和SAR参数的海浪有效波高反演方法  

A retrieval Approach for Significant Wave Height Based on Multi-Layer Perceptron and Parametersacquired by Sentinel-1 SAR

作　　者：高亚飞 王运华[1,3] 张彦敏 姜文正[2] Gao Yafei;Wang Yunhua;Zhang Yanmin;Jiang Wenzheng(Faculty of Information Science and Engineering,Ocean University of China,Qingdao 266100,China;The First Institute of Oceanography,Ministry of Natural Resources,Qingdao 266061,China;Laboratory for Regional Oceanography and Numerical Modeling,Pilot National Laboratory for Marine Science and Technology(Qingdao),Qingdao 266237,China)

机构地区：[1]中国海洋大学信息科学与工程学部,山东青岛266100 [2]自然资源部第一海洋研究所,山东青岛266061 [3]青岛海洋科学与技术试点国家实验室,区域海洋动力学与数值模拟功能实验室,山东青岛266237

出　　处：《中国海洋大学学报（自然科学版）》2024年第2期121-133,共13页Periodical of Ocean University of China

基　　金：国家自然科学基金项目(41976167,52101393);山东省自然科学基金项目(ZR2021MD023,ZR2021QD001)资助。

摘　　要：本文在分析雷达海面图像的多个参数与海浪有效波高相关性的基础上,应用哨兵1A卫星Level-2 SAR数据中的后向散射系数、图像强度归一化方差、截止波长、主波波长、主波波向、图像偏度和峰度参数提出了基于多层感知器(MLP)的海浪有效波高反演方法。详细讨论了输入参数组合不同时,MLP模型反演的整体海浪有效波高(SWH)、风浪有效波高(SWH_(WW))和涌浪有效波高(SWH_(S))的精度。针对极端海况下数据较少导致模型产生的系统偏差,文中采用校正函数减小该误差,从而增强了模型的适用性。与欧洲中期天气预报中心(ECMWF)和Jason3卫星提供的整体海浪有效波高数据进行对比,本文方法所得结果的均方根误差分别为0.471和0.535 m,相关系数为0.923和0.922。与ECMWF提供的风浪和涌浪有效波高相比,本文方法反演所得风浪和涌浪有效波高的均方根误差分别为0.534和0.512 m,相关系数分别为0.898和0.815。Based on the analysis of the correlation between several parameters of radar sea surface image and the significant wave height,the paper proposes a method which is based on a multi-layer perceptron(MLP)model to retrieve significant wave height by applying the backscatter coefficient,normalized variance of image intensity,azimuth cut-off wavelength,peak wavelength,peak wave direction,image skewness and kurtosis parameters from Sentinel 1A satellite Level-2 SAR data.The inversion accuracy of the total significant wave height(SWH),the significant wave height of wind wave(SWH_(WW))and the significant wave height of swell(SWH_(S))by the MLP model with different combinations of parameters is discussed in detail.A correction function is used to reduce the systematic bias in the model caused by the low availability of data in extreme sea conditions,thus enhancing the applicability of the model.When compared with SWH data provided by European Centre for Medium-Range Weather Forecasts(ECMWF)and Jason-3 satellite,the root mean square errors of the method are 0.471 and 0.535 m,with correlation coefficients of 0.923 and 0.922,respectively.When compared with SWH_(WW) and SWH_(S) data provided by ECMWF,the root mean square errors of the method are 0.534 and 0.512 m,with correlation coefficients of 0.898 and 0.815,respectively.

关 键 词：哨兵1A SAR数据 多层感知器 海浪有效波高 校正函数 

分 类 号：P731.22[天文地球—海洋科学]