基于卫星遥感数据的RA-BP神经网络三维海温反演方法  

Three-Dimensional Sea Temperature Inversion Method Using RA-BP Neural Network Based on Satellite Remote Sensing Data

作　　者：万江岳 于华明[1,2] 张辰宇 温琦 Wan Jiangyue;Yu Huaming;Zhang Chenyu;Wen Qi(Sanya Oceanographic Institution,Ocean University of China,Sanya 572024,China;College of Oceanic and Atmospheric Sciences,Ocean University of China,Qingdao 266100,China;Collgeg of Computer Science and Technology,Faculty of Information Science and Engineering,Ocean University of China,Qingdao 266100,China)

机构地区：[1]中国海洋大学三亚海洋研究院,海南三亚572024 [2]中国海洋大学海洋与大气学院,山东青岛266100 [3]中国海洋大学信息科学与工程学部计算机科学与技术学院,山东青岛266100

出　　处：《中国海洋大学学报（自然科学版）》2024年第7期1-11,共11页Periodical of Ocean University of China

基　　金：国家重点研究发展计划项目(2022YFD2401304);三亚崖州湾科技城科技专项(SCKJ-JYRC-2022-101);海南省三亚市海洋生态保护修复工程项目跟踪监测与效果评估项目资助。

摘　　要：本文利用2005—2017年的实测海温数据和卫星观测数据,构建了基于RA-BP神经网络的三维海温反演模型,对2018年印度洋东北部和太平洋西部研究海域海温剖面进行了反演,并与回归分析法和基础BP神经网络反演海温结果进行了对比分析。以均方根误差和Pearson系数作为检验指标,结果显示所有模型反演海温剖面与实测海温剖面的Pearson系数均在0.99以上。对混合层、温跃层、中深层三个水深范围,RA-BP神经网络模型在印度洋东北部研究区域的均方根误差(RMSE)均值分别为0.32、0.76和0.17℃,在太平洋西部研究区域的RMSE均值分别为0.24、1.01和0.24℃,单个剖面的RMSE在0.6℃以下。这表明在研究海域,基于实测海温数据和卫星遥感数据构建的RA-BP神经网络模型是可行的,且具有良好的反演精度。Based on the measured sea temperature data and satellite observation data from 2005 to 2017,a three-dimensional sea temperature inversion model based on RA-BP neural network is constructed in this paper.The sea temperature profile ofthe research areas in the northeast Indian Ocean and the western Pacific Ocean in 2018 is retrieved,and the results are compared with those of regression analysis method and the basic BP neural network.Taking the root mean square error and Pearson coefficient as the test indicators,the results show that the Pearson coefficients of all models′retrieved sea temperature profiles and measured sea temperature profiles are above 0.99.For the three water depth ranges of the mixed layer,the thermocline and the middle-deep layer,the average RMSE of the RA-BP neural network model in the research area of the northeast Indian Ocean are 0.32,0.76 and 0.17℃respectively,and the average RMSE in the research area of the western Pacific Ocean are 0.24,1.01 and 0.24℃respectively,and the RMSE of a single profile is below 0.6℃.This shows that the RA-BP neural network model based on measured sea temperature data and satellite remote sensing data is feasible and has good retrieval accuracy in the research areas.

关 键 词：海表温度 绝对动力高度 回归分析 神经网络 三维海温反演 

分 类 号：P731.11[天文地球—海洋科学]