摘要
为快速获取大范围、准实时的海洋内部结构,海面遥感数据被广泛应用于构建温度剖面垂直结构,但卫星遥感仅能获得较为准确的海洋表面或者近表层数据。为了提高全海深温度剖面的反演精度,本文以水下固定深度处温度为约束,通过径向基函数神经网络生成海表面温度和海平面高度异常等海表遥感数据与温度剖面之间的非线性映射,并对约束深度选取的理论依据进行讨论。南海海域温度剖面的反演结果表明,第1阶经验正交函数系数可以表征温跃层的垂直位移,而第1阶经验正交函数基函数极值点对应深度处的温度与第1阶经验正交函数系数之间具有强相关性。当增加该深度处温度为约束时,温跃层的反演精度比仅使用海面遥感数据约提高0.35℃,反演温度剖面的平均均方根误差约为0.33℃。
In order to quickly obtain a large-scale,quasi-real-time internal structure of the ocean,sea surface remote sensing data are widely used to construct the vertical structure of the temperature profiles,but satellite remote sensing can only obtain relatively accurate ocean surface or near-surface data.In order to improve the accuracy of temperature profile inversion,this paper takes the depth-fixed temperature as the constraint,and the nonlinear mapping between the temperature profiles and the sea surface remote sensing data such as sea surface temperature(SST)and sea level anomaly(SLA)is generated through the radial basis function(RBF)neural network,and discuss the theoretical basis for constrained depth selection.The inversion results of the temperature profiles in the South China Sea show that the first empirical orthogonal function(EOF)coefficient can characterize the vertical displacement of the thermocline.And there is a strong correlation between the temperature at the depth corresponding to the extreme point of the first EOF and the first EOF coefficient.Therefore,when the temperature at this depth is added as a constraint,the inversion accuracy of the thermocline is about 0.35℃higher than that of only using sea surface remote sensing data,and the mean root mean square error of temperature profile inversion is about 0.33℃.
作者
李倩倩
王子文
朱金龙
隽智昊
李琪
罗宇
Li Qianqian;Wang Ziwen;Zhu Jinlong;Juan Zhihao;Li Qi;Luo Yu(College of Geodesy and Geomatics,Shandong University of Science and Technology,Qingdao 266590,China;College of Underwater Acoustic Engineering,Harbin Engineering University,Harbin 150001,China)
出处
《海洋学报》
CAS
CSCD
北大核心
2023年第7期126-136,共11页
基金
山东省自然科学基金面上项目(ZR2022MA051)
中国博士后科学基金(2020M670891)
山东科技大学科研创新团队支持计划(2019TDJH103)
山东省高等学校青年创新团队人才引育计划(卫星定位导航研究创新团队)
山东省自然科学基金(ZR2020MA090)。
关键词
温度剖面
径向基函数神经网络
经验正交函数
海表面温度
海平面高度异常
定深温度
temperature profile
radial basis function neural network
empirical orthogonal function
sea surface temperature
sea level anomaly
depth-fixed temperature
