Reconstruction performance analysis for Basis Function of the sound speed profile
-
摘要: 海水声速剖面通常使用经验正交函数（Empirical Orthogonal Function，EOF）进行稀疏表示，然而基函数会受到数据完备性和数据测量时间的制约，其代表性误差会导致声速剖面重构精度受限。为了提高声速剖面的重构精度，本文利用模糊C均值聚类对BOA_Argo历史数据集进行聚类分析，探讨不同聚类空间的训练集数据对实测声速剖面重构精度的影响。研究表明，声速剖面具有明显的时空聚集特性，聚类后的历史声速剖面集生成的基函数和平均声速剖面具有最优的重构性能。本文研究结果有助于为历史声速剖面训练集的选取提供实际指导意义，进而提高声速剖面重构精度乃至反演精度。Abstract: Empirical Orthogonal Functions (EOFs) are usually used for sparse representation of the sound speed profile (SSP). However, due to the restriction of data completeness and measurement time, the representative error of the EOF will lead to limited accuracy of SSP reconstruction. In order to improve the reconstruction accuracy of SSP, the fuzzy C-means clustering algorithm is used to analyze the BOA_Argo historical data set and the reconstruction accuracy of the measured SSP based on different clustering spaces of data samples is discussed. The results shows that the SSPs are significant temporal-spatial clustering. The EOF and mean SSP generated by the clustered historical SSPs have the best reconstruction performance. The results of this paper are helpful to provide practical guidance for the selection of historical SSP training data and can improve the accuracy of SSP reconstruction.
-
图 4 第II型声速剖面进一步聚类结果
Fig. 4 Further clustering results of the sound speed profiles of Type II
图 5 第II-1型与CTD测量得到的声速剖面
Fig. 5 The sound speed profiles of Type II-1 andmeasured by CTDs
图 6 CTD和不同区域的平均声速剖面
a. 第I型；b. 第II型；c. 第III型；d. 第II-1型
Fig. 6 CTD and mean sound speed profiles in different regions
a. Type I; b. Type II; c. Type III; d. Type II-1
图 9 CTD测量声速剖面的重构误差
a. 全区域；b. 第I型；c. 第II型
Fig. 9 The reconstruction error of sound speed profiles measured by CTD
a. Whole area; b. Type I; c. Type II
图 10 重构声速剖面的均方根误差比较
Fig. 10 Comparison of the RMSE of the reconstructed sound speed profiles
图 12 CTD测量声速剖面的重构误差
a. 第II型；b. 第II-1型
Fig. 12 The reconstruction error of sound speed profiles measured by CTD
a. Type II; b. Type II-1
图 13 重构声速剖面的均方根误差比较
Fig. 13 Comparison of the RMSE of the reconstructed sound speed profiles
图 14 各训练集与CTD提取的第1阶EOF基函数幅度随深度的变化
Fig. 14 The variation of first EOF amplitude extracted from training sets and CTD with depth
图 16 ΔC1随深度的变化（a）、以第II型声速剖面作为训练集时 ΔC1和 ΔC2随深度的变化（b）、以第II-1型声速剖面作为训练集时ΔC1和ΔC2随深度的变化（c）、ΔC 随深度的变化（d）
Fig. 16 ΔC1 changed with depth (a), ΔC1 and ΔC2 changed with depth (the Type II sound speed profiles as the training set) (b), ΔC1 and ΔC2 changed with depth (the Type II-1 sound speed profiles as the training set) (c), ΔC changed with depth (d)
图 17 CTD测量声速剖面的重构误差
a. 第II型；b. 12° × 12°
Fig. 17 The reconstruction error of sound speed profiles measured by CTD
a. Type II; b. Type 12° × 12°
图 18 重构声速剖面的均方根误差比较
Fig. 18 Comparison of the RMSE of the reconstructing sound speed profiles
表 1 不同训练集与CTD基函数的相关系数
Tab. 1 The correlation coefficient between different training sets and CTD EOFs
基函数阶数 第1阶 第2阶 第3阶 第4阶 第5阶 第6阶 第II型 0.84 0.05 0.39 0.38 0.37 0.04 第II-1型 0.90 0.51 0.89 0.49 0.51 0.13 
下载: 导出CSV
表 2 不同训练样本下CTD声速剖面的重构误差
Tab. 2 The reconstruction error of CTD profiles under different training samples
误差 训练集 全区域 第I型 第II型 第II-1型 12° × 12°区域 最大声速误差/（m·s−1） 6.22 7.21 4.81 5.28 4.92 最大均方根误差/（m·s−1） 1.36 1.69 1.04 1.18 0.99 平均均方根误差/（m·s−1） 0.88 1.05 0.67 0.71 0.69
下载: 导出CSV
-
[1] 李大海, 吴立新, 陈朝晖. “透明海洋”的战略方向与建设路径[J]. 山东大学学报(哲学社会科学版), 2019(2): 130−136.Li Dahai, Wu Lixin, Chen Zhaohui. Strategic direction and construction path of transparent oceans[J]. Journal of Shandong University (Philosophy and Social Sciences), 2019(2): 130−136. [2] Wang Xun, Khazaie S, Chen Xue. Linear approximation of underwater sound speed profile: precision analysis in direct and inverse problems[J]. Applied Acoustics, 2018, 140: 63−73. doi: 10.1016/j.apacoust.2018.05.003 [3] 张宝华, 赵梅. 海水声速测量方法及其应用[J]. 声学技术, 2013, 32(1): 24−28.Zhang Baohua, Zhao Mei. Sound speed measurement in seawater and its application[J]. Technical Acoustics, 2013, 32(1): 24−28. [4] 唐军武, 陈戈, 陈卫标, 等. 海洋三维遥感与海洋剖面激光雷达[J]. 遥感学报, 2021, 25(1): 460−500. doi: 10.11834/jrs.20210495Tang Junwu, Chen Ge, Chen Weibiao, et al. Three dimensional remote sensing for oceanography and the Guanlan ocean profiling lidar[J]. National Remote Sensing Bulletin, 2021, 25(1): 460−500. doi: 10.11834/jrs.20210495 [5] 赵航芳, 汪非易, 朱小华, 等. 海洋声学层析研究现状与展望[J]. 海洋技术学报, 2015, 34(3): 69−74.Zhao Hangfang, Wang Feiyi, Zhu Xiaohua, et al. Ocean acoustic tomography: current progress and future prospect[J]. Journal of Ocean Technology, 2015, 34(3): 69−74. [6] LeBlanc L R, Middleton F H. An underwater acoustic sound velocity data model[J]. The Journal of the Acoustical Society of America, 1980, 67(6): 2055−2062. doi: 10.1121/1.384448 [7] 沈远海, 马远良, 屠庆平, 等. 浅水声速剖面用经验正交函数(EOF)表示的可行性研究[J]. 应用声学, 1999, 18(2): 21−25.Shen Yuanhai, Ma Yuanliang, Tu Qingping, et al. Feasiblity of description of the sound speed profile in shallow water via empirical orthogonal functions (EOF)[J]. Journal of Applied Acoustics, 1999, 18(2): 21−25. [8] 张镇迈, 李整林, 戴琼兴. 利用有限深度声速数据重构全海深声速剖面[J]. 声学技术, 2008, 27(5): 106−107.Zhang Zhenmai, Li Zhenglin, Dai Qiongxing. Sound speed profile reconstruction from the data measured in a limit depth[J]. Technical Acoustics, 2008, 27(5): 106−107. [9] Tolstoy A, Diachok O, Frazer L N. Acoustic tomography via matched field processing[J]. The Journal of the Acoustical Society of America, 1991, 89(3): 1119−1127. doi: 10.1121/1.400647 [10] 何利, 李整林, 彭朝晖, 等. 南海北部海水声速剖面声学反演[J]. 中国科学: 物理学 力学 天文学, 2011, 41(1): 49−57.He Li, Li Zhenglin, Peng Zhaohui, et al. Inversion for sound speed profiles in the northern of South China Sea[J]. Scientia Sinica (Physica, Mechanica & Astronomica), 2011, 41(1): 49−57. [11] Piao Shengchun, Yan Xian, Li Qianqian, et al. Time series prediction of shallow water sound speed profile in the presence of internal solitary wave trains[J]. Ocean Engineering, 2023, 283: 115058. doi: 10.1016/j.oceaneng.2023.115058 [12] 李倩倩, 李宏琳, 曹守莲, 等. 基于遥感数据和表层声速的全海深声速剖面反演[J]. 海洋学报, 2022, 44(12): 84−94.Li Qianqian, Li Honglin, Cao Shoulian, et al. Inversion of the full-depth sound speed profile based on remote sensing data and surface sound speed[J]. Haiyang Xuebao, 2022, 44(12): 84−94. [13] 谢骏, 胡均川, 笪良龙, 等. 海洋声速剖面的自动聚类研究[J]. 海洋学报, 2009, 31(2): 34−39.Xie Jun, Hu Junchuan, Da Lianglong, et al. Study on automatic clustering of sound speed profile in the ocean[J]. Haiyang Xuebao, 2009, 31(2): 34−39. [14] 杨帆, 王华, 高文典, 等. 北大西洋深海声速类型区划及会聚区特征分析[J]. 海洋预报, 2021, 38(6): 103−110.Yang Fan, Wang Hua, Gao Wendian, et al. Zoning of sound speed profile types and characteristics of convergence zone in the deep North Atlantic Ocean[J]. Marine Forecasts, 2021, 38(6): 103−110. [15] Liu Yuyao, Chen Wei, Chen Yu, et al. Ocean front reconstruction method based on K-means algorithm iterative hierarchical clustering sound speed profile[J]. Journal of Marine Science and Engineering, 2021, 9(11): 1233. doi: 10.3390/jmse9111233 [16] Jones D C, Holt H J, Meijers A J S, et al. Unsupervised clustering of southern ocean Argo float temperature profiles[J]. Journal of Geophysical Research: Oceans, 2019, 124(1): 390−402. doi: 10.1029/2018JC014629 [17] Li Bo, Xu Jin, Pan Xinxiang, et al. Marine oil spill detection with X-band shipborne radar using GLCM, SVM and FCM[J]. Remote Sensing, 2022, 14(15): 3715. doi: 10.3390/rs14153715 [18] Li Hong, Xu Fanghua, Zhou Wei, et al. Development of a global gridded Argo data set with Barnes successive corrections[J]. Journal of Geophysical Research: Oceans, 2017, 122(2): 866−889. doi: 10.1002/2016JC012285 [19] Shi Jiulin, Xu Ning, Luo Ningning, et al. Retrieval of sound-velocity profile in ocean by employing Brillouin scattering LiDAR[J]. Optics Express, 2022, 30(10): 16419−16431. doi: 10.1364/OE.457095 [20] Wang Hua, Li Yunbo, Li Qinghong, et al. Cluster analysis of deep water sound speed profiles in Indian Ocean[J]. IOP Conference Series: Earth and Environmental Science, 2019, 310(5): 052048. doi: 10.1088/1755-1315/310/5/052048 -
计量
- 文章访问数: 387
- HTML全文浏览量: 142
- PDF下载量: 33
- 被引次数: 0
