基于K-means聚类分析与单经验正交函数回归法的南海声速剖面估计方法研究

基于遥感参数和Argo历史数据对水体声速剖面（Sound Speed Profile，SSP）进行重构，对单经验正交函数回归（single Empirical Orthogonal Function-regression，sEOF-r）法在南海的适用性进行了研究。由于南海动力活动的复杂性，SSP扰动相对复杂，同时海域内SSP样本稀疏，相关的SSP统计学估计方法在南海区域还难以有效应用。文章基于K-means对样本进行聚类分析，讨论南海海域正交经验函数模态的一致性。通过扩大重构实验网格解决样本稀疏的问题。利用经典的sEOF-r对南海SSP进行反演，对重构SSP的误差分析说明了该方法在南海海域应用的有效性。SSP重构的均方根误差为2.341 1 m·s^-1，较大误差主要出现在深度40~200 m，其原因是海域内混合层深度发生变化。实验证明在南海区域内利用遥感参数可以有效地估计SSP。

Sound speed profile inversion in the South China Sea based on K-means cluster analysis and single empirical orthogonal function regression

The sound speed profile (SSP) is reconstructed by remote sensing parameters and Argo previous data in the South China Sea, and the applicability of single empirical orthogonal function regression (sEOF-r) in the South China Sea is studied. Due to the complexity of hydrodynamic activities in the South China Sea, the corresponding SSP disturbance is relatively complex, and meantime the SSP samples in the sea area are so sparse that the related SSP estimation methods are still difficult to be effectively applied. Based on the K-means cluster analysis of samples, the consistency of the orthogonal empirical function modes is discussed in this paper. Expanding the inversion grid can solve the problem of sparse samples. The classic sEOF-r is used to invert the SSP in the South China Sea, and the error analysis of the reconstructed SSP is used to prove the effectiveness of the method. The root mean square error of the SSP reconstruction is 2.341 1 m·s^-1, and the larger error mainly occurs at the depth of 40-200 m. The reason is that the depth of the mixed layer changes in the sea area. The experiment demonstrates that the SSP in the region of the South China Sea can be estimated efficiently by use of remote parameters.