水系划分的剖面相似系数聚类法

提出一种基于温度剖面相似系数的水系划分方法。基本思想是:将各温度剖面视为独立样本,各深度数据为样本变量,先基于划分区域水文特征选取合理数量的水系中心剖面,再利用各样本与各中心剖面相似系数大小进行聚类,得到各水系划分数据集合;利用几何平均求得各水系集合新的中心剖面,重复剖面相似系数聚类过程,直至中心剖面不再变化为止。最后利用国家海洋信息中心发布的中国近海CTD温、盐产品对该方法进行试验,并对聚类得到的各类温度剖面展开讨论。结果较好的反应出各区域温度剖面特征,综合体现出东海各区域温度大小、海流、水团和水深特性。

Water System Division by Clustering Algorithm of Profile Similarity Coefficients

The division of temperature profiles is of great importance to discuss hydrology characteristics and underwater acoustic classification, and in this paper a method of water system division by clustering algorithm of profile similarity coefficients is put forward. The basic thought is： each temperature profile is treated as a sample and data of different layers as variable. Firstly, several reasonable profiles of water system centers are picked out. Then, the clustering algorithm of profile similarity coefficients is used to cluster them and the musters of each water system are obtained. Finally, the geometry average method is used to get the new water system centers, and the cluster steps are repeated until the water system centers are stable. In the end, the CTD data of China Ocean Information Center is used to experiment this cluster algorithm. The cluster results are analyzed, which show reasonable characteristics of the temperature profiles in each area, and generally reflects the features of temperature, ocean currents, water mass and depth in the East China Sea.