海洋涡旋作为一种快速连续变化的海洋现象,如何分析和挖掘其移动特征成为当前海洋涡旋定量研究的重点.本文引入空间数据挖掘的社区网络划分方法,将涡旋过程看作复杂的移动网络,对涡旋移动的聚集性特征进行探索和分析.首先,以网格为统计单元对1992—2011年近20年南海海洋涡旋移动数据进行组织,基于图论模型构建了涡旋瞬时移动(TP),涡旋移动起止点(OD),涡旋最小描述距离的特征点移动网(MDL)和涡旋过程移动再生数据(RSP)4种状态的海洋涡旋的移动网络图;其次,采用基于快速模块度优化的区域划分方法分别得到4种状态下涡旋移动的聚集性区域;最后,利用弦图对区域内和区域间涡旋移动规律进行了可视化分析,发现海洋涡旋的RSP数据能够弥补原始涡旋移动数据在区域划分方法中呈现的数量不足的问题,能够在足够数据量的情况下,有效地发现从起点到终点的主要移动通道和涡旋移动的聚集性区域,这些区域反映了南海涡旋从其产生、发展到结束整个演化过程的聚集性特征.%Mesoscale eddies are a prominent dynamic phenomenon in the ocean with complex and continuous changes during lifecycles.It has become a research focus to investigate eddy propagation characteristics using data driven techniques.From the perspective that eddy trajectories collectively create a spatial network,this study presents a community detection method to uncover latent clusters of eddy activities.The creation of the mobility network is a-chieved by dividing the study area,the South China Sea,into regular grids and projecting eddy trajectories obtained from 1992 to 2011 to this grid-like network with stop locations as nodes and transitions as edges.The trajectories are preprocessed and generalized at four different granularities:trajectory partitions (TP),origin-destination tran-sitions (OD),minimum description length (MDL),and regeneration of strong processes (RSP).Then,a fast un-folding algorithm is applied to the networks created at different granularities to discover potential communities.Fi-nally,the relationships between different communities are visually analyzed using the chord diagram.The results show that the RSP data of eddy tracks overcome the data insufficiency in TP-network,and the method is able to i-dentify major paths and clustering patterns of eddy activities throughout the life from the RSP data.
展开▼
