MFSTR-tree：面向Argo海洋浮标的时空数据索引

针对Argo海洋浮标数据的准实时性、海量性、时空异变性等特点和多种查询应用需求,分析了当前时空索引方法的优势与不足,提出了一种多频率STR-tree索引与格网索引的混合索引结构MFSTR-tree。该方法在首先轨迹束层利用动态轨迹束作为叶节点生成STR-tree结构,将STR-tree索引结构灵活、数据冗余少的优势进一步扩大;接着通过轨迹束的多种频率在采样点层构建格网索引,实现在查询效率上的提升;同时给出了该结构插入算法和查询算法的具体描述。本文以中国Argo实时资料中心提供的2015年海洋浮标数据为例,将该方法与HR-tree和STR-tree方法进行了构建效率和查询效率的对比实验,结果表明该方法在保证了构建存储效率和时间效率的同时,有效改善了原有STR-tree应用于Argo数据中的查询效率问题。

MFSTR-tree：An Spatio-temporal Index for Argo Floats Data

Management mode of relational database and file type database is difficult to support the data of Argo floats data. Because Argo buoy floats freely with sea current and data volume is huge in scale as the mobile object. In view of the quasi-real-time, massive nature, spatio-temporal variation and other characteristics of Argo ocean buoy data, as well as multiple query application requirements, the advantages and disadvantages of the current space-time indexing method are analyzed. The deficiencies of current spatio-temporal indexing method include: (1) The Argo buoy data has large volume and are observed over a long time span. When a STR-tree index is established on the trajectory of a buoy, the over-long trajectory tends to lead to a high overlap ratio between the MBB of the STR-tree, which further leads to the reduction of search efficiency. (2) The Argo floats data are sampled at different frequencies and with relatively stable frequency, but the influence of the update frequency on index structure optimization is often ignored. Therefore, a hybrid index structure called MFSTR-tree with multi-frequency STR-tree index and grid index is proposed. First, the dynamic trajectory beam is used as the leaf node to generate the STR-tree structure in the trajectory beam layer, taking advantage of the flexibility and the less data redundancy of the STR-tree index structure. Then, the improvement of query efficiency is realized by use of the multiple frequencies of the track beam at the sampling point layer according to the construction grid index. The corresponding interpolation algorithm and query algorithm are described in this paper. To verify the construction and query efficiency of MFSTR-tree, a comparison experiment was conducted with HR-tree and STR-tree for Argo floats in 2015 from China Argo real-time data center. The experimental results show that under the premise of guaranteeing construction time efficiency and storage efficiency, HR-tree still maintains natural advantages in single-time query and is much more efficient than the other two. After optimization, the MFSTR-tree had efficiency of 40% higher than the general STR-tree. The query efficiency of the HR-tree decreases significantly with the query window size expanded to 4% of the total range. MFSTR-tree is further optimized on the basis of the original STR-tree, which improves the efficiency of the sampling point selection process in the trajectory bundle. Therefore, the advantage is more obvious, and the verification of the algorithm is realized.