An ocean current-oriented graph-based model for representing Argo trajectories. (September 2022)

• Record Type: Journal Article
• Authors: Cunjin, Xue; Zhang, Tianyu; Xu, Yangfeng; Su, Fenzhen
• Is Part Of: Computers & geosciences. Volume 166(2022)
• Publication Date: 2022-09
• Subjects: Argo trajectory -- Trajectory semantics -- Hierarchical representation -- Graph-based model -- Ocean currents; Environmental policy -- Periodicals; 550.5
• DOI: 10.1016/j.cageo.2022.105143
• Languages: English
• ISSNs: 0098-3004

Abstract

Abstract: Argo (array for real-time geostrophic oceanography) trajectories, generated by ocean currents at different depths, are a vital data source for studying ocean currents. One traditional method for representing the trajectories ‒ which uses a time-ordered series of spatial locations ‒ confuses the Argo trajectories corresponding to the parking depths and the one corresponding to the sea surface. It is, therefore, a great challenge to study ocean currents directly based on the Argo trajectories available in the Argo Global Data Assembly Centers (GDACs). Based on the working principles of Argo floats, in this paper, a hierarchical semantic representation of Argo trajectories is proposed and an ocean current-oriented graph model for representing Argo trajectories, named CoGAT, is designed. CoGAT is based on the following ideas. 1). The complete trajectory of an Argo float is represented by a node named TrajNode, which includes several nodes labeled ParkingTrajNode and labeled SurfaceTrajNode by edges, i.e. IncludingEdges. Each ParkingTrajNode represents the trajectory that an Argo float follows at its parking depth, whereas the SurfaceTrajNode represents the trajectory at the sea surface. 2) By IncludingEdges, each ParkingTrajNode or SurfaceTrajNode includes many sub-nodes, each of which is a ParkingSubTrajNode or SurfaceSubTrajNode, and each of these sub-nodes represents a sub-trajectory within a cycle of an Argo float at a parking depth or at the sea surface. A time-ordered relationship between two ParkingSubTrajNodes or two SurfaceSubTrajNodes is represented by the edge class SequenceEdge. 3) Nodes named LocationNode record the spatial locations of an Argo float, and two LocationNodes make up a sub-trajectory within a cycle; i.e., they constitute a ParkingSubTrajNode or a SurfaceSubTrajNode. A time-ordered relationship between two LocationNodes means a direction of a sub-trajectory at a parking depth or at the sea surface; this relationship is denoted as a ParkingSubTrajEdge or SurfaceSubTrajEdge. 4) The six classes of nodes and four classes of edges described above allow CoGAT to not only separate the Argo trajectories into separate trajectories at different parking depths and at the sea surface but also enable the Argo trajectories to be stored using a temporally granular representation; e.g., monthly, seasonal or annual. In this study, global Argo trajectories for the period January 2000 to April 2020 were stored in a Neo4j-based graph database named GATDB, and comparisons with the spatial Oracle database of Argo trajectories demonstrated the ability of CoGAT to provide a spatial and temporal representation of Argo trajectories at different depths and to store these trajectories. Highlights: A graph-based Argo trajectory model for analyzing ocean currents is developed. A hierarchical semantic representation of Argo trajectories at different parking depths is proposed. Six classes of nodes and four classes of edges are used to describe Argo trajectories. The proposed model performs better than Oracle at representing and storing Argo trajectories.