SLA-Based Orthogonal Parallel Detection of Global Rotationally Coherent Lagrangian Vortices

Abstract

Abstract In this paper, we present a highly effective orthogonal parallel algorithm for identifying global rotationally coherent Lagrangian vortices (RCLVs) in heterogeneous systems and a long-time-scale global sea level anomaly (SLA)-based RCLVs product. First, a many-core parallel computing method is used to accelerate the Lagrangian-averaged vorticity deviation (LAVD) computing process. The computation is approximately 8000 times faster than that of a previous method. Second, the global LAVD field is divided into several regions. These regions are searched with a multiprocess CPU parallel pool to identify simultaneously RCLVs. All the identified RCLVs in these regions are merged seamlessly into a global eddy map. The algorithm improves the global RCLV identification efficiency, making the proposed method approximately 20 times faster than a single-threaded method. The LAVD many-core computing method and the RCLV multiprocess parallel method are orthogonally combined. The resulting algorithm is at least 500 times faster than previous nonparallel methods, making the computing of global RCLVs feasible. Third, the advection of Lagrangian particles in RCLVs and Eulerian eddies is analyzed to demonstrate the material coherence of RCLVs and the reliability of our algorithm. Finally, a global RCLVs product from 1993 to 2019 containing 52 567 eddies is produced with a 90-day time interval. This is the first time that a long-time-scale global Lagrangian eddy product has been presented.