An Improved Winding-Angle Method to More Accurately Identify Mesoscale Eddies

Abstract

AbstractThe winding-angle (WA) method is an automatic eddy detection method based on the geometric characteristics of instantaneous streamlines. The original WA method clusters closed streamlines using a predetermined threshold. It is difficult to obtain a common threshold for accurately clustering various mesoscale eddies with variable shapes and dimensions. Moreover, the original WA method is not suitable for detecting multicore structures. In this paper, an improved WA method was proposed to more accurately identify mesoscale eddies and to detect multicore structures. It does not depend on the previously used clustering threshold; rather, it is based on the spatial relationships of streamlines to detect mesoscale eddies of various types and dimensions. Streamlines are matched with possible eddy centers (PCs), which are then grouped into different ?related groups? according to the containment relationships between them and the outermost streamlines of the groups. Each group represents a vortex structure, and the number of PCs in each group represents the number of eddy cores. The eddy boundaries and eddy cores of multicore structures represented by multi-PC groups are identified by topological relationships of the streamlines. The time requirement of the improved method is higher than that of the original algorithm, although it does not demand additional memory space and utilizes fewer CPU resources. More importantly, the improved method provides more accurate identification results and greatly refines the incorrect identifications from the original method induced by the predetermined threshold. Success metrics for the improved WA method are also more desirable relative to those for the original and other commonly used methods.