An Optimized-Parameter Spectral Clustering Approach to Coherent Structure Detection in Geophysical Flows

In Lagrangian dynamics, the detection of coherent clusters can help understand organization transport by identifying regions with trajectory patterns. Many clustering algorithms, however, rely on user-input parameters, requiring a priori knowledge about flow and making outcome subjective. Building conventional spectral method Hadjighasem et al. (2016), new optimized-parameter approach is developed that automatically identifies optimal parameters within pre-defined ranges. A noise-based metric for quantifying coherence resulting also introduced. The applied to two benchmark analytical flows, Bickley Jet asymmetric Duffing oscillator, realistic, numerically generated oceanic coastal flow. latter case, identified model-based are tested using observed trajectories real drifters. all examples, our succeeded in performing partition domain into minimal inter-cluster similarity maximum intra-cluster similarity. For flow, qualitatively similar over same phase tide different days even years, whereas opposite tidal different.