apply RCLV

[hscannell]

I've replicated the a 2-layer QG model in Zhang et al. 2020 using the same simulation parameters. Lagrangian particles are initialized and allowed to advect for 90 days. The particle history (position, velocity, vorticity, and PV) is captured daily. Coherent eddies are identified by averaging the vorticity deviation is along the Lagrangian trajectories (LAVD) over 60 days. Coherent eddy cores are located where there is a local maximum in LAVD within a minimum separation of 20 pixels between maxima. The material boundary of the eddy is determined using the Coherency Index (CI), which measures the variation in the density of particles originating within the LAVD material boundary. As in Zhang et al. (2020), I've used a CI threshold of -0.75. Based on the sensitivity test in Zhang et al. (2020), this threshold was found to be the critical value below which the contour encloses incoherent particles. They found that "increasing (decreasing) the magnitude of the CI threshold increases (decreases) the number of detected eddies by about 25% but does not significantly affect the statistics of coherent eddy properties (e.g., the eddy radius, zonal propagation velocity, and meridional PV flux)." Finally, the last step is to remove any detected eddies that contain fewer than 200 particles.

Figure Caption: Reproduced Figure 3 from Zhang et al. (2020). Potential vorticity anomaly at the start of 60-day eddies detected from the RCLV method. The initial boundary of the eddies are contoured in red. The black lines show the trajectory of the eddy core. The dots show the particle positions 60 days after originating within the eddy boundary and are colored to show groupings of coherent structures.

[rabernat]

Fantastic work! This looks great! Looking forward to chatting about this tomorrow.