Open rabernat opened 6 years ago
Hi @rabernat, in terms of the robustness of floater
package itself, should it a better choice to make our searching algorithm independent of the coherency index? I mean a certain index is not perfectly defined and has its own strengths and weaknesses.
We've found that using the coherency index by itself tends to detect rather large, non-compact eddies with very large convexity deficiencies. This is due to the normalization by the initial spread of the particles—the search algorithm can get arbitrary small negative negative values of the coherency index by including lots of particles. A pathological example would be simply counting all the particles in a closed or periodic domain as an "eddy". The particles can never leave and, if the flow is non-divergent, they can't bunch up or spread out, so the coherency index remains close to zero for all time. It never seems to get that bad in practice, but I don't see how to avoid detecting large, non-convex "eddies" without falling back on the convexity deficiency.
The current algorithm searches for a level curve of LAVD with a specified convexity-deficiency threshold to define the outer boundary of the RCLV.
As discussed with @pittwolfe and Wenda (can't find his github handle), it would be cool to have the option to use coherency index directly (instead of CD) for the boundary.