Eddy tracking error: "High number of conflict (nb_conflict)"

[boatwrva]

Hi Antoine,
I've been trying to track the identified eddies from Anticyclonic and Cyclonic nc files. I have a large dataset but I'm only using 180 days right now where each identified eddy files is 1 day. The code is able to process them but it results in an empty field and tells me (I think at every timestep through which I'm running) there is a 'high number of conflict: ## (nb_conflict)':

High number of conflict : 46 (nb_conflict)
One observation have 7 links
High number of conflict : 66 (nb_conflict)
High number of conflict : 64 (nb_conflict)
High number of conflict : 57 (nb_conflict)
High number of conflict : 56 (nb_conflict)
High number of conflict : 63 (nb_conflict)
High number of conflict : 59 (nb_conflict)
High number of conflict : 60 (nb_conflict)
High number of conflict : 53 (nb_conflict)
High number of conflict : 54 (nb_conflict)
High number of conflict : 54 (nb_conflict)

I couldn't find this error in the source code and I'm not sure where it is coming from or how to change my parameters to fix it. When I used the INFO logger, a snippet of the output is:

INFO 2021-07-20 13:52:09,384 observation.load_file :
	loading file '/content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20131230.nc'
INFO 2021-07-20 13:52:09,422 groups.filled_by_interpolation :
	498 obs will be filled (unobserved)
INFO 2021-07-20 13:52:09,436 observation.load_file :
	loading file '/content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130101.nc'
INFO 2021-07-20 13:52:09,472 observation.load_file :
	loading file '/content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130102.nc'
INFO 2021-07-20 13:52:09,509 tracking.track :
	/content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130102.nc match with previous state
INFO 2021-07-20 13:52:09,515 observation.load_file :
	loading file '/content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130103.nc'
INFO 2021-07-20 13:52:09,556 tracking.track :
	/content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130103.nc match with previous state
WARNING 2021-07-20 13:52:09,561 observation.solve_simultaneous :
	High number of conflict : 46 (nb_conflict)
INFO 2021-07-20 13:52:09,566 observation.load_file :
	loading file '/content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130104.nc'
INFO 2021-07-20 13:52:09,603 tracking.track :
	/content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130104.nc match with previous state
WARNING 2021-07-20 13:52:09,609 observation.solve_simultaneous :
	One observation have 7 links
WARNING 2021-07-20 13:52:09,614 observation.solve_simultaneous :
	High number of conflict : 66 (nb_conflict)

It would be a great help to know how I could fix this and have an eddy tracker output!

Here is my code: https://colab.research.google.com/drive/1B2rE2vUrjifczql2HkTS0TCrgm-GBVVP?usp=sharing
I've successfully plotted all of the output eddy files, so I don't think it is a problem there. I have tried it on both the anticyclonic and cyclonic files, and it gives the same output (or similar) of high number conflict and an empty dataset. I've also tried the different trackers (ex default and area tracker) but they yielded the same result.

Thanks!

[AntSimi]

Could you share fews indentification outputs, by example:

• /content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130101.nc
• /content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130102.nc
• /content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130103.nc
• /content/drive/MyDrive/Ocean/tracked_eddies/all_time/Cyclonic_week_20130104.nc
  I think there is something to understand in identification

[boatwrva]

Here is a google drive link to the identification outputs: https://drive.google.com/drive/folders/17rStMWDeoeeDJS9kCLMiYLCIT3C5JSoJ

Here is also an example plot of the output at one timestep (with chlorophyll overlayed), which shows the eddy centers and their effective contours

[AntSimi]

After a more accurate reading of log, i think high number of conflict come with default tracker. It seems coherent for me because by default we search next step in a large area and in your case you have lot eddies in tiny area, that why you have a lot of candidate for each eddies. In my opinion you must prefer AreaTracker with virtual option which allow missing detection during N timesteps.

You could get general information about your output dataset like this :

print(eddies_area_tracker)

You could also if you want run Tracking with bash

If you want use GUI, maybe set value of your datas, if matplotlib interactiv is activate you could move time if you move black line in bottom axes:

g = GUI(Acyc_default_tracker=eddies_area_tracker)
g.now = 22000  # Here julian day(day from 1/1/1950) 22000 => 27/3/2010
g.bbox = 0, 9, 36, 40 # Here box area lonmin, lonmax, latmin, latmax
g.adjust()
g.show()

[boatwrva]

Hi - the tracking is working because the print(eddies_area_tracker) does print the information about the tracked eddy dataset. The empty dataset was because I hadn't realized what g.now and g.bbox were so I was plotting an eddy tracking date that didn't exist and also plotting in the wrong lon/lat area box.

However, now I have successfully tracked in python, but the documentation for 'Track in Python') does not actually save the tracked eddy dataset, compared to the 'Tracking' documentation says that the yaml file will create 4 datasets of the tracked eddies. Could you provide code to save these dataset if we are tracking in python?

[AntSimi]

To save you could use write_file method

[boatwrva]

Hi! Lots of delay, but I was able to save the datasets! This seems like a simple and silly problem, but I'm trying to run some of the post-processing files on the eddy tracking datasets.
I've ran most of the post-processing scripts and most of them work (for example, I can display the fields, find tracks that go through an area, plot one track, etc). However, I'm having trouble with the "track animation" where you choose one eddy and track it over time. In the documentation,

from py_eddy_tracker.appli.gui import Anim
from py_eddy_tracker.observations.tracking import TrackEddiesObservations


# Load experimental atlas, and we select one eddy

a = TrackEddiesObservations.load_file("/content/drive/MyDrive/Ocean/tracks/Anti_tracks/Anticyclonic.nc")

eddy = a.extract_ids([1000]).index(slice(0, 35))

a = Anim(eddy, sleep_event=0.1, intern=True, figsize=(8, 3.5), cmap="viridis") # sleep_event = 1e-10
a.txt.set_position((17, 34.6))
a.ax.set_xlim(xmin, xmax)
a.ax.set_ylim(ymin, ymax)
a.show(infinity_loop=False)

However, I just get an empty grid. I have been reading through some of the documentation on saving animations from python, but it seems like the Anim function from py_eddy_tracker isn't intuitive to me and the normal animation saving methods. Is there a way to save the animation to see if it is working but just not showing up?

Let me know what you think! Thanks