Delay write to file after a given particle age

[MDTocean]

Question

I am releasing particles at every timestep (i.e. repeatdt is on) and location in my domain, then deleting particles that don't satisfy a condition. However, the initial timestep of all particles is still saved in the output. To reduce filesize, is there a way of ensuring the first timestep of the particles is not included in the saved output?

[erikvansebille]

Hi @MDTocean, thanks for using Parcels and reaching out with this question.

The simplest solution is to first do a pset.execute() without the output_file argument (depending on which version of parcels you use, you can set dt=0) and after that do your normal pset.execute()simulation with output_file. The first call will delete all the particles so they won't appear in the output file

if this is giving you issues (it may be tricky in some versions), you could also create a new ParticleSet from the positions of the old ParticleSet after the execute without output_file, and then go ahead as you planned

Does that make sense?

[MDTocean]

Hi @erikvansebille
Thank you very much for your response. It may be that I don't understand the functionality of setting dt to 0 (which is new to me), but my understanding of your suggestion is to pset.execute the first timestep without an output file, then pset.execute with a filename from the second timestep onwards. If that's right, then I could see how that would work if I am releasing all particles in the first timestep, but since repeatdt=True I am releasing particles at all timesteps, and so the initial state of those later particles will still be saved to the output. Have I understood correctly? Thanks again, and feel free to let me know if it would help to see my code setup.

[erikvansebille]

Ah, if you use repeatdt, then the second suggestion is more feasible. Something like (rough code; not tested)

# Create a ParticleSet with all Particles
pset_full = ParticleSet(...)

# run with dt=0 so that Particles that don't satisfy the condition are deleted
pset_full.execute(...., dt=0)

# Now create a new ParticleSet with the lon and lat of the full pset
pset = ParticleSet(..., lon=pset_full.lon, lat=pset_full.lat, depth=pset_full.depth, repeatdt=...)

# run that ParticleSet as you would otherwise do
pset.execute(..., outputfile=...)

[MDTocean]

Very useful, I will try this method. While I am running some tests, you mentioned that setting dt=0 will depend on version -- from which version can this be used? I am using version 2.4.2.
Many thanks again.

[VeckoTheGecko]

then deleting particles that don't satisfy a condition

What is the condition? Are you sampling fields? Is it not easier to seed the particles before passing it through to Parcels?

[MDTocean]

Hi @VeckoTheGecko
Thanks for your question. I am uploading a tracer field into the fieldset, which the delete-condition uses to delete particles associated to a below-threshold tracer value. This tracer field associated to the particles depreciates through the simulation, therefore the condition serves two useful purposes: to automate the particle seeding (by deleting anything with low tracer concentration at the start), and also to delete particles later on once their tracer has substantially depreciated. This method further takes advantage of Parcels' interpolation of tracer fields (from any grid) onto particle positions, which I find very convenient for setting up experiments using different tracer fields as initial condition. Does that make sense?

[VeckoTheGecko]

Yes, that makes sense. I think there are a few options:

• @erikvansebille 's above
• Run a short simulation (one step) to sample the field, then decide positions to run another simulation

Hope that helps