Appending multiple particle set outputs to the same .Xarr file

[TomMBIO]

Hi everyone,

I have made some progress with trying to run multiple particle sets, each beginning on a subsequent week:

`
iterations = range(26) # Simulate for 26 weeks
psets = []

cohort = 0
release_time = datetime.datetime(2022, 3, 1, 00, 00, 0)

for i in iterations:
cohort_array = np.full(n_sites, cohort) #create an array of cohort values for each site
# Create a new ParticleSet for each iteration with updated release time and cohort ID
pset = ParticleSet(
fieldset=fieldset,
pclass=Particle,
lat=release_lat,
lon=release_lon,
depth = None,
time=release_time,
cohort=cohort_array
)

psets.append(pset)

kernels = pset.Kernel(Sample_land) + pset.Kernel(Unbeaching) + pset.Kernel(AdvectionRK4) + pset.Kernel(DiffusionUniformKh) + pset.Kernel(SampleT)
output_file = pset.ParticleFile(name=r"<filepath>\mytilus1_6.zarr", outputdt=datetime.timedelta(minutes=60), write_mode='append')

# Execute the simulation for this ParticleSet
pset.execute(
    kernels,
    runtime=datetime.timedelta(days=80), 
    dt=datetime.timedelta(minutes=60),
    output_file=output_file  
)

# Increment the cohort ID and release time for the next batch of particles
cohort += 1
release_time += datetime.timedelta(days=7)

`

Just to clarify, I am doing this instead of using repeatdt because I wish to assign each release event a cohort number.

It appears write_mode='append' does not have functionality with parcels, and I am a bit stuck. One solution could be to assign each output its own file, and combine them in post, but this feels a bit unsophisticated. Do you have any ways to append trajectories to the same xarr file automatically? I assume using multiple particlesets for an experiment is common practice but not sure.

Thank you,
Tom

[erikvansebille]

Hi @TomMBIO, I don't think you need write_mode='append'. Appending is the default (and only) way of writing to a zarr file in Parcels. What happens if you don't give that argument?

[TomMBIO]

Hi Erik,

When I didn't do this, the particle files appeared to be overwritten with each iteration. But I will double check this

[erikvansebille]

Ah, you'll have to move the creation of the ParticleFile object outside of the for-loop.

[TomMBIO]

This appears to have worked. Thank you!