You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi everyone (yes it is me again, unfortunately...)!
I have 20 years of data, and I would like to seed every week every year and let the particles live for 20 years before killing them. I can't run simulations lasting longer than 7 real days on the server I use, so the solution was to run 20 independent jobs corresponding to a different start year for seeding which would cost less computation time and I would combine the outputs when doing post-processing.
In each job, I would seed the particles every week for the first year, and let them advect for 20 years. So my years range from 181 to 200, so that would mean that for year 195 for example, it would seed particles every week during year 195, and then let them advect for 20 years (doing 195-200 and then 181-194, I hope it is clear).
In the beginning, I wanted to create the fieldset so that it followed that order, but as the times shifted from 200 to 181 at a certain moment, Parcels did not like it as it wanted timesteps in chronological order... The weak solution I found was to load the 20 years in chronological order and then assign a time to each particle, corresponding to the targeted seeding year and week. This works fine with year 181, but for the year 198 for example, it means it would have to run double the time (running until year 198 with no seeding, then finally seeding but having to run for another 20 years, gives a runtime of 38 years in total...)
FYI: I use the argument time_periodic=runtime when creating the Fieldset to loop periodically over the years.
I don't know if there is a solution for this kind of problem, changing the time in the files before giving them to parcels for each run is possible but might take a lot of time for each simulation... So I was wondering if anyone already ran into this kind of problem and has a loophole, or did I miss something in one of the tutorials for that?
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
Hi everyone (yes it is me again, unfortunately...)!
I have 20 years of data, and I would like to seed every week every year and let the particles live for 20 years before killing them. I can't run simulations lasting longer than 7 real days on the server I use, so the solution was to run 20 independent jobs corresponding to a different start year for seeding which would cost less computation time and I would combine the outputs when doing post-processing.
In each job, I would seed the particles every week for the first year, and let them advect for 20 years. So my years range from 181 to 200, so that would mean that for year 195 for example, it would seed particles every week during year 195, and then let them advect for 20 years (doing 195-200 and then 181-194, I hope it is clear).
In the beginning, I wanted to create the fieldset so that it followed that order, but as the times shifted from 200 to 181 at a certain moment, Parcels did not like it as it wanted timesteps in chronological order... The weak solution I found was to load the 20 years in chronological order and then assign a time to each particle, corresponding to the targeted seeding year and week. This works fine with year 181, but for the year 198 for example, it means it would have to run double the time (running until year 198 with no seeding, then finally seeding but having to run for another 20 years, gives a runtime of 38 years in total...)
FYI: I use the argument
time_periodic=runtime
when creating theFieldset
to loop periodically over the years.I don't know if there is a solution for this kind of problem, changing the time in the files before giving them to parcels for each run is possible but might take a lot of time for each simulation... So I was wondering if anyone already ran into this kind of problem and has a loophole, or did I miss something in one of the tutorials for that?
Thanks in advance!
Esteban
Beta Was this translation helpful? Give feedback.
All reactions