search

geoschem / geos-chem

GEOS-Chem "Science Codebase" repository. Contains GEOS-Chem science routines, run directory generation scripts, and interface code. This repository is used as a submodule within the GCClassic and GCHP wrappers, as well as in other modeling contexts (external ESMs).
http://geos-chem.org
Other
169 stars 165 forks source link

[QUESTION] How to speeding up nested grid model run? #924

Closed dorosieun closed 3 years ago

dorosieun commented 3 years ago

Ask a question about GEOS-Chem:

Hi :> I am currently attempting a nested model run for Asia with 0.25x0.3125 resolution. My domain for simulation is as follows. lon : 70 - 150E lat : -7 - 50N As a result of my test, it takes about 12h/day to run for the domain. I've seen how to speed up as suggested in #875, and that method has already been applied by me. Is there any other technique to speed up?

I'm running GEOS-Chem with PBS scheduler and available number of threads is 36 per 1 nodes.

yantosca commented 3 years ago

We have a discussion thread about this. Please see: https://github.com/geoschem/geos-chem/discussions/875. Long story short, you may be able to get your nested grid run to go faster if you use 600/1200 for the timesteps in input.geos

dorosieun commented 3 years ago

Thanks @yantosca. I keep looking the discussion #875.

@msulprizio said,

I would therefore recommend that users revert to the 10/20 minute (600/1200 s) timesteps in input.geos for MERRA-2 nested-grid simulations if they are looking to improve run time. I will follow up with the Nested Model Co-Chairs to make sure they are on board with this recommendation and, if so, we can implement this fix in GEOS-Chem 13.3.0.

But, I used 0.25x0.3125, GEOS_FP nested-grid. The default setting of this grid is 300/600s. Will the accuracy of the simulation be reasonable even if the time step is doubled(600/1200s)?

msulprizio commented 3 years ago

@dorosieun That's right. The recommendation has always been to use the 300/600s timesteps when using 0.25x0.3125 GEOS-FP nested-grid simulations because of the higher grid resolution.

Here is also a wiki page with some tips for speeding up your simulations: http://wiki.seas.harvard.edu/geos-chem/index.php/Speeding_up_GEOS-Chem.

yantosca commented 3 years ago

I am going to close this issue...we will post further information about speeding up nested-grid runs to the discussion #875.