search

ESMCI / cime

Common Infrastructure for Modeling the Earth
http://esmci.github.io/cime
Other
160 stars 207 forks source link

Add support for half-degree CAM runs in CESM #2467

Closed billsacks closed 5 years ago

billsacks commented 6 years ago

@tilmes says that the atmospheric chemistry group would really like to have out-of-the-box support for half-degree CAM runs for CESM2. I'm not sure if they need support for fully-coupled or just F compsets. They have the mapping files they need, but these are currently being set in some way other than the standard config_grids.xml mechanism. I'm opening this issue on her behalf, to get these moved into config_grids.

gold2718 commented 6 years ago

I would love to know what is really expected here. This issue is appropriate for the grid alias and mapping files but without initial datasets and possibly land surface data sets, it is not going to do much. Also, the mapping files need to be added to the appropriate places in the inputdata repo. @tilmes, can you communicate what cases need to run with the CAM group via email?

tilmes commented 6 years ago

We only need F-cases to work. We do have initial datasets and land surface datasets available. The mapping files are in /glade/p/cesmdata/cseg/inputdata/cpl/gridmaps/fv0.47x0.63. This is a request for the CAMchem group. We want in particular this compset to work FCSD --res f05_g16. Thanks.

mvertens commented 6 years ago

@fvitt - can you please take this issue on and have it done before the code freeze?

fvitt commented 6 years ago

I will work with Simone on this when I find some time. I will need to figure out how to specify the correct mapping files out of the box for the 1/2 degree resolution. If someone (Bill?) can give me some pointers on how to do this that would be helpful.

billsacks commented 6 years ago

@fvitt the mapping files need to be specified in cime/config/cesm/config_grids.xml

mvertens commented 6 years ago

@fvitt - The key issue is that this needs to be done before May 11 - when the code freeze is scheduled to occur. If we cannot have this in place before then we can target this functionality for cesm2.1 - which will occur in early fall.

On Wed, May 2, 2018 at 11:16 AM, fvitt notifications@github.com wrote:

I will work with Simone on this when I find some time. I will need to figure out how to specify the correct mapping files out of the box for the 1/2 degree resolution. If someone (Bill?) can give me some pointers on how to do this that would be helpful.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ESMCI/cime/issues/2467#issuecomment-386052945, or mute the thread https://github.com/notifications/unsubscribe-auth/AHlxE0KuCx-v0rPxoLL4mrI1agtnHZPVks5tuenYgaJpZM4TSgJy .

fvitt commented 6 years ago

In the case I tried I only had an issue with the clm surface file. By removing the crop option the case was able to run using: surfdata_0.47x0.63_16pfts_Irrig_CMIP6_simyr1850_c170919.nc

jedwards4b commented 6 years ago

Is it an acceptable solution then to define a new compset for this without CROP enabled?

fvitt commented 6 years ago

I don't believe so.

On Fri, May 4, 2018 at 12:18 PM, jedwards4b notifications@github.com wrote:

Is it an acceptable solution then to define a new compset for this without CROP enabled?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESMCI/cime/issues/2467#issuecomment-386690447, or mute the thread https://github.com/notifications/unsubscribe-auth/AI6KpsIwbZ5gG4QiAnMCqU059XWDb_R7ks5tvJuFgaJpZM4TSgJy .

ekluzek commented 6 years ago

That part of this requires a CTSM update. Which corresponds to this issue...

https://github.com/ESCOMP/ctsm/issues/364

fvitt commented 6 years ago

For the f05_f05_mg17 grid the *_DOMAIN_FILE vars are set to "UNSET" which seems to be a problem. Where are these set?

mvertens commented 6 years ago

Have you created these domain files for the new mg17 mask? You would need to run the code in cime/tools/mapping/gen_domain_files.

On Thu, May 17, 2018 at 10:25 AM, fvitt notifications@github.com wrote:

For the f05_f05_mg17 grid the *_DOMAIN_FILE vars are set to "UNSET" which seems to be a problem. Where are these set?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ESMCI/cime/issues/2467#issuecomment-389926659, or mute the thread https://github.com/notifications/unsubscribe-auth/AHlxE7m4wCDeIeyzLJFnAeO1ESZKyAOYks5tzaRwgaJpZM4TSgJy .

fvitt commented 6 years ago

I am learning how to. $ gen_domain -m -o -l

What do I use for ?
Is g17 ? Is 0.47x0.63 ?

ekluzek commented 6 years ago

@fvitt Yes, exactly.

fvitt commented 6 years ago

exactly what? what do I use for filemap?

On Thu, May 17, 2018 at 12:44 PM, Erik Kluzek notifications@github.com wrote:

@fvitt https://github.com/fvitt Yes, exactly.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESMCI/cime/issues/2467#issuecomment-389968744, or mute the thread https://github.com/notifications/unsubscribe-auth/AI6KpruUgdMVWUvSF35ua8onerVTiA85ks5tzcUfgaJpZM4TSgJy .

mvertens commented 6 years ago

@ekluzek - can you please help Francis with this - I am in meetings all afternoon.

ekluzek commented 6 years ago

@fvitt we need to make the mapping file first. I've looked for it in the usual possible places, and can't find it.

@mnlevy1981 you've made mapping files in the past. Could you make this one? Or if not I know there's been issues with mapping in cime, what version of cime should I use to create the mapping file?

mnlevy1981 commented 6 years ago

Catching up on the ticket, it seems like made f05_g16 maps but you still need f05_g17 maps as well as the domain files? Starting from that point, in a cheyenne checkout of the latest CIME master:

  1. Create all the maps
cd tools/mapping/gen_mapping_files/
./gen_cesm_maps.sh -fatm [location of fv0.47x0.63 SCRIP grid file] -natm 0.47x0.63 -focn /glade/p/cesmdata/cseg/mapping/grids/gx1v7_151008.nc -nocn gx1v7 --serial

One the login node of cheyenne, this should load all the modules you need and run the ESMF tools. I tested the script by making f09_g37 maps, and CISL did not kill the process... if mapping between the finer grids uses too much CPU, you can run qinteractive and then drop --serial from the argument list and build the maps (in parallel) on a compute node.

  1. Copy these files into the proper subdirectories of /glade/p/cesmdata/cseg/inputdata/cpl/gridmaps/ (atm -> ocn maps in fv0.47x0.63/ and ocn -> atm maps in gx1v7/)

  2. Create the domain files

$ cd $CIMEROOT/tools/mapping/gen_domain_files
[follow instructions in INSTALL to build executable]
$ ./gen_domain -m [conservative map from gx1v7 -> fv0.5 that you just made] -l fv0.47x0.63 -o gx1v7 [-fv pole option?]

Regarding [-fv pole option?] -- I think you want to run with -p 2 to apply the north-pole fix on the finite volume grid, but look at the README and make sure that's really the case.

Let me know if anything goes awry, or if you need more help with anything.

ekluzek commented 6 years ago

@fvitt you should use the pole fix "-p 2" option to "gen_domain. the latest fv05 file was made by @mnlevy1981 and is here...

/glade/p/cesmdata/cseg/mapping/grids/fv0.47x0.63_141008.nc

I think that gives you everything you need to do this now. Try it and let us know what breaks...

ekluzek commented 5 years ago

This was already put in place and is working now.