search

NCAR / amwg_dev

Repo to store model sandboxes and cases used for CAM development
9 stars 2 forks source link

b.cesm3_cam058_mom_e.B1850WscMOM.ne30_L58_t061.camdev_cice5.026g_tx2_3.004 #251

Open gustavo-marques opened 1 year ago

gustavo-marques commented 1 year ago

Description:

Same as 26g_tx2_3 but with the following modifications:

1) Land/sea mask modifications around Antarctica. Run 26g_tx2_3 had a few single grid points along the Antarctic coast. Because CICE is still using a B-grid, ice pilled up at these points, reaching ~ 300 m in a few locations. The modified land/sea mask eliminates these points;

2) The mapping r05 >> tx2_3 was created using 

eFold        = 1000000.0 m
rMax         = 1000000.0 m

instead of the default values

eFold = smoothing eFold distance in meters (default: 1000000)
rMax  = maximum radius of effect in meters (default: 300000)

The new r05 >> tx2_3 mapping should increase the spreading of liquid and frozen runoff around Antarctica. A different mapping file is used for Greenland gland4km >> tx2_3 and the default eFold and rMax values are used in this case.

Additional notes:

Case directory: Locally (if still available): /glade/p/cesmdata/cseg/runs/cesm2_0/b.cesm3_cam058_mom_e.B1850WscMOM.ne30_L58_t061.camdev_cice5.026g_tx2_3.004

Sandbox: Locally (if still available): /glade/work/gmarques/cesm.sandboxes/cesm3_cam6_3_058_MOM_e_tx2_3

Contacts:

@gustavo-marques and @cecilehannay

adamrher commented 1 year ago

Could you provide definitions of rx0, eFold and rMax? How are these made regionally / ice sheet specific? In particular, this statement: "A different mapping file is used for Greenland gland4km >> tx2_3". Sorry, I'm the questions guy.

gustavo-marques commented 1 year ago

No worries. These are good questions!

eFold and rMax are input parameters for the mapping file generation tool. These were already defined above, but I am adding the definition again here. 

eFold = smoothing eFold distance in meters (default: 1000000)
rMax  = maximum radius of effect in meters (default: 300000)

rx0 (r05) is the runoff grid used in B-compsets. I could not find more info about this grid other than the ASCII file below. Perhaps @billsacks can tell us more about r05 :smile:

/glade/p/cesm/cseg/inputdata/lnd/clm2/rtmdata/rdirc.05.061026

Liquid runoff to the ocean (ROF >> OCN) is spread via ROF2OCN_LIQ_RMAPNAME, which is the r05 >> tx2_3 mapping in this case. There is also a ROF2OCN_ICE_RMAPNAME mapping file, which controls how frozen runoff (icebergs) is spread. We usually use the same mapping file for ROF2OCN_LIQ_RMAPNAME and ROF2OCN_ICE_RMAPNAME. In this run GLC is active only in Greenland (this grid is called gris4) and there are liquid and frozen fluxes that go directly from GLC >> OCN. These are spread via GLC2OCN_LIQ_RMAPNAME and GLC2OCN_ICE_RMAPNAME. Again, we use the same mapping for the latter two. Because we have liquid/frozen fluxes to OCN coming from both GLC and ROF, we can use different mapping files for them.

Hope this helps clarify.

adamrher commented 1 year ago

Thanks for clarifying this for me. I never really understood the nuances of how runoff is mapped to the ocean. I'm inferring based on the naming conventions I'm aware of:

SMAPNAME <---mapping of scalar fields (bilinear) FMAPNAME <---mapping of fluxes (conservative) RMAPNAME <---mapping river fluxes (super-duper special mapping that takes the fluxes at the river "mouth" and spreads over some finite distance of ocean points, set by parameters eFold and rMax).

I didn't realize the GLC component was active in our B-case runs. I recall @Katetc switched all our F-cases to stub glacier a while back, I think because of weird conflicts between ATM_GLC and NUOPC. Looking through the sandbox Gustavo is using, it looks like the B-cases are still set to CISM2%GRIS-NOEVOLVE mode. So I guess we fixed the glitch or something ...

Anyhow. Kate, Bill or @whlipscomb I had thought the SMB contribution to runoff was included in the river fluxes to the ocean. Maybe not? Could you clarify the purpose of the GLC2OCN mapping files in these runs? It's my understanding that in NOEVOLVE mode there is no explicit calving flux available on the CISM grid, so I'm just not sure what's being mapped in this configuration.

jedwards4b commented 1 year ago

The ATM_GLC issue may have been simply that your runs were shorter than the GLC timestep - that's been fixed in more recent versions of CIME (it may still error out but it gives an error message that explains how to correct the situation.)

Katetc commented 1 year ago

I didn't realize the GLC component was active in our B-case runs. I recall @Katetc switched all our F-cases to stub glacier a while back, I think because of weird conflicts between ATM_GLC and NUOPC. Looking through the sandbox Gustavo is using, it looks like the B-cases are still set to CISM2%GRIS-NOEVOLVE mode. So I guess we fixed the glitch or something ...

In default B cases CISM should always be in NOEVOLVE mode. In CMIP6 cism was NOEVOLVE except for our specific interactive ice sheet cases. We just need a specific model set up for fully coupled ice sheet runs that isn't accepted for the deck runs yet. The F-cases got converted to stub largely because cism takes a long time to build and that was annoying in the cam aux testing suite, and there's no scientific reason to include any sort of active ice sheet in an F case.

Anyhow. Kate, Bill or @whlipscomb I had thought the SMB contribution to runoff was included in the river fluxes to the ocean. Maybe not? Could you clarify the purpose of the GLC2OCN mapping files in these runs? It's my understanding that in NOEVOLVE mode there is no explicit calving flux available on the CISM grid, so I'm just not sure what's being mapped in this configuration.

I could always be wrong here... but in my understanding CISM liq and ice run off is routed through the river model when it is in NOEVOLVE mode, as it looks like this B case is. There have been experiments in ocean-ice sheet coupling in fully active mode, and these would likely be required to run active Antarctica. That coupling is mostly for temperature and salinity at the base of the ice shelves in Antarctica. This isn't really a requirement in Greenland, though.

Sooooo, I think that Gustavo shouldn't need GLC2OCN mapping files here, and if they are specified in a NOEVOLVE situation then they probably aren't doing anything or being used. Feel free to tell me I'm wrong though.

billsacks commented 1 year ago

Responding to a few recent points:

adamrher commented 1 year ago

Thanks Kate and Bill for this additional information. The handling of the runoff fluxes is much clearer to me now.

I would also like to follow-up to understand what grid / resolution is this rx1. I'm only familiar with r05 which is 0.5deg grid.

@billsacks thx for the pointer the GLC2OCN documentation. Not that you need more work to do, but section A.1.4.3 is having users modify cime/cime_config/cesm/config_grids.xml, which is outdated for ~cesm2.3 tags or greater.

billsacks commented 1 year ago

Thanks for pointing out that issue with the documentation. I'll at least open an issue for it.

References I see to rx1 are for the data runoff model; I don't know anything more about it.

gustavo-marques commented 1 year ago

Thanks @billsacks and @Katetc for the clarifications!

The ROF grid is indeed r05 and not rx0. Sorry for the confusion.

I assumed the GLC2OCN mapping was being used because the model failed before I updated the mapping file using the new ocean grid. We have started a discussion via email about keeping GLC and ROF fluxes separately in the ocean model. This confusion is one more reason for keeping these fluxes separate.