Missing cice files for compset F

[odoherty@ucsd_edu]

Hi All,I am running CESM 1.2.0, with compset F on yellowstone.  I am trying to generate iop files from CAM to use with SCAM.  I have built and compiled CESM 1.2.0, however when run an error message is generated, which originates because no prescribed cice file is specified.  I am wondering where I can find the appropriate file, and how I would go about configuring the model to look for it in the right place.  Hopefully relevant error messages and model output follow.  Thanks in advance for any help that can be provided.Regards,
Owen Doherty-------------------- 0:(shr_sys_abort) ERROR: ice_open_nc: Cannot open /glade/p/cesmdata/cseg/inpu
tdata/share/domains/UNSET
   0:(shr_sys_abort) WARNING: calling shr_mpi_abort() and stoppingThe file: ./Buildconf/cice.input_data_list is empty.
Correspondingly ./Buildconf/ice_in reads:&grid_nml
 grid_file = '/glade/p/cesmdata/cseg/inputdata/share/domains/UNSET'
 grid_format = 'nc'
 grid_type = 'latlon'
 gridcpl_file = 'unknown_gridcpl_file'
 kcatbound =  0
 kmt_file = '/glade/p/cesmdata/cseg/inputdata/share/domains/UNSET' Relevant information from env_build.xml:
    


    


    


    


    


    


    


    

 

[odoherty@ucsd_edu]

Hi All,
In an attempt to circumvent this stumbling block, I set:
grid_file = '/glade/p/cesmdata/cseg/inputdata/ocn/docn7/domain.ocn.1x1.111007.nc'kmt_file = '/glade/p/cesmdata/cseg/inputdata/ocn/docn7/domain.ocn.1x1.111007.nc'
This was done by altering the cice namelist file before building to include:grid_file = '/glade/p/cesmdata/cseg/inputdata/ocn/docn7/domain.ocn.1x1.111007.nc'
kmt_file = '/glade/p/cesmdata/cseg/inputdata/ocn/docn7/domain.ocn.1x1.111007.nc'
This selection was made to use the same file as the prescribed sea ice files:stream_domfilename = '/glade/p/cesmdata/cseg/inputdata/ocn/docn7/domain.ocn.1x1.111007.nc' However, this was a poor choice of dataset to select as the model generates the following errors, suggesting the problem is in cice:   0:CalcWorkPerBlock: Total blocks:   128 Ice blocks:   128 IceFree blocks:     0 Land blocks:     0
   0:(shr_sys_abort) ERROR: latlongrid: ni,nj not equal to nx_global,ny_global
   0:(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping

The question is how do I identify the proper dataset to use for grid_file and kmt_file?

 

[odoherty@ucsd_edu]

Hi All,
In an attempt to circumvent this stumbling block, I set:
grid_file = '/glade/p/cesmdata/cseg/inputdata/ocn/docn7/domain.ocn.1x1.111007.nc'kmt_file = '/glade/p/cesmdata/cseg/inputdata/ocn/docn7/domain.ocn.1x1.111007.nc'
This was done by altering the cice namelist file before building to include:grid_file = '/glade/p/cesmdata/cseg/inputdata/ocn/docn7/domain.ocn.1x1.111007.nc'
kmt_file = '/glade/p/cesmdata/cseg/inputdata/ocn/docn7/domain.ocn.1x1.111007.nc'
This selection was made to use the same file as the prescribed sea ice files:stream_domfilename = '/glade/p/cesmdata/cseg/inputdata/ocn/docn7/domain.ocn.1x1.111007.nc' However, this was a poor choice of dataset to select as the model generates the following errors, suggesting the problem is in cice:   0:CalcWorkPerBlock: Total blocks:   128 Ice blocks:   128 IceFree blocks:     0 Land blocks:     0
   0:(shr_sys_abort) ERROR: latlongrid: ni,nj not equal to nx_global,ny_global
   0:(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping

The question is how do I identify the proper dataset to use for grid_file and kmt_file?

 

[odoherty@ucsd_edu]

Today I revived the documentation for the T42_T42 resolution set, which suggests that each model compenent is functioning on the same grid. ATM_GRID 64x128 ATM_NX 128 ATM_NY 64 T42 is Gaussian grid: LND_GRID 64x128 LND_NX 128 LND_NY 64 T42 is Gaussian grid: ICE_GRID 64x128 ICE_NX 128 ICE_NY 64 T42 is Gaussian grid: OCN_GRID 64x128 OCN_NX 128 OCN_NY 64 T42 is Gaussian grid:

Feeding a grid with the required resolution into the cice grid variables, yields the same error message as previous attempts. (Note: cice_in more or less unchanged from yesterday.)

0:CalcWorkPerBlock: Total blocks: 128 Ice blocks: 128 IceFree blocks: 0 Land blocks: 0
0:(shr_sys_abort) ERROR: latlongrid: ni,nj not equal to nx_global,ny_global
0:(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping

Any feedback regarding appropriate troubleshooting techniques or suggestions as to what datasets to link to would be appreciated. I am curious if other people have had similar problems to this. I would imagine that running basecompset -F with T42_T42 resolution is fairly common and am curious as to why this did not work "out of the box."

Thanks for any comments or help in advance.

Respectfully Submitted,

Owen Doherty

 

[odoherty@ucsd_edu]

Today I revived the documentation for the T42_T42 resolution set, which suggests that each model compenent is functioning on the same grid. ATM_GRID 64x128 ATM_NX 128 ATM_NY 64 T42 is Gaussian grid: LND_GRID 64x128 LND_NX 128 LND_NY 64 T42 is Gaussian grid: ICE_GRID 64x128 ICE_NX 128 ICE_NY 64 T42 is Gaussian grid: OCN_GRID 64x128 OCN_NX 128 OCN_NY 64 T42 is Gaussian grid:

Feeding a grid with the required resolution into the cice grid variables, yields the same error message as previous attempts. (Note: cice_in more or less unchanged from yesterday.)

0:CalcWorkPerBlock: Total blocks: 128 Ice blocks: 128 IceFree blocks: 0 Land blocks: 0
0:(shr_sys_abort) ERROR: latlongrid: ni,nj not equal to nx_global,ny_global
0:(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping

Any feedback regarding appropriate troubleshooting techniques or suggestions as to what datasets to link to would be appreciated. I am curious if other people have had similar problems to this. I would imagine that running basecompset -F with T42_T42 resolution is fairly common and am curious as to why this did not work "out of the box."

Thanks for any comments or help in advance.

Respectfully Submitted,

Owen Doherty

 

[rneale]

Owen, do you have a particular attachement to T42? I would recommend doing the 2 deg FV resolution fv19_fv19 and I would also recommend using an FAMIP compset to see if that works (may also work with T42).

 

[rneale]

Owen, do you have a particular attachement to T42? I would recommend doing the 2 deg FV resolution fv19_fv19 and I would also recommend using an FAMIP compset to see if that works (may also work with T42).

 

[odoherty@ucsd_edu]

Hi Richard, Thanks so much for taking the time to share your insight.  I am not attached to T42 resolution.  At one point I recall reading that iop files for SCAM needed a resolution of T42 with a eularian dynamics core, but perhaps I am mis-remembering?  I will give it a go with the f19_f19 resolution as suggested.  Thanks! Owen

 

[odoherty@ucsd_edu]

Hi Richard, Thanks so much for taking the time to share your insight.  I am not attached to T42 resolution.  At one point I recall reading that iop files for SCAM needed a resolution of T42 with a eularian dynamics core, but perhaps I am mis-remembering?  I will give it a go with the f19_f19 resolution as suggested.  Thanks! Owen

 

[eaton]

It is still true that you must run with the Eulerian dycore when using SCAM.I duplicated the error you found with an F compset using T42_T42 resolution.  However I had no problem running that same configuration from a CAM standalone script.  If you do it that way then you'll see that the datasets used by docn and cice are all on the T42 grid. 

 

[eaton]

It is still true that you must run with the Eulerian dycore when using SCAM.I duplicated the error you found with an F compset using T42_T42 resolution.  However I had no problem running that same configuration from a CAM standalone script.  If you do it that way then you'll see that the datasets used by docn and cice are all on the T42 grid. 

 

[odoherty@ucsd_edu]

Hi Brian,
It was my intention to run CAM standalone from the get-go so please allow me to ask a new user question.  At the CESM Tutorial this summer and in the documentation, the F compset was described as CAM standalone mode, to be invoked using the ./create_newcase script.  However, I think you are suggesting that I call and run the CAM standalone scripts in:/glade/p/cesm/releases/cesm1_2_0/models/atm/cam/bld
The documentation (http://www.cesm.ucar.edu/models/cesm1.2/cam/docs/ug5_3/c136.html) for this approach is missing for both "sample run scripts" and "examples" so I will try and hack through this.  If there are any suggestions on how to approach this or any examples I could modify that would be appreciated greatly.
Thank you for your thoughts and response.
Owen 

 

[odoherty@ucsd_edu]

Hi Brian,
It was my intention to run CAM standalone from the get-go so please allow me to ask a new user question.  At the CESM Tutorial this summer and in the documentation, the F compset was described as CAM standalone mode, to be invoked using the ./create_newcase script.  However, I think you are suggesting that I call and run the CAM standalone scripts in:/glade/p/cesm/releases/cesm1_2_0/models/atm/cam/bld
The documentation (http://www.cesm.ucar.edu/models/cesm1.2/cam/docs/ug5_3/c136.html) for this approach is missing for both "sample run scripts" and "examples" so I will try and hack through this.  If there are any suggestions on how to approach this or any examples I could modify that would be appreciated greatly.
Thank you for your thoughts and response.
Owen 

 

[eaton]

Yes, the scripts in atm/cam/bld are what I was suggesting. They are just templates, but very simple, so you can just look through them and customize to your own needs. Since you're running on yellowstone the run-yellowstone.csh script will be very close to what you need. The template script is set up to run on 1 yellowstone node. To run an F compset configuration at T42 resolution you should give configure the arguments "-dyn eul -hgrid 64x128 -phys cam4". Also, you need to do the build on the login node, so the first invocation of the script should look like% ./run-yellowstone.csh buildOnce you have a successful build then submit the script to the queue:% bsub < run-yellowstone.csh

 

[eaton]

Yes, the scripts in atm/cam/bld are what I was suggesting. They are just templates, but very simple, so you can just look through them and customize to your own needs. Since you're running on yellowstone the run-yellowstone.csh script will be very close to what you need. The template script is set up to run on 1 yellowstone node. To run an F compset configuration at T42 resolution you should give configure the arguments "-dyn eul -hgrid 64x128 -phys cam4". Also, you need to do the build on the login node, so the first invocation of the script should look like% ./run-yellowstone.csh buildOnce you have a successful build then submit the script to the queue:% bsub < run-yellowstone.csh

 

[odoherty@ucsd_edu]

Hi Brian,
I used your approach, which worked like a charm.  I was able to output a number of days, starting in January (I assume ).  Thanks very much.
  I am having trouble producing a series of data for June.  With this scripted based approach the default case (I think) is some sort of climatology.  I have tried specifying -use_case for the default and 1850-2005_cam4 along with start_ymd  (in the namelist) of values lik 20000601, 19990601, 00010601.  I cannot seem to select the appropriate time.  The error is always the same (with sometime different values listed):
CAM build-namelist - ERROR: No default value found for ncdata
user defined attributes:
key=ic_ymd  val=00010601
Can you please point me in the proper direction for specifying a start of the run?  I suspect the solution is trivial, but I can't quite seem to figure out the details.Thank you, Owen

 

[odoherty@ucsd_edu]

Hi Brian,
I used your approach, which worked like a charm.  I was able to output a number of days, starting in January (I assume ).  Thanks very much.
  I am having trouble producing a series of data for June.  With this scripted based approach the default case (I think) is some sort of climatology.  I have tried specifying -use_case for the default and 1850-2005_cam4 along with start_ymd  (in the namelist) of values lik 20000601, 19990601, 00010601.  I cannot seem to select the appropriate time.  The error is always the same (with sometime different values listed):
CAM build-namelist - ERROR: No default value found for ncdata
user defined attributes:
key=ic_ymd  val=00010601
Can you please point me in the proper direction for specifying a start of the run?  I suspect the solution is trivial, but I can't quite seem to figure out the details.Thank you, Owen

 

[eaton]

You'll need to customize your namelist by specifying the initial file that you want to use via the ncdata namelist variable.  There is no default initial file to start a run from 1 June.  You need to generate that yourself.  The most common method is to start a run from 1 Jan and run it for 5 months before writing an initial file.  The inithist namelist variable gives you control over how often initial files are written by the model.

 

[eaton]

You'll need to customize your namelist by specifying the initial file that you want to use via the ncdata namelist variable.  There is no default initial file to start a run from 1 June.  You need to generate that yourself.  The most common method is to start a run from 1 Jan and run it for 5 months before writing an initial file.  The inithist namelist variable gives you control over how often initial files are written by the model.

 

[odoherty@ucsd_edu]

I realized I never responded to this.  Just wanted to say thank you, your proposed solution worked!