Regional run failed: ERROR in UrbanParamsType.F90

[adrienD]

Hi everyone, I am trying to run a regional case (I've created my own SCRIP and mapping files) with CTSM on the new supercomputer mistral (I've been able to make a default run without problems). I know it's a very specific case, but I hope someone can help me.

I've successfully created the surfdata and domain files, and the model building is working fine. However, I have an error when I try to run the model. You can find the main problem from the lnd.log file below and the different log files are attached.

Reading in urban input data from fsurdat file ...
(GETFIL): attempting to find local file
surfdata_218x200_hist_78pfts_CMIP6_simyr2000_c210309.nc
(GETFIL): using
/work/aa0049/a271098/CTSM/inputdata/lnd/clm2/surfdata_map/surfdata_218x200_hist
_78pfts_CMIP6_simyr2000_c210309.nc
Opened existing file
/work/aa0049/a271098/CTSM/inputdata/lnd/clm2/surfdata_map/surfdata_218x200_hist
_78pfts_CMIP6_simyr2000_c210309.nc 458752
UrbanInput
/work/aa0049/a271098/CTSM/inputdata/lnd/clm2/surfdata_map/surfdata_218x200_hist
_78pfts_CMIP6_simyr2000_c210309.nc
UrbanInput: parameter nlevurb= 5
does not equal input dataset nlevurb= 10
ENDRUN:
ERROR: ERROR in UrbanParamsType.F90 at line 477

Thanks in advance,
Adrien

 

[jedwards]

The data set in
surfdata_218x200_hist_78pfts_CMIP6_simyr2000_c210309.nc
has 10 urban types but somewhere you are specifying only 5. These need to match.

 

[erik]

Yes, jedwards is correct. I think the problem you are running into is this one...

New urban raw dataset currently inconsistent with CTSM code and urban temperature streams file · Issue #1252 · ESCOMP/CTSM

Brief summary of bug The CTSM currently crashes when using a newly generated surface dataset. Mksurfdat points to a new urban raw dataset that puts nlevurb = 10 in surface datasets, while the CTSM ...

github.com

You can update to a newer version of CTSM where we fixed that issue (ctsm5.1.dev023).

 

[adrienD]

Yes, jedwards is correct. I think the problem you are running into is this one...

New urban raw dataset currently inconsistent with CTSM code and urban temperature streams file · Issue #1252 · ESCOMP/CTSM

Brief summary of bug The CTSM currently crashes when using a newly generated surface dataset. Mksurfdat points to a new urban raw dataset that puts nlevurb = 10 in surface datasets, while the CTSM ...

github.com

You can update to a newer version of CTSM where we fixed that issue (ctsm5.1.dev023).

Thanks for your help, once again! I have updated to the ctsm5.1dev029 version and the regional run is working.

However, I have huge issues with the real-time length of simulations. With ctsm5.1.dev023, I was able to run default 5-years simulations (--res f19_g17 --compset I1850Clm50Sp) in less than 3 hours and now I am not able to finish before 8 hours. With the exact same set-up. I try to increase the numbers of cpu/nodes without success.

After 8 hours, the slurm job stops due to time limit, even if I set walltimemax="24:00:00". But I suppose it's due to our supercomputer slurm installation.

Any help on that?

 

[oleson]

I tried a short simulation with that tag and compset/res and the performance was about what I would expect.
I'm not sure what to suggest, maybe one of our software engineers has a suggestion....

 

[adrienD]

Sorry I meant "With ctsm5.1.dev013, I was able to run default 5-years simulations (--res f19_g17 --compset I1850Clm50Sp) in less than 3 hours..."

 

[jinping F]

Yes, jedwards is correct. I think the problem you are running into is this one...

New urban raw dataset currently inconsistent with CTSM code and urban temperature streams file · Issue #1252 · ESCOMP/CTSM

Brief summary of bug The CTSM currently crashes when using a newly generated surface dataset. Mksurfdat points to a new urban raw dataset that puts nlevurb = 10 in surface datasets, while the CTSM ...

github.com

You can update to a newer version of CTSM where we fixed that issue (ctsm5.1.dev023).

Hello everyone,

I have encountered the same issue. Could someone please tell me where to manually modify the mksurfdat to point to the old raw dataset? I have searched for a long time but have not found where to make the change. I went into /home/fengjinping/project/cesm2.0/components/clm/tools/mksurfdata_map/src/mkvarctl.F90 and manually changed character(len=256), public :: mksrf_furban = ' ' ! urban data file name to character(len=256), public :: mksrf_furban = '/glade/p/cesm/cseg/inputdata/lnd/clm2/rawdata/mksrf_urban_0.05x0.05_simyr2000.c120621.nc' ! urban data file name. When I use the command:
./mksurfdata.pl -r usrspec -usr_gname $GRIDNAME -usr_gdate $CDATE -l $CSMDATA -debug
the mksrf_furban is still set to '/home/fengjinping/scratch/cesmlab/lnd/clm2/rawdata/mksrf_urban_0.05x0.05_simyr2000.c170724.nc'.

Any reply would be valuable. Thank you in advance!

 

[slevis]

I haven't used mksurfdata_map in years and may not remember correctly. I seem to remember at some stage in the process you end up with a namelist. If you then modify the contents of the namelist, you should accomplish what you're attempting.

 

[jinping F]

Thank you for your reply, but the attempt I made still did not succeed. Do you or anyone else have any ideas?
Here is what I tried:
cd /home/fengjinping/project/clm5/tools/mksurfdata_map
export CSMDATA=/home/fengjinping/scratch/cesmlab
export GRIDNAME=sihuliuyu0.1dgress
export CDATE='240530'
export REGRID_PROC=30
./mksurfdata.pl -r usrspec -usr_gname $GRIDNAME -usr_gdate $CDATE -l $CSMDATA -debug

At this point, I modified the generated files surfdata_sihuliuyu0.1dgress_hist_78pfts_CMIP6_simyr1850_c240827.namelist and surfdata_sihuliuyu0.1dgress_hist_78pfts_CMIP6_simyr2000_c240827.namelist by setting mksrf_furban="/home/fengjinping/scratch/cesmlab/lnd/clm2/rawdata/mksrf_urban_0.05x0.05_simyr2000.c120621.nc".

However, when I ran the command:./mksurfdata.pl -r usrspec -usr_gname $GRIDNAME -usr_gdate $CDATE -l $CSMDATA
the mksrf_furban was still set to "/home/fengjinping/scratch/cesmlab/lnd/clm2/rawdata/mksrf_urban_0.05x0.05_simyr2000.c170724.nc".
I've tried many different approaches, but none have worked. Could you please help me further?

 

[oleson]

I think you'd have to run the executable directly with the namelist as an input, e.g., something like:

./mksurfdata_map < surfdata_sihuliuyu0.1dgress_hist_78pfts_CMIP6_simyr1850_c240827.namelist

I can't verify this as the mksurfdata_map tool has not been ported to our hpc.

 

[jinping F]

Thank you @oleson for the precise instructions, and thank you @slevis for the guidance!

That's great! This command works! I've learned a lot from this process!

However, there's still one thing I don't understand. The mksrf_urban_0.05x0.05_simyr2000.c120621.nc file is labeled as simyr2000. Why can it be used for creating surface data files for the year 1850?

 

[oleson]

Urban data is only available for year 2000.