CTSM - CLM5 - CESM2.1.3 - results of a quick run

[wvsi3w]

Hello,
I have used CLM5 for my project. The configuration files are attached.

I did the newcase with the following command:
./create_newcase --case /home/meisam/scratch/cases/may1h --compset I1850Clm50Bgc --res f19_g16 --machine narval --walltime 02:00:00 --run-unsupported
"narval" is the machine that we configured it and it can be seen in config_machine.xml file that is attached here.

the input data for this compset and this resolution is about 300 GB!

After doing the ./case.setup ./case.build and ./case.submit the model finishes simulating it around 40 mins (wallclock time) for STOP_N=3 years.

The problem is when I look into the results (run files, .nc files), it is only showing one month (last month) of simulation results, which in this case is from Dec of 3rd year to Jan of 4th year.

I used ncview to take a look at these results and it doesn't have all the results for 3 years of simulation. I also downloaded these .nc files (clm2.nc, cism.nc, mosart.nc, ...) and they don't have the 3 years, only 1 last month is shown here.

Do you know why?

--------------------------------------------------
my env_run.xml has this info: ("STOP_OPTION" value="nyears") & ("STOP_N" value="3") & ("STOP_DATE" value="-999"). So basically I put 3 years for the run and the STOP_DATE is negative, which means this id is off by the model definition. So do you think I should change that negative value? on the other hand, it says "STOP_DATE"=Alternative date yyyymmdd date option, sets the run length with STOP_OPTION and STOP_N, a negative value implies off". Does this mean I don't need to change the STOP_DATE value, and it would start from year 1 and end the simulation in year 3 ???
-------------------------------------------------
why when I check my env_run.xml file I see this (entry id="RUN_STARTDATE" value="0001-01-01") when basically the compset is saying that it will start from pre-industrial (1850)? should I change the RUN_STARTDATE to "1850-01-01"?
-------------------------------------------------
the compset I used doesn't have any scientifically supported grids and I used (res=f19-g16). Does it matter what res to choose for this compset; as it is not mentioned in the documentation file ??
-------------------------------------------------

Moreover, I saw that in my config_workflow.xml there is nothing for my machine "narval". Is it important to have it in that file? because there are not many machines included in that file.

 

[slevis]

Hello,
I have used CLM5 for my project. The configuration files are attached.

I did the newcase with the following command:
./create_newcase --case /home/meisam/scratch/cases/may1h --compset I1850Clm50Bgc --res f19_g16 --machine narval --walltime 02:00:00 --run-unsupported
"narval" is the machine that we configured it and it can be seen in config_machine.xml file that is attached here.

the input data for this compset and this resolution is about 300 GB!

After doing the ./case.setup ./case.build and ./case.submit the model finishes simulating it around 40 mins (wallclock time) for STOP_N=3 years.

The problem is when I look into the results (run files, .nc files), it is only showing one month (last month) of simulation results, which in this case is from Dec of 3rd year to Jan of 4th year.

I used ncview to take a look at these results and it doesn't have all the results for 3 years of simulation. I also downloaded these .nc files (clm2.nc, cism.nc, mosart.nc, ...) and they don't have the 3 years, only 1 last month is shown here.

Do you know why?

--------------------------------------------------
my env_run.xml has this info: ("STOP_OPTION" value="nyears") & ("STOP_N" value="3") & ("STOP_DATE" value="-999"). So basically I put 3 years for the run and the STOP_DATE is negative, which means this id is off by the model definition. So do you think I should change that negative value? on the other hand, it says "STOP_DATE"=Alternative date yyyymmdd date option, sets the run length with STOP_OPTION and STOP_N, a negative value implies off". Does this mean I don't need to change the STOP_DATE value, and it would start from year 1 and end the simulation in year 3 ???
-------------------------------------------------
why when I check my env_run.xml file I see this (entry id="RUN_STARTDATE" value="0001-01-01") when basically the compset is saying that it will start from pre-industrial (1850)? should I change the RUN_STARTDATE to "1850-01-01"?
-------------------------------------------------
the compset I used doesn't have any scientifically supported grids and I used (res=f19-g16). Does it matter what res to choose for this compset; as it is not mentioned in the documentation file ??
-------------------------------------------------

Moreover, I saw that in my config_workflow.xml there is nothing for my machine "narval". Is it important to have it in that file? because there are not many machines included in that file.

I will start by answering your specific questions:
1) I agree with you that you should not have to change STOP_DATE.
2) I don't think it matters whether you change RUN_STARTDATE. We default it to 0001-01-01 because this is a "time-slice" simulation, i.e. running for 1850 only and not for a range of years, even if the simulation itself is multi-year. This way users don't misunderstand the output from the time-slice simulation as representing specific years in history.
3) It should not matter. The choice is yours.
4) This is also your choice. Feel free to add your machine to that file.

Now back to the question about your output:
1) You can control the output with namelist variables. See the user's guide for help.
2) There's a chance that the output that you're looking for exists and that you're looking in the wrong directory. When simulations complete, the output moves to an /archive directory located two levels up from your simulation's /run directory.

 

[wvsi3w]

I will start by answering your specific questions:
1) I agree with you that you should not have to change STOP_DATE.
2) I don't think it matters whether you change RUN_STARTDATE. We default it to 0001-01-01 because this is a "time-slice" simulation, i.e. running for 1850 only and not for a range of years, even if the simulation itself is multi-year. This way users don't misunderstand the output from the time-slice simulation as representing specific years in history.
3) It should not matter. The choice is yours.
4) This is also your choice. Feel free to add your machine to that file.

Now back to the question about your output:
1) You can control the output with namelist variables. See the user's guide for help.
2) There's a chance that the output that you're looking for exists and that you're looking in the wrong directory. When simulations complete, the output moves to an /archive directory located two levels up from your simulation's /run directory.

Dear slevis,
Thanks for your great response.

I checked the lnd/hist directory before there was nothing there but when I changed the <entry id="DOUT_S_SAVE_INTERIM_RESTART_FILES" value="FALSE"> to "TRUE" , after that, I could find the archive files of all 3 years (monthly).

May I ask what should we do if changing that RUN_STARTDATE is not useful? I mean if I want to run it for a range of years that starts from say 1850 (or any other range of years)?

I have looked into the Tutorials that are available for CLM5 and could not find my answer. They are all doing a quick run or a test (not real projects).

 

[oleson]

If you want to run for "real" years then you would choose a different compset, a historical compset, e.g., IHistClm50BgcCrop. That compset is setup to start in 1850 and can be run through 2014.