How can I output more fields from a run which has already been finished

[chen63@illinois_edu]

I am just wondering whether I could output more fields from a run which has already been finished? I would like to investigate some fields which are listed in the Master Field List but not output in my original history files. Is there any way I can output more fields without running the case again?Thanks!

 

[strandwg]

The only way to generate more data from a finished run without rerunning is to use the data you saved and derive, if possible, additional data.Otherwise, you must rerun your simulation to save additional output. 

 

[hannay]

If don't need the extra output for the length of teh run, you can do a branch run starting from the end of your completed run.Say you have a run from year 1 to year 200. Then you realized you want to output extra fields. You can start a branch from year 200 and output extra fields using FINCL namelists (see namelist documentation for FINCL)

 

[chen63@illinois_edu]

Hi, Hannay, Thanks for your reply. If I want extra output during year 1 to 200, can I do a branch run starting from a time point between year 1 to 200?  I am running CESM1.0 with CAM5.1. I have several basic question regarding the branch run, which maybe silly... Does the new run take anything from the old run other than the restart file? By this I mean, how can the new run maintain the bit-for-bit reproducibility of the old run? As instructed by http://www.cesm.ucar.edu/models/ccsm4.0/ccsm_doc/x2933.html, I have removed all the files related to the old run or change the name of the directories of the old run. Under this situation, how can the model get things from the old run just by the namelist variable REFCASE? How about the initial file? I have a specified ncdata file in the cam namelist and I keep this setting in the namelist for the new branch run. Will the model still use this initial file to start or will it use the restart file obtained from the old run as specified by REFCASE and REFDATE? In my case, I am running with offline meteorology. When configure and build the new branch run, shall I keep the namelist variable met_data_file the same as the old run because this indicates when the run starts and can be difference from REFDATE. Will the model automatically use the meterological data starting from the REFDATE?  Thanks a lot!

 

[hannay]

When you do a branch run you have to copy all the restart files from the previous into your run directory. This is where the model will find teh information about how to start.
Let's say you have done a first run called "run1" and you want to branch from there to create a run called "run2". Say you want to branch from year 10 of run1.1. you create a run called run2 (create_newcase -case run2 ...)2. you go in the case directory and set run2 as a branch of run1../xmlchange RUN_TYPE=branch,RUN_REFCASE=run1,RUN_REFDATE=0010-01-01,GET_REFCASE=FALSE3. you set up./cesm_setup4. you build./run1.build5. Then you go into the run directory of run2 and you copy all the restart files from run1 for year 10 into the run directory of run2.On yellwostone:cp /glade/scratch/$LOGNAME/archive/run1/rest/0010-01-01/* /glade/scratch/$LOGNAME/run2/run/THis is from these files than CESM will get the information how to do a restart.6. You submit your job.

 

[chen63@illinois_edu]

Thanks, and this is the same step I followed except I used the same case name for run1 and run2. This is also permitted and does not affect the simulation, right?If the model get the information on how to do a restart through the previous restart files from run1, why can't I simply create run2, set RUN_TYPE = startup, but CONTINUATION_RUN = TRUE, and copy the restart files from run1 into the run directory of run2? Regarding the initial file and start date of the offline meteorological file, my understanding is the original namelist setttings do not matter, as the model will restart from the restart files from run1 and use REF_DATE as the start date. Am I correct?Thanks!

 

[chen63@illinois_edu]

I have tried to use the same namelist variables for the offline meteorological data, but it seems like we need to change the name of the first met data file to when run2 starts.

 

[mai]

Did you set BRNCH_RETAIN_CASENAME to TRUE in the env_run.xml file?

 

[chen63@illinois_edu]

Yes, I did, and it is actually in the env_conf.xml file in my version of CESM

 

[mai]

Which version of CESM are you using?

 

[hannay]

"If the model get the information on how to do a restart through the previous restart files from run1, why can't I simply create run2, set RUN_TYPE = startup, but CONTINUATION_RUN = TRUE, and copy the restart files from run1 into the run directory of run2? "If you set the run as startup it will NOT ready any restart files. This is the way the model works.If you want to read restart files you need to set to branch or hybrid.

 

[chen63@illinois_edu]

I am running CESM1.0 with CAM5.1. 

 

[mai]

With CESM 1.0 you need to do something like the following:  % xmlchange -file env_conf.xml -id RUN_TYPE -val 'branch'
  % xmlchange -file env_conf.xml -id RUN_REFCASE -val 'your case name here'
  % xmlchange -file env_conf.xml -id RUN_REFDATE -val 'date for restart set e.g. 1948-01-01'
  % xmlchange -file env_conf.xml -id BRNCH_RETAIN_CASENAME -val 'TRUE'
  % xmlchange -file env_conf.xml -id GET_REFCASE -val 'FALSE'

% ./configure -clean% ./configure -case% $CASE.clean_build% $CASE.build Of course you also need to add the new variables to the namelist in Buildconf/cam.buildnml.csh or whatever component you want to output more fields in. Check that CONTINUE_RUN is FALSE in env_run.xml and make sure the other controlling items in env_run.xml are what you want. Check the $CASE.run script, since it has also been re-made.

 

[chen63@illinois_edu]

Thank you very much, Mai and Hannay.  The new branch run case runs successfully and I appreciate your help!

 

[chen63@illinois_edu]

Could it be possible that the branch run generate history files with different size from the old run, when I did not make any changes to the namelist variables about the output fields such as nhtfrq, mfilt and fincl in the branch run? Thanks!