differences between initial files and restart files

[yaozhixiong@foxmail_com]

Hi, I am running CESM1.2.0. It will generate initial files and restart files for CAM component. These two files include different variables. I would like to set up a specific initial condition for cam component, so I let 'ncdata' equal to an initial file in user_nl_cam:ncdata = *.cam.i.*It will not work if ncdata equals to restart file.My question is, what is the difference between initial files and restart files? CAM component have initial files and restart files, but other components only have restart files. Why? Thanks very much.Yao

 

[eaton]

Roughly the initial file contains a minimal set of fields to start a simulation while the restart file contains the complete state of the model needed to continue a run and produce the exact simulation which would have been produced if the model had not been stopped.  It is unfortunate that a restart file can't be used for an initial run (at least this is true for the FV dycore, not sure about other dycores).  The reason is that initial and restart runs use different code paths for reading the input file and there are some differences in the variable names expected in the two files.  I suspect this wouldn't be hard to work around by using something like the NCO operators to make some variable name changes in the restart file.

 

[thibaut_lurton@cnrs-orleans_fr]

Hello.Another question related to this restart/IC files difference:I know how to perform restarts based on the *.r* files. However, I noticed that on 1st January, some *.i* files (IC) were generated.If I want to perform a restart from this 1st January date onwards, what are the steps to follow? Need I specify the *.cam.i.* as a new ncdata? Need I also r-point onto the *.r* files?Thanks.Thibaut.

 

[eaton]

You can't do a restart run using the initial files.  You can only use them for an initial run.  The ncdata variable is the correct variable for specifying the new initial file. 

 

[thibaut_lurton@cnrs-orleans_fr]

Hello,So, whenever I conclude a whole-year run, what is best for the follow-up? Continue onto the next year as a standard branch run with the *.r* restart files, or make a new initial run with the *.i* files?Thanks very much,Thibaut.

 

[eaton]

There is more than one way to continue a run so that you can do what best suites your needs.  If you want to make some change to how the simulation is moving forward then continuing the run using the initial file from the previous run allows the most flexibility in making changes.  I you only want to change some characteristic of the history output (more frequent output for example) then a branch run will allow that without changing the climate realization.  If you just want to continue the previous climate realization and have the results be identical to a run that was never stopped then the restart mechanism allows that.