how to set the Initialization Time in CESM

[CGL]

Hi,all!

I look through the page and I want to know what's Initialization Time? For FHIST_BGC, the Initialization Time value is HIST. Does means the timeseries of inputdata is 1850-2000?

If I want to set the Initialization Time by my own, the parameter is "RUN_STARTDATE"? Where can i find the parctice example or guidance？I barely know the protocol for setting this.

---

The initialzation time of F2000climo is 2000. The value is fixed, it means I can't change the initialization time? like I change this value to 1850, What will happen？

If I set the initialzation time to 2000 in FHIST. Now,What's difference between F2000climo and FHIST?

 

[sacks]

Good question. I skimmed through our documentation and agree that we should document this better, and I just created an issue for this (https://github.com/ESCOMP/CESM/issues/226).

You're right that the compset's forcing time period will set RUN_STARTDATE, and then you can modify RUN_STARTDATE from there. But typically, you should start by choosing an appropriate compset, since there are other settings that depend on this compset time period. For example, even though RUN_STARTDATE is the same for an 1850 or HIST compset, these are set up in very different ways: a HIST compset uses transient datasets for various forcings such as landcover, etc., so HIST is appropriate when you want to do a transient run from 1850 through the 20th century, whereas 1850 is appropriate for a fixed-time-period run, e.g., for spinning up the model. Compsets with 2000 or 2010 also use (relatively) fixed forcings that are appropriate for modeling near-present-day climate. I would not recommend your suggestion of starting with F2000climo and then changing RUN_STARTDATE to 1850, because this would give you an inconsistent state, since the defaults for various forcing files depend on the time period from the compset (e.g., landcover, etc.).

HIST compsets currently have forcings that run from 1850-2015.

To see more details, I'd suggest setting up cases with a few of the compsets you have mentioned here, running case.setup then preview_namelists, and then comparing the namelists generated in CaseDocs – then you'll see what kinds of things differ for these different compsets.

 

[CGL]

Good question. I skimmed through our documentation and agree that we should document this better, and I just created an issue for this (Improve documentation of compset forcing time period · Issue #226 · ESCOMP/CESM).

You're right that the compset's forcing time period will set RUN_STARTDATE, and then you can modify RUN_STARTDATE from there. But typically, you should start by choosing an appropriate compset, since there are other settings that depend on this compset time period. For example, even though RUN_STARTDATE is the same for an 1850 or HIST compset, these are set up in very different ways: a HIST compset uses transient datasets for various forcings such as landcover, etc., so HIST is appropriate when you want to do a transient run from 1850 through the 20th century, whereas 1850 is appropriate for a fixed-time-period run, e.g., for spinning up the model. Compsets with 2000 or 2010 also use (relatively) fixed forcings that are appropriate for modeling near-present-day climate. I would not recommend your suggestion of starting with F2000climo and then changing RUN_STARTDATE to 1850, because this would give you an inconsistent state, since the defaults for various forcing files depend on the time period from the compset (e.g., landcover, etc.).

HIST compsets currently have forcings that run from 1850-2015.

To see more details, I'd suggest setting up cases with a few of the compsets you have mentioned here, running case.setup then preview_namelists, and then comparing the namelists generated in CaseDocs – then you'll see what kinds of things differ for these different compsets.

Thanks,@sacks. The difference between FHIST and F200climo is the fixed initial time. If I change the initial time of FHIST(like I set the initial time as 2000), The model will use forcing data of 2000 to run.Then, The F2000climo will consisit with FHIST.
If I set the initial time as 1850 for F2000climo, the F2000climo will unstable and can't understand. Do I get your point right?

 

[slevis]

@CGL I don't think that the F2000climo will have trouble understanding or be unstable; however, it will generate output labeled with years 1850, 1851, etc. while using input data (boundary conditions, initial conditions, ...) corresponding to the modern day. I prefer to set the start date to 1 for time-slice simulations and to a real year for transient simulations.

 

[sacks]

Sorry for the slow reply. To add to what @slevis said: there are more differences between FHIST and F2000climo than just the initial time. Again, I would recommend setting up both cases and comparing the namelists.