Questions about hybrid and RUN_REFDATE

[yugong]

Hello everyone, I am new to CESM. I have learned a lot from this forum, but I still have some questions about "hybrid" and "RUN_REFDATE".

When i create a new case with ./create_newcase --case B1850 --compset B1850 --res f09_g17, and then ./xmlquery RUN_REFTYPE
I expected it to return "startup", since it's a new case. But it returns "hybrid", which really confuesd me.

After further check, i found this new case has a RUN_REFCASE: b.e21.B1850.f09_g17.CMIP6-piControl.001_v3hist

1. So my first question is: Does B1850 is actually created based on another case? Under what conditions can one create a "startup" case?

2. my second question is that i found REFDATE=0134-01-01, what does it mean? Does it indicate that the case has run for 134 years, and B1850 was created using the output from the 134th year? Additionally, RUN_STARTDATE=0001-01-01, what does it mean? why it is not 1850-01-01? What would happen if I executed ./xmlchange RUN_STARTDATE=1850-01-01

3. I know CESM2 Component Sets Definition gives a very brief description about compsets. Is there some ways i can know more details of each compset?

Could anyone provide some answers? my version is release-cesm2.1.5-0-g7a6c5b0 Thank you very much!

 

[slevis]

1) Your B1850 case is not "created based on another case" but uses initial conditions from that other case. I think that the model would NOT read initial condition files for all the components if you changed to "startup". So changing to "startup" may initialize some components without spun-up initial files.

2) REFCASE and REFDATE tell the model what initial condition files to look for. RUN_STARTDATE 0001 or 1850 means that the year labels of your output files will start with 0001 or 1850. We prefer to use 1850 (or any other "real" year) when doing transient simulations and 0001 when doing time-slice simulations.

3) I don't know... Maybe someone else can chime in here.

 

[yugong]

1) Your B1850 case is not "created based on another case" but uses initial conditions from that other case. I think that the model would NOT read initial condition files for all the components if you changed to "startup". So changing to "startup" may initialize some components without spun-up initial files.

2) REFCASE and REFDATE tell the model what initial condition files to look for. RUN_STARTDATE 0001 or 1850 means that the year labels of your output files will start with 0001 or 1850. We prefer to use 1850 (or any other "real" year) when doing transient simulations and 0001 when doing time-slice simulations.

3) I don't know... Maybe someone else can chime in here.

Thank you for your reply! I still have some questions that I would like to clarify：

1) Do you mean "startup" should be used when i want to run an "ideal" model. For example, if i download dataset such as ERA5 as inital files, and i increase SST over tropical Pacific to see what will happen in the next 10 years. Under this conditon, i should use "startup"?

2) i create a FWHIST case, the REFCASE is "f.e21.FWHIST.f09_f09_mg17.cesm2.1-exp011.001_v2", both REFDATE and STARTDATE are 1979-01-01。After i builded the case, it automately downloaded the initial files. However, I found there is only a "1979-01-01" folder available to download, see https://svn-ccsm-inputdata.cgd.ucar...21.FWHIST.f09_f09_mg17.cesm2.1-exp011.001_v2/ Does this mean i can only achieve initial files on 1979-01-01？What RUN_STARTDATE is used for? If I change RUN_STARTDATE to 1980-01-01, the model still uses initial files on 1979-01-01? But i leaned from tutorial that in a hybrid run, start date can be modified relative to the reference case, see Hybrid, Branch and Startup — CESM Tutorial This really confuses me.

I would greatly appreciate your response!

 

[zhangmeixin]

Thank you for your reply! I still have some questions that I would like to clarify：

1) Do you mean "startup" should be used when i want to run an "ideal" model. For example, if i download dataset such as ERA5 as inital files, and i increase SST over tropical Pacific to see what will happen in the next 10 years. Under this conditon, i should use "startup"?

2) i create a FWHIST case, the REFCASE is "f.e21.FWHIST.f09_f09_mg17.cesm2.1-exp011.001_v2", both REFDATE and STARTDATE are 1979-01-01。After i builded the case, it automately downloaded the initial files. However, I found there is only a "1979-01-01" folder available to download, see https://svn-ccsm-inputdata.cgd.ucar...21.FWHIST.f09_f09_mg17.cesm2.1-exp011.001_v2/ Does this mean i can only achieve initial files on 1979-01-01？What RUN_STARTDATE is used for? If I change RUN_STARTDATE to 1980-01-01, the model still uses initial files on 1979-01-01? But i leaned from tutorial that in a hybrid run, start date can be modified relative to the reference case, see Hybrid, Branch and Startup — CESM Tutorial This really confuses me.

I would greatly appreciate your response!

In my understanding, but I'm not sure if it's correct:
I think if you use the initial time of 1979, then you should better change the startdate to 1979, if you change it to 1980 may be the running is not transient.
And "i can only achieve initial files on 1979-01-01?" I think it is true, if you want to get another initial time, you can run the model from 1979 to output the initial data.
About question 1, I dont konw too.

 

[zhangmeixin]

1) Your B1850 case is not "created based on another case" but uses initial conditions from that other case. I think that the model would NOT read initial condition files for all the components if you changed to "startup". So changing to "startup" may initialize some components without spun-up initial files.

2) REFCASE and REFDATE tell the model what initial condition files to look for. RUN_STARTDATE 0001 or 1850 means that the year labels of your output files will start with 0001 or 1850. We prefer to use 1850 (or any other "real" year) when doing transient simulations and 0001 when doing time-slice simulations.

3) I don't know... Maybe someone else can chime in here.

HI! But I see "all model components are initialized from the basic default initial conditions." from tutorial about "startup", I am a little confused.

 

[slevis]

Assuming I understand your confusion, I will try to clarify:
Think of "basic default" initial conditions in startup cases as in contrast to "spun up" initial conditions in hybrid cases.

 

[zhangmeixin]

Assuming I understand your confusion, I will try to clarify:
Think of "basic default" initial conditions in startup cases as in contrast to "spun up" initial conditions in hybrid cases.

I will check it and thanks for your reply!

 

[yugong]

HI! But I see "all model components are initialized from the basic default initial conditions." from tutorial about "startup", I am a little confused.

Thanks! I have understood the question about "STARTDATE". Could you please explain more specifically about what is "basic default initial conditions" and "spun up conditions"? Does "spun up conditons" mean that the system achieve radiative balance, which is downloaded from REFCASE? But what is "basic default initial condition"?

 

[slevis]

You should not assume that the "spun up" initial conditions that we provide are in "radiative balance" unless there's explicit documentation stating this. If you need such initial conditions, you are best off creating them yourself for the purposes of your research.

In contrast to "spun up" initial conditions, the "basic default initial conditions" are simply sufficient to get a simulation started but may, for example, start the whole planet with the same temperature.

 

[yugong]

You should not assume that the "spun up" initial conditions that we provide are in "radiative balance" unless there's explicit documentation stating this. If you need such initial conditions, you are best off creating them yourself for the purposes of your research.

In contrast to "spun up" initial conditions, the "basic default initial conditions" are simply sufficient to get a simulation started but may, for example, start the whole planet with the same temperature.

Thank you! My last question is about "REF_STARTDATE".

In a new created BHIST case, the default "RUN_SRARTDATE" is 1979-01-01. I want to conduct a simulation from 2013-01-01 to 2013-01-07. So I replace the atmosphere initial conditon with ERA5 of 2013-01-01. Under this condition, should I set RUN_STARTDATE=2013-01-01? The simulated result of 2013-01-07 would be identical to that of 1979-01-07 if i do not change RUN_STARTDATE?

If the result of 2013-01-07 and 1979-01-07 are identical, this means RUN_STARTDATE is just a time label?
The length of the initial file on time dimension is 1 (only has date on 1979-01-01), what do we do whith "RUN_STARTDATE"?

Any comment would be appreciated!!!

 

[slevis]

RUN_STARTDATE is just a label if you do time-slice simulations. So if you run for 1850 or for 2000, it doesn't matter if you label your history files 0001 or something else.

I think that RUN_STARTDATE is more than a label if you run transient simulations because in that case you may get different values of CO2 and other inputs depending on your choice.

The best way to find out for sure is to try things and see what happens.

 

[yugong]

RUN_STARTDATE is just a label if you do time-slice simulations. So if you run for 1850 or for 2000, it doesn't matter if you label your history files 0001 or something else.

I think that RUN_STARTDATE is more than a label if you run transient simulations because in that case you may get different values of CO2 and other inputs depending on your choice.

The best way to find out for sure is to try things and see what happens.

Thank you for your patience!