Scheduled Downtime
On Tuesday 24 October 2023 @ 5pm MT the forums will be in read only mode in preparation for the downtime. On Wednesday 25 October 2023 @ 5am MT, this website will be down for maintenance and expected to return online later in the morning.
Normal Operations
The forums are back online with normal operations. If you notice any issues or errors related to the forums, please reach out to help@ucar.edu

Suggestions for modifying the start date (CESM2)

acgvar

Alvin C.G. Varquez
Member
Hi everyone! Learning everything based on the tutorial. Really helpful to port to a new machine in Tokyo. Now reached to the point of being able to run the simulation.

This time, I modified the years to 5 following the tutorial by Dr. Shields with the case setup:
./create_newcase --case ~/cases/b.day2.0 --res f19_g17 --compset B1850 --machine tsubame
xmlchange STOP_OPTION=nyears,STOP_N=5,JOB_WALLCLOCK_TIME=12:00:00
Also modified the frequency output for ATM (CAM) to every 2 weeks..

Now it's running for more than 20 minutes. I noticed that the top contents of atm.log shows:
********** Time Manager Configuration **********
Calendar type: NO_LEAP
Timestep size (seconds): 1800
Start date (yr mon day tod): 1 1 1
0
Stop date (yr mon day tod): 6 1 1
0
Reference date (yr mon day tod): 1 1 1
0
Current step number: 0
Current date (yr mon day tod): 1 1 1
0
************************************************

If my understanding is correct setting --compset B1850 means having an initial condition of the year 1850. What does the start date mean from the log above? Does this mean its running from January 1, 1850 or January 1, 0001? Is it necessary to modify the start year prior to ./case.build for the model to actually start at the year 1850?

I'm sorry for this confusion in my part.
Also, it's difficult to interpret the timing folder to judge how fast the model is actually running.

The model is really fascinating and its exciting to learn about it at a short time.
 

hannay

Cecile Hannay
AMWG Liaison
Staff member
Replies to your questions are below:

What does the start date mean from the log above? => In B1850 compset, you are cycling over the year 1850. It means you are repeating the year 1850 over and over. The start date is when you start the model. The is the year one of your run (0001-01-01)

Does this mean its running from January 1, 1850 or January 1, 0001? => you are starting with initial condition from January 1, 1850 but with the compset B1850, you are repeating the year 1850 over and over. This is what we call a pre-industrial control run.

Is it necessary to modify the start year prior to ./case.build for the model to actually start at the year 1850? => YOu would do this if you do a BHIST (historical run that runs from 1850 to 2014) but not for a B1850.
 

acgvar

Alvin C.G. Varquez
Member
Thanks Hannay. I see!

So this means, regardless of modifying the start time in a B1850, it will always start with an initial condition utilizing January 1, 1850?
So January 1, 0001 does not mean anything in a B1850 case. If I change the start time in B1850 to 2000, January 1, 2000 will have the same result as January 1, 0001?

If I'm actually interested in having the model start at the initial condition of year 2000, I should modify the compset to BHIST.

Is my understanding of the above correct?
Also, I can't seem to find a detailed documentation of the above.
If you have any links, I would like to check it out.

Thank you very much!
 

hannay

Cecile Hannay
AMWG Liaison
Staff member
Your interpretation is correct.

If you want to modify the initial condition, you can modify the variables:
RUN_REFCASE
RUN_REFDATE
RUN_TYPE

Let say you want to start a run from year 2000. You have somewhere initial conditions coming from another run.

Let's say I have a previous run named: b.e21.BHIST.f09_g17.CMIP6-historical.003 than have restart for the year 2000.

To use the initial condition condition from the year 2000, you would set:
RUN_REFCASE = b.e21.BHIST.f09_g17.CMIP6-historical.003
RUN_REFDATE = 2000-01-01
RUN_TYPE = hybrid

Now, to answer the question whether you should use a B1850 or a BHIST, depending which forcing you want.
The B1850 will give you forcing from the 1850
the BHIST will give you forcing from the historical period 1850-2014.

In your case, you may want to cycle over the year 2000 (?)

There is some documentation at:
 

acgvar

Alvin C.G. Varquez
Member
Hi Hannay,

Thank you so much for your clear explanations. I am slowly understanding this but I'm sorry for these followup questions.

1. I hope you don't mind if I followup on your statement, "the BHIST will give you forcing from the historical period 1850-2014.".. Is this an ensemble of the year 1850 to 2014 or does it have actual "annual" forcings from the period 1850 to 2014? If I use BHIST and set RUN_REFDATE to 2000, will it run and use the forcing of 2001, and then 2002 for the until 2014 for the following years, in this BHIST compset or will it cycle forcings for 2000?

2. In your example RUN_REFCASE = b.e21.BHIST.f09_g17.CMIP6-historical.003, and RUN_REFDATE = 2000-01-01, this won't run if year 2000 isn't inside your REFCASE, right?

3. Are there online databases (e.g. CMIP6 as in your example) with existing REFCASE which I can branch off from?

4. Thank you for sharing the documentation, I will see if I can find detailed descriptions of the COMPSETs inside.

Once again, thank you so much. This thread should be quite useful for multiple new users such as myself.
 

hannay

Cecile Hannay
AMWG Liaison
Staff member
1. I hope you don't mind if I followup on your statement, "the BHIST will give you forcing from the historical period 1850-2014.".. Is this an ensemble of the year 1850 to 2014 or does it have actual "annual" forcings from the period 1850 to 2014? If I use BHIST and set RUN_REFDATE to 2000, will it run and use the forcing of 2001, and then 2002 for the until 2014 for the following years, in this BHIST compset or will it cycle forcings for 2000?

=> This is correct. If you use BHIST and set RUN_REFDATE to 2000, it will use use the forcing of the year 2000, then 2001, then 2002.
If you want to cycle over a single year (the year 2000 for instance), you need to set the namelists to do that. You can look how it is done in the B1850 compset.

2. In your example RUN_REFCASE = b.e21.BHIST.f09_g17.CMIP6-historical.003, and RUN_REFDATE = 2000-01-01, this won't run if year 2000 isn't inside your REFCASE, right?

=> This is correct, if you set
RUN_REFCASE = b.e21.BHIST.f09_g17.CMIP6-historical.003, and
RUN_REFDATE = 2000-01-01,
you need to have restart files for the year 2000, otherwise it will not run.

3. Are there online databases (e.g. CMIP6 as in your example) with existing REFCASE which I can branch off from?

These are some initial conditions that comes out of the box. Unfortunately, we don't put restarts on ESG.
If you have access to the NCAR computers (cheyenne or casper), you can go to
/glade/campaign/collections/cmip/CMIP6/restarts
 

acgvar

Alvin C.G. Varquez
Member
Hi Hannay,

Happy new year! Thank you very much for your detailed and very clear responses. I really appreciate it! Will do further tests based on the knowledge acquired thanks to you. The next challenge would be how to create an account to Cheyenne while being in Japan.
 

cgolja

Colleen
New Member
Hi all!

I've been referencing this thread regarding spin up/ restart files as I begin my first runs with CESM 2.1.3. I've also set out to use the B1850 compset and wanted to inquire about the advised "way" of using this compset. I have found the restart files referenced within this thread and the linked thread; is it common practice to initialize any new experiment using the B1850 compset from the end of a longer PI control run (e.g. year 1000 of b.e21.BW1850.f09_g17.CMIP6-piControl.001)? Is there a common scenario in which using the out of the box initial conditions is preferred?

Thank you for all your help, this thread has been incredibly useful!
 

cgolja

Colleen
New Member
Replying to my own comment here: I see now the the RUN_REFCASE and RUN_REFDATE suggest this run is defaulting to a hybrid run starting from year 300 of b.e20.B1850.f19_g17.release_cesm2_1_0.020!
 

ucas_qs

qiushi Zhang
Member
This is the requirements to get an acccount

Being in Japan, I don't think that you quality.

If you don't have access to cheyenne, I can also provide restarts on a ftp server. You would need to tell me which year you need.
Hello, Dr. Cecile Hannay!
I am running the BHISTcmip6 case, I want to get the restart file in 1850,1950 and 1978 , but I don't have access to cheyenne. I have registered on Globus but Can you upload to FTP? Thank you very much
 
1. I hope you don't mind if I followup on your statement, "the BHIST will give you forcing from the historical period 1850-2014.".. Is this an ensemble of the year 1850 to 2014 or does it have actual "annual" forcings from the period 1850 to 2014? If I use BHIST and set RUN_REFDATE to 2000, will it run and use the forcing of 2001, and then 2002 for the until 2014 for the following years, in this BHIST compset or will it cycle forcings for 2000?

=> This is correct. If you use BHIST and set RUN_REFDATE to 2000, it will use use the forcing of the year 2000, then 2001, then 2002.
If you want to cycle over a single year (the year 2000 for instance), you need to set the namelists to do that. You can look how it is done in the B1850 compset.

2. In your example RUN_REFCASE = b.e21.BHIST.f09_g17.CMIP6-historical.003, and RUN_REFDATE = 2000-01-01, this won't run if year 2000 isn't inside your REFCASE, right?

=> This is correct, if you set
RUN_REFCASE = b.e21.BHIST.f09_g17.CMIP6-historical.003, and
RUN_REFDATE = 2000-01-01,
you need to have restart files for the year 2000, otherwise it will not run.

3. Are there online databases (e.g. CMIP6 as in your example) with existing REFCASE which I can branch off from?

These are some initial conditions that comes out of the box. Unfortunately, we don't put restarts on ESG.
If you have access to the NCAR computers (cheyenne or casper), you can go to
/glade/campaign/collections/cmip/CMIP6/restarts
Hi Cecile, following this thread I wanted to ask what is the way to get BHIST restarts from 2000 for people outside NCAR ?
 

strandwg

Moderator
Staff member
Hi Cecile, following this thread I wanted to ask what is the way to get BHIST restarts from 2000 for people outside NCAR ?

Talk to me. What year(s) would you prefer? I can provide them via the NCAR Data Sharing Service, which requires a Globus endpoint at your institution.
 
Talk to me. What year(s) would you prefer? I can provide them via the NCAR Data Sharing Service, which requires a Globus endpoint at your institution.
I just saw on the ESGF node that there is a 10 member BHIST ensemble is it possible to get the 2000 restarts from each of the members?
 

strandwg

Moderator
Staff member
I just saw on the ESGF node that there is a 10 member BHIST ensemble is it possible to get the 2000 restarts from each of the members?

Actually, there are 11 ensemble members. The January 2000 (only) restarts from all 11 members are available from the following Globus URL:


The total volume is ~218 GB.

Let me know of any problems.
 
Top