Different results of iCESM

[XinhaoSuo]

Hi,

When I run the iCESM1.2, I was puzzled by its' results.
First I ran a simulation of iCESM.1.2 from 1975 to 2005(FAMIPC5CN), but something went wrong with the machine, and I only got the result from 1995 to 1998.
So I ran a new simulation from 1995 to 2005, and want to make the results consistent.
However, when I compared the results of the two simulations, I found that their TMQ(Total precipitable water) was inconsistent in 1998 or other years. Is it because of insufficient preheating(start-up time) time?

 

[nusbaume]

Hi Xinhao,

When you say "inconsistent", do you mean that one of the simulations is physically unrealistic, or just that the runs don't exactly match each other? If they don't exactly match each other, did you change anything between the two runs other than the run length, like the forcing files, or PE layout (e.g. NTASKS)? If not, can you compare the SST outputs from the two runs to see if they match? Basically there can be multiple reasons why one simulation may not be exactly equal to another run even if both simulations are physically valid (and thus trustworthy), so I am mostly just asking to see if there is a specific reason that caused them to mis-match.

However, if one of the simulations looks physically unrealistic then please let me know, as that could indicate a "real" problem.

Thanks, and have a great day!

Jesse

 

[XinhaoSuo]

Hi Jesse,

I didn't change anything between the two runs other than the run length and start-year(one started in1976,the other started in 1995 ).
I also compared the air temperature, and they were also different. Now I will run two simulations, one is from 1995, and the other is from 1996. I will compare their air atmosphere temperature and TMQ. I'll let you know as soon as I have results.
Thank you for your help!!!

Wish you have a nice day!

Best regards!

Xinhao

 

[nusbaume]

Hi Xinhao,

I believe that if you start the simulations from different years, then they will start with different initial conditions (at least for SST and sea ice), which will result in the simulations being inconsistent with each other, although still physically valid. So if you want to make sure the runs match each other you'll likely need to start them from the same year (unless you are branching one run from the other).

Also, I believe the surface air temperature between the two runs will be different if they are start-up runs beginning at different years. Instead you'll want to specifically look at SST or sea ice, which are prescribed and thus should be the same between the models regardless of the start year.

Anyways, I hope that helps, and good luck with your test simulation!

Jesse

 

[XinhaoSuo]

Hi Jesse,

Thank you for your help!!
I'm going to rerun the simulation to make sure the data is consistent.What you said really helped me a lot.

In addition, I have two more questions:

First of all, is it possible to get reevaporation data? (The amount of liquid water in the atmosphere evaporating or ice sublimating into gaseous water) . Can I simulate it?

The second question is: If I want to simulate water vapor markers in future scenarios and still use F-CASE, what do I need to do? (Replace the SST data and emissions data? )


Wish you have a nice day!

Best regards!

Xinhao

 

[sysong]

Hi Jesse,

I'm having a similar problem.

There is a slight difference in the results of two model simulations with the same forcing file.
The only difference is the PE layout (e.g. NTASKS).

I want to know how this difference in machine settings could make different results.

Thank you in advance for your help!

Best regards,
Se-Yong

 

[cacraig]

There is a coupler namelist called BFBFLAG which guarantees bit-for-bit answers when task counts are different. As it runs slightly slower, this flag is FALSE by default. It is documented here:

CESM2 Driver Namelist Definitions

Namelist definitions for Driver in CESM2

www.cesm.ucar.edu