Temperature tendency budget in Held Suarez simulations

[shiduan]

Hi all,

I'm getting started with CESM2.1.5 and running a held-suarez test with the default settings (Eulerian spectral-transform dynamical core, T42z30_T42_mg17). I'm trying to understand the temperature tendency terms. Should I expect the total temperature tendency (given by TTEND) to be the sum of dycore tendency (DTCORE), diffusion (DTH) and relaxation (QRS)? I've done a quick check but these terms do not sum to the total tendency. Maybe I’m misunderstanding the tendency budget. Could someone help me with this?

Additionally, is my understanding correct that dycore tendency is related to advection?

Thanks!

 

[islas]

It is possible that DTCORE also includes the temperature tendency due to diffusion. Have you tried just adding up DTCORE and QRS and seeing whether they equal TTEND (i.e., excluding DTH)? I'm not certain but looking at the code it seems like DTCORE might be calculated as the difference in temperature between after the dycore and before the dycore, in which case I think it would also include the diffusive tendencies.

 

[aherring]

I think Isla is correct, that the vertical diffusion (DTH) is not active in Held-Suarez, and that the only physics tendencies are QRS and the energy fixer EFIX. DTCORE is diagnosed as the state at the beginning of the physics at the current time-step minus the state at the end of the physics of the previous time-step, and represents the T-tendency due to the dynamical core.

 

[islas]

I think the energy fixer should be off in Held-Suarez. That was definitely a decision that was made before. If it's on, I think that's unintentional. I suspect it's not on.

 

[aherring]

You're right Isla, EFIX is only called for the adiabatic option of simply physics, and so not Held Suarez. Sorry for the confusion!

 

[shiduan]

It is possible that DTCORE also includes the temperature tendency due to diffusion. Have you tried just adding up DTCORE and QRS and seeing whether they equal TTEND (i.e., excluding DTH)? I'm not certain but looking at the code it seems like DTCORE might be calculated as the difference in temperature between after the dycore and before the dycore, in which case I think it would also include the diffusive tendencies.

Here is a quick check on middle level (level 15) using DTCORE+QRS-TTEND:


I assume there should be another term for this residual, although the values are pretty small.

 

[shiduan]

I just noticed there is another variable TTEND_TOT, which seems to be the sum of DTCORE+QRS.
I wonder what's the difference between TTEND_TOT and TTEND. When updating the temperature in the model, is the TTEND_TOT used or TTEND?

Thanks!

 

[aherring]

TTEND should be similar to DTCORE -- it's the EUL dycore tendency computed in the dynamics, whereas DTCORE is the dycore tendency diagnosed on the physics side of the code. TTEND_TOT is the total T-tendency (dycore+physics) diagnosed on the physics side.

 

[shiduan]

It is possible that DTCORE also includes the temperature tendency due to diffusion. Have you tried just adding up DTCORE and QRS and seeing whether they equal TTEND (i.e., excluding DTH)? I'm not certain but looking at the code it seems like DTCORE might be calculated as the difference in temperature between after the dycore and before the dycore, in which case I think it would also include the diffusive tendencies.

Hi Islas, could you also explain the PTTEND (physics tendency). I'm getting the same PTTEND as QRS. Why are these two terms the same?

 

[aherring]

PTTEND and QRS should be the same, as PTTEND is the T-tendency due to the sum of all physics processes, and in Held Suarez there is only one physics process for temperature, QRS.

 

[shiduan]

PTTEND and QRS should be the same, as PTTEND is the T-tendency due to the sum of all physics processes, and in Held Suarez there is only one physics process for temperature, QRS.

Thanks! Now I'm clear about these terms.

Besides, I'm wondering if DTV (temperature tendency due to vertical diffusion) and DTVKE (temperature tendency due to vertical diffusion of kinetic energy) somehow available as output variables?

 

[aherring]

There is no vertical diffusion of temperature in Held Suarez, only newtonian relaxation (QRS). Are you referring to another configuration?

 

[shiduan]

There is no vertical diffusion of temperature in Held Suarez, only newtonian relaxation (QRS). Are you referring to another configuration?

Thanks Adam.

I've also checked the temperature before and after physics (TAP, TBP). Should we expect (TAP-TBP)/dt equals to QRS (PTTEND)?
ATM_NCPL equals to 72 and NCPL_BASE_PERIOD is day. So I saved instantaneous output at 20-min interval. I was assuming QRS * 60 *20 should equal to TAP-TBP, but I found it is QRS *60*40=TAP-TBP. Is there something wrong with my calcualtion/understanding?

 

[aherring]

You're running the EUL dycore, which uses leapfrog time-stepping. The way the integration works is that at each 20 minute interval dt, it steps back to the previous state t-dt, and integrates over 2*dt to produce the state at t+dt. So (TAP-TBP)/(2*dt) is what the QRS tendency represents.

Our more modern dycores in CAM (FV, SE, MPAS, FV3) don't use leapfrog time-integration.