search

DataWaveProject / CAM

Community Atmosphere Model
Other
1 stars 0 forks source link

Examine the effect of running the model with zero for unimportant tendencies #17

Open jatkinson1000 opened 2 months ago

jatkinson1000 commented 2 months ago

As discussed in meetings, the emulator does not generate tendencies of q or t that the physics-based scheme does. These are apparently uncommon and not important. However, they are used in CAM so need to be supplied in some form - they are used immediately in the energy and momentum checkers etc.

Advice from Joan is to run the emulator alongside the physics scheme and use the results from the physics scheme. This is what is currently done in #14

On 2024/09/26 it was suggested by Pedram and Qiang that we should instead try setting these to 0.0. To examine if this is a viable approach run the physics-based scheme with zeros for these tendencies and observe the effect.

jatkinson1000 commented 2 months ago

An inspection of the gw_draw_prof() routine we are emulating indicates the following intent(out) variables:

Of which only utgw and vtgw are forecast by the net. I will set all to 0.0, except tau which will be unchanged from its input value to gw_draw_prof().

This is implemented on branch zero-tend in commit https://github.com/DataWaveProject/CAM/commit/af38a0d0b4212819badc7d059a5986da865c8f06

jatkinson1000 commented 2 months ago

I am running the FMTHIST compset at resolution ne30pg3_ne30pg3_mg17 as recommended by collborators.

The build logs are located at /glade/derecho/scratch/jatkinson/CAM_GW_zero-tend/bld/ on Derecho. The run logs are located at /glade/derecho/scratch/jatkinson/archive/CAM_GW_zero-tend/ on Derecho.

The atmospheric log specifically can be found at /glade/derecho/scratch/jatkinson/CAM_GW_zero-tend/run/atm.log.6220081.desched1.240928-121643, and the cesm log at /glade/derecho/scratch/jatkinson/CAM_GW_zero-tend/run/cesm.log.6220081.desched1.240928-121643.

The NetCDF files are at /glade/derecho/scratch/jatkinson/CAM_GW_zero-tend/run/. This is not the standard results as this is not yet a scientifically verified case. Since we have not yet been provided with any means of examining the output (post-processing scripts or metrics) I am unable to proceed further so will pass to collaborators.

Note that the results of the baseline case (tag cam6_3_139, with no modifications) can be found at /glade/derecho/scratch/jatkinson/CAM_GW_base_run/run/ for comparison.

jatkinson1000 commented 1 week ago

We are currently waiting on feedback from @yqsun91 as to how these compare.

We have implemented the request to set to zero in #14 but the evaluation is a scientific question.