Request for an explanation of vertical transport in GEOS-Chem.

[garvitagarwal290]

Name and Institution (Required)

Name: Garvit Agarwal Institution: TCS Research, India

Confirm you have reviewed the following documentation

• [ ] Support guidelines
• [ ] User manuals
• [ ] Debugging GEOS-Chem and HEMCO errors
• [ ] Current and past Github issues

Description of your issue or question

Hi, I had a few questions about the vertical transport in GEOS-Chem. Some of them are probably coming from a lack of good understanding of atmospheric processes so I hope its not too much to ask.

1) I assume that the module 'vdiff_mod.F90' in GeosCore is for vertical diffusion (correct me if thats wrong). However in 'gc_environment_mod.F90' line 569, I see that Vdiff is initialised only if PBL mixing and use_non_local_pbl are turned on in geoschem_config.yml. Does that mean that if these 2 options are not turned on, simple diffusion doesnt happen at all?

2) I have discussed with @yantosca that even if everything else including PBL mixing is turned off, except Transport, even then there is some vertical transport component computed at https://github.com/geoschem/geos-chem/blob/main/GeosCore/tpcore_fvdas_mod.F90. However looking at the code I didnt really understand what is causing the vertical transport. I only picked up "Compute vertical mass flux from mass conservation." or "...mass flux...in the vertical direction...from the hydrostatic relationship". Is it simply diffusion?

3) I understand that PBL mixing would cause major vertical transport, but in my understanding, it would be limited to layers close to the surface right? So for very high vertical layers the vertical transport is happening only due to convection and TPCORE's contribution right?

[yantosca]

Thanks for writing @garvitagarwal290.

I assume that the module 'vdiff_mod.F90' in GeosCore is for vertical diffusion (correct me if thats wrong). However in 'gc_environment_mod.F90' line 569, I see that Vdiff is initialised only if PBL mixing and use_non_local_pbl are turned on in geoschem_config.yml. Does that mean that if these 2 options are not turned on, simple diffusion doesnt happen at all?

For these settings:

  pbl_mixing:
     activate: true
     use_non_local_pbl: true

VDIFF non-local PBL mixing is called.

For these settings:

  pbl_mixing:
     activate: true
     use_non_local_pbl: false

TURBDAY full PBL mixing is called.

See: https://wiki.seas.harvard.edu/geos-chem/index.php/Boundary_layer_mixing

I have discussed with @yantosca that even if everything else including PBL mixing is turned off, except Transport, even then there is some vertical transport component computed at https://github.com/geoschem/geos-chem/blob/main/GeosCore/tpcore_fvdas_mod.F90. However looking at the code I didnt really understand what is causing the vertical transport. I only picked up "Compute vertical mass flux from mass conservation." or "...mass flux...in the vertical direction...from the hydrostatic relationship". Is it simply diffusion?

I think there is some vertical diffusion but this is outside of my wheelhouse. I would refer you to the following references:

• Lin, S.-J., and R. B. Rood, 1996: Multidimensional flux form semi-Lagrangian transport schemes, Mon. Wea. Rev., 124, 2046-2070.
• Lin, S.-J., W. C. Chao, Y. C. Sud, and G. K. Walker, 1994: A class of the van Leer-type transport schemes and its applications to the moisture transport in a General Circulation Model, Mon. Wea. Rev., 122, 1575-1593.

Also tagging @sdeastham

I understand that PBL mixing would cause major vertical transport, but in my understanding, it would be limited to layers close to the surface right? So for very high vertical layers the vertical transport is happening only due to convection and TPCORE's contribution right?

Yes, PBL mixing is done up to the top of the PBL, which is specified by the PBLH met field quantity.

[garvitagarwal290]

Thanks. So both VDIFF and TURBDAY routines strictly compute the boundary layer turbulence only right? I thought that VDIFF included computation of vertical (gradient based) diffusion because of its name.

If diffusion is not computed in VDIFF, then where is it computed? From my look at the source code, the only other likely place is the TPCORE module.

[sdeastham]

Hi @garvitagarwal290,

I don't know how much I can help but I'll try! Neglecting processes which move chemical species relative to the air, almost all vertical transport in GEOS-Chem occurs via:

1. Bulk motion (resulting from divergence in the wind field, and handled within TPCORE);
2. Convective transport (handled within the various convection routines); and
3. PBL mixing (VDIFF or TURBDAY).

Diffusion is considered to be negligible, in part because the real diffusion rate will be exceeded by the numerical diffusion rate.

[garvitagarwal290]

Thanks @sdeastham for your comment but I don't understand how horizontal winds can cause vertical transport? Also, could you share a good resource for me to understand numerical diffusion and/or an introduction to how TPCORE numerically computes transport? Tagging @yantosca if he can also help in this. Thank you.

[sdeastham]

Hi @garvitagarwal290 - vertical air motion is the natural result of divergence in the wind field. This is easiest to think of in two dimensions (horizontal and vertical). If you have converging wind fields, the air is forced to move vertically to satisfy mass conservation.

To better understand all this I'd strongly recommend Modeling of Atmospheric Chemistry by Brasseur and Jacob: https://www.cambridge.org/core/books/modeling-of-atmospheric-chemistry/88C5AEAD7C28EA3E17FFA6D2CE92DE06

[garvitagarwal290]

Thanks a lot @sdeastham !

[garvitagarwal290]

Hi @sdeastham, sorry to bother you again, but going through relevant parts of the book and trying my best to understand from other sources as well, I just have a few questions/confirmations:

1) Can you confirm if TPCORE uses a semi-Lagrangian numerical scheme in the east-west direction and the Eulerian numerical scheme in north-south and vertical direction? 2) if the above is true then, this means that in the north-south and vertical direction, numerical diffusion is non-negligible and in east-west direction it is close to negligible (because of semi-Lagrangian) ? 3) I couldnt find a discussion comparing the strengths of molecular diffusion, numerical diffusion and advection in the vertical direction in the book. Could you help me convince myself better about why molecular diffusion is ignored?

[sdeastham]

1. IIRC, TPCORE uses a semi-Lagrangian scheme when the CFL is expected to exceed 1. I would refer to Lin and Rood (1996) for the details though.
2. No. Conventional Eulerian and semi-Lagrangian schemes both produce numerical diffusion.
3. At some level, the best I can do here is recommend that you try calculating the rates of diffusion - molecular and numerical - and verify that the latter is greater than the former. Articles such as Rastigejev et al. (2010) do a nice job of showing how numerical diffusion in a 3D CTM is much larger than you'd expected based on the order of accuracy of the solver alone.