🐛 [BUG] - model_EMC.f90 - RMOHO_STRETCH_ADJUSTMENT does nothing

[lsawade]

Description

The Par_file says changing the RMOHO_STRETCH_ADJUSTMENT will change the depth of fictitious moho. That does not seem to be the case. Am I misunderstanding the statement?

Affected SPECFEM3D_GLOBE version

3f24cd9

Your software and hardware environment

linux cluster, a100

Reproduction steps

1. Set `USE_LOCAL_MESH = .true.` and adjust `RMOHO_STRETCH_ADJUSTMENT` in constants.h.in
2. Fictitious Moho still at 40km no matter what value `RMOHO_STRETCH_ADJUSTMENT` takes.

Screenshots

No response

Logs

stdout of `xmeshfem3D`

 simulation time step size:
   DT determined =    5.5000000000000000E-002
   Par_file: user overrides with specified DT =    2.5000000000000001E-002

 using local mesh layout:
   number of layers in crust  =           10
   number of layers in mantle =           60

   number of doubling layers  =            1
            doubling layer at =           11
   fictitious moho at depth   :    40.0000000     (km)

### OS

Linux

[Evans-Onyango]

Hi Lucas, try changing the USE_LOCAL_MESH flag to false. Also, take a look at these slides on LOCAL_MESH true vs false, they show power of local_mesh flag. I hope this helps.

[danielpeter]

this also depends on what model you have chosen. from the info above, it's not clear which one your using as the output provided is incomplete.

note that the RMOHO_STRETCH_ADJUSTMENT parameter so far only made sense for Earth models that want to honor a moho surface when overimposing a 3D crustal model on top of a mantle model, for example by putting CRUST1.0 on top. the adjustment then helps to tune the stretching and honoring of the element boundaries.

however, meshing for a model like EMC is mostly oblivious of moho variations and only requires some simple mesh gridding. in that case, the parameter has no effect - no matter what USE_LOCAL_MESH is set to - as there is no 3D moho variation to honor. i think to see a change, you could try with models like S40RTS, s362ani, sgloberani_aniso, spiral, and so on, anything that uses a 3D crustal model.

[lsawade]

Hi Lucas, try changing the USE_LOCAL_MESH flag to false. Also, take a look at these slides on LOCAL_MESH true vs false, they show power of local_mesh flag. I hope this helps.

The slides are great!

however, meshing for a model like EMC is mostly oblivious of moho variations and only requires some simple mesh gridding. in that case, the parameter has no effect - no matter what USE_LOCAL_MESH is set to - as there is no 3D moho variation to honor.

I see. So given @Evans-Onyango's slides, the best way to use local_mesh is working around the prescribed 40km Moho and setup appropriate doubling layers, to optimize execution speed without loss of fidelity. Just FYI, I have run the various local_mesh setups, and compared the waveforms to global PREM, and 1 chunk PREM automatic setups, waveforms look identical. Phew.

Feel free to close, since I just misunderstood the Par_file comment!