Open yuanjianz opened 2 weeks ago
You may well be correct. my memory is that I introduced the weighted sum for MFX/Y at the request of @sdeastham who had some data to indicate that this was needed.
I know that there was some evidence that the units claimed by GEOS were not correct for some associated quantities.
If you have a version that seems to give you good answers, just update this PR and we'll roll it in. GCHP is the only user of the feature - so they are free to define what they need to happen. (Sorry to duck* - I've not given it thought in a long while.)
Hi @tclune, thanks for the reply. My remaining question is the dynamic timestep of GEOS-FP and GEOS-IT. Many thanks again for your input.
@bena-nasa @mathomp4 Can you answer the question above from @yuanjianz regarding time step?
I'll check the systems tomorrow on Discover. @mathomp4 A reminder!
@yuanjianz From what I see, both GEOS-FP and GEOS-IT run at a 450 sec heartbeat. Obviously the dynamics might substep, but as far as History would know, it is 450 seconds.
@mathomp4 Thanks, I will modify the scaling factor accordingly.
Hi all,
I have several questions regarding the horizontal flux regridder and its fix #2056.
Currently variables using HFLUX regridder in GCHP are mass fluxes MFX/Y and courant number CX/Y.
MFX/Y
As mentioned in https://github.com/GEOS-ESM/FVdycoreCubed_GridComp/issues/174, GEOS archived MFX/Y are in
Pa m+2 = kg * g
. Another evidence is in FV3fv_tp_2d
, it seems mfx/y does not need to multiply by edge length to be used.Also in GEOS-5 file glossary, MFX/Y are already multiplied by edge length:
Thus, isn't it better to rollback to the "before-fix" HFLUX regridder that simply
MFX_12 = MFX_1 + MFX_2
instead of the "fixed" version thatMFX_12 = (MFX_1*dy_1 + MFX_2*dy_2) / d12
? E.g. MFX is uniform2 kg * g
in C180. If regridded to C90, intuitively thinking, 4 kg of air enter the C90 grid. Thus, MFX should be4 kg * g
instead of still being2 kg * g
.CX/Y
Courant number CX is
u*dt/dx
if I understand correctly. If the velocity component U is the real "flux" that needs the HFLUX regridder, thenCX_12 = [ (CX_1*dx_1) * dy_1 + (CX_2*dx_2) * dy_2 ]/ (dy_12 * dx_12)
. I am not quite sure about this part, please correct me if I am wrong.Timestep
Since MFX and CX are both dt dependent, I am curious what's the dynamic heartbeat of GEOS-DAS products like MERRA-2, GEOS-FP and GEOS-IT? Is it 450s like mentioned in https://github.com/GEOS-ESM/FVdycoreCubed_GridComp/issues/174#issuecomment-1218222853? If so, we might be using the wrong scale factor for MFX/Y in GCHP(600s). It should be 1.3333333 instead of 0.6666667. And it help explains why our imported mass flux are always two times smaller than wind derived mass flux.
Tagging @lizziel @sdeastham for GCHP related issues Tagging @tclune for the Horizontal Flux Regridder implementation