Units reported for horizontal mass flux appear to be incorrect

[sdeastham]

The horizontal mass flux units are reported as Pa m+2 s-1, but are listed as accumulated mass flux (presumably accumulated over the time step):

FVdycoreCubed_GridComp/DynCore_GridCompMod.F90

Line 1014 in b331e2a

UNITS = 'Pa m+2 s-1', &

Reading the routine fv_tp_2d, it also appears that mfxxyz should indeed be the accumulated mass flux, which should have units of Pa m+2 (Pa m+2 = kg * g) rather than Pa m+2 s-1. If this is correct, then the vertical mass flux calculated in fv_computeVerticalMassFluxes is receiving the accumulated flux rather than the average mass flux, and will be outputting accumulated vertical flux (kg m-2) rather than the average flux (i.e. kg m-2 s-1). This makes sense given that the resolution to a prior issue (#159) means that the time step is no longer used in fv_getVerticalMassFluxes, so an accumulated horizontal flux can only yield an accumulated vertical flux.

The solution is easy - the units of the MFX and MFY exports should either be changed to Pa m+2, or (preferably) the exported data should be divided by dt, as should the data provided to fv_getVerticalMassFluxes. This would ensure that the archived data are truly fluxes and not accumulated fluxes. Flagging @wmputman and @mathomp4 who I think will be able to evaluate whether this is correct or not, and flagging @LiamBindle as it's relevant to a reviewer question.

[sdeastham]

@tclune - tagging you as discussed!

[mathomp4]

I'll probably need to defer to @wmputman for any opinion. I'm guessing you are right since you seem to have thought about this! 😄

[wmputman]

These fluxes are indeed accumulated over all the small time steps inside of FV3, and thus they represent the total mass transported across the cell interface over the long time-step of the model (HEARTBEAT = 450s). So I "think" I agree that the units should not have the per second in them. And that should be consistent with what we did for MFZ.

@tclune was going to see if he can verify these units from tp_core on up to the exports.

[tclune]

I've done my first pass now, and I get units Pa m^2. I'll follow up later with permalinks to what I think are the relevant steps in the calculation. Fairly simple actually other than the 3 layers of code and some variable name changes.