Bug in vertical integrals?

[ilopezgp]

The water paths are defined as the integrals of the corresponding specific humidity multiplied by density. They have units of [M][L]^{-2}.

It seems that the current implementation is an arithmetic sum of the grid-mean values in the column, which would mean that we are missing the product with respect to the local dz, and the result has units [M][L]^{-3}.

If this is not a bug and there is some undecipherable hidden integral in these expressions, we should still change the current implementation to something more human-readable.

TurbulenceConvection.jl/driver/compute_diagnostics.jl

Lines 183 to 187 in 5a27c77

	cent = TC.Cent(1)
	diag_svpc.lwp_mean[cent] = sum(ρ_c .* aux_gm.q_liq)
	diag_svpc.iwp_mean[cent] = sum(ρ_c .* aux_gm.q_ice)
	diag_svpc.rwp_mean[cent] = sum(ρ_c .* prog_pr.q_rai)
	diag_svpc.swp_mean[cent] = sum(ρ_c .* prog_pr.q_sno)

[trontrytel]

I think it's some ClimaCore magic?

This is the commit that changed it from explicitly having dz there
c7581fe

[ilopezgp]

@trontrytel has found an earlier version where we did multiply by dz, following the rectangle rule of integration: https://github.com/CliMA/TurbulenceConvection.jl/blame/25457ecfbbeb030193b778f057ead1e48b17fc04/diagnostics/compute_diagnostics.jl#L160

[ilopezgp]

I think it's some ClimaCore magic?

This is the commit that changed it from explicitly having dz there c7581fe

If this is not a bug, it is not great design to call an integral a sum. We should fix the notation.

[charleskawczynski]

We define sum over fields to be, effectively, volume integrals (see here for weighted jacobian). See CC#748 and CC#693 for existing issues to define things more precisely / correctly for what we want.

So, it's technically an issue, but (I think) it should happen to be correct at the moment.

[ilopezgp]

We define sum over fields to be, effectively, volume integrals (see here for weighted jacobian). See CC#748 and CC#693 for existing issues to define things more precisely / correctly for what we want.

So, it's technically an issue, but (I think) it should happen to be correct at the moment.

Great, thank you! I will add a comment tracking this just above in the TC code. Here is the definition used, https://github.com/CliMA/ClimaCore.jl/blob/main/src/Fields/mapreduce.jl#L24