Compute SHF using SF, do not re-compute offline

[juliasloan25]

Purpose

Since evaporation is no longer included in the SH computation in SF.jl, we now use the output of sensible_heat_flux provided directly by SF.jl's call to surface_conditions. This will make fluxes calculated in atmos and in land more similar, which is a good to prepare for coupled runs.

Our fluxes differ from those calculated in SurfaceFluxes still in these ways:

• land components usually require r_sfc, an additional resistance term, in the evaporation calculation.
• for the bucket and snow models, this is the only change
• for the soil model, we do not allow for condensation or frost. Again we modify the evaporation using a surface resistance. We also compute SF twice, once for liquid water, and once for ice. This requires us to combine the SH_ice and SH_water into a single number also.
• for the canopy, the surface resistance affects SH and E

To-do

Content

Note that this change required an increase in the threshold for passing in one of our tests (for the canopy temperature jacobian contribution from SHF). Previously, we used only cp_m of the atmos in the SH flux, which is indep of T_canopy. now, the coefficient in the Jacobian depends on cp_m_sfc. This does depend on T_canopy, because it depends on q_canopy, which we approximate as q_sat(T_canopy). We hypothesize that this is why the error in the SH approximation grew with this PR. @trontrytel

Previous: SHF = -rho_sfc cp_m_atmos (T_atmos - T_canopy)/resistances
Now: SHF = -rho_sfc/resistances * [stuff indep of T_canopy ..... - cp_m_sfc(T_canopy)*T_canopy ....]

Since we already have an issue open to improve the jacobian approximation, we are accepting this for now. @juliasloan25

For good measure, though, I added the derivative of rho_sfc with respect to T_sfc

Comparisons

This PR followed by implicit canopy PR long run:
Long Runs at 1/2 year:
Canopy Temp:


ET:


SWC:

Canopy Alone Convergence:
this PR vs current main


Errors:

SoilCanopy Convergence:
this PR vs current main


Errors:

---

• I have read and checked the items on the review checklist.

[Sbozzolo]

Let's regenerate the Manifests before merging.

[akshaysridhar]

Two small issues
(1) Variable naming in canopy_boundary_fluxes.jl
(2) Neutral stability tolerance kwarg in energy_hydrology.jl

[akshaysridhar]

Thanks @kmdeck ! Another note: ts_cpu.png and ts_gpu.png plots in the buildkite output (static map albedo tests) are blank; since these are visualisations of time-dependent vars maybe we can remove them entirely from static map cases.

[kmdeck]

Thanks @kmdeck ! Another note: ts_cpu.png and ts_gpu.png plots in the buildkite output (static map albedo tests) are blank; since these are visualisations of time-dependent vars maybe we can remove them entirely from static map cases.

We have some cleanup to do in our buildkite runs which will be tackled in another PR. I will add this to the list! Thank you!

[juliasloan25]

Maybe we could explain what it does if r_sfc !=0 too

[kmdeck]

I am trying to work too quickly today and being a little sloppy - merged before I saw this. I'll fix this in my next PR.