Ambaum 2020 saturation vapor pressure + phase handling

[dcamron]

Dependent on #3725.

In a Christmas Eve sprint, I implemented the analytic saturation vapor pressure from Ambaum 2020, per recommendations from #626. Lots of updated test values, and these may change further, but I wanted to get the changes out there quickly so we could evaluate their impacts, as well as get the changes into the full test matrix.

All departures are small in relative magnitude, but some callouts for potential scrutiny:

• test_lfc_ml
• test_el and its extensions
• any test involving cape_cin (mentioned in issue above as impacted, and needs LCL changes to see full picture.)
• test_dewpoint_specific_humidity_xarray (I reduced the precision instead of redefining the test data)

test_equivalent_potential_temperature_masked worked before. The calculations work on the individual masked array elements but for some reason it blows up on the full array. numpy.ma.exp offered no help. If anyone has any insight here it had me pulling my hair out a bit. Marked this test, not the first time we've run into it in other contexts.

I'm not touching image tests until the other PR goes in and we tackle a review here, and maybe not even until LCL. Done.

To support LCL validation and future support of automatic LCL/LDL/LFL handling, I've built on (and credited) some previous work to enhance certain moisture calcs to support phase handling. In this case, saturation_vapor_pressure now supports water vapor equilibrium partial pressure over liquid and ice, and can automatically do so across the triple point temperature. I've plumbed this into moisture quantity calculations that depend on saturation vapor pressure as well.

I plan to follow up these changes with some tutorial and cookbook content in late Spring or early Summer.

Happy holidays! 🎅

• Tests added
• Fully documented
• Closes Update saturation_vapor_pressure #508, Saturation Vapor Pressure #532, computation of relative humidity not sensible for upper troposphere #2320, Draft for adding support for saturation water vapour over ice #2323, Update e_s to a more accurate formulation and add e_i (based on Murphy and Koop, 2005) #2480

[dopplershift]

I think it makes sense to do this by itself. I don't understand why the test is failing with overflow in exp(). I can get the test suite to fail, but I can't get it to warn when running in a notebook.

EDIT: It's definitely overflowing, and we can just ignore in the test like we do on some others. In a notebook, it appears once and is then suppressed.

[dopplershift]

After reading through section 5, I don't think we should rely on our constants here, but use the tuned value for $C_{pl} - C_{pv}$ that's used in the paper, 2180 J / kg / K (eq. 14). Our values are essentially assuming an ideal gas for water vapor, and nominally for the triple point; the discussion in Ambaum 2020 explains how their value minimizes errors against empirical data from IAPWS95 in the 0-100 degC range.

[dopplershift]

I'm open to the idea that we allow users to override this value.

[dcamron]

Will attach a figure summary of constant differences here later to justify neglecting this.

[dcamron]

At this point we arrive at green, though I'm adding a few tests for dependent calcs now.

@dopplershift it looks like behavior tested in test_moist_lapse_failure no longer fails. Would you prefer I leave some version of that test in for consistency through time? Or just remove it?

[dopplershift]

Ooof, that test looks like it's completely flaky given how many warning filters/xfails we have there. I think that was just checking that we enable some error catching/whatnot. Let's just go ahead and kill it.

[dopplershift]

🎉