Regression function-based surface albedo parameterization (broadband)

[LenkaNovak]

Following Jin et al., 2011, as required in issue CliMA/ClimaCoupler.jl#577

Components

• add functions
• implement unit test
• compare with Jin et al 2011
• integrate in driver
• test for other types of radiation
• docs
• clean up

Notes

• long runs are non-deterministic (expected due to the random seed), but in some cases they can lead to failing conservation errors (e.g. compare this build v this build

QA

Long runs

• high res slab ocean aquaplanet with RegressionFunctionAlbedo passes build

Unit test comparisons

Direct

w=0 (note for latitude of 90deg this doesn't quite converge to 1 due to the accuracy of the coefficients provided in the paper, but this is currently sufficient for our purposes):

Diffuse

slabplanet after 20d

direct:

diffuse:

[szy21]

30-100 day average albedo in the longrun. The values are a bit too high compared with observations, but the spatial pattern looks pretty good! We can just calibrate the parameters later.

[LenkaNovak]

30-100 day average albedo in the longrun. The values are a bit too high compared with observations, but the spatial pattern looks pretty good! We can just calibrate the parameters later.

Nice, thanks for checking! Is this the direct or diffuse component btw?

[szy21]

This is the combined albedo (defined as sw flux up / sw flux down). We don't save direct and diffusive fluxes separately in RRTMGP anymore, but can separate them if needed for debugging.

[LenkaNovak]

That's good to know, thank you!