Make soil evaporation scheme modular

[kmdeck]

We will change our evaporation scheme to that of #732 , but eventually add back in this scheme
#735.

They seem to perform similarly, but the former is slightly more stable.

Coarse sand evaporation site:
New scheme:

Old scheme with beta:

After 334 days:
New scheme:

Old scheme with beta:

[braghiere]

Hi Kat, do you know why (i) the model drop in E happens 3 days before the data and (ii) the minimal E is 0 in the model and 1 in the data? It makes me think that something is wrong either with the model, or with the parameters?

[kmdeck]

@braghiere Im not sure, but a couple of general comments:

• both evaporation schemes decrease evaporation dramatically as the soil surface moisture drops. The issue is that then we have a resolution dependence. so, thicker layers take longer to dry out (at the same evaporation rate), so that changes the timing of the "drop". I tried to implement something that was "resolution independent" (linear interpolation to the surface) in the end.
• the drop time is definitely parameter dependent, as it depends on when soil moisture drops below a critical level, which depends on soil type/van Genuchten parameters
• the previous version of the model did not drop evaporation drop to zero as the soil dried. but then we ran into the issue where evaporation was continuing even when the soil became very dry, and at some point it should drop to zero. This seems to be an issue with this model (Swenson and Lawrence), where the soil resistance becomes a finite number as the soil dries.
  SwensonLawrence2014.pdf
• What we currently merged into main is described here:
  lehmann2018.pdf

I 100% agree we should look into this more! And should aim to increase the agreement here. But I hesitate to read to much into this one example at this time given that we dont know the soil parameters exactly.

[braghiere]

Thanks! That makes sense. Would it be helpful to find an example of observations in a place we know exactly the soil parameters? Shouldn't be too hard to find?

[kmdeck]

this would be very helpful! I think we could also ask Dani for his opinion/help, because this example is in one of his papers.