Integrate Woodard et al. (2021) permafrost code and parameters

[bpbond]

This PR integrates the permafrost code from Woodard et al. 2021 by @dawnlwoodard based on original code by @ashiklom . It has been entirely re-implemented for Hector v3, so while it follows the same logic there are some differences:

• We use Boost for lognormal numerics instead of R - see #512
• static_c isn't tracked separately; we simply use the f_static value to implicitly calculate it in respiration calculations

Performance for SSP 2.45

Corresponding figure (not exactly, but many of same outputs; note this is using an older version of the model and the RCPs) from Woodard et al.:

Permafrost loss (a couple hundred Pg C) and temperature change (~0.4 K) look pretty good.

Thaw curve:

Other

• Permafrost is off (i.e. permafrost_c = 0 currently. However, test_hector.sh confirms that the model runs with permafrost active.
• All parameters are exposed in the R interface.
• A copy of the 2001 paper permafrost code is archived in misc/.

Still to do

Where will existence of permafrost be documented in v3?

[github-actions[bot]]

“leeyabot down“

[dawnlwoodard]

@bpbond I remember there being a reason we separated static_c out into a pool originally and didn't just do a fraction but now I'm not remembering the justification for that. But I do know it's something we went back and forth on. Do you remember that rationale? and is it no longer relevant?

[bpbond]

🤷‍♂️ I don't remember. I couldn't see a good reason to keep it numerically, but at the same time that's part of why I am committing your code: so we have a close at hand, easy reference point in case we want to re-implement static_c.

[github-actions[bot]]

“leeyabot down“

[dawnlwoodard]

Okay so when I go in and remake my pf paper figure (just with 2 SSPs for simplicity), I get some pretty different results for the lowest scenario (SSP126). Do we know why this is? And do we believe that it should happen so quickly? It's refreezing much faster than it did in my paper. I know refreezing was the trickiest dynamic we had to deal with in first designing it, so it's one I'm particularly looking at to make sure the behavior makes sense in this version. (Variable labels are not plotting for whatever reason - should be Permafrost C, Thawed Permafrost C, Atmospheric CH4, Atmospheric CO2, Global Mean Temperature)

Also, I don't see a way to get output for the different RH variables (rh_thawedp, rh_ch4, rh_soil, rh_det) that I used in my permafrost paper analysis. Probably not critical but wanted to flag this as I did find them useful outputs to be able to look at to double check behavior.

[dawnlwoodard]

Also, I get a pretty different ssp126 result if I run without biomes (the first figure was made by duplicating my paper methodology with a permafrost and non-permafrost biome). Which makes me wonder why the refreeze stops in the first, but not in this. That doesn't necessarily make sense to me.

[github-actions[bot]]

Differences in Hector outputs

Hello, this is leeyabot! 🤖

The current pull request's outputs do not differ from 3.0.0 (4c41d6e).

[github-actions[bot]]

“leeyabot down“

[github-actions[bot]]

Differences in Hector outputs

Hello, this is leeyabot! 🤖

The current pull request's outputs do not differ from 3.0.0 (4c41d6e).

[github-actions[bot]]

Differences in Hector outputs

Hello, this is leeyabot! 🤖

The current pull request's outputs do not differ from 3.0.0 (4c41d6e).

[bpbond]

@kdorheim Draft permafrost vignette created in 01d9db83d7b0034163733336db9f3686752bf974 and is here: https://rpubs.com/bpbond/994470

[github-actions[bot]]

Differences in Hector outputs

Hello, this is leeyabot! 🤖

The current pull request's outputs do not differ from 3.0.0 (4c41d6e).

[dawnlwoodard]

@kdorheim @bpbond the permafrost vignette doesn't mention running with biomes, which I believe is important to getting accurate permafrost results. Can this be added? I can share examples of how to do this if helpful.