@jenniferholm
So, the stuff to do with vcmaxse, jmaxse etc. is the photosynthetic acclimation code that must have been put into CLM after FATES branched off and before ELM did so. The changes are documented in @danicalombardozzi 's paper, wherein the variability in the 'se' terms as a function of growing temperature is explained.
https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1002/2015GL065934

Essentially they are derived from a Kattge and Knorr 2007
https://onlinelibrary.wiley.com/doi/full/10.1111/j.1365-3040.2007.01690.x

There is some uncertainty about whether this is really the way forward for acclimation as, for example, it doesn't have any representation of genetic limitations, nor is there any 'cost' associated with the plants being able to shift to photosynthesizing OK at higher temperatures.

I think that this is something that @alistairrogers was potentially considering implementing in FATES anyway. Perhaps by consolidating it with ELM you'd perhaps be doing them a favour. However, it might be more useful to have the acclimating vs not acclimating versions as alternatives with a switch, which might be more involved than you were anticipating.

Is this a thing you've looked at in MAAT @walkeranthonyp @alistairrogers?

Also adding @Qianyuxuan to this discussion since she is the one working/learning how to update the code to implement the changes.

Currently she is working on Medlyn. @jenniferholm probably a good idea if Cherry could chat with you while we are away to discuss her progress and issues with namelist/handshake. Would that be possible? I haven’t had a chance to sit with her in a week but my understanding is she is at the point of needing some assistance with model logic and you may be able to give her some insight

Also as for options to turn on/off different schemes, that was indeed the intention to provide compatibility with past runs but also test different hypothesis with FATES, starting at the site and then larger scales.

Thanks @alistairrogers and @serbinsh

One question is whether we have any reason to want to keep the Kattge & Knorr version? As Alistair mentions it is problematic at high temperatures. My feeling is that we should probably -not- bother re-implementing into FATES if we have plans to add a philosophically better alternative.

Counter-arguments include that it might be interesting to look at a bunch of different acclimating terms, but that might be overcomplicating things for no good scientific reason?

@rosiealice good point. I think the idea of providing various hypothesis to test is a good one, thus in favor of implementing KK, but if that means a boatload of additional effort it may not be worth it, as you state.

My impression is that we would have to support a more complicated param file, namelist read(?) and perhaps more internal tracking of variables. Also perhaps more effort to make a connection between both HLMs. Not really sure of the overhead that would add (management, numerical). Worth considering before we make the leap.

Also as mentioned it’s easy to run some tests in MAAT first just to experiment with the logic against our measurements and then consider what we might want do.

That said, our hope is that by introducing Cherry into the code base and making mods via an example like Medlyn, It should be easier for her to work on other updates we see fit based on these discussions. So we have a little time before she moves on to acclimation

This is a great discussion. I thought those equations were from Kattge and Knorr. I agree with @alistairrogers that if we know this method has some problems, and there is a better fix, maybe we don’t include KK.

@serbinsh I’ll get in touch with Cherry and talk with her. Good idea for us to chat and I can ask her where she is, and what she thinks, about putting Medlyn and/or Kumarathunge et al 2019 like was mentioned above. @Qianyuxuan @walkeranthonyp i believe also had ideas with implementing Medlyn, and testing with MAAT. Multiple people working on this, so this is good to coordinate :)

As for the parameter file, I’m working with Ryan on this. There is a bunch of new camp related parameters, and I think adding in a couple more should be ok.

Thanks all!

** opps. Not camp parameters, but “cnp parameters”

Hey again,
I also noticed (I think maybe for the first time!) that actual photosynthesis and the three limited photosyntheses are not outputs in FATES. Right?
Of course there is GPP, which is really similar, but GPP is more the net photosynthesis. With these current nutrient updates it seems like a good idea to add in gross photosynthesis, also as listed in FATES as carbon assimilated in the leaf layer (umolC/m2/s). This would be the equivalent of "FPSN" in the HLMs. And then get to the co-limiting ones afterwards, as needed. It seems like going beyond GPP and looking at the CO2 that has been fixed is good for things like PPFD or spectrum of solar radiation available to produce photosynthesis. Am I missing this output somewhere?

Before I go down this path, I wanted to check with the group -- has anybody already started on adding photosynthesis as a History file output? And open to others thoughts on this.
For now, I'm only adding photosynthesis to the outputs.....which looks to me like "psn_out" which then becomes "psn_z" to take into account each canopy layer, leaf layer, and pft.

@jenniferholm
Sounds good! And no, no overlap. I am more and more suspicious that my numbers from Walker et al 2014 are probably not suitable for a global model. But the code structure will be very useful and we should be able to update the parameters with a much more extensive dataset soon.

Re Kattge n Knorr I vote for leaving it there for the time being and @Qianyuxuan can deal with it in the course of her work. But I suggest we don't need to keep it once replaced. As far as I understand it the equation structure is similar with Dushan's analysis, primarily Dushan has updated the parameters and is valid across a greater range of T. But I could be wrong, I need to read Dushan's paper more closely.

@Qianyuxuan give me a shout if you need some help with getting medlyn into the code.

I'll try n do a comparison of Kattge n Knorr n Kumarathunge in MAAT fairly soon. Anytime anyone would like to see comparison of photosynthetic models just let me know and I'll try to knock it out quickly.

@jenniferholm not 100 % sure what you're asking re output.

No, there's no output of either the gross photosynthesis (i.e. without the N limitation) nor the three co-limiting rates (noting that we should be using the no-colimitation theta values from FATES not ELM). Gordon added output for the wp, wc and wj at some point to CLM4.5, but I'm fairly sure that was after the FATES division.

You could add them, but there will be some work to scale them to the cohort level and then to the patch level. In the photosynthesis routine everything is on a per m2 leaf basis, and is scaled in
ScaleLeafLayerFluxToCohort subroutine, so adding those outputs will be a little more involved. Also, because they are going to end up being averaged across leaf layers and cohorts they might be a little less useful than in a simpler mode. The limiting rates are likely to be different in different leaf layers and thus the gross fluxes will be a little hard to interpret. We'd need to think pretty hard about what it was we wanted to look at exactly, I think, before going to the bother of implementing it.

As an aside, I'm actually not sure what the ESM nutrient down-regulation of photosynthesis does any more. Which theory are you planning to implement? Does it allow leafN and thus Vcmax to vary? Apologies I'm out o f the loop on this!

@jenniferholm following up on @rosiealice's point -- FATES in principle does have the machinery to output history variables resolved by leaf layer, or by leaf layer x canopy layer, or by leaf layer x canopy layer x PFT; we added all that when we were concerned that the canopy light profiles might be off. All of these do average across all the cohorts (or at least all the cohorts of a given PFT). But, as @rosiealice says, it really depends what science question you are after for deciding if or how one might best output this type of thing.

@jenniferholm et al. I had a look at the Kumarathunge paper. The general form of the instantaneous temperature response functions are the same as Kattge and Knorr 2007 - they're the modified Arrhenius. But there is a wrinkle in that the form of the temperature acclimation functions that they recommend are different for the JV ratio and DeltaSJ (or ToptJ), specifically they're a function of Thome which is the temperature from the home environment of a species.

That's going to be a little tricky to represent. In the paper this is interpreted as an adaptation term. But reading through the paper it could also be perhaps a term that represents acclimation at a different timescale, e.g. year(s). If that is a valid interpretation it might be easier to implement. @alistairrogers thoughts?

@Qianyuxuan , @walkeranthonyp -- Cherry, I wanted to touch base about this again and see where you are with implementing any updates from the Kumarathunge et al. 2019 paper on temperature acclimation? Because I have some time to work on it this week, and I'd like to begin implementing it and some tests.
This is because I've finally gotten to a good place with the nutrient spinup procedure, and I've also added in the code changes I worked on last year regarding the photosynthesis, leaf level constraints. These photosynthesis runs are successful but still need a lot of actual science testing to make sure the results make sense.

See https://github.com/jenniferholm/fates-1/tree/parteh-acnp-withlive-photo (but this is a much older version before it was decided to remove the TPU terms.) I'm working off of a newer nutrient branch from Ryan, and will update my photosynthesis branch later this week, or next week.

Before going anything further, I wanted to check with you -- have you started any work on a new temperature acclimation approach?
I have just done 2 simple tests, and they give very divergent results.

1. the original FATES code, with the "se" terms from the parameter file (vcmaxse, jmaxse)
2. using the KK2007 equations -
   vcmaxse = 668.39_r8 - 1.07_r8 * min(max((temp_a10-tfrz),11._r8),35._r8) !Kattge & Knorr 2007
   jmaxse = 659.70_r8 - 0.75_r8 * min(max((temp_a10-tfrz),11._r8),35._r8) !Kattge & Knorr 2007
3. next is bring in the Kumarathunge et al. 2019 approach, with the updated coefficients and using a function of Thome which is the temperature from the home environment of a species in the "se" terms and J/V ratio. And there also looks to be an update to the activation energy term of vcmax. (Thank @walkeranthonyp for your last comment above!)

@Qianyuxuan , @serbinsh, @Claire zarakas -- thanks for your response on what you've been working on, great!

I have very embarrasingly fallen behind on the work you, Claire, Shawn, Anthony et al. have/are worked on, like this work with temperature acclimation, already implementing Kumarathunge 2019, and like in today's FATES modeling call on smoothing paramters. Great work. I'm not going to do anything with implementing temperature acclimation. I'll try to do a better job staying up to date. Could I get included on active email chains? Only if it makes sense, I'd like to help where I can. And not duplicate, or sorry get behind on current development/testing.

This was way back in 2020, but here is my branch on adding in basic nutrient constraints on photosyntheis -- https://github.com/jenniferholm/fates-1/tree/parteh-acnp-withlive-photo

When testing nutrient spinups I've since been woring in @rgknox parteh branches, and using his latest developments. As you guys probably know, he's been doing good work with the cnp scalars, nutrient uptake, etc. So I paused the leaf level nutrient stuff.
Now that those branches are in a good spot, I can work on updating my photosynthesis branch and merging with the latest development. But I don't want to duplicate efforts. This isn't anything with stomatal conductance, TPU, or smoothing. Just bringing in Anthony's 2014 equations on leaf level N and P on vcamx25top and jmax25top. (You can see there was some older code bringing in the KK2007 equations, but those can be removed in a next iteration since that effort is more advanced by Cherry and Claire, sorry).
!based on doi: 10.1002/ece3.1173
@walkeranthonyp - I'd love to update these parameters from your 2014 paper to your most recent findings, if you would like.

What is the best way to move forward on coordinating and collaborating on this?

Sounds great @jenniferholm ! As far as I know no-one else is working on this right now.

While I think the approach in my 2014 paper is a decent way to bring n and p limitations on photosynthesis together, I do think the parameter values are not ideal. A paper is in review that shows a general underestimation of gridded global vcmax using my empirical model (Remi Luo, Trevor's previous postdoc is lead on that paper).

I'll reach out to see if we can use parameter values from a much broader analysis of aci curves.

Longer term I'm thinking that the Nick Smith, Han Wang et al approach might be best to set vcmax and then use an inversion of the 2014 model or something else to help determine leaf nutrient demand.

There are a couple items in this #768 pull request that I would love to get some expert opinion on....

@walkeranthonyp -- The "kp25top" term (initial slope of CO2 response curve) has also been updated to account for the newly varying vcmax25top based on leaf N and leaf P, but only when using PARTEH. Otherwise (c-only FATES and non-PARTEH) kp25top is still being derived in

I would appreciate any feedback on this change, and also make sure I'm not double charging kp25top twice.....once at the beginning of the simulation in FatesParameterDerivedMod (kp25top(ft,iage) = 20000._r8 * vcmax25top(ft,iage)) using the static, default vcmax25top value from the parameter file, and then again in the photosynthesis code when using case PARTEH. I think in the current setup I am accidentally multiplying kp25top twice......but I wanted to hold off on correcting this until I got some expert opinion on if a nutrient dynamic vcmax25top value should be used (??)

Along the same lines, what about also updating the factor of "1.67", from Medlyn et al (2002) Plant, Cell and Environment 25:1167-1179 in the current jmax25top equation "jmax25top(ft,iage) = 1.67_r8 * vcmax25top(ft,iage)" to also take into account the dynamic vcmax25top based on leaf N and leaf P? I did not include this "1.67" factor in my code changes.

Anthony -- I'm still using your 2014 paper as the current, status quo approach. BUT, have you heard more about using parameter values from a much broader analysis of aci curves? I think once we have the okay to test with those values, they should be easy to implement since the general structure is there in the PR.