Add option to use the Johnson & Berry formulation for the response of electron transport to irradiance

[alistairrogers]

Describe the changes

Julien Lamour is leading a manuscript where we have used a multi-biome analysis of light response curves to demonstrate that the Johnson & Berry (2021) (JB) formulation for the response of electron transport to light has a better goodness of fit to gas exchange data than the original Farquhar von Caemmerer and Berry (FvCB) formulation that is currently implemented in FATES. Julien’s work also showed that the JB model could be simplified to a form that resembles a rectangular hyperbola. The goal with this issue to to add a switch to enable use of the simplified JB model formulation in FATES. I’ve attached my recent AGU talk deck so you can see the equations and hopefully follow the logic.

The current FvCB formulation is a non-rectangular hyperbola (NRH). When the convexity parameter (theta_psii) approaches zero the response can be described by a rectangular hyperbola with the same form as the simplified JB formulation. Note that Julien’s work supports the elimination of the convexity parameter and adoption of a RH formulation.

There are four parameters to consider in a RH formulation

Irradiance

Alpha (leaf absorbtance)

I’ve had a noodle around in FATES and it looks like FATES delivers absorbed irradiance (qabs) to the currently implemented NRH FvCB formulation. So we just need to simply the proposed equation to account for irradiance being expressed as absorbed irradiance

Phi (Quantum yield)

Quantum yield is not explicitly called out in FATES but is (1-f)/2. F (fpns) is hard coded (0.15_r8)

Sidebar on f. This could potentially be pulled out into a parameter file and it is dependent on a number of factors, f typically decreases with stress. The default fpns (0.15) is relatively low and is what you might expect to observe in a dark-adapted unstressed leaf. Note other land surface models use much higher values for f (see table 1 in Rogers et al. 2019).

Jmax (FvCB)

Calculated from Jmax25 which is calculated from Vcmax25 using a growth temperature dependent ratio (jvr).

The JB model does not include Jmax but includes a new parameter, the maximum conductance through the cytochrome b6f complex (Vqmax). Julien’s work also presents an equation to convert Jmax to Vqmax (with some assumptions about the irradiance used to estimate Jmax).

What I think needs to happen is the following

Addition of a switch to flip between the existing FvCB formulation and the simplified JB formulation
Addition of an equation to convert PFT specific Jmax to Vqmax (slide 22, with possible simplification)
Addition of the simplified RH JB formulation (slide 13)
It would be nice to tidy up quantum yield and address the f issue but not necessary at this point. We can stay with the FATES default of (1-f)/2

The anticipated net result of this will be to decrease modeled photosynthesis at intermediate irradiance.

Is your request related to a problem?

Rogers_B23L-05_Tuesday3pm.pptx

Alternative options

No response

Additional context

No response

[rosiealice]

On the issue with 'f' @alistairrogers might this explain why it has always been so hard to get the model to produce sensible GPP fluxes using the observed (quite high) Vcmax's for high latitude vegetation? That was a big problem in CLM5 and I note that @adrifoster 's calibration of FATES-SP mode returns a quite low vcmax for boreal trees, for example.

[adrifoster]

On the issue with 'f' @alistairrogers might this explain why it has always been so hard to get the model to produce sensible GPP fluxes using the observed (quite high) Vcmax's for high latitude vegetation? That was a big problem in CLM5 and I note that @adrifoster 's calibration of FATES-SP mode returns a quite low vcmax for boreal trees, for example.

yes we really need to bring down GPP in SP mode! Some of the "calibrated" vcmaxes are quite low, and it worries me.

[rosiealice]

So I ran an SP run with f set to 0.35, and just fwiw there is a large impact. I think this should perhaps be a parameter we can modify? (sorry about the massive colorbars, trying to figure out where their size is set in the diagnostics package is eluding me right now).

[JessicaNeedham]

I can make a start implementing these changes. @rgknox guessing it makes sense to add them on top of #1262 ?

[adrifoster]

@rosiealice and @JessicaNeedham yes it would be great to do all this (esp. making f a parameter! (pft-specific?)).

I still haven't re-run my calibration ensembles so it would be really good to get this in before I kick those off.

[alistairrogers]

Hi Rosie, Adrianna,

Yes. high (observed) Vcmax at high latitudes will increase photosynthesis at high irradiance. Increasing f (lowering quantum yield) will reduce photosynthesis at non-saturating irradiance (see fig 6 in my 2019 paper linked above for a visual). Pulling f out as a tunable parameter would be useful but fwiw I don't think it should change with PFT (unless those PFTs differ in stress tolerance). In the long run, I think changing it with stress, would be a better approach perhaps analogous to btran tuning of Vcmax and stomatal slope. Fig 5 in the 2019 paper shows the effect of low temperature on "f" and how much of an outlier the CLM4.5 assumption of f=0.15 is relative to other LSMs, something that is exacerbated at low temperature. Another paper you might want to check out is Xueli's JGR paper where she applied an empirical tweak on quantum yield based on my 2019 paper and reduced GPP at high latitude to better match benchmarks.

It would be very exciting to see this change in FATES.

Alistair

[rosiealice]

How does it change with stress? Does it dynamically adjust or is it more of an adaptation thing (more defence=less light absorption ?) Do we have a good handle on that from observations? Interesting...

[alistairrogers]

I'd say it's an area for research. Stress is one factor, there's a long history of cold stress reducing quantum yield (increasing f) but also high temperature stress. But yes when the products of the dark reactions of photosynthesis can't be used you need to turn down the supply of electrons from the light reactions. Stress that limits photochemistry should be well correlated with increasing f. Of relevance to LSMs photoprotective mechanisms would operate on the scale of hours to a day, photodamage could be days. But overall f=0.15 is optimal and rarely realized in the field.

[glemieux]

FYI, @JessicaNeedham and I are bringing in a namelist option for this soon via E3SM-Project/E3SM#6918 and ESCOMP/CTSM#2904.

[rosiealice]

Just wondering whether this should in fact be PFT specific as we might actually think that even our current default PFTs differ in their stress tolerances? Depending on the stress, but boreal NTs are for sure more tolerant of cold stress, etc etc...

[alistairrogers]

I think it would be valuable. Even within the Boreal NTs there is significant variation in cold stress tolerance so being able to add PFTs in the future that could include this variation would be useful

[JessicaNeedham]

So the fnps parameter should be pft-specific. Do we want the switch between the FvCB or the JB models to be PFT dependent or should we assume that we'll use one model for all PFTs?

[alistairrogers]

It seems to me that making fpns PFT specific would give the most flexibility in the future and of course they could all have the same parameterization to start with. The switch between FvCB and JB should be for all PFTs - one model for all.

[JulienLamour]

Thanks for adding me to the repo and for tagging me Alistair!

As Alistair said, switching from the FvCB to the JB model requires two things: 1) Implementing a rectangular equation in FATES instead of the nonrectangular hyperbola. An alternative is to set Theta_psii to a very low value (near 0), so the non rectangular hyperbola is numerically close to a rectangular hyperbola. 2) converting the Jmax parameterization for the JB model (called Vqmax in JB). We have an analytical equation to derive the new parameter Vqmax from Jmax. So, if you have the list of Jmax for all the PFTs, I can send you the equivalent Vqmax parameters (Vqmax will be between 1 and 2 times Jmax).

By definition, the quantum yield is the initial slope of electron transport with light in the FvCB model. Or, in other words, it corresponds to the slope of electron transport with light in the dark. This slope is key to describing the photosynthesis response to light. Mathematically, this initial slope does not depend on Theta_psii. However, the slope at low light and intermediate light depends on Theta_psii. For practical applications, it is very difficult to distinguish Theta_psii from the quantum yield because both numerically impact the slope at low light. So, in practice, Theta_psii also has an impact on the light response to low light and on the "apparent" quantum yield. My point here is that using the JB model (i.e, Theta_psii close to zero) will change numerically the response of photosynthesis to low light and the apparent quantum yield. In our analysis, using Medlyn et al. 2002 parameterization for the quantum yield (i.e f = 0.26 instead of 0.15) led to better fits for the light curves. But the most important effect was to change Theta_psii, changing f had a lower impact. Medlyn's parameterization for f has been used in other TBMs such as CABLE, MedFATE or JSBach. Note that converting Jmax to Vqmax depends a little bit on f.

For practical questions about light, it would be very interesting to know when photosynthesis is limited by electron transport (Aj) or when it is limited by rubisco (Ac). This information is key to understanding ecosystems' productivity and their potential response to higher CO2 and temperature since the response of Ac or Aj to these factors is different. So my question: is it possible to have this information with FATES simulations? Is this already an output of FATES simulations? I think it would be quite novel to have this info, and also important when analysing the effect of implementing the JB equation.

Medlyn BE, Dreyer E, Ellsworth D, Forstreuter M, Harley PC, Kirschbaum MUF, Le Roux X, Montpied P, Strassemeyer J, Walcroft A, et al. 2002. Temperature response of parameters of a biochemically based model of photosynthesis. II. A review of experimental data. Plant, Cell & Environment 25: 1167–1179.

[JessicaNeedham]

Thanks @JulienLamour , as a part of this PR I can add a history variable that tracks whether photosynthesis is limited by electron transport or rubisco.

[alistairrogers]

Thanks @JulienLamour to me it would make more sense to add your analytical solution to convert Jmax to Vqmax. That way FATES can take the PFT value of Jmax and convert it when using the JB option. This would future-proof the JB option from changes in Jmax in the PFT file. We'd need to think about the best irradiance for the analytical solution. Charlie ran a quick test with theta_psii ~0 before I started this issue (significant impact on GPP). Hopefully adding the RH as an option is easy but if not yes we could use this shortcut + your analytical solution for evaluation purposes. It would be nice to do it right though.

[rosiealice]

I think the psn limitation variable would be super useful from the perspective of debugging. Esp if it had a vertical dimension option. If we reported it as a fraction it will suffer from averaging issues (at night etc) like albedo does. We had this option in an old version of CLM that Gordon implemented and I will try to remember how he got around this problem.