Skip to content

Lake-model PSII fluorescence: Gu et al. (2019) ​

Run the example, or view its panels in the gallery. Parameters, units and the digitized figure values are in gu_2019.toml.

Gu, Han, Wood, Chang and Sun (2019, New Phytologist 223:1179–1191, doi:10.1111/nph.15796) write the fate of PSII excitation as a competition between four first-order rates (fluorescence , constitutive heat , regulated heat and photochemistry ) and replace the rates by three quantities of PAM fluorometry under the lake model, in which all antennae share one exciton pool:

With      the yields are

and their ratio does not contain NPQ. With PSII excitation   , internal PSII emission   and linear electron flow  :

An observed flux relates to the internal PSII emission by  : is the probability that a PSII photon emitted inside the leaf reaches the observed flux, and the PSII share of that flux (Gu et al. set it to one; their Eq. 2 excludes PSI). Both refer to the same spectral and angular quantity as and depend on wavelength; Eq. 21 as written is for a hemispheric, spectrally integrated photon flux, so a narrow-band directional radiance has to be converted by the optical model first. The PSII share of absorbed light enters only through , which fluorescence and electron flow share, and so does not appear in Eq. 21.

What is implemented ​

PaperPackage
, LakeModelFluorescence
Eqs. 14, 16 (yields from and NPQ), Eq. 17 after multiplying by lake_model_yields
Eq. 16 solved for NPQ from , and , then Eq. 14fluorescence with PSIIOpenFractionInputs or CoupledPhotosystemInputs
Eq. 21 ( from fluorescence), with and electron_flow_from_fluorescence with ObservedFluorescenceInputs
Eq. 23 (assimilation at the chloroplast from )electron_limited of carbon_assimilation with that

LakeModelFluorescence is a fluorescence component like RateCompetitionFluorescence and CytochromeFluorescence: the same fluorescence call evaluates it, and it accepts the inputs that fluorescence_inputs returns for CytochromePhotosynthesis.

julia
lake = LakeModelFluorescence(; maximum_photochemical_yield = 0.83,
    heat_to_fluorescence_ratio = 19.0, yield_tolerance = 1e-12)
inputs = fluorescence_inputs(cytochrome_photosynthesis, site, excitation)
fluorescence(cytochrome_fluorescence, inputs)   # Johnson & Berry (2021)
fluorescence(lake, inputs)                      # Gu et al. (2019), PSII only
electron_flow_from_fluorescence(lake, LeafPhotosynthesis.ObservedFluorescenceInputs(;
    photon_flux = 2.0, escape_probability = 0.8, psii_fraction = 1.0, psii_open_fraction = 0.5))

The model has no closure: is an input, and NPQ follows from it and the electron flow. Eq. 24 of the paper (diffusion from ambient air to the chloroplast with three resistances) is not a package function; the example solves its   case in closed form and the tests check it against carbon_assimilation.

Not implemented: the C4 forms (Eqs. 25–27), the chloroplast-envelope resistance of Eq. 24, a predictive model of , PSI emission, and escape ( is an input).

Relation to the Johnson & Berry (2021) model ​

For connected PSII units with exciton transfer rate , CytochromeFluorescence uses a photochemical yield that tends to the lake expression above as   and to the separate-units (puddle) expression at  . The lake model is therefore the fully connected, PSII-only limit of the PSII part of CytochromeFluorescence; the tests confirm that the two return the same PSII yields and NPQ for the same inputs when is very large. The two share the rate physics and differ in the antenna assumption and in what closes the system:

Gu et al. (2019)Johnson & Berry (2021)
PSII antennalake ( )connected units,    (Table 1)
Open fractioninput ( measured or modeled elsewhere)predicted from Cytochrome capacity and PSI light (Eq. 22)
NPQinput, or implied by and solved so that PSII delivers the electron flow the dark reactions use
PSI fluorescenceexcludedincluded; weighted in the PAM observer
Electron flowfrom fluorescence (Eq. 21) or from light with and NPQ (Eq. 17)from light, , , CO₂ and O₂
Maximum PSII yieldmeasured (0.83)rate constants (0.882); the apparent is lower (0.821) because of PSI

The figures of the paper ​

With and NPQ digitized from Fig. 1c (Vitis vinifera, eight irradiances) and the caption values  ,  :

  • Eq. 16 reproduces the plotted of Fig. 4c within 0.002 at every point.

  • Eq. 14 gives between 0.0098 and 0.0122. The plotted in Fig. 4c, as digitized here, is 0.0124 to 0.0163, 26–33 % higher. This apparent inconsistency between figure and equation is consistent with the figure showing       , Eq. 14 without the term   in the bracket, which the digitized points follow within 1.5 %; a single different (it would have to be 14.0 to 14.9) fits less well, and with the plotted values the four yields sum to 1.06–1.18. The cause is not established: calibration of the digitization or a different convention in the figure remain possible. One input is not independent: the fourth NPQ marker of Fig. 1c is hidden and was reconstructed from the combination plotted in Fig. 4c. The paper's qualitative statement holds either way: varies by a factor 1.24 over the light response while halves.

  • Eq. 21 with series diffusion (Eq. 24 at  , solved by the package's gas_exchange for a prescribed electron flow) reproduces the five curves of Fig. 2a at their right end (SIF = 5) within 0.07 μmol CO₂ m⁻² s⁻¹.

Flexas et al. (2002) Figure 2 ​

In the lake model the normalized steady-state fluorescence of PSII is

for an unchanged probing light, PSII cross-section and optical weighting between the dark reference and the light-adapted state. A prediction of therefore needs two of , and NPQ from somewhere. With estimated after Oxborough & Baker, the same relation holds identically between the detector levels , , , of a fluorometer, whatever PSI adds to them, when is the apparent : applied to PAM quantities it is an identity and cannot fail.

The example keeps the gas exchange, the drought schedule, the PSII and PSI rate constants and the PAM observer of the Flexas benchmark and changes only the fluorescence formulation. All three rows below are compared with the measured signal through the same observer (PSI supplies 39 % of dark in it):

FormulationClosureRMSE, 49 lit points of Fig. 2RMSE, Inset (PPFD 750)
CytochromeFluorescence, , and PSI from one model,  0.0620.038
LakeModelFluorescence, model the same , and PSI; lake antenna0.2440.324
LakeModelFluorescence, transferred apparent and of the benchmark; of Gu et al. Fig. 1c at every stage0.4330.615

Without the PSI level (PSII emission alone against the measured signal) the first two rows score 0.180 and 0.290.

A fourth evaluation uses no photosynthesis model: and NPQ of Fig. 1c in the first form above with  . It follows the well-watered curve ( ; RMSE 0.044 over the seven irradiances inside that curve's range, peak 1.44 against the observed 1.45), and the   curve less well (0.123). By the identity above this is a consistency check of the digitized PAM quantities, not a prediction. Gu et al. do not state which treatment of Flexas et al. their Fig. 1c shows; the agreement suggests a leaf close to the well-watered ones.

The second and third rows are sensitivity experiments on how the lake relations might be closed; neither is a formulation Gu et al. propose. What they show:

  1. Connectivity, with the open-center population held fixed. Johnson & Berry's (their Eq. 22) is a population of open centers and can be placed in a lake antenna. With the dark-balanced PSII share of the benchmark, their light-limited electron flow is   per PSII photon, where is the PSI/PSII electron ratio and its value at the CO₂ of the dark balance. A lake then emits   from PSII: within 0.02 of one at 21 of the 49 points and below it at the others, never above 1.0002, while the observations rise to 1.45. The same flow and population in connected units (  ) need less quenching (intrinsic PSII NPQ 2.2 against 4.2 at  , PPFD 516), and rises. The antenna assumption, not the population, makes the difference: the lake that a PSII-only fluorometer would report for the connected units,       with     (0.28 where  ), put into the lake relations with the same returns the connected-units fluorescence and NPQ exactly (tested). The Flexas walkthrough shows how the fit depends on between these limits and how it trades off against the PSI weight (connectivity and the dark level).

  2. Consistency between and . With the electron flow fixed,    and in Eq. 21   (tested); at fixed NPQ the sensitivity of is the smaller . The transferred and the benchmark's electron flow do not belong together: at   and PPFD 729 the benchmark's apparent is 0.26 (PSII alone 0.33), the PAM value of Fig. 4c at that light about 0.45. The row therefore misses even the well-watered curve (0.80 against 1.45). Which electron flow is right for the Flexas leaves cannot be decided with the data in this repository; a PAM electron transport rate above the gas-exchange estimate is common and can reflect electron sinks other than carboxylation and photorespiration.

  3. Drought at one irradiance (Inset, PPFD 750). Every row shares the electron flow and falls with conductance, as observed. From   to   the observed falls by a factor 1.61, CytochromeFluorescence by 1.71, the lake with by 1.40, and the lake with the transferred by 2.9. Conditional on the benchmark's apparent electron flow, the open fraction that reproduces the observations falls from 0.21 to 0.12, where the transferred curve has 0.39 at every stage; read the other way, Eq. 21 with the transferred returns from the observed 1.8 ( ) to 3.3 ( ) times that apparent flow.

The comparison does not establish that the hypotheses of CytochromeFluorescence describe these leaves: its row carries chosen biochemical parameters, a prescribed non-stomatal drought schedule and a PSI detector weight transferred from another instrument and species (see the chosen inputs). It shows that the lake relations need treatment-specific or NPQ, or an equivalent electron flow, and that the result is sensitive to connectivity and to the observation model. A discriminating test needs the PAM record of each treatment ( with , NPQ and electron transport rate against PPFD) with matched gas exchange; Flexas et al. measured these, but they are not digitized here. Sustained quenching or photoinhibition under drought, which would lower , is held fixed in every row (Flexas et al. report dark of 0.80–0.83 across the leaves).