Leveraging leaf-level optimality processes with explicit acclimation improves global GPP representation in an individual-based DGVM (LPJ-GUESS v4.1.1)

Forrest, Matthew; Dantas de Paula, Mateus; Gomes de Almeida, Filipe; Harrison, Sandy P.; Prentice, I. Colin; Hickler, Thomas

doi:10.5194/egusphere-2026-3273

Preprints

https://doi.org/10.5194/egusphere-2026-3273

Preprints

25 Jun 2026

| 25 Jun 2026

Leveraging leaf-level optimality processes with explicit acclimation improves global GPP representation in an individual-based DGVM (LPJ-GUESS v4.1.1)

Matthew Forrest, Mateus Dantas de Paula, Filipe Gomes de Almeida, Sandy P. Harrison, I. Colin Prentice, and Thomas Hickler

Abstract. Vegetation models are indispensable tools for investigating and projecting the terrestrial carbon cycle, both as standalone models and embedded in global climate models. However, current models vary widely in their representation of ecosystem processes and consequently in their projected future carbon dynamics. Eco-evolutionary optimality (EEO) approaches, which derive and test hypotheses about optimal plant behaviour under specific environmental conditions as a consequence of natural selection, have been proposed as a means to improve the reliability of vegetation models and the robustness of their future projections. Here we embed EEO-derived models for photosynthesis and leaf dark respiration, and their acclimation to changing conditions, into the widely used LPJ-GUESS vegetation model. We evaluated the simulated gross primary production (GPP) patterns against remotely-sensed GPP derived from sun-induced fluorescence and found that the EEO configurations improved the spatial distributions (a mean reduction in error of 15 % across gridcells) and global interannual variability (a mean reduction in error of 32 % after accounting for differences in global totals) compared to the standard version of LPJ-GUESS. Evaluation against GPP fluxes from eddy flux covariance measurements also showed improved performance, the R² of 5-day GPP increased from 0.45 to 0.48 (averaged across 147 sites). The simulated global carbon pools, fluxes, burnt area and biome distributions were not impacted substantially. The improvements were achieved with no alteration to processes except photosynthesis, respiration and plant water uptake, and with no recalibration or tuning. The EEO configuration also reduced model run time and eliminated the need for poorly-constrained PFT-dependent parameters governing the temperature response of photosynthesis. As well as being a tangible improvement to LPJ-GUESS, this study further confirms the usefulness of EEO approaches to improve global vegetation models.

Received: 12 Jun 2026 – Discussion started: 25 Jun 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 3390 KB)

Supplement (91 KB)

Download & links

Matthew Forrest, Mateus Dantas de Paula, Filipe Gomes de Almeida, Sandy P. Harrison, I. Colin Prentice, and Thomas Hickler

Status: closed

RC1:
'Comment on egusphere-2026-3273', Anonymous Referee #1, 20 Jul 2026
General comment
The authors implemented eco-evolutionary optimality principles into the LPJ-GUESS vegetation model, by merging the model with an adapted version of the P-model for photosynthesis. They then evaluated the performance of the model against in-situ eddy covariance fluxtower measurements and global data products of GPP, as well as existing global biome distribution maps. While the metrics show only a small (but significant) improvement compared to the standard version of LPJ-GUESS, these new simulations are based on highly reasonable optimality constraints, which allow the model to use less (often poorly constrained) parameters and assumptions. I think this is a very interesting approach and a very useful and meaningful contribution to the development of LPJ-GUESS and dynamic vegetation models in general. I do have a few (minor) questions and concerns, especially regarding the different values for leaf-to-canopy scaling factor and comparison with a nitrogen-limited version of standard LPJ-GUESS. However, I could see the article being published close to its current form.

Specific comments
L194-216: These lines are sligthly confusing at first read, because it seems Eqs. (3) and (4) should be connected like Eqs. (1) and (2) are. However, when puzzling everything together further in the section it becomes clear that the relationship between GPP_indiv and I_abs,indiv is more complex in LPJ-GUESS. Maybe consider rephrasing a bit to make this clear earlier on? Maybe starting from Eq. (4) could help, and then introduce Eq. (3) only later when it's needed. And maybe don't mention Eqs (1) and (2) at all, because they are not really used? Just a suggestion.

L342: Why is this extremely rare? Aren't such drought conditions rather common, especially when you do global simulations? Is this also extremely rare in Standard LPJ-GUESS? To which degree does the P-model capture the diversity in stomatal regulation (e.g. isohydry)?

L349-350: Regarding the SMART scheme producing somewhat better results, is this in general, or in your simulations? What do you have to back up this statement? E.g. if in general, do you have a reference for this?

L359: Correct me if I'm wrong, but as far as I know the leaf-to-canopy scaling factor (α_a) can have quite a large impact on the simulated GPP and carbon stocks in LPJ-GUESS. I understand that you set this parameter a-priori (ie. you did not tune it afterward to get a good result) even though the new value is still "an initial guess that appears more plausible" (L668-669). Still, is it fair to compare two versions of LPJ-GUESS that have a different value for this? How much of the model difference is due to this parameter, compared to the difference introduced by the EEO approach alone? See also my next comment for a similar concern.

L371-373: The standard simulations have N limitation activated, but your EEO-model does not. Similarly to the previous comment, I wonder how this affects the model comparison, e.g. how much of the difference in model output can be explained by this? Doesn't this difference obscure the effect of the EEO approach on its own? I understand that you want to compare the "default versions" of both models (ie. the standard model has N-limitation by default, and your model does not) and that is totally valid (same for alpha_a). However, it also seems a bit "unfair" to compare a model with N-limitation to a model without it? See also next comment. I also see you discuss this in Sect 4.5 where you give good reasons to doubt the causal relationship between soil N and V_cmax, and you provide future pragmatic approach to include N limitation. But this also makes me wonder whether it would make sense to (also) compare your model against a version of LPJ-GUESS where N-limitation on V_c,max is turned off?

L444: Similar to the previous comment, the standard model version produces a higher annual total, even though it is N limited. I wonder what N limitatation would have on the P-model? Maybe it will underestimate GPP? See also Table 2 (and text from L534) which shows that C stocks, leaf area and NPP were overall a bit lower in EEO version, despite not having N limitation.

L690: Is this really a fundamental part of the P-model on its own?

Technical corrections
L124-125: Maybe rephrase? A bit hard to read in one breath

L37,L135: The term "mitochondrial respiration" is only used twice in this paper, while it's called dark respiration otherwise. Consider harmonising this terminology throughout the paper.

L145: Typo in "LPJ-GUES"

L165: It could be nice to refer to the three EEO hypotheses in the later part of this section, so we know which part of the theory (which formula) corresponds to which hypothesis, ie. from line 196

L218: What is ξ? It has not been defined before. I do see that you define ξopt in eq. (6) however. Or did you maybe mean the ratio χ on this line, as in L174-175?

L324: Word and maybe a comma missing?

L473-474: ...but for many of these biomes still better than the standard model, right? (from Table B2)

L476: The reference to Fig. 6 comes before the first reference to Fig. 5 as far as I can tell. Maybe reorder the figures?

L804: I guess the bold values in Table B2 mean the "better" value. Maybe add that to the caption.

L551: "lower left and upper right quadrants"?
Citation: https://doi.org/10.5194/egusphere-2026-3273-RC1
- AC1: 'Reply on RC1', Matthew Forrest, 05 Oct 2026
  
  Thank you for the review, responses in the attached .pdf.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3273-AC1
RC2:
'Comment on egusphere-2026-3273', Anonymous Referee #2, 16 Sep 2026

The authors applied an eco-evolutionary optimality (EEO) approach for calculation of assimilation rates in the LPJ-GUESS DGVM. Additionally, an acclimation process was implemented. This alternative approach has two major advantages: It saves computation time and gives better agreement between observed and modelled GPP. They used for LPJ-GUESS mode in individual mode and see an improvement in model performance and better agreement with observational data. In general, the paper is well written and structured and fits into the scop of the journal.
The model is run with modelling individual trees. it might be interesting to see the performance in run time and quality of the average individual model mode. This mode is significantly faster and is best suited for coupling to Earth system models. `The authors state, however, that acclimation must be calculated on an individual level. A short discussion how this approach can be applied to average individual models should be added.
The authors compare the LPJ-GUESS model with EEO photosynthesis and acclimation to the unchanged model. It would be interesting to see the impact of different acclimation times on the model performance to see how important the addition of acclimation is.
More specific remarks:
Line 292: Soil moisture is denoted as Theta, identical to the co-limitation parameter Theta in Table 1.
Line 279: One question. Does this correspond to the original photosynthesis model setting the co-limitation factor Theta=1? Are there any justifcati9on to assume this?
Please add unit of GPP and R_d in the figure captions.
The authors use 500 yrs of spin up of the model. The carbon and nitrogen pool should be in equilibrium after the spin up period It might be in particular for the boreal and polar regions that the pools are in disequilibrium because of the low temperatures leading to very low decomposition rates This might not be important for evaluation of GPP values bit for soil pools it is important. Also nitrogen limitation might change.
The author claims that an error in the fAPAR values propagate directly to the GPP value due to its linear relationship. The fAPAR dynamics is also affected by the phenology. LPJ-GUESS has an annual update of the carbon pools distributing the assimilated carbon to all vegetation carbon pools. The phenology function determines the intra-annual the intra-annual variability of leaf carbon changing the fAPAR value. A discussion should be added.
The authors claim to have a faster model using the EEO approach. Pleas gives some numbers for a typical configuration on an HPC cluster.

Citation: https://doi.org/10.5194/egusphere-2026-3273-RC2
- AC2: 'Reply on RC2', Matthew Forrest, 05 Oct 2026
  
  Thank you for the review, responses in the attached .pdf.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3273-AC2

Status: closed

RC1:
'Comment on egusphere-2026-3273', Anonymous Referee #1, 20 Jul 2026
General comment
The authors implemented eco-evolutionary optimality principles into the LPJ-GUESS vegetation model, by merging the model with an adapted version of the P-model for photosynthesis. They then evaluated the performance of the model against in-situ eddy covariance fluxtower measurements and global data products of GPP, as well as existing global biome distribution maps. While the metrics show only a small (but significant) improvement compared to the standard version of LPJ-GUESS, these new simulations are based on highly reasonable optimality constraints, which allow the model to use less (often poorly constrained) parameters and assumptions. I think this is a very interesting approach and a very useful and meaningful contribution to the development of LPJ-GUESS and dynamic vegetation models in general. I do have a few (minor) questions and concerns, especially regarding the different values for leaf-to-canopy scaling factor and comparison with a nitrogen-limited version of standard LPJ-GUESS. However, I could see the article being published close to its current form.

Specific comments
L194-216: These lines are sligthly confusing at first read, because it seems Eqs. (3) and (4) should be connected like Eqs. (1) and (2) are. However, when puzzling everything together further in the section it becomes clear that the relationship between GPP_indiv and I_abs,indiv is more complex in LPJ-GUESS. Maybe consider rephrasing a bit to make this clear earlier on? Maybe starting from Eq. (4) could help, and then introduce Eq. (3) only later when it's needed. And maybe don't mention Eqs (1) and (2) at all, because they are not really used? Just a suggestion.

L342: Why is this extremely rare? Aren't such drought conditions rather common, especially when you do global simulations? Is this also extremely rare in Standard LPJ-GUESS? To which degree does the P-model capture the diversity in stomatal regulation (e.g. isohydry)?

L349-350: Regarding the SMART scheme producing somewhat better results, is this in general, or in your simulations? What do you have to back up this statement? E.g. if in general, do you have a reference for this?

L359: Correct me if I'm wrong, but as far as I know the leaf-to-canopy scaling factor (α_a) can have quite a large impact on the simulated GPP and carbon stocks in LPJ-GUESS. I understand that you set this parameter a-priori (ie. you did not tune it afterward to get a good result) even though the new value is still "an initial guess that appears more plausible" (L668-669). Still, is it fair to compare two versions of LPJ-GUESS that have a different value for this? How much of the model difference is due to this parameter, compared to the difference introduced by the EEO approach alone? See also my next comment for a similar concern.

L371-373: The standard simulations have N limitation activated, but your EEO-model does not. Similarly to the previous comment, I wonder how this affects the model comparison, e.g. how much of the difference in model output can be explained by this? Doesn't this difference obscure the effect of the EEO approach on its own? I understand that you want to compare the "default versions" of both models (ie. the standard model has N-limitation by default, and your model does not) and that is totally valid (same for alpha_a). However, it also seems a bit "unfair" to compare a model with N-limitation to a model without it? See also next comment. I also see you discuss this in Sect 4.5 where you give good reasons to doubt the causal relationship between soil N and V_cmax, and you provide future pragmatic approach to include N limitation. But this also makes me wonder whether it would make sense to (also) compare your model against a version of LPJ-GUESS where N-limitation on V_c,max is turned off?

L444: Similar to the previous comment, the standard model version produces a higher annual total, even though it is N limited. I wonder what N limitatation would have on the P-model? Maybe it will underestimate GPP? See also Table 2 (and text from L534) which shows that C stocks, leaf area and NPP were overall a bit lower in EEO version, despite not having N limitation.

L690: Is this really a fundamental part of the P-model on its own?

Technical corrections
L124-125: Maybe rephrase? A bit hard to read in one breath

L37,L135: The term "mitochondrial respiration" is only used twice in this paper, while it's called dark respiration otherwise. Consider harmonising this terminology throughout the paper.

L145: Typo in "LPJ-GUES"

L165: It could be nice to refer to the three EEO hypotheses in the later part of this section, so we know which part of the theory (which formula) corresponds to which hypothesis, ie. from line 196

L218: What is ξ? It has not been defined before. I do see that you define ξopt in eq. (6) however. Or did you maybe mean the ratio χ on this line, as in L174-175?

L324: Word and maybe a comma missing?

L473-474: ...but for many of these biomes still better than the standard model, right? (from Table B2)

L476: The reference to Fig. 6 comes before the first reference to Fig. 5 as far as I can tell. Maybe reorder the figures?

L804: I guess the bold values in Table B2 mean the "better" value. Maybe add that to the caption.

L551: "lower left and upper right quadrants"?
Citation: https://doi.org/10.5194/egusphere-2026-3273-RC1
- AC1: 'Reply on RC1', Matthew Forrest, 05 Oct 2026
  
  Thank you for the review, responses in the attached .pdf.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3273-AC1
RC2:
'Comment on egusphere-2026-3273', Anonymous Referee #2, 16 Sep 2026

The authors applied an eco-evolutionary optimality (EEO) approach for calculation of assimilation rates in the LPJ-GUESS DGVM. Additionally, an acclimation process was implemented. This alternative approach has two major advantages: It saves computation time and gives better agreement between observed and modelled GPP. They used for LPJ-GUESS mode in individual mode and see an improvement in model performance and better agreement with observational data. In general, the paper is well written and structured and fits into the scop of the journal.
The model is run with modelling individual trees. it might be interesting to see the performance in run time and quality of the average individual model mode. This mode is significantly faster and is best suited for coupling to Earth system models. `The authors state, however, that acclimation must be calculated on an individual level. A short discussion how this approach can be applied to average individual models should be added.
The authors compare the LPJ-GUESS model with EEO photosynthesis and acclimation to the unchanged model. It would be interesting to see the impact of different acclimation times on the model performance to see how important the addition of acclimation is.
More specific remarks:
Line 292: Soil moisture is denoted as Theta, identical to the co-limitation parameter Theta in Table 1.
Line 279: One question. Does this correspond to the original photosynthesis model setting the co-limitation factor Theta=1? Are there any justifcati9on to assume this?
Please add unit of GPP and R_d in the figure captions.
The authors use 500 yrs of spin up of the model. The carbon and nitrogen pool should be in equilibrium after the spin up period It might be in particular for the boreal and polar regions that the pools are in disequilibrium because of the low temperatures leading to very low decomposition rates This might not be important for evaluation of GPP values bit for soil pools it is important. Also nitrogen limitation might change.
The author claims that an error in the fAPAR values propagate directly to the GPP value due to its linear relationship. The fAPAR dynamics is also affected by the phenology. LPJ-GUESS has an annual update of the carbon pools distributing the assimilated carbon to all vegetation carbon pools. The phenology function determines the intra-annual the intra-annual variability of leaf carbon changing the fAPAR value. A discussion should be added.
The authors claim to have a faster model using the EEO approach. Pleas gives some numbers for a typical configuration on an HPC cluster.

Citation: https://doi.org/10.5194/egusphere-2026-3273-RC2
- AC2: 'Reply on RC2', Matthew Forrest, 05 Oct 2026
  
  Thank you for the review, responses in the attached .pdf.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3273-AC2

Matthew Forrest, Mateus Dantas de Paula, Filipe Gomes de Almeida, Sandy P. Harrison, I. Colin Prentice, and Thomas Hickler

Supplement

https://doi.org/10.5194/egusphere-2026-3273-supplement

Matthew Forrest, Mateus Dantas de Paula, Filipe Gomes de Almeida, Sandy P. Harrison, I. Colin Prentice, and Thomas Hickler

Viewed

Total article views: 439 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
263	128	48	439	63	41	47

HTML: 263
PDF: 128
XML: 48
Total: 439
Supplement: 63
BibTeX: 41
EndNote: 47

Views and downloads (calculated since 25 Jun 2026)

Month	HTML	PDF	XML	Total
Jul 2026	47	20	4	71
Aug 2026	22	23	6	51
Sep 2026	97	38	17	152
Oct 2026	97	47	21	165

Cumulative views and downloads (calculated since 25 Jun 2026)

Month	HTML	PDF	XML	Total
Jul 2026	47	20	4	71
Aug 2026	22	23	6	51
Sep 2026	97	38	17	152
Oct 2026	97	47	21	165

Viewed (geographical distribution)

Total article views: 406 (including HTML, PDF, and XML) Thereof 406 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 28 Oct 2026

Short summary

Global photosynthesis on land is a large flow of carbon in the Earth system which underpins food production and ecosystems. It is particularly important to know how it responds to climate change, but current vegetation models do not agree on this. We implemented a new photosynthesis model based on recent theoretical advances into a vegetation model. We found that it improved the representation of photosynthesis when compared both to global data from satellites and detailed site measurements.


Total:	0
HTML:	0
PDF:	0
XML:	0