Revised nitrogen balance for plant photosynthesis and growth in the dynamic global vegetation, hydrology and agriculture model LPJmL (5.10)

Heinke, Jens; Schaphoff, Sibyll; Wirth, Stephen Björn; Müller, Christoph

doi:10.5194/egusphere-2026-4549

Preprints

https://doi.org/10.5194/egusphere-2026-4549

Preprints

28 Sep 2026

| 28 Sep 2026

Status: this preprint is open for discussion and under review for Geoscientific Model Development (GMD).

Revised nitrogen balance for plant photosynthesis and growth in the dynamic global vegetation, hydrology and agriculture model LPJmL (5.10)

Jens Heinke, Sibyll Schaphoff, Stephen Björn Wirth, and Christoph Müller

Abstract. The balance between the supply of reactive nitrogen to plants and their nitrogen demand determines the extent to which plant growth, and the associated terrestrial carbon and water cycles, are limited by nitrogen (N). While developing recent versions of the Lund-Potsdam-Jena managed land (LPJmL) model, we identified a legacy scaling error in the nitrogen demand formulation that artificially inflated plant leaf N demand by a factor of approximately two. Correcting this structural inconsistency fundamentally altered the modeled terrestrial N balance and effectively eliminated N limitation on simulated vegetation growth. To restore process realism, we present LPJmL version 5.10, which resolves this legacy formulation by implementing a comprehensive re-evaluation of plant uptake, recycling, and phenological dynamics. Specifically, this model update introduces: (1) a corrected leaf N demand calculation that removes inappropriate daylength scaling; (2) updated root uptake kinetics for separate nitrate (NO₃^-) and ammonium (NH₄⁺) uptake based on experimental data; (3) changes to N recovery and turnover in plants; and (4) revised tree phenology to improve seasonal canopy dynamics. We evaluated the resulting model behavior using the International Land Model Benchmarking (ILAMB) tool, spatial vegetation patterns against satellite observations, and comparisons with global literature values for N cycle components. LPJmL 5.10 preserves aggregate land-surface performance for energy and hydrological diagnostics while establishing a more self-consistent and realistic internal N cycle. Relative to its predecessor, global vegetation N stocks and N uptake decreased by ~35 %. This reduction in plant demand shifted mineral soil nitrogen toward a larger nitrate pool, which subsequently increased immobilization, leaching, and denitrification fluxes. These altered pathways bring global N losses into closer agreement with absolute empirical literature estimates. As the final release of the LPJmL 5 series, version 5.10 documents the exact model configuration used for the TRENDY simulations in the Global Carbon Budget 2025 and provides a structurally sound foundation for future model development in the LPJmL 6 series.

Received: 28 Jul 2026 – Discussion started: 28 Sep 2026

Competing interests: At least one of the (co-)authors is a member of the editorial board of Geoscientific Model Development.

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

Jens Heinke, Sibyll Schaphoff, Stephen Björn Wirth, and Christoph Müller

Status: open (until 23 Nov 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2026-4549', Anonymous Referee #1, 22 Oct 2026 reply
Heinke et al. present their recent developments of terrestrial N cycling in the LPJmL model. Following a bug fix in plant N demands, the authors further revised and re-tuned several N-related processes and parameters, and evaluated the updated model mainly against ILAMB. I appreciate authors’ efforts to improve the model, and the manuscript is generally well written and easy to follow. However, I am not fully convinced that the updates are sufficiently constrained by observations. Compared with the previous version, the overall ILAMB performance is similar or even slightly worse, while the substantial reduction in plant N uptake raises further concerns. I have no objection to the bug fix itself, but the authors should clarify whether the degraded model behaviour after the bugfix is mainly due to inappropriate parameter values, or instead reflects more fundamental issues in the parameterization structure, C-N coupling, or other missing processes. Furthermore, additional analyses are needed to demonstrate that the revised version produces more realistic ecosystem responses to environmental change.

Here are my major comments:
I am highly concerned that the removal of daytime scaling systematically reduces simulated leaf N content by about 50%, given the current structure that Nleaf is diagnosed from the optimized Vcmax. There are now abundant site-level trait observations that could be used to constrain leaf N content or leaf C:N ratios. Have the authors evaluated both model versions against those observations, and which version agrees better with the data? More fundamentally, if simulated Nleaf is highly sensitive to the potentially optimized Vcmax, while the optimized Vcmax itself is difficult to constrain, have the authors considered revisiting the formulation of Nleaf rather than re-tuning other N-related parameters to recover the reasonable global C and N budgets?

A weakness of the current model structure is that the C and N allocation is simply based on the prescribed ratio, whereas plants can dynamically adjust biomass allocation and tissue stoichiometry under nutrient limitation. Such acclimation should be part of the ecosystem response to N limitation, rather than only representing N limitation through its effect on photosynthetic capacity Vcmax. The authors should discuss to what extent the current formulation can capture these whole-plant adjustment strategies.

I am also confused by f(NC) calculation of plant N uptake. The plant N:C ratio is calculated from all plant tissues, whereas the reference thresholds are derived from leaf C:N ratio. Because woody tissues generally have much lower N content than leaves, I wonder whether this function can robustly constrain plant N uptake across different nutrient conditions and PFTs.

In Table 2, the fN,turnover (1- N resorption ratio) is re-tuned based on Sophia et al. (2024). However, this study suggests the N resorption ratio of roughly 50% in tropical trees, increasing toward about 70% in temperate and boreal forests. I am therefore unclear why there are many fN,turnover values areset to 0.7 (indicating the N resorption ratio is only 30%)? Why there seems no differences between the temperate and boreal forests?

Several important diagnostics of model performance are currently missing. First, relative to a N-unlimited simulation, what is the spatial pattern and magnitude of the N limitation effects on C, and how does this differ between the old and new model versions? Second, can the new version better reproduce ecosystem responses observed in manipulation experiments? In particular, I think it would be valuable to evaluate the simulated responses of ecosystem C fluxes to N addition. Such analyses would provide a more direct test of whether the revised model represents N limitation more realistically, rather than only producing plausible global C and N budgets.

Minor comments:
Line 44: Actually, the lack of a relationship linking carboxylation capacity back to leaf nitrogen content should be a very important issue in current model structure.

Lines 202 -212: All of the parameter-tuning look arbitrary and hard to follow. Could authors add one more table to summarize these changes and justifications for them?

Lines 302-303: What are the drivers of the increased mineral N pools? Is it because of the less N uptake? Why NO3 pool size now is almost doubled than NH4? Can authors provide any evidence on it?

Figure 2. Not clear and hard to compare old and new versions. Could it be one single N variable per sub-plot and compare the difference? Besides, what I fail to understand is that the PVN and LU should be both ‘spinup’ stage according to authors description (Lines 224-228), why in this figure they can cover 1901-2024 and have differences?

Line 347-379: It is surprising that even though the plant N demand changed a lot, the BNF budget remains the same. Or not surprising as the maximum BNF rate is tuned to be 50 times higher.

Table 5. Could authors also provide the pft-level comparison?

Reply
Citation: https://doi.org/10.5194/egusphere-2026-4549-RC1
RC2: 'Comment on egusphere-2026-4549', Anonymous Referee #2, 24 Oct 2026 reply

The manuscript by Heinke et al. presents a bug fix for the latest subversion of the land surface model LPJmL v5 and evaluates its effects on model performance. The identified bug concerned the calculation of nitrogen (N) demand, and correcting it necessitated additional revisions to the representation of N cycling processes. Although these revisions provide a more realistic representation of N cycling, the authors acknowledge that further improvements remain necessary.
The authors benchmarked the revised model against observation-based datasets and compared its performance with that of the previous, unrevised version. Overall, the revisions did not improve model performance with respect to carbon, water and energy budgets or crop yields. However, the revisions changed simulated ecosystem N budgets. Unfortunately, the simulated N fluxes were compared with empirical estimates only at the level of global totals. Simulated plant N uptake and vegetation N stocks decreased substantially in the revised model. In contrast, global biological N fixation (BNF) remained unchanged, at values substantially above the range of data-driven estimates. Simulated soil organic N stocks also remained unchanged, which the authors explained with increased microbial immobilisation of mineral N compensating for reduced litter N inputs. Reduced plant N uptake nevertheless resulted in greater N losses through leaching and denitrification, bringing these fluxes closer to existing estimates, at least in terms of global totals. Changes in simulated volatilisation and N₂O emissions were minor.
Overall, this model revision and evaluation could lead to a publication in GMD. However, major revisions are needed before the manuscript can be accepted. Although the manuscript clearly describes the model changes, the motivation and justification for several of them require clarification (see my general comments). More importantly, the evaluation of simulated N fluxes against observation-based benchmarks and estimates from other models could be strengthened (see my major comments).

Major comments:
In Section 3.6 and Table 6, the authors compare their simulated N fluxes to other estimates only at the level of global totals. I am sure that they could make a much more detailed comparison accounting also for global spatial patterns and differences between types of vegetation.
Note that for many of the N fluxes, there are spatially explicit estimates available, for instance in the Bouwman et al. paper referenced in Table 6. For leaching, Beusen et al., 2015 could be interesting: https://doi.org/10.5194/gmd-8-4045-2015. For denitrification, Feng et al., 2023 have produced a benchmark: https://doi.org/10.1038/s41467-023-38803-z. The authors could further use SoilGrids2.0 as benchmark for simulated soil N stocks (https://doi.org/10.5194/soil-7-217-2021) and WISE30sec for soil C/N ratios (https://doi.org/10.1016/j.geoderma.2016.01.034). In addition, they could compare their simulations to outputs of N-MIP models (https://nmip.net/), even if those are following different simulation protocols.
Finally, I would like see a better assessment of how the globally reduced plant N demand affects nitrogen limitation of primary production. Would it be possible to quantify by how much primary production is constrained by N limitation, for instance by comparing against simulations with no N limitation (unlimited N availability)? More importantly, why doesn’t the reduced plant N demand decrease BNF? Are rates of BNF independent of soil N availability in LPJmL? Would it be possible to compare simulated BNF rates against literature values?

General comments:
L16 : What does the « absolute » mean? Is it necessary?
L21-24: That reads a bit weird. Maybe think about splitting into two sentences and rephrase the latter part to something like: “Consequently, the representation of the N cycle in models also depends on how these mechanisms are explicitly or implicitly represented.”
L42: Clarify what you mean by “additional variables with substantial changes”?
L61-64: What about other management practices besides irrigation and rainwater harvest? Did they already play a role for carbon stocks and fluxes in version 4?
L69: That refers more specifically to LPJmL v5, as earlier versions did not represent the N cycle?
L72: “as” => “by”
L81. What exactly do you mean by “growing season”? Length and timing (sowing dates)?
L98: “leaf nitrogen concentration” would here be more precisely “leaf nitrogen to carbon ratio”?
L163-171: Could you shortly explain the motivation to move from the formulation by Zaehle et al. to that of Smith et al.? Did Smith et al. demonstrate a better performance for their formulation?
L170: That midpoint is CN_leaf,med in Table 2?
Table 2: All crops have the same parameter values? Which crop types does that refer to? Could you add the definitions of the parameter (column headings) as footnote to the table?
L195-199. How were these changes motivated? Can you give some references for empirical studies showing that for instance leaf-on period is maximum 180 days per year in summer-green forests?
L206-208: The initial value for nfixpot seems very low compared to the new value. Could you briefly mention here how that initial value used in von Bloh et al. was selected? The parameter is unitless?
L208-209: What is the unit of the turnover rates?
L209-2012: What are the units of these parameter values?
Section 3.1: I think the modeling protocol should be part of the Methods and Materials section.
L305-306: Would you say that the reduction in demand also leads to a decrease in N limitation?
L309: Why is there an increase in fire emissions?
L346-349: Were you able to prove that the performance of BNF simulation improved? If yes, how?
L350-351: Could the simulated immobilization rates be compared against synthesis of observations, like Elrys et al., 2022? https://doi.org/10.1111/gcb.16202
L351-355: What are the actual mechanisms that caused the SOC increase in permafrost regions?

Reply

Citation: https://doi.org/10.5194/egusphere-2026-4549-RC2

Jens Heinke, Sibyll Schaphoff, Stephen Björn Wirth, and Christoph Müller

Viewed

Total article views: 508 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
311	154	43	508	43	53

HTML: 311
PDF: 154
XML: 43
Total: 508
BibTeX: 43
EndNote: 53

Views and downloads (calculated since 28 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	107	28	10	145
Oct 2026	184	108	30	322
Nov 2026	20	18	3	41

Cumulative views and downloads (calculated since 28 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	107	28	10	145
Oct 2026	184	108	30	322
Nov 2026	20	18	3	41

Viewed (geographical distribution)

Total article views: 508 (including HTML, PDF, and XML) Thereof 508 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 07 Nov 2026

Short summary

Nitrogen is essential for plant growth, and vegetation models must capture how its availability limits it. We found a long-standing error in one such model, LPJmL, that had overestimated plants' nitrogen needs by about a factor of two. Correcting it removed this limit almost entirely, which required further adjustments to how the model handles nitrogen uptake, recycling, and plant leaf timing, to restore a realistic level of nitrogen limitation and overall model behavior.


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