Dear Reviewer,

Thank you very much for taking the time to provide your detailed and insightful feedback on our manuscript. We greatly value your input and are working diligently to address your comments in our revisions.

Improving the Technical Description of the Models & Coupling

Comment: "As the first try, I understand the complexity (and difficulty) of coupling two well-established models, and why the authors chose to use the data files to exchange the information between these two models and the interactions between vegetation biomass and herbivores. However, I still want to see more details of the model mechanisms and processes that drive the vegetation and trophic dynamics through animal foraging."

We acknowledge that providing more details about the processes in both models would enhance the understanding of the coupling mechanism and its effects. We acknowledge that providing more details about the processes in both models would enhance the understanding of the coupling mechanism and its effects. In the revised manuscript, we will improve the model descriptions for both LPJ-GUESS and Madingley:

• For LPJ-GUESS, we will include detailed information about all processes directly or indirectly impacted by herbivory damage. This will also address the questions raised in the third paragraph of your review.
• For Madingley, we will elaborate on the model formulations governing feeding processes, reproduction, and dispersal among grid cells. Additionally, we will clarify how dispersal influences the herbivory feedback, as noted by Reviewer #2.

Comment: "As I can get from the method section, the browsing of herbivores from the Madingley model acts as a disturbance to vegetation leaf biomass. So, I think it would be necessary for the authors to give more details how the demographic and growth processes respond to the reduction of leaves. For example, do they have a “compensation” mechanism to grow new leaves? Are all the trees affected by animal foraging samely (i.e., the same leaf reduction ratio) or these effects are size-dependent? (because the taller trees have low probability of being eaten by animals compared to the seedlings). Also, how is the size structure and regeneration of vegetation are affected?"

Comment: "Lines 98 ~ 122, LPJ-GUESS model, I’d like to see how the demographic processes and cohort dynamics are simulated in LPJ-GUESS, because they can be affected by herbivore foraging."

Following the previous reply,  we will improve the model descriptions and include detailed information about all processes affected by the coupling. We will cover processes capable of compensating damage to leaves and defining how animals forage.

Animals in Madingley do not choose between candidates when selecting autotroph biomass to consume. In fact, animals in Madingley do not even experience individual trees but an aggregated biomass stock from which they can access a fixed fraction.

Comment: "Lines 140~141, why the evergreen leaves have longer damage effects than deciduous leaves? Why is it related to the leaf lifespan? New leave can grow out anyway."

In LPJ-GUESS, leaf regrowth is restricted to a single event per year. Shedding old leaves and initiating new growth is simulated on the first day of each year. For deciduous trees, which drop all their leaves prior to regrowth, the carbon mass of newly regrown leaves is primarily determined by the tree’s productivity and water stress experienced during the preceding year. In contrast, evergreen trees shed only one-third of their leaves annually to accommodate their three-year leaf lifespan. As a result, the carbon mass of regrown leaves in evergreen species is also influenced by the leaf carbon mass from the previous year. This information will be included in the methods section of the revised manuscript.

Comment on Line 238 "“no shifts in vegetation compositions”. Savanna can be affected by browsing. Are the PFTs and different sized trees are uniformly affected by animals?"

Yes, animals have no possibility to select specific individuals or PFTs. They are unaware of neither the PFTs that contributed to the biomass stock they feed upon nor the traits of said PFTs. We will mention this in the revised methods section.

Additional Analysis

In response to your suggestions, we will be reporting on the results of two additional analyses in a revised manuscript:

1. Analysis of the Coupling Effects on the Animal Populations

Comment: “The authors analyzed the vegetation productivity and biomass changes in response to animal foraging. For the trophic model (Madingley), it should have the biomass (or density) of herbivores, omnivores, and carnivores. Is it possible to present their biomass/density as well?”

In response to your comment and similar concerns raised by reviewer #2, we have analysed overall biomass densities for herbivores, omnivores, and carnivores across the entire model domain. In the revised manuscript, we will compare these values at a site level for the “uncoupled,” “one-way coupled,” and “fully coupled” states of our coupling work.

1. Sensitivity Study

Comment: “In the model, there should have at least two types of key parameters, consumption coefficients and regrowth rates, that are used to describe browsing effects and vegetation responses. […] So, I think it needs a comprehensive uncertainty analysis in the mode-data evaluation with a ensemble runs at different key parameters."

Preferential feeding is an active area of development in Madingley and the LPJ-GUESS/Madingley coupling but is beyond the scope of this manuscript.

Comment: "Section 2.5 Study setup. I think it would be helpful if the authors conduct a set of ensemble runs with different parameter values at site level and then report the changes in the biomass at different trophic levels. The large spatial scale runs cannot give such details."

Comment on Lines 229~231: " Also, I think it is extremely high for the browsing effects on boreal forests. That is why we need to conduct site-level runs with different parameter values. Or, the setting of eatable evergreen leaves needs to be examined carefully."

We will perform a sensitivity analysis of key parameters across different simulation sites. While conducting an ensemble of these simulations and identifying all relevant parameters is a time-intensive process, we are committed to investigating their effects at the site level.

In response to your suggestion, to address limitations in the model system, we will add a subsection at the end of the Methods   section in the revised manuscript. This will include discussions of the concerns you raised, specifically regarding browsing vs. grazing and leaf accessibility based on vertical reach.

We will also link the changes made to address your remaining comments to the specific lines of the manuscript to ensure that all of those concerns are thoroughly addressed.

Thank you again for your thoughtful and constructive feedback. If you have any additional comments or concerns, please do not hesitate to reach out to us.

Dear Reviewer,

Thank you for your thorough and constructive review of our manuscript. We greatly appreciate the time and effort you have invested in providing very detailed feedback. We are committed to addressing your comments comprehensively and have outlined our planned changes in the revised manuscript as follows:

1. Coupling Design and Spatial Explicitness

Comment: "First, in contrast to the earlier work (Krause et al.), it is important in the coupled version that the processes simulated in Madingley are appropriately fed forward and fed back. Now, Madingley has things such as animal movement, but LPJ-GUESS is running on a 0.5° (lat, lon) grid, and hence spatially explicit effects from Madingley are neither driven correctly by LPJ-GUESS, nor can the feedbacks be treated appropriately in LPJ-GUESS (this is not a criticism, but just a fact). How important are these processes in Madingley, and what are the implications (possible bias) when they are ignored or misrepresented in the present model application? In other words, how can the authors assure that the two-way coupling is not leading to ‘pathological’ feedback that leads to bias in the simulation results? For example, the huge impact of moose in boreal forests appears grossly exaggerated."

and Comment on Line 153: "I am not sure that the coupling between the two models is done in a correct manner, from a systems theoretical point of view. It appears that both models are discrete-time models, and hence the state of L(t) [LPJ-GUESS at time t] should influence the state of M(t+1) [Madingley at the next time step], and NOT M(t), and vice versa. I am fully aware that reality is different, because reality does not know about a time step. Yet, model formalisms are clear and unequivocal (cf. Zeigler 1976, _Theory of modelling and simulation_. Wiley)."

We appreciate your thoughtful concern regarding the coupling order from a systems-theoretical perspective and your observations on the spatial explicitness of each model.

The primary aim of the coupling was to implement herbivory as a recurring defoliation event within LPJ-GUESS and to explore the impacts this has on vegetation growth and carbon cycling. To achieve this, we designed the coupling to treat Madingley as an external herbivory module. This module operates on Madingley’s monthly time step, requiring LPJ-GUESS to provide the monthly average vegetation biomass by running the simulations for the corresponding month in advance. At the end of each month, herbivores damage the vegetation, with animal movement modelled afterwards. Animal Movement is handled after the damage is passed to LPJ-GUESS. This sequencing ensures that damage remains spatially explicit and does not affect adjacent grid cells  .

We recognise the need to explain the coupling loop and simulation setups more clearly - especially with Madingley and LPJ-GUESS using the same grid resolution. In the revised manuscript, we will expand the Methods section to include a more detailed description and carefully revise the technical illustration to enhance clarity.

We agree that the impact of our coupling is likely overexaggerated in the most northern regions of the model domain. This is based on evergreen trees suffering more from herbivory than deciduous trees and a direct consequence of the methodology used to implement damage to the PFT’s leaf carbon mass. We will elaborate on this in the revised methods section.

Comment on Line 68-82: "I have elaborated on the scaling problem in the coupling between LPJ-GUESS and Madingely above. As a matter of fact, the issue is hidden rather well in the manuscript. We learn only much later (l. 191) that LPJ-GUESS is running on a 0.5° grid. I think the authors should be more explicit and exhaustive in the technical description here, and address the scaling problem."

Madingley and LPJ-GUESS both operate at a grid resolution of 0.5°. In the coupled simulation, Madingley generates the grid list, which is then passed to LPJ-GUESS at the start of the run. We acknowledge the need to clarify the spatial explicitness of both model domains, and we will address this in the revised manuscript.

Comment on Line  99-107: "The description of this (very complicated) spin-up system is not explained well here, and it is not congruent with Fig. 1, neither in terms of the terminology nor the time scales (e.g., a 10,000-yr period is mentioned in the text, but cannot be found in Fig. 1). This must be improved."

We will revisit the technical illustration to reflect the described spin-up phases better.

Comment on Line 118-120: "It seems a very poor assumption that all carbon consumed by herbivores is returned to the litter pool immediately. A more reasonable assumption would be that 90% of the C is respired by the herbivores, and only 10% is returned to the litter pool. What is the substantiation for this entirely unrealistic assumption (which may have considerable implications for the simulation results)?"

We chose to transfer all carbon mass to the litter directly because a detailed tracking of the carbon and nitrogen cycle is currently the subject of another development branch. However, we acknowledge the importance of assessing the effects of this implementation. As suggested by Reviewer #1, we will conduct a sensitivity analysis of key model parameters at the site level. This analysis will include a comparison between our current assumption for the litter pool and an alternative assumption similar to yours.

Comment on Line 134: "At this point, we have no idea what a “grid cell” is (cf. further above). Be more explicit and upfront with these (potentially important!) things."

This will be included in the methods section of the revised manuscript.

Comment on Line 134-153: "Most endothermic animals (such as moose in boreal forests) do not have access to most leaf biomass, because the point of trees is to grow tall (competition for light), which removes a lot of the ecosystem leaf area from the reach of ungulates. It appears that this is ignored when assuming that very sizeable fractions of leaf area are accessible to herbivores, such as one third of all evergreen leaf biomass. What are the effects of this (totally unrealistic) assumption? Could this be the cause that the model produces excessively high reductions of leaf area particularly for boreal forests? However, such effects should also be visible elsewhere in the simulated domain. This needs elaboration, explanation and clarification."

Madingley generally favours large animals for a multitude of reasons. The lack of a vertical structure – or generally simplified accessibility of leaves - is a clear contributor to this advantage. While we recognize this as a key limitation of the Madingley model, implementing a vertical structure for both the vegetation stocks and animal cohorts is well beyond the scope of this study. Nevertheless, we will add a detailed paragraph covering this limitation of the Madingley model in the revised manuscript.

1. Coupling Stages Comparison and Impacts on Animal Biomass Densities

Comment: "Third, although the paper is about the coupling of LPJ-GUESS with Madingley, not a single result from that latter model is shown. All we are presented with are results of LPJ-GUESS. However, in a coupled system it would be important to learn about both counterparts. The manuscript would gain a lot if also the animal side of things was presented for the European and African domains, and put into context with the available literature on tropic chains in the different biomes/ecosystems. This would also be an important element to learn about and better understand the results of LPJ-GUESS (cf. the moose example further above – is it true that Madingley is simulating large amounts of _this_ Animal Functional Type indeed?)"

and Comment on Line 250: "Figure 2 is occupying a lot of space, but in the text very little is made out of it. Perhaps instead of showing all these details without really picking them up in the text, omit part of it and show Madingley results (cf. general comments and further above)?"

We will include a comparison between the uncoupled Madingley, the one-way coupled Madingley & LPJ-GUESS (Krause et al. 2022), and the fully coupled Madingley & LPJ-GUESS. Additionally, we have created domain-wide maps showing the coupling’s effects on herbivore, omnivore, and carnivore biomass densities. This enhancement aligns with the recommendations of both you and Reviewer #1. Madingley provided only animal biomass & abundance densities as output in all simulations. Detailed cohort output for a model domain of this size would not only require multiple terabytes of storage but would also slow down the simulation significantly. However, we will include detailed animal output in the sensitivity analysis described in the upcoming section:

1. Sensitivity Analysis

In response to feedback from Reviewer #1, we will also perform a sensitivity analysis of key parameters at site level. This will provide a deeper understanding of how variations in these parameters influence model outcomes. However, identifying the relevant parameters, as well as running multiple ensembles of the simulations at different sites is a time-intensive process.

1. Manuscript Revision

Comment: "Second, it is not principally a problem that LPJ-GUESS is simulating potential natural vegetation, whereas most data sources used to benchmark the uncoupled and coupled models are from managed ecosystems. If a clear expectation was brought forward how the models should deviate from the data (also for which region – savannas and boreal forests may be subject to different expectations, for example), this would be quite helpful for the assessment. However, the manuscript does not actually contain a clear set of research questions at the end of the Introduction, but we are learning about the goals of the paper much later on (e.g. lines 287, 303, 359-360)."

To enhance clarity, we will define the research question earlier in the manuscript.

Comment: "Fourth (and least importantly), the writing of the paper needs to be improved considerably. I have already noted the absence of research questions (or hypotheses, or goals). Also, the organization of the text, particularly in the Results and Discussion sections, is not satisfactory. Paragraphs contain multiple topics, and the reasoning is going back and forth. This makes for very rough reading at least in portions of the manuscript. Lastly, the manuscript needs careful checking for grammar mistakes and general “elegance” of writing. It is a rough piece at the moment."

We will thoroughly review the manuscript to identify and correct formal errors and reorganise paragraphs to improve readability. We will also restructure the text parts and figures of the results section.

Comment on Line 93-98: "This is technical jargon that is very hard to understand for “outsiders”; for example, why is the ability to switch the “temporal and spatial grid cell handling loop” (what on Earth is that??) necessary? I was totally lost with sentences like this one."

This will be a cornerstone of revisiting the manuscript and improving readability. We will try to avoid technical jargon and make the text accessible to a broader audience.

Comment on Line 361: "I would not combine limitations with the outlook (Conclusion?) of the manuscript. Rather have a section that deals with methodological limitations, and then add a section Conclusions that brings home the major messages of the paper (based on the research questions, which currently are lacking)."

Limitations will be included in a subsection of the methods in the revised manuscript.

1. Technical Comments

When submitting the revised manuscript, we will provide a detailed list of all changes made in response to your remaining specific comments (referenced by line number) to ensure each of those comments is thoroughly addressed.

Thank you again for your valuable feedback, which greatly helps us improve the quality and clarity of our work. Please feel free to reach out if you have additional comments or suggestions.