Dear Referee #1,

Thank you for constructive feedback on our manuscript. Please find our response below with the original comments in plain font and our replies in bold.

Dear Editor and Authors,

Thank you for the opportunity to review this paper.

The authors present and innovative model with a generally clear methodology description,  model evaluation and scenario example. The CIBUSmod is a useful addition to the library of biophysical models. I have some comments to improve the article and model.

In the introduction you discuss the larger global or regional models but what about existing national or sub-national models beyond FABLE?  It would help to discuss these to further justify the need for CIBUSmod.

We have now added a short sentence with reference to some of the existing national-scope models that we deem fall into the same family of models as CIBUSmod:

“In addition, several other national-scope biophysical models have been developed, often with a specific set of research questions in mind (see e.g. van Selm et al., 2023; Duffy et al., 2022; Karlsson and Röös, 2019).”

There are however obviously many more models/papers that could be mentioned here (depending on selection criteria for model types etc.). But, we want to refrain from going into too much detail here as it is beyond our scope here to do an extensive review of existing models and we want to avoid a lengthy introduction.

The “pre-defined initial state” is unclear to me (line 61) , is it only crops area and animal numbers or does it also include e.g., human diets? This could be explained more.

Thank you for highlighting this ambiguity. We revised the first sentence in this paragraph to make it explicit that demand is calculated directly from user-input:

Original:

“CIBUSmod satisfies an exogenous national demand for crop and animal products (including exported products) with domestic agricultural production by distributing crop areas and animals over a number of user-defined sub-national regions.”

Revised:

“CIBUSmod uses user-input on e.g. human diets, population size, and processing conversion efficiencies to calculate national demand for crop and animal products (including exports) and then meets this demand by distributing crop areas and livestock across a number of user-defined sub-national regions.”

We also changed the order of the following sections such that the one on calculating demand comes before the section describing the optimisation procedure that distribute crop areas and livestock. We hope that this makes it clearer that human diets and demand are not variables in the optimisation.

If domestic supply of a by-product doesn’t meet demand it’s imported and if excess is available it goes to the waste management which requires precise balancing (line 128). Does the model have built in flexibility to exchange similar products? For example if domestic supply of rapeseed meal exceeds demand, but imports of soybean meal are needed, could the model exchange soybean meal and rapeseed meal to reduce overall imports and reduce the rapeseed meal going to waste management? If it can’t this seems like a limitation and should be discussed.

You are right in that this is a current limitation of the model, and it is an area where we are currently working on developing the modelling framework. This is already discussed in the discussion section, but to make it clear already here we added a reference to the discussion:

“This approach has some clear limitations, which are further discussed in section 4.”

The approach that will be implemented in future versions is based on including feed rations for all animal categories as variables in the optimisation model to allow those to adjust to available by-products in different scenarios. See Wanecek (2025) cited in the manuscript.

In table two the livestock systems are described, why are cattle considered one animal production system? Dairy cows and suckler cows have very different nutrient requirements and diets. These should be split these into two different systems (same as the poultry system).

We received a similar comment from referee #2. Our answer here is thus formulated also with that comment in mind.

Table 2 is organised according to the modules used to handle different livestock systems. So, dairy herds (dairy breeds) and suckler cow herds (beef breeds) are handled with the same module in the model (CattleHerd), but those are parametrised very differently with regards to e.g. growth rates, cow lactation (only covering calf needs in the suckler systems), replacement rates, etc. Also, the parameters for equations to calculate energy required can be set independently for different breeds. As such, the same underlying architecture can handle both dairy herds and suckler cow herds while accounting for the drastically different feed requirements for dairy vs suckler cows and also the other differences in herd structure (calving intervals, slaughter ages and weights, growth rates, etc.) affecting feed requirements (and production) for the young stock.

We added a sentence just before Table 2 that clarifies this:

“The livestock production modules thus allow for a high degree of flexibility in parametrising different livestock production units (e.g. separate units representing dairy and suckler cow herds with their differences in feed requirements and production).”

We also extended the section on calculating cattle energy requirements:

“For cattle, metabolizable energy intakes are calculated based on the methods presented in Spörndly (2003), which include equations to calculate energy requirements for maintenance, growth, lactation, and gestation that can be parametrised to represent different breeds and production systems.”

I also have concerns basing feed requirements on energy or dry matter alone? What prevents the model (or user) feeding high energy low protein crops to cattle and high dry matter low quality crops to sheep?  How can you ensure that the diet will also meet protein requirements?

The current version of the model relies on the user to supply nutritionally adequate feed rations. This is a limitation and requires that users have some level of animal nutrition competence. This also relates to the development presented in Wanecek (2025), referred to above. We have now highlighted this aspect in the section of the discussion addressing these limitations and describing plans for future development by adding:

“In addition, the current approach requires the user to supply balanced feed rations for the different animal categories with regards to e.g. protein-to-energy ratio and energy density, which requires animal nutrition competence to ensure nutritionally adequate feed rations.”

The N/P/K excretion is fixed for some animals and uses a mass balance for others, the mass balance approach is more accurate and prevents an imbalance between nutrient intake and  excretion. Why can’t you use a mass balance approach for all animal species?

A very relevant question. As you say it would be preferable (not least from a consistency perspective) to use this approach for all species. However, in the current version, the sub-models for the poultry, horse and sheep herds are less detailed than those for the cattle and pig herds. To use the mass balance approach, we need accurate modelling of live weight gains across all animal categories and life stages in the herds. This is done for cattle and pigs and involves detailed accounts of growth rates and mortality for the different animal categorise and at different life stages.

The live weight gains are stored in a dedicated output table for each animal production unit, and the model automatically uses the mass balance approach if these data are available. As such, if the other animal herd modules are developed to include live weight gain calculations this approach would also apply to them.

However, at the time being we have not deemed the additional accuracy and consistency of this approach being worth the additional data requirements. For example, in Sweden cattle and pig manure accounts for around 80% of nitrogen in applied animal manure, while poultry, horses, and sheep together account for the remaining 20%.

We have added a short motivation for this also in the manuscript:

“The mass balance approach (Figure 2) is used for livestock modules that explicitly calculates live weight gains for each animal category. Currently that includes cattle and pigs, which are the two largest contributors to manure excretion in many EU countries (Köninger et al., 2021).”

Other comments:

Line 30: Van Selm et al. 2022 uses a feed allocation model and not the CiFoS model.

Thank you for highlighting this. Our ambition was to cite some of the earlier work on which this model is built. But, we have now replaced Van Selm et al. (2022) with van Zanten et al. (2023) which is the first publication to use the CiFoS name.

Line 109: the python solver package and gurobi solver should be mentioned in the model description. Potentially you could also add a required software sub section to the model description. Also, I believe the gurobi solver requires a licence.  

We are not quite sure what you mean by “in the model description” here. The package and solver are mentioned in the manuscript. In addition, the repository includes a requirements.txt that details all Python packages required to run the model as well as a user guide that explains the installation procedure (including steps needed to acquire and set up a free academic licence for Gurobi). The repository is available at Zenodo (https://doi.org/10.5281/zenodo.15300943) and GitHub (https://github.com/SLU-foodsystems/CIBUSmod).

Line 113: Are the flexible constraints only limited to two or can you add more?

It is possible to add any number of the two flexible constraints, and we have also tried to make it relatively easy to programmatically add other constraints that do not follow the generic format of C8 or C9. We have revised to make this a bit clearer:

“The model also includes two categories of flexible constraints (C8 and C9 in Table 1) of which any number may be included in a model run to manually constrain solutions depending on specific research questions. In addition, it is also possible to programmatically construct additional constraints without changing the main code of the model as long as the problem remains convex.”

Line 248: Is the use of conventional manure in organic systems currently allowed in practice?

Yes, it is allowed under EU organic regulations with the exception of manure originating from “factory farming” (Commission Implementing Regulation (EU) 2021/1165). In Sweden, KRAV (the organisation responsible for developing rules used for most Swedish organic production) specifies this in more detail and e.g. prohibits the use of poultry manure from broiler production where the stocking density is higher than 21 kg/m2, pig manure from slaughter pig herds producing more than 50 pigs per year (except if on deep litter), and manure from caged birds and other animals.

However, in CIBUSmod we do not account for these specifics, but assume that conventional manure can be used on organic cropland if the manure from organic animals is not enough to cover requirements.

We also know that this is the case today as manure application on organic cropland according to national statistics exceeds availability of organic animal manure. This has also been shown for other regions in Europe (see e.g. Vergely et al. (2024) cited in the manuscript for the French case).

Also the description and use of non-animal manure sources is unclear.    

We have now revised the section describing the allocation of non-manure organic fertilisers across crop areas to hopefully make it clearer:

“After accounting for TAN and long-term N release from applied manure, residual plant available N requirements are used to allocate other organic fertilisers (e.g. biogas digestate) across crop areas. The total amount of organic fertiliser available is calculated in the WasteAndCircularity module (see section 2.9). The allocation procedure follows a similar logic as for animal manure: (1) organic fertilisers produced in each region are first applied to organic crops within that region; (2) any surplus is then distributed nationally to organic crop areas; (3) remaining amounts are allocated to conventional crops regionally; and (4) any final surplus is distributed nationally to conventional areas.”

We have also added a sentence in section 2.9 with reference back to this description of the allocation procedure:

“The allocation of generated organic fertilisers is handled in the PlantNutrientMgmt module (see section 2.6.2).”

Line 509: If national emission factors for leaching aren’t available, what about using an emission factor from a country in a comparable climate zone?  

Thank you for this constructive suggestion. We have now changed to use a leaching factor from Finland’s national inventory report (which unlike Sweden uses the IPCC default method). This factor was much lower than the IPCC default value (0.144 compared to 0.24), which resulted in our estimates for indirect N2O emissions from leaching agreeing more closely with the Swedish national inventory report. Therefore, we have now revised the figures (mainly Figure 9 (3.D) and Figure 13 (NO3-)) and text sections that present and discuss these results:

(Lines 526-528) “Estimated indirect nitrous oxide emissions from deposition and leaching are comparable to what is reported in the national inventory. While the national inventory, uses a sophisticated process-based model to estimate leaching, CIBUSmod estimates leaching as a fixed shssare of N inputs, according to IPCC (2019) Tier 1 methodology. A leaching factor of 0.144 kg N leached per kg N input, developed for Finland’s national inventory report, was used across all crops.”

(Lines 676-678) “In its current version, CIBUSmod models nutrient leaching from agricultural land only in terms of N leaching, using a basic method that does not directly account for management practices that reduce leaching. Future work will aim to incorporate improved methods to estimate both N and P leaching and runoff, incorporating regional differences in soil and climate while accounting for different management practices.”

Dear Referee #2,

Thank you for constructive feedback on our manuscript. Please find our response below with the original comments in plain font and our replies in bold.

Dear Editor and Authors,

This is an interesting and very clear paper providing a detailed description of an innovative biophysical agri-food model at national and sub-national (territorial) level. The methodological detail provided in the paper and the open-source nature of the model are very welcome and valuable in terms of fostering transparency and possible appropriation.

I understand that this paper is more methodological in nature and am convinced that the model constitutes a solid and valuable tool to assist in exploring more sustainable food systems. Yet, I feel the authors could expand more on this last point : how can CIBUSmod contribute to advancing food system transitions? Why do we need the model? For whom has it been developed? Is it only to be used by researchers or also by/in collaboration with food system actors (farmers, value chains, policy...)? This seems particularly important since the authors point out that one of the main interests of the model resides in its national/sub-national nature. Maybe this could be expanded a little bit more in the introduction and the discussion.

Thank you for this useful comment. We have now extended the opening paragraph of the introduction to more clearly set this work into the context of foresight studies and scenario analysis:

“Foresight studies and scenario analysis are valuable tools for studying different visions for future food systems, and identify synergies and trade-offs that can guide policy decisions under high levels of uncertainty (Reilly and Willenbockel, 2010; Woodhill et al., 2025). While scenarios can be purely qualitative, presented as narratives of possible futures, they are often combined with computational models to quantitatively assess their implications for key outcome variables (Riera et al., 2025; Reilly and Willenbockel, 2010). To support such analysis, numerous land and food system models have been developed, varying in scope, levels of detail, and complexity.”

We have also added a paragraph in the discussion section that deals with the model’s potential application together with stakeholders and in teaching:

“Because of its relatively high level of detail, CIBUSmod is primarily designed to be used with direct involvement of researchers. However, the ambition is for it to be a valuable resource also in education and collaborative foresight projects where alternative future scenarios are jointly developed and explored. This can be done, for example, through an iterative story-and-simulation process (Volkery et al., 2008; Karlsson et al., 2018), where stakeholders draft scenario narratives that researchers translate into quantitative model inputs. Alternatively, simplified Excel workbooks that contain a limited set of adjustable parameters, along with explanations of what each parameter represents, can be constructed. This enables participants to directly modify inputs and test different scenarios without the need for a comprehensive understanding of the model in its entirety. This latter approach has already been applied in PhD-level education, where student teams created narratives of future food systems and translated them into quantitative parameter changes. The model was then run, and results analysed to see whether outcomes matched expectations. Students were then able to investigate which levers had the strongest effect on key outcome variables as well as cases where their prior assumptions diverged from the model’s behaviour.”

Below a series of more targeted comments throughout the paper.

Line 24 :  Here you could refer to papers and work discussing the general use of foresight and scenario studies for food and farming systems. Some references below:

- de Lattre-Gasquet, M., Rostom, F.Z., Hazoume, T., 2023. Guidance Document on Foresight Processes for Food Systems Transformation through Agroecology. FAO, CIRAD, GIZ.

- Reilly, M., Willenbockel, D., 2010. Managing uncertainty : a review of food system scenario analysis and modelling. Philos. Trans. R. Soc., B 365 (1554), 3049–3063. https://doi.org/10.1098/rstb.2010.0141

- Riera, A., Duluins, O., Antier, C., Baret, PV. 2025. Which types of quantitative foresight scenarios to frame the future of food systems? A review. Agricultural Systems. https://doi.org/10.1016/j.agsy.2025.104270

- Woodhill, J., B. Peters, J. Dengerink, N. De Paula. 2025. Using Foresight for Food Systems Transformation: A guide for policymakers, practitioners and researchers. Foresight4Food and UN Food Systems Coordination Hub.

- (in French) Duru, M., Aubert, P.-M., Couturier, C., Doublet, S., 2021. Scenarios de systemes alimentaires `a l’horizon 2050 au niveau europeen et français : Quels eclairages pour les politiques publiques ? Agronomie, environnement & soci´et´es 11 (1). https://doi.org/10.54800/ssa050.

Thank you for this useful list of references. See response above.

Lines 29-30 : Mention at what scale the models cited here operate. Also at global level?

We now mention the spatial scope of these models; global for BiOBaM-GHG and SOL-m and EU for CiFoS.

Line 36: Not sure of the wording "larger regional-scope". Here regional = supra-national or sub-national?

Revised to “… global-, or larger regional- (e.g. EU) scope …” to make it clear that we refer to supra-national.

Lines 45-46: Split sentence for improved clarity : "To address this, CIBUSmod is introduced. Developed within the Swedish Mistra Food Futures program, it constitutes a relatively light-weight biophysical food system modelling framework designed to incorporate detailed sub-national data and knowledge."

Thank you for this suggestion. We have now revised the text to:

“To address this, CIBUSmod is introduced. It constitutes a relatively light-weight biophysical food systems modelling framework designed to incorporate detailed sub-national data and knowledge.”

Line 55: "user guide"; "focuses“

Revised accordingly.

Lines 64-64: What do you refer to by "different levels of specificity"? Is this only spatial?

Thank you for highlighting this ambiguity. We have now expanded this section slightly to make it clearer that this does not only refer to spatial detail:

“A key module of the model is the ParameterRetriever, which provides a flexible way to input data at different levels of specificity depending on availability, with the model automatically using the most specific data provided. Specificity refers both to spatial detail (i.e. from national average values to region-specific values) and to how parameters are defined across different categories, allowing them to be applied broadly or distinguished with finer resolution. For example, the parameter controlling the share of animal manure handled with different manure management systems could be specified broadly per animal species (e.g. the same distribution for all pigs) or with finer detail per production system and animal category (e.g. a specific distribution for organic sows) if data is available.”

Presentation of model sequence (section 2) :

- While reading the paper, I wondered what the most logical order of presentation would be and whether it would not be clearer to start by presenting section 2.2 (calculation of demand) as this is the start of the calculation sequence. Yet, as other factors such as feed requirements (presented in 2.4) also determine crop areas, changing the order of sections might in fact not be necessary. However, I do think it would be good to insist a little bit more on presenting the general sequence of the calculation. For example, you could specify (at lines lines 58-59 or 119-120) that you focus on the share of the demand that is met by national production. In other words, if I understand correctly, you "recreate" agricultural production departing from national demand, uses and exports. E.g. I assume the model does not estimate and distribute a production area for banana's based on their consumption level in the Swedish diet ? I guess this is dealt with through the constraints from Table 1 (e.g. C7 related to climate suitability) ?

Thank you for your careful reading and constructive suggestion. Following this comment and a similar comment from referee #1 we have revised the first sentence of section 2 to read:

“CIBUSmod uses user-input on e.g. human diets, population size, and processing conversion efficiencies to calculate national demand for crop and animal products (including exports) and then meets this demand by distributing crop areas and livestock across a number of user-defined sub-national regions.”

This hopefully makes it a bit clearer that we first estimate demand for domestic production based on user-inputs and then “(re)-create” the agricultural production system that could meet this demand. We have also reordered section 2.1 and 2.2 such that the description of the module handling the calculation of demand comes first. However, we chose to keep the description of the optimisation step responsible for regional distribution before the description of the other modules (crops, livestock) even though (as you note) it relies on calculations performed in the subsequently described modules.

A minor note: The constraint on climate suitability (C7) is not used to avoid production of e.g. exotic fruits in the Swedish case study presented. Instead, this is handled by specifying a 100% import share for food items that are not produced in Sweden. But the climate suitability constraint applies to e.g. rapeseed and grain legumes which are not cultivated in the colder regions in the north.

- Line 88 : Again, for clarity purposes, I would recall that the that the distribution of crop areas and animals numbers are calculated to meet demand.

We revised the first sentence here to:

“Crop areas and animal numbers needed to meet calculated demand for agricultural products are distributed across regions by the GeoDistributor module”

We hope that this, together with the restructuring described above, is sufficient to make this clear.

Line 102: drop "will penalise these relative deviations, which"

We see that this is not strictly needed here and that dropping it would reduce the complexity of the sentence. However, we think that this provides relevant information on how the scaling power (sp) affects the optimisation goal function, and thereby the resulting solution in terms of land use. Therefore, we have decided to keep this as is.

Table 1: C1 is this constraint applied at national or sub-national level ?

This constraint applies on national level. We have added this to the description in the table.

Table 2: Cattle : Does the model make no distinction between dairy cows and suckler cows? I assume it does but then they should be mentioned separately given the differences in practices (even if calculation methods are the same).

We received a similar comment from referee #1. Our answer here is thus formulated also with that comment in mind.

Table 2 is organised according to the modules used to handle different livestock systems. So, dairy herds (dairy breeds) and suckler cow herds (beef breeds) are handled with the same module in the model (CattleHerd), but those are parametrised very differently with regards to e.g. growth rates, cow lactation (only covering calf needs in the suckler systems), replacement rates, etc. Also, the parameters for equations to calculate energy required can be set independently for different breeds. As such, the same underlying architecture can handle both dairy herds and suckler cow herds while accounting for the drastically different feed requirements for dairy vs suckler cows and also the other differences in herd structure (calving intervals, slaughter ages and weights, growth rates, etc.) affecting feed requirements (and production) for the young stock. 

We added a sentence just before Table 2 that clarifies this:

“The livestock production modules thus allow for a high degree of flexibility in parametrising different livestock production units (e.g. separate units representing dairy and suckler cow herds with their differences in feed requirements and production).”

We also extended the section on calculating cattle energy requirements:

“For cattle metabolizable energy intakes are calculated based on the methods presented in Spörndly (2003), which include equations to calculate energy requirements for maintenance, growth, lactation, and gestation that can be parametrised to represent different breeds and production systems.”

Table 2: Feed requirements of pigs : Add "NE" (= Net Energy?) in the table legend.

Thank you for noticing this. “NE = Net Energy” has been added to the table legend.

Line 127: I would replace "that" by "the demand" as the sentence is otherwise quite unclear.

We agree that this sentence is quite unclear. However, inserting “the demand” there would not be correct. This balancing of dairy fat utilisation is a small (but still fairly important) technical detail in the model. Therefore we want to describe it, but not spend too many lines of text on it. We have now revised this paragraph to:

“For dairy products, the model balances supply and demand for dairy fats. Domestic demand for high-fat products such as butter and cream may exceed the milk fat obtained as a by-product of producing low-fat products. In that case, the model increases low-fat (skimmed) milk powder exports by an amount sufficient to meet the excess demand for dairy fats.”

Lines 248-249: Is it realistic (e.g. in the Swedish context) to consider that manure from conventional systems is first used on organic crops before conventional ones ?

Good question.

Based on national statistics we know that the total animal manure application on organic cropland is around 22,000 tonnes N* plus an additional 4,000 tonnes from non-animal manure organic fertilisers. However, these statistics do not separate between animal manure originating from conventional or organic animals. We do however know the number of animals in organic production and can calculate the manure excretion based on this. With CIBUSmod we calculate that 11,000 tonnes N available to spread is generated from organic animals. Thus, there is a considerable “shortage” of animal manure that needs to come from conventional animals.

Based on this, we designed the rule-based procedure described, which seem to give results that align fairly well with what we can infer from national statistics. So, while this procedure is rough and based on limited data, we consider it fit for purpose. It also ensures that we do not end up with nutrient shortages in the organic production systems as we can’t “top up” with mineral fertiliser application, which is what the model does for the conventionally farmed areas. This also aligns with previous studies that have found considerable flows of manure from conventional to organic areas (see e.g. Vergely et al. (2024) cited in the manuscript for the French case).

* Note: This is higher than what we currently estimate in CIBUSmod, mainly because we underestimate demand for organic crop production (and thereby organic crop areas). This underestimation is likely due to e.g. assuming that cereals used for exports and bioenergy are only conventional and that horses (used for riding) only consume conventional fodder.

Line 279: "are" instead of "or"

Thank you for noticing this typo. We have now corrected this.

Line 361: Do I understand correctly that you estimate demand for organic foods based on the share of organic in production ? Or do the national statistics inform you on the share of organic in consumption ?

Thank you, this was poorly described in the original manuscript. Input data refers to the share of organic in consumption. These shares were mainly sourced from a report that is published yearly by the large actors in the Swedish organic food industry. However, for some foods (especially meat), these shares were adjusted to align organic production with national agricultural statistics for organic production. We have now expanded this description and included a reference to the industry report:

“Organic proportions in consumption for most foods were obtained from Swedish organic food industry statistics (Swedish Organic Farmers Association et al., 2022), and in some cases adjusted to match production data according to national agricultural statistics (The Swedish Board of Agriculture, 2024).”

Lines 443 and 452: Figure 6 instead of Figure 3

Thank you for spotting this. The figure references have now been corrected.

Section 4: In the discussion, in lines 655-661 you discuss the addition of a wider set of output indicators. While you mention "social sustainability" in line 559 and provide some possible examples in the following sentence, I feel you could expand a bit more on the integration and importance of discussing model outcomes in the light of socio-economic aspects, as this is often an important remark (critic) that is made to such biophysical models.

Thank you for raising this point. We have now added a short paragraph in the discussion that highlights that this type of model cannot be used to assess socio-economic effects of scenarios or the policies/socio-economic conditions required to reach different scenarios. But that they may be combined with other methods and models to perform such analysis:

“It is important to note that CIBUSmod is a strictly biophysical mass-flow model. It cannot assess the policies or other socio-economic conditions required to realise a given scenario, nor the socio-economic impacts of that scenario. However, biophysical food system models have previously been combined with economic models to analyse the policies needed to achieve specific outcomes (Röös et al., 2022) – an approach that could also be applied with CIBUSmod.”

Dear Referee #3,

Thank you for constructive feedback on our manuscript. Please find our response below with the original comments in plain font and our replies in bold.

Dear authors, this is a well written and accessible paper on the CIBUS-model – so I do not have major comments – for some smaller points, please see below.

However, I may highlight the following two aspects (for details, see below).

* Please add somewhat more explanation on the basic formulae used to determine land areas

Thank you for raising the lack of clarity in the section describing the optimisation goal function. We have now revised this section (see answer to the specific comments below).

* Please use additional data, where available (as you mention at various places in the manuscript), to calibrate the baseline, or explain why you don’t do so, albeit partly potentially relevant discrepancies between the model baseline and the observations may occur.

Thank you for this comment. Our philosophy in assigning values to the different parameters in the model was to base these on available national statistics or values from peer-reviewed or grey literature. This was an iterative process where we compared a number of key model output variables against available statistics (e.g. cropland areas, nitrogen fertiliser use). If large discrepancies were found, we revisited data sources to see if there were other (better) data available that could reduce these discrepancies.

However, we avoided 'model calibration' in the sense of tuning parameters until the model matched the validation data acceptably. The model includes many parameters that could be used to calibrate certain output variables. For example, the area of ‘Winter wheat’ could be calibrated by adjusting diets, animal feed rations, imports/exports, yields, process conversion factors, waste levels, etc. Therefore, it is a risk that we would calibrate the “wrong” parameter, which would influence scenario analysis (e.g. if we would calibrate baseline cereal areas by increasing waste factors the potential for reducing environmental impacts with waste reduction may be overestimated). Therefore, we deemed it safer to accept and describe the discrepancies remaining after exhausting available data sources, rather than finetuning parameters.

Detailed comments:

Line 90: I do not fully get the rationale of the exact formulation of the terms in the formulae, with the normalization gterm R and the term f, Why, for example, not just defining f_ij = (1/x0_ij)^sp? This would also deliver the minimization of absolute differences for sp=0 and it would deliver the minimization of relative ones for sp=1, I think. Please add some 2-3 sentences to in detail explain the rationale behind the exact choice of this formula.

Line 93: please provide the exact formula for rn_i,j

Line 96: please explain what x0 with a bar on top is

Thank you for these useful comments. We have now revised this section to show the exact formula for rn and added a sentence to motivate why this procedure was used rather than the more straight forward approach that you suggest above:

“This normalisation was introduced to account for the different units of measurement across features (e.g. hectares for crops, square meters for greenhouse crops, and headcounts for animals) and to prevent features with larger numerical values from being given disproportionately higher weight in the model.”

We have also made sure to explicitly state the meaning of each symbol used in each formula.

Line 155: why there are no goats covered? – Goats belong to the animals with larger numbers in many countries and are often reported as are sheep – so why not cover goats? Also when covering horses – for food systems, goats are more relevant than horses in most cases, I guess.

The livestock modules included has been partially informed by the first use case of the model (i.e. Sweden). There, goats are not common while horses, albeit not playing a large role in food production, are important user of agricultural land and other resources. As such, horses are important to consider when modelling the agricultural system. We have added a couple of sentences to explicitly state the modules currently included, and highlight that it is possible to add additional livestock modules if required for a specific use case.

“Currently, the model includes modules for cattle (dairy and beef), pigs, poultry (layers and broilers), sheep, and horses. Additional livestock modules may be developed to represent other species if needed for a specific use case.”

Line 424/425: wouldn’t this be an argument to adjust the feeding ration in the model to better reflect this in the baseline? Or would it be very difficult to adjust this? Based on the description of the model and the assumption of national level feeding rations, I would assume that it should not be challenging to implement such change in assumptions – and given the data, it would also be well based on the available data.

Please see answer to general comment above.

Line 452: underestimation of total cropland in the model. – Shouldn’t this be corrected by default? – Just to have a baseline that is as close as possible to the reported data? Or why did you decide to run the model with such underestimation in the baseline? Please either adjust or add some explanation and please also add which consequences for the results and their interpretation this may have.

Please see answer to general comment above.

Line 506 ff: similar as above, why not using the level of discrepancy to adjust the parameters – this applies also to the N2O emissions from leaching: given that the fixed share of allied N from IPCC 2019 Tier 1 results in too high emissions in comparison to the more sophisticated calculations done in the inventory, this could be used to adjust this emission factor, couldn’t it? Then, the model would be aligned with the baseline and the argument to adjust this factor would be very sound.

Thank you for this suggestion. This was also raised by Referee #1. We have now changed to use a leaching factor from Finland’s national inventory report (which unlike Sweden uses the IPCC default method). This factor was much lower than the IPCC default value (0.144 compared to 0.24), which resulted in our estimates for indirect N2O emissions from leaching agreeing more closely with the Swedish national inventory report. Therefore, we have now revised the figures and text sections that present and discuss these results (also see our answer to Referee #1; https://doi.org/10.5194/egusphere-2025-2625-AC1).