Using a 2-D spatially explicit model, Rötzer et al. investigate how spatial arrangement of soil particles and quantity/quality of organic matter inputs affect carbon (C) and nitrogen (N) fluxes and storage. The topic is suitable for SOIL and for the special issue “Advances in dynamic soil modelling across scales”. While the model framework is not new, here N is added for the first time, allowing to consider organic matter quality in terms of C:N ratio, and its effects on C allocation by soil microbes (mostly via a C overflow mechanism under N limitation). The methods are appropriate, though their description should be improved (see below). Results are interesting, but my impression is that those related to quantity/quality of inputs are expected (stemming from the model assumptions) and less interesting than those related to soil structure (which I think could be expanded). Language should be improved—see some examples of common issues below. My main concerns are listed first, followed by specific comments.

Main concerns

Language. Section structure should be streamlined and harmonized, as now paragraphs are sometimes indented and sometimes not, several paragraphs contain a single sentence, and use of spaces between paragraphs is inconsistent. This makes it difficult to follow the narrative within each section. Please check citation format, as now it does not always follow standard practice (“Author (year)” for in text citations and “(Author, year)” for citations in brackets), e.g., L38, 174, 185, 286. Several sentences are hard to understand because of missing commas. To give one example, please see L199-202, where I would suggest adding four commas as follows: “Initially, POM particles with a concentration of 0.48 gCcm−3 and a C/N ratio of 100, and necromass with a concentration of 0.32 gCcm−3 and C/N ratio of 10, amounting to a volume fraction of 5% of the solid area, were randomly added to the pore space, corresponding to 60 POM and 60 necromass particles, each with a size of 6-10 μm in diameter.” In many instances word choice is not appropriate (though I am not a native speaker, so my impressions might be wrong), e.g., “Respired” instead of “Respirated” in Figure 1, L133 “outflows” instead of “sinks”, L159 “Monod” instead of “Michaelis-Menten” (kinetics don’t involve enzymes in this model), in some figure “amount” is actually a “content” (mass/mass). In other instances, the meaning of the chosen term is unclear, e.g., L44 “microbial C/N efficiency”, L71 “experimental limitation”, L401 “optimal” (what is optimized?), L558 “agitation”, L570 “exclusive” (why are some pores exclusive?). These are just some examples. A thorough proof-reading is necessary, perhaps with help from a native speaker.

Methods structure. L74-93 present the model structure, so should be moved to the Methods. L260-270 do not present any result, but rather explain how model data is analyzed, so they belong to the Methods.

Model implementation. It is not clear how the differential equations at the core of the model are solved in the hierarchical structure shown in Figure 4. Runge-Kutta method is used to solve the mass balance equations through time within a day, and then spatial fluxes are added, if I understand correctly. But if that is the case, is the solution converging numerically without successive iterations to feed back spatial flux information into the mass balance equations? Is the one-hour time step sufficiently short to ensure stability with an explicit method? Moreover, it is not clear how Eq. A2 is implemented. The inequalities compare partial derivatives on the left-hand side and contents on the right-hand side, which is not physically meaningful (units are different).

Model setup. Some important assumptions make the simulations hard to generalize. First, the soil is assumed saturated for the duration of the simulations, but in such conditions, over 500 days anoxic conditions would develop, leading to rather different processes and controlling factors for C and N dynamics. While perhaps more difficult computationally, considering partly saturated conditions would make simulations more realistic (then one could argue that oxygen is not a limiting factor), and also allow comparing the effects of particle arrangement and pore water content on C and N dynamics. Which one dominates? Second, POM is defined as particles with a size between 6 and 10 microns, while it is operationally defined as particles larger than about 50 microns. I agree that for this kind of numerical experiments, smaller POM particles are appropriate, but I would mention that they are smaller than according to the usual definitions. More important, the C:N of POM is set to 100 g C/g N, which is reasonable for fresh conifer litter (or wood fragments), but much higher than most other plant residues. As a result, the simulated soil is almost always N limited, except at the time of root exudation, which provides inputs with lower C:N. I wonder if it should be the opposite—low organic matter C:N and high root exudate C:N?

Results presentation. Model simulations provide numerical values for fluxes and contents, but the values themselves are in general not particularly important and can be read from the figures. Now the Results section reports several numerical values that, in my view, distract from the main messages—what are the important trends and patterns we should look for in the figures? My suggestion is to streamline the Results section so that the answer to the research question (in this case a general aim rather than a question) emerges more clearly.

Analysis of results. One of the most interesting results is briefly described in L373-383. There the interaction between spatial structure and organic matter stoichiometry emerges. I would expand this analysis beyond a qualitative statement based on visual inspection of a figure, making it quantitative. What can we say in general about these interactions across model runs within a scenario, or across scenarios?

Specific comments

General: typically, mass units referring to a specific element are expressed as e.g., g C or g N, not Cg or Ng

L45: please define CUE

L73, 94: aims are repeated; I would suggest merging these sentences so all aims are in the same place

Figure 1: leaked nitrogen—is it including organic and inorganic N?

Figure 2: following the logic of the manuscript, Figure 2 might actually become Figure 1, as it illustrates the physical environment described by the model. Also, in current Figure 2, high C:N is associated to the rhizosphere, while in the model simulations POM has high C:N and root exudates have generally lower C:N; that is a bit confusing (note also that the labels “max” and “0” on the color bar for C:N are too small and can hardly be read)

L105 (and elsewhere): I wonder if it is necessary to cite a long list of articles adopting a similar approach; perhaps it would be enough to cite the last one, or the one that sets the stage for this work

L106: I would also indicate here that the model simulates a 2-dimensional domain

L110: what does “they” refer to?

L114: if POM is attached, it will become less bio-available, as it is associated to soil minerals

L116: convoluted sentence

L132: dissolved organic or inorganic N, or both?

L134: “opposed” in what sense?

L136: equations on page 8 allow for variability in microbial necromass C:N (two independent equations without constraints on the C:N ratio); with these equations, necromass is homeostatic only if microbial and necromass C:N are exactly the same and do not change through time

L157: convoluted sentence

L172: if I understand correctly, extra-cellular enzymes are assumed not limiting, not “ubiquitous” (i.e., they could be homogeneously distributed but still limiting)

L189: just a curiosity—what happens if two soil particles end up overlapping when randomly placed in the domain?

L224: are these rates expressed per unit root surface or per unit surface in the simulation domain?

L242-243: repetition of paper citations already made earlier

Section 2.4: scenario names are not very intuitive, e.g., “3Pulse” refers to a scenario with 10 pulses. Maybe I would rename them “dynamic”, “static”, “CN”, “10 pulse” without a number to keep it simple (the numbering matters less than the actual treatment you apply in each scenario)

Figure 5 (and elsewhere): panel titles are generally placed on top of their panel, line styles are too similar to be able to distinguish the lines, colors for dissolved substrate and old necromass look the same in the bottom panels

L294: what does “This” refer to?

L326-329: I am not sure why this point regarding microbial biomass nearing equilibrium is presented here when showing results for N fluxes; the same argument holds for C fluxes

L373: not clear what “event” refers to

L437: parameters were not fitted but chosen to obtain reasonable results, which is not conceptually very different

L464: plant residues in agricultural fields are likely much richer in N than assumed here

L471: this sentence states what “equilibrium” means; I am not sure I understand where the argument is going

L475-482: this result is a direct consequence of the model assumptions, so it does not seem particularly insightful

L520: exudate “favorable C:N” does not translate to “exudate diversity”—you could have residues from a single species (no diversity) with very low C:N ratio (favorable C:N)

L521: but structural changes of the magnitude experienced in agricultural fields are not included in this analysis…

L524-527: these last sentences distract from the main message of the manuscript by speculating on future work, while it would be more impactful to conclude with L523, where a key result is summarized (just my personal opinion)

L528: it would be nice if a version of the code to run a basic simulation could be provided

L543: at this point in the text the term “attractive” is not clear (it will be explained later)

L561: what do you mean by “microbial… concentration can spread”? if microbes colonize neighboring cells, they spread, while concentration per se does not spread (it might be just a terminological issue)

This paper presents a very interesting modeling exercise using an agent-based model that couples soil carbon dynamics driven by microbial activity with the dynamic formation of soil structure, resulting from the interplay between soil biological activity and physicochemical properties. Moreover, adding root exudates to the system increases the model's realism, enabling exploration of scenarios under varying conditions, including different root exudate qualities and quantities. The paper is well written, and I enjoyed reading it. And I feel it is a good concept that warrants further exploration. Although I really like the paper, I feel some conceptual aspects are not entirely clear and could have strong implications for the conclusions drawn from this modeling exercise. I will start pointing out the conceptual questions I have for the authors, and later, I will provide some minor comments on specific lines. 

Conceptual questions: 

1. Conceptually, the model describes two main external carbon inputs, root exudates and fresh POM, which are converted to DOC and later to necromass via microbial death. The idea of POM as one of the main "solid" forms of carbon could be a model assumption, but I wonder whether there could be a bit more discussion of what that means. I feel that litter will be the initial C input into the soil, and it will first be partitioned into POM, but it could also contribute to the MAOM (mineral-associated organic matter) as a more recalcitrant form of carbon, or even to the DOC pool. I understand the authors do not explicitly model MAOM, but necromass could be assumed to be the more stable carbon form, whereas POM is the more readily degradable form. But as it is in the model, my understanding is that there is little difference in the degradation rates of necromass and POM. This is very evident from the results, where both pools increase over time, and I fail to see whether the system has a more stable or more easily degradable carbon pool. Also, in the model, POM and necromass are treated similarly (with the same degradation equations and the same first-order degradation rates), with the main difference in stoichiometry. It would be nice if the authors could reflect a bit on these model assumptions and not only describe their choices, but also whether this makes sense mechanistically. I would, for example, expect necromass to degrade far more slowly than POM. I can also imagine fractions of litter or root exudates being sorbed and quickly protected, thereby contributing to a more stable carbon fraction, not only to the POM or DOC pool. Root exudates also contribute to the POM fraction. However, these relationships are not part of the model. It probably isn't necessary to include them and re-run the model, but some explanation in the discussion and methods is important, especially as some model limitations. It could also be interesting to expand a bit on how carbon fractions are represented in different models. The introduction goes directly into modifications of the CAM framework, which is fair, but other frameworks for representing carbon dynamics in soil could at least be briefly mentioned.  
2. I wonder if the authors could explain a little bit about the practical relevance of your scenarios. This would be very important because in some cases, parameters are taken from different studies, and I do not know if they all match the system the authors are trying to simulate. I fully understand that it is not possible to find everything in one study, but an initial explanation of which system (conventional arable land, organic, grassland, etc.) the authors are simulating would help us better understand whether the parameters are realistic and which ones might be a bit of a stretch. This is also important during the discussion. There is mention that the simulations match this and that study, but I am not sure whether the study's conditions match the scenarios.  
3. Along the same lines as the model scenarios, I am really hesitant about how the root exudation process is represented. I understand you will not include different types of exudates, but to me, exudation is a continuous process, not really occurring in single pulses here and there. It is clear that there will be moments when exudation is higher, triggered by environmental conditions or developmental stages, but in the simulations, it seems random—or at least I did not see a justification. I wonder if the authors could add something about this in the methods, such as why this root exudation mode was selected, and maybe discuss a bit more about a more realistic root exudation process and its potential consequences. 
4. In several parts of the paper, there is mention of “gluing spots”, but the authors briefly describe them, I believe, in the discussion as EPS substances, for example. The model, however, does not explicitly represent the production of these substances by microbes or the costs associated with this investment, which will be reflected in the CUE, so I wonder what the authors mean by this conceptually and in the model. 
5. The conclusions are a bit weak, as new ideas are included that, to me, do not necessarily represent the paper's main point. For example, I'm not really sure whether you have explicitly slow-cycling or fast-turnover carbon pools in the model, since they are treated similarly. It could be that I am wrong, so if you can provide a supplementary figure of the decay of both fractions, POM and necromass, that could be interesting. The scale at which you present the two of them in the current figures does not allow for any distinction. Also, the conclusions about management practices in lines around line 520 are correct (“Conversely, in soils that are subject to frequent structural changes—such as agricultural croplands—microbial activity is notably higher due to the regular exposure and decomposition of previously inaccessible organic matter. This increased accessibility enhances microbial biomass formation, leading to higher elevated CO2 emissions”), but I do not see how you can derive them from this study. Unless I am totally wrong, I see your dynamic model as a natural progression of the soil structure rather than a man-made intervention. Is this accurate? This means that, with or without human intervention, the dynamic nature of soil structure leads to greater emissions, so it needs to be accounted for in models. I think this is the main conclusion. I could derive these conclusions if you had a well-mixed model that could simulate the field after tillage, for example.

Minor comments: 

1. Line 10: What does µCT data mean? 
2. Line 354: What do you mean by “favorable conditions”? I assume this is based solely on C concentration because the model has no temperature or moisture dependencies. 
3. Line 524: Maybe delete “elevated”