The manuscript entitled “Underestimation of Anaerobic Decomposition Rates in Sphagnum Litterbag Experiments by the Holocene Peatland Model Depends on Initial Leaching Losses” by Teickner et al tested decomposition rates from the Holocene Peat Model against litterbags decomposition experiments.

The approach is interesting as it combines litterbag data and a Bayesian approach to improve the parametrisation of a model. Improving the use of litterbag data into peat decomposition models is timely and useful to further advance peat decomposition models. The manuscript is well structured. Unfortunately, major concerns regarding the scope, methods and some assumptions should be addressed before considering this manuscript for publication.

General comments:

The motivation, overall approach and scope of the study are not clearly presented. Was the purpose of the study to identify an improved range of values for Wopt and c2?

It is obvious that some interesting experiment has been done. Unfortunately, it is extremely difficult to judge from the current version of the paper if it is sound and whether is constitutes a significant contribution to the field.

Also, it would be helpful to clarify what the difference is, in terms of contribution, between this paper and “A Synthesis of Sphagnum Litterbag Experiments: Initial Leaching Losses Bias Decomposition Rate Estimates” by Teickner et al., as the conclusions seem quite similar.

The experimental design and methods need to be well justified. The experimental design includes a series of assumptions that need to be justified in more detail, so that the reviewer can assess of the quality and soundness of the study. Please see specific comments for details.

Methods should include more detail to allow the reader to reproduce the experiment. Some areas of the methods are well covered. However, it remains unclear what has been done with the litterbag data and with the model. More detail on the models, including HPM, as well as the Bayesian tools used are needed.  

The experiment should be presented as a stand-alone in the paper. Information from Teickner et al. 2024 and its supplementary material was useful to get some sense of what has been done here but it should not be necessary to read another paper to understand this study. If the paper draws from the other paper’s conclusions, then maybe summarize the previous conclusions and explain the link and contribution of this paper.

Assessing the validity of the decomposition functions of the model is interesting. It does not constitute a test of the validity of the whole model. There are therefore limitations around the conclusions that can be drawn from this experiment. The limitations of the study should be clearly presented.

Terminology: some technical terms are not consistently or not appropriately used throughout the paper, which brings confusion: e.g. model, prediction, equation, confidence level. Please check usage throughout the text. Also, some statements mention that the decomposition module was tested in this study and some statements indicate that this is not what was done here.

Readability: In many places, the statements are difficult to follow, which makes the paper hard to read. For example, sentence L4-6 should be rephrased to improve clarity.

Conclusions: The gradient from oxic to anoxic conditions in HPM is designed for long-term peat decomposition within a dynamic simulation. Is it therefore possible to draw conclusions on its suitability from this experiment?

Specific comments:

L1: The title needs to be revised as the litterbag experiments were not conducted with the HPM per se. Also, it might be worth considering a title in line with the general scope of the study.

L35: Could you please justify and explain this approach? Considering only a section of the model does not seem to be an adequate method to test the validity of a model.

L73: Could you please give more detail for clarification: Are the decomposition rates predicted by the HPM obtained from HPM simulations?

L81: What kind of information does that deliver? Identifying new estimates to be used in future HPM versions would seem a better aim?

L92: Could you be more specific? It seems that HPM’s decomposition module was not tested here.

L115: It is unclear how this can be done while excluding the interactions with the other modules of the model. Could you please give more details?

L139: Would it not be more appropriate to reject this study as it brings a lot of uncertainty in the results?

L146: Could you please explain why this is needed and how this differs from the approach taken in HPM?

L158: It is unclear why 2 PFTs were assigned. Could you please give more details on your approach?

L160: Could you please specify what is included or modified in the model versions? How are these model versions linked to the decomposition equations from Frolking et al. 2010?

L169: Could you please explain how they are combined?

L203: Could you please explain why uncertainties in water table depths can be considered negligible?

L208: Model descriptions need more details. A separate section could be appropriate.

L254: Table 3, inaccuracy: different versions of the decomposition module combined with other tools. Also, could you give details as to where details can be found on the 53 litterbag experiments?

Figure 1: 0.3 -0.4 values seem high. Could they be related to S. angustifolium or other species that are seen in a wide range of habitats and might not always be typical of hummocks?

L289: Figure 2, water table depth are more than 25 cm above litterbag in some cases. Could you please explain how it can be the case?

L350: Was this done within this study? If so, could you please add details in the methods section?

L396: Cannot be estimated: do you mean in this study?

L407 and L443: please refer to more recent versions of HPM

L443: HPM was first published (Frolking et al. 2010) with an annual water balance calculation. This has been modified to a monthly water balance calculation a few years later, e.g. Treat’s HPM-Arctic . Please consider looking at the latest available HPM code to ensure your conclusions are in line with the current state of development of the model.

L428: For each site-specific simulation, c2 is adjusted to better represent site-specific conditions. Which value of c2 do you refer to when making this statement?

L467: How does S12 support this statement?

L477: As mentioned above S. angustifolium might not be appropriately classified here. However, this brings to light that a more detailed classification could be useful to avoid misinterpretation of the HPM PFTs.

L494: Would some of the results not be beneficial to other users or help enhance wider knowledge?

Technical comments

L3: missing blank space

L4-6: Long and complicated sentence, would benefit from being rephrased.

L174-182: Some sentences are unclear.

L214: typo that

L223: typo from

L254: model behaviour of HPM: unclear, could you rephrase?

L370: references seem to be in the wrong place

L397: Figure7, confidence interval

L504: The last sentence is unclear. Could you please rephrase it?

S1: model equations: are they not rather sample distributions?

This manuscript describes a study in which Sphagnum moss litterbag data was used to evaluate whether the decomposition module of the Holocene Peatland Model (HPM) can predict correctly Sphagnum litter decomposition rates. In between the experimental data and HPM, there was a litterbag model (hereafter I'll use my own acronym LBM) that was used to estimate decomposition rates from the data. The LBM is described by the authors in another manuscript that is also under review in Egusphere.

Several different versions of the model test set-up were used. Firstly, decomposition rates from HPM ran with the standard parameter values were compared against decomposition rates from standard LBM. In addition to this, the decomposition rates were estimated with Bayesian statistics by letting different parameter sets of HPM vary. The authors found that the standard HPM underestimates anaerobic decomposition rates for many moss species and predicts too steep decrease of decomposition rates from oxic to anoxic conditions.

The topic is useful and relevant, and the authors seem to have done lots of work, but the methods and materials should be described in more detail and clarifications are needed in the manuscript. In my opinion, in its present form, it is difficult to follow the text, and some information is missing - although I have to admit I'm not especially familiar with the Bayesian methods. In particular, I find the role of the LBM and the Bayesian modelling unclear. Below are my specific comments and questions. I focused on the Methods section since I think it is important to modify it first.

1) Abstract

- The abstract text does not mention the LBM (=the litterbag model) at all

- l. 3: Please correct "conditionsand"

- l. 13-14: Is this a correct sentence?

2) Introduction

- l. 29-30: An unclear sentence (do the equations cause observed discrepancies less reliably, or do the tests identify things less reliably?)

- l. 50: Please clarify what you mean with "HPM uses litterbag data". Were the standard parameters derived from litterbag data?

- l. 66-68: Are you referring to your own LBM, or are there many models compatible with HPM? Please clarify.

- l. 87-88: In my opinion, you need to tell here, and especially in the Methods section, more about your LBM. It remains unclear what kind of it is and what is its purpose. Why is a separate model needed to derive the decomposition rates from the litterbag data? 

- l. 89: Please define what is "initial leaching loss".

- l. 94-95: What do you mean by "Our test": this current work or an earlier study?

Table 1: Can you please add here also the other parameters that you are optimizing, alpha and l₀. It would make it easier to follow the text, if all the parameters could be checked from here.

3) Methods

-It would be great to have a schematic picture that illustrates the relationship between the different models and data that you use. If possible, it could include some more general information of the HPM. There apparently are also other parts in the model, as you focus only on a small part of it (the decomposition module). Are there other PFT's than Sphagnum?

- Please add a section describing the LBM, there's no information about it now.

- l. 99-105: Please divide this sentence into smaller pieces, it is difficult to read. I do not completely understand what you did and for what you used only the data of the latter reference list. Did you compare the results to the same data into which you had fitted the models?

- l. 109-111: Is this now the same list of input as in the Introduction l. 38-39? Porosity is missing here. Also, is "the fraction of mass already lost" the same as "initial leaching loss"? 

- l. 112-115: Please divide this sentence into two, it's slightly complicated.

- l. 115-116: The logic of this sentence is not clear. Do you mean that one can't directly use litterbag data as a comparison for the HPM, because HPM doesn't predict masses but decomposition rates?

Section 2.2.2: Please report how many replicates/data points there were for each species, niches, etc. It is relevant information but not clearly mentioned anywhere.

- l. 145: Why did you estimate k_i,0 only in these two set-ups?

- Section 2.2.3: Where was this information (degree of saturation) needed, was it input for the HPM? 

- l. 154: Does this mean you set the mass loss to zero in the model?

- l. 163: "Minimum water content at the surface" appears here for the first time. Is it an additional input parameter for HPM?

- l. 166-168: So how did you determine the litterbag decomposition rates in this case?

- l. 169-173: I was trying to think how to name the model versions so that one doesn't need to always check the differences from the table. Perhaps it would help if you explained here why you named the models like this. E.g. where does the LE part come from? 

- l. 172: Please show the formulas you used for l₀.

- l. 174-178: Please explain this part more thoroughly. What did the Bayesian model do exactly? I assume you optimized different parameter sets of HPM, using the default parameters of HPM as priors. But what was the role of the LBM here? Did you simultaneously optimize parameters of the LBM, and if yes, which parameters were they? Why did you assume that the default parameters of HPM are better priors than something based on the litterbag data? Could you also have used some standard parameters of the LBM as the starting point?

- l. 180: How did you obtain the uncertainties?

- l. 182-183: How do you estimate the uncertain peat properties? Do you mean you assumed the decomposition rates from the LBM are correct and the HPM parameters are correct, and then analyzed what kind of peat properties were needed to achieve these?

- l. 184: Only parameters or also the peat properties?

- l. 189-190: I would have thought this was vice versa: weren't the leaching losses estimated more independently in the l₀ version?

- l. 200-201: The message of this sentence is unclear.

- l. 202-203: Is it so that the Granberg model was not included in the Bayesian system, it had some fixed parameters? Is the minimum peat water content at the surface is from Granberg?

- l. 209-210: Again I'm a little bit lost with what the Bayesian model did. Can you please explain if the estimates from HPM and LBM were somehow optimized together, or if they were separate. I suppose that estimation of the decomposition rates from the litterbag data was done with the LBM?

- l. 211: Should it be "large probability indicates..."?

- l. 218-219: Please define what is a one-pool decomposition rate? What pool?

- l. 224-225: Is it possible to re-formulate this. E.g. it might be clearer to say that in the folds, you included only data of those species for which there was data from several sites.

- l. 232: This kind of prior knowledge sounds relevant - were there many this kind of restrictions? Perhaps they need to be listed.

- l. 239: Where was this priorsense package?

- l. 250 and Table 3: To me it looks like the RMSE_test is smaller for HPMe-LE-peat_l0.

4) Results

- l. 257: Didn't you also adjust the peat parameters in HPMf-LE-peat?

- l. 263-264: Where are the estimates of initial leaching losses needed?

- l. 271-272: Could this result be affected by the priors? Also, to me it seems (Fig. 2) that especially for S. angustifolium, these estimates are very similar - but perhaps I misunderstood the plot.

- l. 258: Relations of l₀ and what?

- Figure 2: I don't understand why the "HPM=No" values are different for different model versions, if they were not predicted by the different versions of HPM but just estimated from the litterbag data. What are the error bars?

- Figure 3: Why especially HPMf vs. the LBM is on the 1:1 line although these were not from the same Bayesian model? Why weren't the Hagemann and Moroni estimates tested against the HPM? What are the error bars?

- Figure 4b: Are these results for one species or some kind of summary?

5) Discussion

- l. 315-316: How significant is it to estimate the initial leaching loss, considering the long-term HPM results about peat accumulation?

- l. 340: I would think it is possible that a more accurate fitting of these parameters to litterbag data is not the only key to better understanding of the C accumulation, but there are other factors affecting the processes and differences between different litterbag experiments.

- l. 343: How do you conclude this? W_opt was related only to aerobic respiration (Table 1).

- l. 344 and 347: there's perhaps a bit repetition. 

- l. 351-354: This explanation is complicated. Can you please write it more clearly. How many different runs did you do exactly?

- Figure 6: Please explain the confidence levels.

- l. 370: What do you mean by the "same study"?

One question about Supporting information. You write:

"The litterbag decomposition model combined with each modification of the HPM is obtained

from Teickner et al. (2024) (model 1-4, see the supporting information to Teickner et al.

(2024) for details). Here, we describe the modules which were added to this model in the

different modifications of the HPM"

- Do you mean that you added some modules in this study, which you didn't have in the other work? Please add also this piece of information in the main text, when describing the LBM.

General comments

The authors tested whether the Holocene Peatland Model (HPM) can predict Sphagnum decomposition rates from litterbag experiments. They used a series of modifications of the model’s decomposition module in combination with Bayesian methods to get posterior distributions of the model parameters. They focused on the decomposition module of the HPM as a strategy to reduce uncertainty and get more precise model predictions. As much as this study has relevance for the advancement of peatland biogeochemistry modelling, major restructuring of the manuscript should be done before publication.

Generally speaking, the writing needs to be improved before this study is published. Sentence construction is often confusing. Furthermore, sentences often do not connect well with each other, creating confusing paragraphs. This hinders the quality of the study and makes it hard to understand, which from the perspective of manuscript-revising is counter-productive. As a general recommendation, try writing shorter sentences with concise ideas (a lot of the sentences here could be split into 2 or even 3 sentences).

I don’t find that the title reflects the contents of the paper. In the title they state that initial leaching losses are important to determine decomposition rates. However, in the discussion, this is only brought up briefly in the last paragraph. Most of the discussion revolves around the quality of the litterbag studies, and water table depth parameters c2 and Wopt. Thus, it is hard to see how they came to this conclusion, when the focus of the manuscript is not there. Additionally, it’s not clear that HPM underestimates decomposition rates. They should either change the title accordingly, or restructure the manuscript to make sure their point is coming across.

The abstract does not provide a complete summary of the study. Generally, this section lacks a clear definition of what are the original contributions of this work and what are precedents from literature. There should be a description of the aims of the work and methods in this section. I suggest that you structure this section the same way the paper is structured, covering the introduction, methods, results and discussion sections in the abstract.

It is important that the authors clarify what is the difference between the current work and another work currently under revision (“A Synthesis of Sphagnum litterbag Experiments: Initial Leaching Losses Bias Decomposition Rate Estimates”). The study is only briefly described in the Introduction section, but there are big interactions between that study and the current one. I suggest giving more details about that previous study in the introduction, methods and discussion sections. This will help the reader understand what makes the two studies different, how they were combined, and ultimately what are the original contributions of the current study.

One of the main confusions comes from the so-called HPM module and the decomposition model. They are both introduced in the Methods section, but it is not clear what the distinction between them is. Formulas for both models should be defined clearly in this section. And if they’re reported on the supplementary material, instead, that should be stated earlier (not 205 lines into the text). Also, since the HPM module is in itself a decomposition model, it is confusing that the other model is just called decomposition model and not something else to distinguish them. Finally, there should be a clear description of the combination of both models. As it is now, the study is not very reproducible from the information presented in the manuscript. The information on the supplementary material (S1 and S2) is full of parameters and formulas, everything written with initials and numbers, but most of them lack any explanation or description of what they are.

Just an idea, the names of the models are complicated to read and remember. So, I suggest you give them more friendly names like HPM-standard, HPM-peat, HPM-all HPM-leaching, and HPM-outlier, or maybe just use less initials. This is only a suggestion, but I encourage you to do it if you really want readers to remember the differences between each model without having to decipher long combinations of initials.

Another main point is that throughout the introduction, results and discussion sections they mention many times aerobic/anaerobic or oxic/anoxic decomposition rates. However, it is not declared in the Methods section how this is considered in the models. From context, I figured out it has to do with water table depth, but specifically how this is incorporated in calculations is not defined. Please, make sure this is clear since it seems to be a big point for your study.

The Results section could be improved. As it is now, it’s not very well structured, there are too many very specific sub-sections. Also, results in the figures are discussed in many sections without a clear order. So first, I suggest making fewer sub-sections. Second, try making sub-sections that correspond to specific figures or tables. This will make it easier to read and to find the data in the manuscript.

Specific comments

Abstract

Line 4: This sentence is not clear. I suggest rephrasing it, maybe try splitting it into two shorter sentences. Also, how are the “large uncertainties” allowing the fitting of the model?

Line 14: The sentence starting with “Based on previous…” is not clear. Who did the previous analysis, the same group of researchers or others? Are the updated parameter estimates part of the results of this work?

Introduction

Line 19: Is “litter that does not decompose fast even under more conditions facilitating microbial decomposition” supposed to refer to litter quality/traits? I suggest rephrasing this part, since the English is not quite correct, and the scientific concepts are not quite clear.

Line 29: The “first” problem as it is written is not clear. Who are “they” (citation?)? Maybe flip the order of the sentence to put the actual problem of the model in the focus.

Line 41: Could you explain why this approach is only useful for short term predictions?

Line 48: How is this proposal different from what was done before as mentioned in line 45?

Line 62: This sounds very similar to the scope of the current study. How are they different?

Line 68: What are the 5 parameters?

Line 84: Both hypotheses here seem more like predictions.

Methods

Line 99: Could you specify if the samples were buried or on the surface?

Line 105: What were the criteria for including these studies in your work? Is this an exhaustive list? It’s not clear in which part of the study did you include the papers with water table depth data and in which part did you include the rest of the studies. How and for what purpose did you test one set of k0 against the other set?

Line 109: when you speak of “module”, is this a part of the HPM? If so, elaborate on the model and its so-called modules.

Line 128: what does PFT stand for?

Line 136: can you elaborate on the criteria of species classification? How was this criterion decided upon?

Line 140: hummock species decompose slowly, I suppose? If assumptions like this are made, they need to be stated clearly and with citations to back them up.

Line 145: here you bring up for the first time the names of two models, but you have not introduced them yet in the text. Maybe consider describing the models first.

Line 147: Describe the Granberg model’s formula and parameters.

Line 163 and 168: what is the “decomposition model” here if not the HPM module?

Line 210: decomposition rates predicted by HPM correspond to each version of the HPM model or just the standard one? And how exactly were the rates estimated from the litterbag data if not from the HPM model?

Line 211: Shouldn’t a high probability indicate a misfit? If the difference is not significatively different from 0, then doesn’t it mean that the HPM predicts well decomposition rates from the litterbag data?

Line 230: In the supplementary information S3 you mention R was used, but in this section, you only mention Stan. Please, give a detailed description of the software used, including which methods were done with each software.

Line 239: For disclosure, I am familiar with Bayesian Methods, but not with cross-validation and power-scaling.

Results

Figure 1: What type of error is used in this figure? Also, shouldn’t the values predicted from litterbag data always be the same?

Line 252: You say that all k were undersestimated for this species but in the Figure the points not predicted with HPM are present only in one of the graphs. Why is that?

Line 263: why is it incompatible?

Figure 2: Why are the values for the k not predicted from HPM different in every frame? Shouldn’t they be constant for each species and have only the values from the HPM vary with each model version? Also, what do negative and positive values of water table depth mean if the are relative to the litterbag?

Figure 3: How do the data points in the litterbag model vs HPMf plots exactly show a linear relationship? From this it does not seem that the more complex models make better predictions. Also, what is the meaning of the “-“ in “(mass-%)”?

Line 274: But in Fig. 3 you see that the best predictions are achieved with the HPMf version?

Line 278: this is not quite visible in Fig. 2.

Line 280: none of the results discussed in this paragraph are shown in Fig. 2 or S3.

Figure 4: it is not possible to differentiate between the species in panel a. Use different colors. Remind the reader in the figure caption what are the other HPM parameters. Apply the same recommendations for the equivalent figures in the Supplementary material, since they are similar.

Line 299: Could this be because the standard value is an average for all hummock species?

Line 302: Please, rephrase this, it’s not clear what you mean.

Figures 5, 6 and 7 are not mentioned in the results section whatsoever. The results therein should be described and discussed just as with the rest of the figures and tables.

Discussion

Line 314: I would not say this is a result of your study. This is maybe a consequence of the results in the study. How would you describe these litterbag experiments that allow to estimate initial leaching losses more accurately?

Line 320: is the steep gradient in decomposition rates really a sign of bad model fit or could it be just that decomposition changes with environmental factors such as water table depth?

Line 349: This should be in the Results section.

Lines 384-391: This should be in the Results section.

Line 433: Since you’re making inferences based on another paper (Quillet et al., 2013a), please give more details about this study.

Line 438: I don’t understand how larger anaerobic decomposition may result in higher water table levels.

Line 482: I believe this is the first time you have brought up the leaching argument in the Discussion, but somehow leaching is mentioned in the title as a central result. The focus of the manuscript is not on leaching at this moment, but most of the attention is on c2 and Wopt. I suggest changing the title accordingly.

Line 483: Fig. 2 does not show clearly that the estimates of k are larger in the HPMe-LE-peat-l0 as you suggest. If anything, they look similar or even lower.

Line 484: How do you suggest better estimates of leaching losses can be achieved in future experiments?

Supporting information

S1: Most of the variables and parameters in this section are not defined. I suggest presenting this section in table form with 3 columns, where one of the columns has human-friendly names for each variable/parameter/model. Making the reader go read three other papers to understand the formulas is not very mindful.

S2: same as S1. Also, S2 does not seem to be referenced in the main manuscript, but if it’s relevant to the manuscript, it should be mentioned.

S3: when you say “all other computations were done in R”, which ones weren’t? Please, detail all the software used in this study.

S4: This section does not seem to be referenced in the main manuscript nor do the individual figures in it, but if this is relevant to the manuscript, it should be mentioned.

S5: In the caption of Fig. S9 you wrote twice HPMe-LE-peat-l0, but one of them should be HPMe-LE-peat. The axis titles are switched (k should be x axis and Species should be y axis). Also, I suggest separating vertically points corresponding to each model, and for the different depths of Sphagnum spec. as well. This way you will avoid the overlap.

S6: I suggest differentiating species with colors in Figs. S10 and S11, and defining what parameters are in the captions of Figs. S10, S11 and S12.

S7: If I understand correctly, data points on the positive end of the x axis are covered in water? If so, I would expect leaching to be higher under those conditions, but somehow it does not seem like it. I find this interesting, could you discuss this in the main manuscript?

S9: Which one of the two panels uses the standard value for 𝑊𝑜𝑝𝑡 or the 𝑊𝑜𝑝𝑡 value estimated by HPMe-LE-peat-l0?

S11: You say decomposition was simulated either under a degree of saturation of 0.6 L/L or 20 cm below the water table. But in the figure, you included -20 cm and 10 cm, so please make sure the description coincides with the figure.

S12: This section does not seem to be referenced in the main manuscript, but if this is relevant to the manuscript, it should be mentioned.

Technical corrections

Line 3: separate “conditions and”.

Throughout the text citation style is inconsistent. I suggest not using parentheses for the year if the citation is already in a parenthesis. Use a comma instead, like you have already in many parts of the text. Correct the citations accordingly in lines: 27, 32, 33, 66, 375, 421, 422, 434, 453.

Line 99: correct “use” for “used”.

Line 117: add “and” before “water table depths”.

Line 223: change “form” with “from”.

Line 328: It should be “as a consequence”.

Line 366: after “:” the next letter should be lowercase, otherwise change “:” for “.”.

Line 475: I think another word like consider, believe, think, etc. is a better alternative for expect here.

Supplementary information S11: change the “,” in S. fallax.