The authors present an assessment of the impact of temperature and precipitations on methane emissions by Northern European wetlands as simulated by 6 process models. This study makes use of a set of ecosystem models, atmospheric inversions and fluxes deduced from eddy-covariance measurements to gain insights on the behavior of the 6 models with regard to temperature and precipitations.

General comments

The protocol of the study, with the various inter-comparisons, is very good, making use of all possible/relevant ways to assess the behavior of the 6 targeted models. It will be very useful to a wide community, working on process models but also on other models and even atmospheric inversions.

Nevertheless, because it should be usable by people with different backgrounds and probably also due to the many co-authors from different approaches, the structure of the manuscript, the order in which the ideas and information is delivered must be revised. Inside many sections and subsections, there are missing links between the ideas or pieces of information. This is not helped by the use of verbs in the past tense which should probably be revised by a native speaker. Even though I am not a native speaker, I was confused by this point about which information is from what was done before the study or not in the scope of the paper, what has been done for the study during the preparatory phase and what has been learned from the study.

Specific comments

Abstract

I think the abstract may be rewritten to better represent the quality and important results of the study: some key sentences are too vague or general, as detailed below.

- the whole abstract: why not write in the present tense?

- l.31: "compared to multi-model mean": a multimodel mean = one reference?

- l.34-35: "how the inversion attempts to change the prior fluxes in the posterior": clumsily stated, not very clear for readers outside inversions. Please rephrase, possibly on the lines of how the assimilation of atmospheric data correct the fluxes.

- l.36: "to move emissions": i.e. some of the emissions or all emissions?

- l.36: "in posterior": this is probably too specific a word for an abstract of a paper not aimed at the inversion community. Please rephrase.

- l.36-37: "in general", "often": this is too vague, please quantify or at least, indicate in space, in time...

- l.37-38: "This was not the case for the warm and dry period of summer 2018": with in general and often in the previous sentence, this sentence does not bring any information.

- l.39-40: "varied from May to September": from one year to another? From one model to another? Please clarify.

- l.41:"balanced": what is a balanced response? Please define or rephrase.

- l.42,43: "significant":  define/quantify what is significant in the context. Or avoid this word outside a clear statistics framework.

- l.43-44: "it is essential to pay more attention to the magnitude, composition, annual cycle and climate driver responses of wetland emissions in different regions." I guess everybody working on process-models knows about this. Isn't there a more practical or precise conclusion from this work?

 

Introduction

 As such, the introduction does not make it very clear how models use input information and what the issues are with temperature and precipitations. 

Regarding the order in which the pieces of information are delivered, first, a presentation of the various types of wetlands in the real world would be useful - it is done partly in the Results, when listing what types are taken into account in some models; then, a presentation of what is done in the models (for example, ignore lakes).

Overall, is not made clear enough what happens in the real world and what happens in the models, so that it's difficult to understand the challenges this study deals with. Examples: it is difficult to understand why the emissions of process models have not already been validated regarding temperature and precipitations dependencies; the way inversions can inform process models is presented in a misleading way: too optimistic if it's flux inversion, not clear if it's process models' parameters which are inverted. 

- l.46: "the second most important greenhouse gas": ANTHROPOGENIC!

- l.48: "peatlands": please list the types of wetlands before going into details for each of them.

- l.52: "There are accurate peat-land maps": what about the mineral lands, mentioned just before but never again discussed?

- l.57: "this feature is badly represented": this seems contradictory with the "accurate peat-land maps" of the previous paragraph. Explain more clearly what is know accurately and what is poorly known.

- l.67: "significantly": by how much?

- l.68: "more climate-oriented": i.e. instead of using maps of the extent? Please explain more clearly.

- l.69: "to emphasize the regional approaches.": what does it mean?

- l.70: "to study the responses of the emissions to air temperature and precipitation": the process models are built to take temperature and precipitation into account, aren't they? Haven't they be validated on these points? Make the challenge(s) clearer!

- l.74: "arises": is this the right word?

- l.75: "inundation": please define.

- l.81: "according to atmospheric inversion modeling": this is too vague. Atmospheric inversions as such are not able to point at the causes/drivers of the corrections they apply to fluxes. So probably something more has been done than raw inversions by Thompson et al.

- l.85-86: "provide a top-down view of the responses of methane emissions to climate drivers, attempting to detach them from the underlying prior assumptions.": not clear, even for somebody working on flux inversion. What is inverted here? Methane fluxes? If so, insights on climate drivers are not easy to obtain and the prior assumptions play a large part!

- l.87-88: "atmospheric inversion models can be used to inform process models on how they should improve their emission estimates and climate responses": not so simple, if it's flux inversions: they can indicate where fluxes should be larger or smaller than what the process models compute but neither why they are too small or too large nor how to correct the process models. This last point is provided by inversion not of fluxes but of parameters of the process models. Please be very precise on the inversions you are dealing with.

- l.89: "to study the responses of the emissions to air temperature and precipitation": I would think this is part of the evaluation/validation of process models! Please make the challenges clearer.

- l.91-92: "the ensemble f models from the Global Carbon Project (GCP) 2020 estimation": of which the Crescendo models are not part?

- l.93-96: please cut up this sentence into several manageable shorter ones.

 Materials and methods

At least one model is in the ensemble used for comparison: please explain how this is not an issue for this study.

At the end of each description, or better still, in an overall summary (maybe in a table), it would be very useful to have the limitations (e.g. neglected processes, poorly known inputs) and advantages of each model compared to the others (not need to list what they all do the same way) with regard to methane emissions. Maybe it would even be possible to state what is expected from each model e.g. performs better in a given region or season. I'd expect the impact of the spatial resolution to be large - but I don't run process models.

Please make the paragraphs of the various models homogeneous, with the same structure e.g. first the general history of development, then the relevant points for methane emissions and in the end, the sets of emissions used on this study with the name used in the remainder of the text.

Subsections 2.1 to 2.7 : it is not clear how many runs or set-ups from each model are used in the study, please provide an overall summary, maybe as a table, with consistent tags. An assessment of the differences between the runs/set-ups and how they are expected to impact methane emissions and their response to temperature and precipitation is expected by the reader.

In several sections, the order of presentation is misleading i.e. the inputs are described after (part of) the description of the results. Please re-order.

Figure 1: provide information on stations, at least their full name and network.

- l.102: "utilised": is this the right word?

- l.104: "further developed in the recent H2020-CRESCENDO project": is there a reference for this e.g. a deliverable and/or report of the project?

- l.107: "in Global Carbon Project": the ensemble to use as a reference includes at least one of the 6 models targeted here: isn't is an issue?

- l.130: "was set to": based on what data/information?

- l.133: "manuscript": what does this mean? Currently a draft? Submitted?

- l.145: "suitability for peatland growth conditions": what does this mean?

- l.158: "in general": do you mean it's not the case here? In some set-ups? Or not on some parts of the domain?

- l.167-169: "CLM4.5 etc": this information does not seem relevant at this point. Maybe put this at the end of the paragraph and in the overall comparison that I suggest above.

- l.178, l.185: "are also used": so that there are two sets of methane emissions by CLM5/JULES in this study?

- l.188: "recently": does this mean it is what is used in this study?

- l.195: "of which": does this mean the ensemble is larger but here a subset is used? If so, why keep in the sub-ensemble models which are targeted in the study? Explain why it is not an issue: maybe because the set-ups are different enough?

- l.196: "GCP-diag": the full ensemble or the subset?

- l.197: "8 models": same remark about the models which are part of the ensemble and targeted in this study.

- l.200: "inversion models": this does not look like the right name for these: any (chemistry-)transport model can be embedded in an inversion framework.

- l.203: "the share": what does it mean?

- l.203-208: "In addition to GPC...": what is the logical link/relevancy of these sentences to the previous explanation? Re-order the description of the inversions to include the information on the priors where the reader can understand which priors go into which set of inversions.

- l.217: "processed and analysed on a 1x1 degree grid": how? Is this expected to lead to some issues?

- l.218: "remapped by bilinear interpolation onto 1x1 degree grid": is this different from the regridding of previous sentence?

- l.224: "mostly": but not all?

- l.231: "were better constrained over that larger region than 1x1 degree given the limited number of surface stations": probably not very clear for readers who are not into flux inversion... Maybe simply state that the uncertainty on the retrieved fluxes is too large at the pixel's resolution but is satisfying when aggregating over the whole region.

- l.231-235: "In Northern...": put the description of the inputs of the inversion before the description of the results i.e. the posterior (or retrieved) emissions.

- l.242-243: "grid-wise mean of the three emission maps available for years 2013 and 2014.": why this choice?

- l.252: "significantly": define.

- l.253-254: "creating confidence in the validity of the CRU-JRA climate data approach": not clear: to me, it shows that the uncoupled approach is OK for the targeted scientific question but "confidence in the validity of the approach" is too strong, it may give the idea that the coupling is not necessary at all. Please also define confidence in this context.

- l.254-256: "In general, CRU gridded datasets are found to be suitable for vegetation analyses and well comparable to e.g. MERRA-2 and ERA5-Land reanalysis datasets, performing well even in remote areas with few observations (Zandler et al., 2020)": put in the description of CRU? The order in which the items of information are presented in the text is very confusing for the reader. Please review the whole structure of the sections to clearly first present the inputs and then their use and finally comment on the impact on methane emissions.

- l.258: "as their mean temperatures were always above zero": why not have a look at negative temperatures? It looks like April is also positive, according to Fig. 2.

- l.261: "those growing season months": = May to Oct? Please define the growing season in reality vs in the models.

- l.263: "correlating": do you mean to use here "correlate"?

Results

Same general remark on the order in which the information is delivered: it must be revised.

Section 3.2: I  don't understand this section at all. The first paragraph announced work on two models but it's not what is done in the following paragraphs. The messages are not clear at all, the text is too descriptive and lacks synthesis/clear messages. I think the whole section is to be built again/rewritten.

Figure 3: not very easy to read with a different color scale for each model. At least try to have only two of them e.g. one for the large emitters and one for the smallest ones. And have them be easy multiples of one another. Another solution: plot some indicator, normalized, without unit, that would be in the same range for all models. It is always possible to put the details with tailored scales in the supplementary material. Show on the panels, e.g. with lines or rectangles, the regions of high temperatures, low precipitations, etc, which are used in the text.

Figure 4: same remarks as for Fig. 3

Figure 6: it is not clear what GCP-post is, the ensemble? 

Tab.S1: multiplier = ratio? 

Figure S3: should probably be used in the rewriting of section 3.2 as it seems very informative about the maximum of the seasonal cycle...

- l.271: "Natural wetland fluxes, including those from peat-lands and wet and dry mineral lands as well as inundated lands": list of types of wetlands is required at the beginning of the paper!

- l.272: "below for": what does it mean?

- l.274-275: "the temperature and precipitation responses": how can this change in the posterior if it's flux inversions? The correlation change because the post emissions change but nothing is said by flux inversions on the relations between temperature/precipitations and emissions (see also previous remarks).

- l.275-276: "The seasonal cycle is also compared to up-scaled eddy covariance flux observations.": please re-order, this information should not be at the end of this paragraph.

- l.279: "highest emissions": show on the color scale what are the highest emissions: the 10th highest percentile%?

- l.279: "high temperature": show on the axis what high temperatures are.

- l.278-282: what is the scientific message of this paragraph?

- l.285: "significant": define.

- l.285: "generally weaker": precise how much, how often?

- l.287: "Multiple regression...": put in the description of the method, not in the results.

- l.284-288: what is the message on these results? What does it mean we must study in the following?

- l.290-291: Already stated before, remove from here.

- l.292: "In total": = over the whole region during the whole period?

- l.293: "bringing the flux estimates closer together": maybe a bar plot would be useful to display these results.

- l.298: "multipliers": = ratios?

- l.299: "above 92%": why 92? Why not 90?

- l.300: why 64%? Is there an idea of Gaussianity somewhere?

- l.300-302: "The highest increase was proposed for July 2014 with second highest mean temperature of 16.7 °C. However, the July 2018 record high heatwave with mean temperature

of 17.2 °C was not among the highest posterior increases.": This suggests that July 2014 is not well represented by the models (prior way too low) whereas July 2018 is (prior already OK so no large correction after inversion). The question then arises of what causes this difference in the models? What do they capture right in 2014 and wrong in 2018? Is it linked to the temperatures or to another parameter that differs between the two heat waves? 

- l.302: "43 mm in July": 2018? Is this part of the answer to the previous comment?

- l.303-305: "Some of the highest...": I understand that some high precipitation emissions are too low in the priors: so what is wrong with them? Moreover, what about the rest of the high precipitation months?

- l.308: "(above unity, i.e. above 88% percentile of all values)": what does this mean?

- l.308: "(51% percentile)": therefore in almost half the cases?

- l.309: "significantly": how much?

- l.312: "relatively": how much is it?

- l.313: "2018 (and 2014 and 2006)": this lumps together different cases: 2018 ends up with average emissions but 2014 and 2006 end up with high emissions. And what about 2010 and 2005 regarding precipitations?

- l.315: "otherwise": I don't understand the logical link.

- l.316: "July 2018 did not show high posterior fluxes here, while": consistent with JSBACH-H in the previous paragraph but the way it's stated suggests that it is a particularity of these simulations... Please clarify.

- l.318: "balanced": what does it mean? What is a balanced / an unbalanced prior?

- l.318-321: "The temperature and...": the message is not clear at all. Cut up the long sentence into several shorter ones and re-formulate.

- l.325: "and also the different model components.": what is meant?

- l.325: "Generally": please clarify.

- l.325: "total": = the whole region during the whole period? Or do you mean wetland fluxes = fluxes from these various categories?

- l.327: "LPJ-GUESS": should it be LPX-Bern here?

- l.330-334: How is this a result of this study? It looks more like an explanation of how LPX-Bern works.

- l.336-339: "Same remark as the previous one. Maybe the separate explanations of each model should go in the suppl?"

- l.341-347: Very descriptive paragraph: what is the message?

- l.349-353: what is the link of this paragraph with the rest of the section?

Discussion

I think there is an issue in the text with separating clearly what we know of the real world and what happens in the models' world(s). This makes the discussion unclear, particularly since what is known before hand and what is learned through the study is not clear (linked but not totally due to the issue about the times of the verbs).

- l.355: "According to process models": and not in the real world?

- l.356: "they comprise": what does it mean?

- l.358: "could be linked": we should know from how the models were built, shouldn't we?

- l.359-360: "Precipitation has a dual role: it presumably increases the wetland area by wetting dry upland soils and raises the water table in the permanent wetlands": in the real world or in the models or both?

- l.360: "constant/neglecting": what is meant? That it's as bad to use a constant value or to neglect the process?

- l.363: "being at largest after prolonged precipitation and generally in autumn (September- October) when the evapotranspiration had already decreased from high growing season levels.": in the model's or in the real world?

- l.364-370: "According to observations...": what is the link with the discussion on the models' results in this study?

- l.372-381: I guess the idea of this paragraph is to compare the findings of this study to others but it is not made clear at all because of the order in which the information is given (and the times of the verbs). Please reorder and reformulate.

- l.372-374: "A modeling study by Poulter et al. (2017) concluded that in boreal regions CH 4 emissions were best correlated with wetland area, followed by temperature and precipitation (as applied with one-month delay). However, methane emissions were highly correlated with temperature in some models (e.g. JULES) which had a high temperature sensitivity.": does this mean this study is not in agreement with Poulter's? What would this imply?

- l.374: "In general": please be more precise.

- l.376-377: "noting that the co-limitation of temperature and precipitation would emerge for the more southern climate zones.": what is the link with this study?

- l.383: "towards August": instead of July?

- l.387-388: "In our work, many models had their seasonal maxima in July or August, notable exceptions being CLM4.5 and CLM5 (bias towards spring) and LPX-Bern (bias towards autumn).": put in the first sentence, where the results of the study are summarized (prior to comparison to the literature).

- l.388-390: "The dominance of mineral land over peatland emissions may delay the month of the maximum emissions, as well as using a large wetland extent in late summer. ": is this the reason why the models give a maximum in July instead of August?

- l.390-391: "Placing more peatlands in the southern parts of the region (like in GCP-diag) or having a weak temperature response could bring an earlier and longer seasonal emission maximum.": is this a suggestion of how to modify models?

- l.391-392: "A pronounced inundation period after snow melt could induce large methane emissions in spring.": in the real world? What is the link to the previous and next sentences?

- l.393: "from satellite observations": not clear to me. Is this an explanation of why the inversions put the maximum in August? But they don't assimilate satellite data. I don't understand the link between this piece of information and the rest of the paragraph.

- l.393-394: "According to flux measurements at boreal peatlands, the month of highest emissions was July or August depending on the year (e.g. Rinne et al., 2020).": does this mean that models put the maximum in July, inversions in August but actually, we don't know which is closer to the real world?

- l.399: "anomalous": please define.

- l.402-403: "Rinne et al. (2020) also noted that methane emissions in four out of five Fennoscandian wetland sites were decreased in 2018 due to a decrease in water table levels. The summer months with high precipitation often resulted in high posterior emissions.": is this put here to suggest that the inversions are consistent with independent measurements? Shouldn't this be stated when presenting the inversions as a reference for comparison?

- l.404-405: "The year 2011 with observed high methane emissions from upland soil in northern Fennoscandia (Lohila et al., 2016) did not stand out in posterior emissions": this suggests that the models do not well represent the high precipitation months. Why?

- l.404-406: "The year 2011 with observed high methane emissions from upland soil in northern Fennoscandia (Lohila et al., 2016) did not stand out in posterior emissions, but large increases were assigned to high precipitation periods in e.g. August 2008, 2016 and late summer 2007": is this the same message as the sentence "The summer months with high precipitations..."?

- l.406-407: "August was also the month of the average seasonal precipitation maximum, while the average temperature maximum was in July.": this sentence is the last in the paragraph. This suggests to the reader that it is very important. Nevertheless, it is actually only a description of the input data of temperature and precipitation, to state when their max is. Please, reorder the logical "progression" of the sections/paragraphs to provide the information in a logical progression.

Conclusion

The conclusion depends on how well the introduction presents the challenges tackled by the study. Therefore, in the current state of the text, seemingly trivial messages appear, e.g. that "it is important to study the overall responses of the emissions to air temperature and precipitations". The reader may winder that it is not already done and if this might due to the fact that models take only C stocks and water table depth into account. The conclusion will have to be re-written after taking into account the comments on all the previous sections.

- l.412-413: "agreed on the month of maximum emissions": this is not clear from the discussion.

- l.413: "balanced": must be defined, see previous remarks.

- l.413: "significantly": see previous remarks.

- l.415: "to move emissions of both in posterior towards co-limitation of temperature and precipitation.": not clear, please reformulate.

- l.415-417: "The set-up of different emission components (peatland emissions, mineral land fluxes) had a significant role in building up the response patterns": how is this not trivial? What is the link with the inversions

- l.419: "multi-year average response patterns": please explain.

- l.420: "anomalous cases of severe droughts with significant water table drawdown in pristine peatlands": not obvious from the discussion. Pristine wetlands, "anomalous" and "significant" not defined in the text.

- l.420-421: "corresponding reductions in methane emissions": Corresponding reductions are simulated by the models? The sentence seems to lack a verb.

- l.421-423: "Depending on the model, wet mineral soil and inundated land emissions can modify the seasonality of methane emissions together with peatland emissions. Therefore, it is essential to pay more attention to the role of the individual emission components, their magnitude, annual cycle and spatial extent in different regions": how is this not trivial?

- l.423-424: "how the fluxes should be scaled up from site to region": isn't it how people working on process model work? What does this study suggest they change in their work?

Technical corrections

- check the acronyms: they need to be fully explicit only at their first occurrence.

- check the language: consistency between US/GB (s/z), accurate vocabulary outside of the strictly scientific (utilise/use) and the time of the verbs.

The authors investigated how simulated CH4 emissions from six ecosystem models respond to temperature and precipitation in Northern European regions. Then, simulated CH4 emissions from two ecosystem models exhibiting contrasting response patterns, and the ensemble mean of GCP atmospheric inversions, were used as priors of wetland emissions in an inversion. By doing this, they explored how the inversion changes the fluxes and alters the response of CH4 emissions to temperature and precipitation.

General comments:

Climate forcing used by the six ecosystem models are different. Previous studies have shown significant variations in process-based LSM outputs stemming from the choice of climate forcing. Four out of the six models were driven by CRUNCEPv7, one was driven by CRU-JRA and another one was driven by CRU-HARMONIE.

If my understanding is correct, temperature and precipitation data from CRU-JRA were applied to all six ecosystem models and the mean of GCP models, when studying the responses of simulated CH4 emissions to temperature and precipitation (Fig. 1). Why not use the specific climate inputs that drive each model?

The representation of wetland area and inundation were various among models. Some models used a static and prescribed wetland area, some models used prescribed but time-varying wetland area, and other models dynamically simulated inundation and wetland area. It would be very useful to have a figure to show the wetland area used by each model.

While the authors acknowledged the importance of wetland area in determining boreal regions CH4 emissions in both the introduction and the discussion section, the comparison of models’ outputs and the analysis of precipitation and temperature responses didn’t address the impact of wetland area.

Figure 7 shows substantial difference among models in terms of both the magnitude and the seasonal cycle of CH4 emissions. The influence of climate forcings and wetland area on these differences should be explicitly discussed.

Though the authors showed both prior and posterior wetland CH4 emissions from the inversion model in Figure 6, as a reader not familiar with inversion models, I still find it challenging to grasp how the prior wetland CH4 emissions are adjusted. To achieve changes in both the magnitude and seasonal cycle of prior wetland CH4 emissions as shown in Figure 6, what other sources or sinks of CH4 have been altered?

Specific comments:

L266: explained CH4 emission -> CH4 emission

L338: Fig 1 should be Fig 2

L392: to -> of?

Figure 3: A figure for upscaled flux observations could be added to show observation-based temperature and precipitation responses of wetland CH4 emissions.

Figure 6: The seasonal cycle of in situ atmospheric CH4 observations could be included in this figure to aid in understanding why the CTE-CH4 inversions shifted the monthly flux maximum towards August.

Figure 7: It is difficult to distinguish lines for LPX-Bern and LPJ-GUESS, lines for the four GCPs, due to similar colors.