This manuscript presents a valuable cross-site comparison of methane fluxes measured by eddy covariance (EC) and chamber methods across 10 wetland and upland sites. The work addresses an important gap in understanding how these two common measurement approaches compare across multiple sites and temporal scales. The dataset is substantial, the statistical analyses are appropriate, and the findings have significant implications for combining multi-scale flux data in syntheses and modeling studies. The multi-site and temporal scales approach, practical recommendations and transparency about limitations are significant strengths.

However, the manuscript would benefit from major revision to improve clarity, address methodological concerns, and better contextualize the findings.

Key issues to resolve:

1. Ebullition circular reasoning between methods and discussion
2. Lack of wetland vs upland comparison despite title emphasis
3. Figure and table readability and formatting (axis overlap, interpretive captions)
4. General language review for sentence structures, removing redundancies and making the text more concise and cohesive

Section-specific comments and key issues:

Introduction

Line 90-91: “Thus, chambers provide a greater spatial site-level representation than EC sites and are needed to fill the missing data gaps.” – contradicting your statement about area representation in lines 80-81

Line 99: “Plot and ecosystem-scale FCH4…” : You mention in your line 78 that different ecosystem and hydrological subtypes within the EC footprint are defined as “plots” (see also line 80), but here it seems to me that by “plots” you mean chamber studies and chamber plots? To avoid confusion, consider re-wording/defining.

Line 113: “…than the true, mean ecosystem-scale FCH4”: this insinuates that the EC ecosystem-scale measurement is the “true” FCH4, but in reality, we do not know this for sure, as EC also comes with its own uncertainties.

In the introduction, you could also discuss the different chamber systems used in GHG studies, i.e. manual sampling with gas chromatography analysis, or in-situ measurements with portable gas analyzers. This distinction affects comparability and should also be discussed in the limitations section.

Methods:

Key issue: Inconsistencies in ebullition treatment

Most sites removed ebullition events from chamber data (Text A1). However, ebullition is mentioned repeatedly as potentially explaining EC > chamber fluxes, and the authors suggest EC captures ebullition better than chambers (lines 477-482, 491-494, 497-499, 569-570). This is circular reasoning: chambers are processed to remove ebullition, then the authors express surprise that EC is higher. Recommendation: Clarify whether ebullition removal from chamber data is appropriate for this comparison or acknowledge it as a limitation. Discuss implications: if ebullition is real and important, should it be removed from chambers when comparing to EC?

Table 1: I wonder if Table 1 could be better presented in the final publication? Perhaps landscape orientation would make the column titles and words not be cut off.

Line 215: I’d like more information about seasonality of measurements and when they were conducted. I see this in supplementary material (Figure B1), but I think this is important information for the main text. Perhaps integrate to Table 1?

Results:

Line 310: What do you mean by large CV? Please quantify.

Lines 313-315: I think it is good practice to put units behind all numbers, even if you are repeating them/listing them.

Figure 3:

• Caption reads like interpreting results rather than figure description. I would say that figure captions should describe content, not provide conclusions.
• Add R² and RMSE in addition to Spearman's ρ – they are more relevant for linear regression.
• The labels on both axes could be more sparse (they nearly overlap and are hard to read) and x axis labels could be written at an angle for better readability.
• Try to keep only one Obs. count legend.

Figure 4:

• Similar comment about the caption as Figure 3.
• Add R² and RMSE in addition to Spearman's ρ – they are more relevant for linear regression.
• Consider adding a regression line onto Figure 4, to show the deviation from y=0 (the blue) line.

Figure B6-B7: without reading the captions, there is no way to understand what these figures are illustrating.

Table 3: Formatting needs improvement - currently difficult to follow. Try to make more compact.

Discussion

Key issue: The title of the manuscript emphasizes "wetlands and uplands" but there is minimal comparison between these ecosystem types in the results and discussion. Lines 592-595 briefly mention this, but given the title's prominence, this deserves substantially more attention. How do the ΔFCH4 patterns differ between the two ecosystem types? Do different processes drive differences in these contrasting systems? If data are insufficient for robust comparison, consider revising the title.

Line 450: Avoid starting immediately with “contrary to our hypothesis.” Start by summarizing key findings, then contrast with hypothesis.

Lines 457-460: long sentence, consider splitting into two.

Lines 463-474: This section could use some citations to back up your claims and relate to previous findings.

Line 475 and onwards: however, you could also discuss here that chamber artifacts can increase chamber fluxes: e.g. disturbance of soil/water surface and thus causing a pulse (especially for manual sampling).

Lines 481-485: long sentence, consider splitting into two.

Lines 485-486: Yes, but Meijide et al found the opposite trend – that chamber fluxes exceeded EC fluxes. Can you make direct comparisons with your results and how would your interpretation change?

Lines 486-489: So what may be the implications of this poor detection of differences at low fluxes? How may it compromise/influence your results?

Lines 489-492: again, consider splitting into two sentences

Line 496: consider if statistics such as p-values are relevant for discussion or if they should be kept only in results.

Lines 495-499: This is a good thought about PA, however, you bring out that your results show decreasing ΔFCH4 with higher PA at the weekly scale, which is not something relevant for ebullition events (very short term). I wonder if you could elaborate on this point.

Lines 501-502: but consider that low u* values are typically removed from EC data if using standard protocols – how would this influence your results’ interpretation?

Line 504: change ‘may be’ to ‘was’. I think overall the word ‘may’ is used excessively in the discussion. Try not to use ‘may’ if you are referring to your own results: your results are certain, not ‘maybe’; the implications of your results and how they relate to previous findings, that is where the word ‘may’ comes in.

Line 511: ‘As expected’ replace with ‘As hypothesized’, or similar

Lines 517-519: You talk about soil CO2 respiration hotspots and hot moments and then bring those same conclusions to explain CH4 hotspots and moments. I think CO2 respiration hotspots do not directly translate to CH4 dynamics. You could try to find studies describing CH4 hotspots/hot moments with chamber measurements; or better integrate the CO2 ideas.

Line 521: By overcome do you mean exceed, i.e. EC FCH4 is higher? Overcome is a strange verb to use here.

Line 530-531: Good point about using representative chamber patches, but discuss the issue of how to best determine these?

Line 535: add ‘the significant effect of uWD…’

Lines 542-544: Rewrite the final sentence of this paragraph; it is confusing. What is ‘significant April’?

Lines 546-548: Why would chambers not capture the release of stored CH4 below the ice and snow cover? Freezing and thawing dynamics also occur inside chambers. I also struggle to see very specific “higher ecosystem-scale FCH4 at CN-Hgu and US-Ho1 in cooler months” in Figure B14. But maybe this is an issue with figure readability.

Lines 550-554: the chamber measurement system issue seems to not belong together in the same paragraph with seasonality issues. Consider re-arranging. Would chamber system issues better be placed in limitations? Perhaps you could also integrate the use of different chamber flux analysis methods (GC vs portable analyzers) and create a new paragraph.

Line 556: I’d suggest removing ‘As we expected’ and just staring the sentence with ‘Diel analyses revealed…’

Line 559: What do you mean by ‘higher diurnal EC FCH4 than chamber FCH4’? Do you mean higher diurnal variation? Or higher daytime flux overall?

Line 573: this sentence needs a citation.

Lines 599-601: This sentence is confusing, consider re-writing.

Limitations:

Could the specific instrument for chamber measurement (e.g. Licor, Picarro etc) also play a role in method discrepancies?

Acknowledge that lack of winter measurements at most sites is a major limitation and may bias annual comparisons.

Supplementary materials

Try to keep figures and figure captions on one page.

Just suggestions: Confusion in Supplementary Material/Appendices naming: Text A1-A3 and figures B1-B19 and Tables C1-C14. They make sense once you make it all the way down to the supplementary material and realize that the letters refer to Appendix A, Appendix B and Appendix C, but when they first pop up in-text, it is confusing. Also, instead of Text A1-A3, you could do Supplementary Methods 1 etc. maybe, as “Text” is ambiguous.

Text A1 regarding QC/QA of chamber measurements on each site: I’d like to see a more uniform description of these. Currently, some sites mention R2 filtering, some not. Some say ebullition was not significant and thus it was removed, some say ebullition was not significant and thus it was *not* removed. Perhaps you could combine Text A1 into a table with different QC/QA criteria for better comparison (i.e. R2 threshold, dark/transparent chamber, GC vs PGA instrumentation, pressure vent and fan, H2O correction, ebullition removal…). Similar to Table C1.

Review for Määttä et al, Biogeosciences, January 2026

Määttä and co-authors conducted a metaanalysis comparing chamber and eddy covariance based methane (CH4) flux measurements in peatlands. They then explore parameters that can explain the difference observed between these two measurement methods.

This is an important and timely contribution that falls well within the Biogeosciences scope. The authors have carefully compiled a unique dataset which holds the potential for important contributions. Unfortunately, I find that the statistical approach is not very mature (see comments below), and I’m therefore not sure if all the reported conclusions are sufficiently supported. The manuscript is clearly written and reads well.

Major comments:

1. In the introduction I’m missing a discussion of the representativeness of chamber and especially EC measurements for peatland surface-averaged FCH4. I’m thinking about ecosystem models here, which delineate peatlands, and then ascribe an average flux to the peatland surface, which is compared to empirical measurements. EC measures the average flux in the footprint, which is different from the delineated peatland. Is there a bias introduced by placing the tower in the center of a peatland, potentially under-sampling lower-flux locations at the edge of the site, and does this introduce a systematic error to model/data comparison?
2. ΔFCH4, as studied in this manuscript, is the combined product of (i) true difference (plot vs. ecosystem flux have different drivers), (ii) measurement uncertainty and artefacts in each method, and (iii) human behaviour (e.g. where do researchers place chambers). I feel like (iii) could be better discussed in the manuscript, and information on chamber placement strategies/goals could provided in the appendix. Also should also be discussed that (iii) is prone to change over time, possibly in response to this manuscript. It is therefore possible that the manuscript, in exploring predictors of ΔFCH4, itself changes these predictions.
3. The authors deal with a challenging dataset: ΔFCH4 is clearly non-normal and crosses 0, as are both chamber and EC flux data, and the data analysis requires mixed effects models to account for repeated measurements at each site. The authors address non-normality through data transformation (power transformation). However, such transformation fundamentally changes model structure: log-transformation turns additive models into multiplicative models, and more complex transformations leave models without clear interpretability (like, random intercepts in MEMs mean something very different after this transformation but I cannot intuit what that means in the real world). Why would e.g. a one-degree temperature difference have distinct effects on low, high, or negative fluxes that follow a certain (but not intuitive) mathematical function? I highly encourage the authors to explore statistical methods that do not rely on such complex transformations, which leave it almost impossible to draw meaningful conclusions from the results.
4. ΔFCH4 vs. mean FCH4: analysis seems to run into some artifacts, especially at half-hourly and hourly level. I highly recommend the authors show Fig B3 instead of Fig 4. B3 shows how this apparent correlation of ΔFCH4 vs. mean FCH4 is driven by sites where either only chambers or only EC showed high CH4 emissions (see also fluxes in B8 and B9 where chambers show no CH4 emissions at all. Unless there’s some conversion error here, this seems to be driven by chamber placement on non-emitting sites in the tower footprint?). B8-B12 show very distinct relationships between chamber and EC fluxes, and I wonder if it would be better to analyse these relationships for each site first and then compare the outcomes rather than pool the datasets and then try to identify overall drivers.
5. Similarly, the dataset is highly unbalanced, including hundreds of entries from some sites but only a few from other sites. This means that any relationships identified in the dataset will be highly driven by a few sites, while relationships at other sites are barely taken into account. Some of this can be addressed by allowing for random slopes rather than just random intercepts in MEMs and/or by downsampling over-represented datasets.
6. Generalization: I’m not quite sure what aim of the statistical analysis is, and specifically, what ‘general population’ do the authors want to make inference about? (i.e., ultimately, what kind of prediction do they want to support?). My guess is that they will respond ‘explain the observed data’ but statistics cannot explain data. Is this exercise meant to provide a means for comparing (or converting between) EC- and chamber-based flux estimates? Is it such a comparison/conversion meant to be valid beyond to sites outside this dataset (i.e., should this help us compare the chamber fluxes from one site to the EC fluxes at another site)? In that case we’re talking about predicting chamber/EC flux the ‘general population’ the authors are trying to make conclusions about is the population of (potential) flux stations, i.e., if I conduct chamber and EC measurements at a new site, what will ΔFCH4 based on various predictors? If that is the case, there’s a lot of pseudoreplication going on throughout the study, as different measurements within a site should not be considered as independent. The mixed effect models used somewhat alleviate this in some analyses but not everywhere.
7. Validation: the authors could test the robustness of their results e.g. with leave-one-site-out analyses. This would provide some additional robustness to the results.