| 25 Sep 2026
A mid-cost automated floating chamber system for high-frequency CO2 and CH4 flux measurements: design, calibration, and comparison with infrequent measurement strategies
Abstract. Small inland waters are important greenhouse gas (GHG) sources; however, most field campaigns rely on daytime-only measurements collected at multi-week intervals, potentially biasing flux estimates. We present the Buoyant Emission Recorder for Temporal High-Frequency Analyses (BERTHA), a mid-cost automated floating chamber system for high-frequency measurements of CO2 and CH4 fluxes on multiple spatial scales. We describe the system’s design and test how empirically derived pressure and water vapour correction factors for the mid-cost CO2 and CH4 sensors increase the accuracy of reported concentrations and derived flux estimates. Here, we found a 17–37 % (CO2) and 4–9 % (CH4) higher mean flux estimate with pressure compensation during the 97-day measuring period. Compensation for water vapour resulted in much smaller differences (~1–2 % increase). Additionally, we calculated the mean CO2 and CH4 emission using high-frequency measurements during 97 days and simulated alternative measurement scenarios based on the full dataset and evaluated how well these ‘reference means’ could be approached by 1) single, daytime-only chamber flux measurements at daily to monthly intervals, and 2) short-term (2–3 days) automated chamber deployments every two–three weeks combined with machine learning-based gap-filling. Results of the simulation analysis showed that daytime-only measurements consistently and strongly underestimated the reference CO2 mean. CH4 emissions matched the reference mean well for daily, but not for monthly measurements (>90 % and ~23 % of simulations deviated <20 % of the reference mean, respectively). Short-term automated chamber campaigns, gap-filled using machine learning closely reproduced the CO2 reference mean, with 89 % and 48 % of means being within ±20 % of the reference mean, for campaign intervals of two and three weeks, respectively. We conclude that high-frequency chamber measurements using mid-cost GHG analysers combined with empirical pressure corrections provide accurate emission estimates, a robust benchmark for evaluating reduced measurements strategies and gap-filling approaches, and potential for improved process-based understanding of GHG fluxes when combined with ancillary high-frequency measurements.
Review of: A mid-cost automated floating chamber system for high-frequency CO2 and CH4 flux measurements: design, calibration, and comparison with infrequent measurement strategies
General comments
The authors present a wealth of data on a novel automatic GHG floating chamber, which is in the mid-cost sensor range and thus is of considerable interest to the research community. The results for CO2 in particular seem compelling though there are limitations with CH4 and ebullition. It would help to clear up the sections describing the methods of ebullition/diffusive flux and make it clear what fluxes are being presented in the text, and being transparent about how much of the flux might be missed/poorly estimated or not.
There is a lot of supplementary material which needs to be distilled down and re-examined how it is currently presented to ensure that only relevant information is included and is presented in a usable way. Similarly, the text could do with shortening, there is a bit of repetition and some sections could be made more succinct.
The authors need to ensure they are clear throughout about the environments that their instrument is applicable for, and particularly in the introduction be careful to constrain to aquatic/appropriate aquatic environments when making generalised statements. It would be nice to see some information about the wider use of BERTHA in other aquatic settings/stability e.g. in the wind.
It would also be helpful to know also about any biofouling over the 97 day experiment? There is no mention of this, yet I would expect there could be algae growth. Please could the authors comment on whether any growth occurred on the chambers/parts of the equipment, and if so the impact on results. If there was no growth, were any methods employed to prevent this/do the authors think that this could be an issue in other environments?
Overall the work warrants publication after revisions.
Specific comments
Abstract: Be good to say here the environment that BERTHA was deployed in for the 97-day measuring period, as this gives context to the findings on differences in CO2, CH4 depending on time of day/frequency of sampling
Line 26: Good to spell out the exact frequency you mean by ‘high frequency’ for CH4 and CO2.
Line 31: Can you give any figures here to back this statement up, in terms of emissions per surface area in comparison to other environments?
Line 43: suggest change to ‘can introduce large…’
Line 47: Do you mean solely on small inland waters? Since there are a lot of flux towers on land making very high frequency measurements
Line 49: Are you talking across all aquatic systems or just ‘small in land waters’? It would be good to make this clear throughout, to make clear what environments your findings are applicable to, and to remove any ambiguity from statements in introduction/discussion etc.
Line 53: Add in ‘aquatic’ here, and check this throughout that you are being precise as to the environments you are discussing
Lines 55-57: Good to add in a line here about why these low cost sensors have been taken up – i.e. need for more measurements, more spatial and temporal coverage across these poorly quantified small inland water bodies
Line 68: change ‘would’ to ‘could’, or ‘are likely’
Lines 72-75: Be good to have a statement in here about the importance of this, not only process-based understanding but also being able to account for these gas fluxes from these water bodies in national and global GHG inventories to understand the overall GHG balance and temporal changes
Line 79: Can you give more information about the ditch, dimensions, depth?
Line 90: Define RH before abbreviating
Line 110-114: this sentence is quite long and hard to decipher, consider splitting into two for clarity
Line 111: Figure 2 and S1 do not relate to this, they show the set up of BERTHA. Please check and correct. Do you mean figure S6?
Line 118: Not clear in the diagram where this stream of calibration gas is being introduced?
Line 154: You give a lot of figures in the supplementary text, but do not show the data for the noise introduced by the membrane pump? It would be helpful to have some indication of what level of noise we are talking about, what was the standard deviation around the mean pressure for each chamber closure?
Line 159: Were you able to assess in any way how good the temperature compensation is of the CO2 and CH4 sensors? I imagine how good this is varies between sensors and could also impact whether low-mid cost sensors are suitable or not.
Lines 171-174: Please split this sentence into two for clarity
Line 179: be good to give the RMSE for transparency
Line 184-189: As above, please split this sentence up for clarity
Lines 196-198: Would these trapped bubbles, then suddenly released into the chamber not still represent the ebullitive flux? It would just be accumulated and released into chamber in bursts, rather than single bubble at a time?
Line 199-201: What do you mean by ‘part of the CO2 measurements’?
Line 210: It seems a useful feature that the system would detect this problem of a chamber frozen open or closed, and the ability to reset. Though you then exclude all the data after this, despite normal closures occurring? This would imply that if the chamber gets stuck then all the data after are not trustworthy. It is not clear to me, why there would be issues follow this? How long is ‘prolonged’? Surely, even if the vegetation has changed in a way different to the ‘natural’ situation, the data would still be useful to assess the limitations of low vs high frequency measurements? It would also be good to state clearly, how long a prolonged closure would necessitate discarding all data afterwards.
Section 2.5: Whilst some information here is useful and particularly gives information on the reliability of this mid-cost instrument, I would suggest that some of this is shortened/made more succinct, particularly those that were not due to instrument failure.
Line 229: You say in line 200, that you calculate total CO2 flux, like total CH4 and do not (that I can see) say how you calculate CO2 ebullition? Equation 7 is for the diffusive flux
Line 231: What kind of concentrations would you expect for ebullition events? It would be helpful to give some indication of what % of the flux you might be missing and whether the sensor range is appropriate for accurate measurement of the fluxes.
Line 236: Why/how were there duplicate time stamps?
Line 240: Is there a reason for the 10ppm threshold for this? Please explain.
Line 246-248: What % of measurements were over the 100ppm limit? This is important to understand how much information is not being captured via the presented method
Line 269 – 272: Since you do not present the ph and O2 data just remove this information
Line 356: how does figure S5 give this information on the % for CO2 ebullition? As noted above please clarify the methods for detecting and quantifying co2 ebullition, and how this can be achieved whereas not for CH4?
Line 359: is this 27% the mean, median?
Figure 3: Why do you swap units in the y axis between a and b? same for figure 4? Be consistent with either moles or grams
Line 380: (and ref fig S6) Are the data in chamber 1 still trustworthy? At what point do the pressure corrections become too large to in introducing too much uncertainty? Did you find any evidence of clogging on retrieval of the chamber?
Section 3.5: It is not clear to me, why it is possible to do the figure 7 type comparisons for CH4 yet you say in the methods that the periodic multi-day measurement comparison is not possible for CH4 because of the ‘aforementioned limitations of our CH4 measurements’. Do these not limit the ability to compare daytime only sampling also? Please expand on this and clarify the limitations of the dataset
Line 488: specific in small aquatic water bodies/ in a drainage ditch
Line 490: related to comment above, I’m not sure this statement is entirely fair, since when the chamber froze and was open /closed for prolonged periods, even though this was detected automatically and restarted, the data were discarded. This statement therefore is perhaps a bit misleading? Similarly, perhaps it needs to be clearer here, and throughout about diffusive vs ebullition and what you are able to quantify accurately for CH4 and CO2.
Line 499 : give numbers in brackets here to back up significance
Line 508: Are most sensors for CO2 NDIR? I wonder how big a problem in the literature this pressure compensation is, i.e. are most people using sensors of this kind and thus the ideal gas law approach is not appropriate, or is this sensor type the minority?
Line 516: Why do you think this difference occurs between the different sensors?
Line 525-530: Can you be sure that the water vapour correction for CH4 is small given what you say in the lines above about the CH4 concentration used in your water vapour compensation procedure being too low? Is this something you are able to repeat at a more appropriate CH4 concentration in order to provide a better estimation of the impact of water vapour concentration on flux estimates?
Line 537-538: please give some indication of what time frame you mean by long and short
Line 546-547: I’m not clear what you are saying here, please can you clarify. I think you are saying that differences are not due to calibration since they were calibrated with the same gases
Line 589: What sizes were these lakes?
Line 601-602: I appreciate you are probably working on validation if this revised design to capture ebullition events, but I think since you do not present data to back up this statement that it ‘now enables accurate concentration of high concentration ebullition events’, it would be better to phrase this as an upgrade/revision that is currently in development and awaiting full experimental testing?
Line 603-605: Maybe good to mention here that there would be a power need for these additional sensors which would need to be factored into longer term deployments and may impact measurement frequency
Section 4.5: It would be good to see some lines on how applicable BERTHA is to other settings, not just ditches, e.g. stability in lakes with exposure to large winds? Do you plan to test their stability in these settings. Be helpful to know how applicable they are to other settings.
Supplementary material
There are a lot of figures and photos here and I recommend condensing this and distilling this to what is necessary and brings added value. For example, I do not think figures S2 and S3 are necessary. Please also check that the order that you reference supplementary material in the text reflects their number ordering, i.e. S1 is the first mentioned
Line 34: Caption not overly clear – and links to lack of clarity in total/diffusive/ebullition fluxes for the to gases. You say CO2 not calculated same way as CH4, but then say CO2 fluxes were excluded based on first and last 6 measurements like the CH4 method? Please look at these sections and try to simplify/clarify to make your intended message clearer.
Figure S7: Please state if this is for all chambers
Figure S9: this figure is not overly helpful with current colour scale and legend labels. All that is apparent is that most measurements are somewhere between maybe +/- 5 and 40ppm in terms of internal H2O change. Please improve this plot, or consider presenting the data in a different way e.g. give mean, SD internal H2O change etc within the text
S12 b:This figure is not overly clear, what do the arrows represent? Also, could you colour the axis or label better. Presume residuals is the right hand y axis of figure B. for B) the orange ‘fit’ doesn’t seem much of a fit, the data points are very scattered
S17: this is not very informative, nor do I think necessary, would suggest to remove