From the journal Environmental Science: Atmospheres Peer review history
Laboratory and field evaluation of a low-cost methane sensor and key environmental factors for sensor calibration
Round 1
Manuscript submitted on 05 8 2022![](https://www.rsc-cdn.org/pubs-core/2022.0.146/content/NewImages/icon-letter.png)
![](https://www.rsc-cdn.org/pubs-core/2022.0.146/content/NewImages/chevron-down.png)
10-Nov-2022
Dear Ms Lin:
Manuscript ID: EA-ART-08-2022-000100
TITLE: Laboratory and Field Evaluation of a Low-cost Methane Sensor and Key Environmental Factors for Sensor Calibration
Thank you for your submission to Environmental Science: Atmospheres, published by the Royal Society of Chemistry. I sent your manuscript to reviewers and I have now received their reports which are copied below.
I have carefully evaluated your manuscript and the reviewers’ reports, and the reports indicate that major revisions are necessary.
Please submit a revised manuscript which addresses all of the reviewers’ comments. Further peer review of your revised manuscript may be needed. When you submit your revised manuscript please include a point by point response to the reviewers’ comments and highlight the changes you have made. Full details of the files you need to submit are listed at the end of this email.
Please submit your revised manuscript as soon as possible using this link:
*** PLEASE NOTE: This is a two-step process. After clicking on the link, you will be directed to a webpage to confirm. ***
https://mc.manuscriptcentral.com/esatmos?link_removed
(This link goes straight to your account, without the need to log on to the system. For your account security you should not share this link with others.)
Alternatively, you can login to your account (https://mc.manuscriptcentral.com/esatmos) where you will need your case-sensitive USER ID and password.
You should submit your revised manuscript as soon as possible; please note you will receive a series of automatic reminders. If your revisions will take a significant length of time, please contact me. If I do not hear from you, I may withdraw your manuscript from consideration and you will have to resubmit. Any resubmission will receive a new submission date.
The Royal Society of Chemistry requires all submitting authors to provide their ORCID iD when they submit a revised manuscript. This is quick and easy to do as part of the revised manuscript submission process. We will publish this information with the article, and you may choose to have your ORCID record updated automatically with details of the publication.
Please also encourage your co-authors to sign up for their own ORCID account and associate it with their account on our manuscript submission system. For further information see: https://www.rsc.org/journals-books-databases/journal-authors-reviewers/processes-policies/#attribution-id
Environmental Science: Atmospheres strongly encourages authors of research articles to include an ‘Author contributions’ section in their manuscript, for publication in the final article. This should appear immediately above the ‘Conflict of interest’ and ‘Acknowledgement’ sections. I strongly recommend you use CRediT (the Contributor Roles Taxonomy, https://credit.niso.org/) for standardised contribution descriptions. All authors should have agreed to their individual contributions ahead of submission and these should accurately reflect contributions to the work. Please refer to our general author guidelines https://www.rsc.org/journals-books-databases/author-and-reviewer-hub/authors-information/responsibilities/ for more information.
I look forward to receiving your revised manuscript.
Yours sincerely,
Dr Claudia Mohr
Associate Editor, Environmental Science: Atmospheres
************
![](https://www.rsc-cdn.org/pubs-core/2022.0.146/content/NewImages/icon-comments.png)
![](https://www.rsc-cdn.org/pubs-core/2022.0.146/content/NewImages/chevron-down.png)
Reviewer 1
This manuscript describes the evaluation of low-cost methane sensors. The sensors were tested in an urban environment, though presumably the lessons learned in this study could also be applied to using these sensors in rural environments (e.g., for monitoring natural gas wells). Overall the paper is well written and relevant to the journal. Please see my comments below.
The laboratory evaluation of the sensors was used in part to quantify potential interferences from other pollutants (CO and NOx) and environmental conditions (T and RG). Supplemental Figure 3 is supposed to show that the sensors do not respond to NO or NO2. However the absolute value of the best-fit slopes in panels C and D are larger than the slope for CH4 in panel A. Maybe the sensors do respond to these species but the typical range in the atmosphere is small enough that the response is rarely or never important? I think the explanation in this section could therefore be clearer.
The sensors have major performance differences in summer versus winter. E.g., Line 321 "Given the low CH4 variance conditions observed in the summer season" - is the problem that there is less variation in methane in the summer (there are clear peaks), or that the sensors perform poorly in summer? There is evidence of similar seasonality in performance of electrochemical NO2 sensors - they perform reasonably well in winter and worse in summer when there is more potential for ozone interference.
Figure 4 focuses on the diurnal patterns using the model that includes the hour of day (HOD) as a parameter. I am a bit skeptical of this model because it could "get the right answer for the wrong reason" (e.g., because HOD is correlated with some sort of typical variation in methane concentrations). Are the diurnal patterns similar for the model that does not include HOD?
Line 499 - it's not immediately clear that you account for VOC interference. It's not shown directly in your lab or field tests. Perhaps this is addressed in a previous paper, but it does not seem to be directly addressed here.
Your study has the longest averaging time in Table 3. This perhaps suggests that you see nominally better performance because of signal averaging and hence noise reduction. What is the relevant timescale for monitoring urban methane? Is 1 hr sufficient, or de we need 1 minute data?
Methane was only evaluated up to 3 ppm. Why was that upper limit selected? The winter data in Figure 3 seem to miss a few peaks. Maybe this is a consequence of only testing up to 3 ppm?
I think a big question in the sensor community is simply "are these good enough for ambient methane measurement?" I would like to see the authors comment on that a bit. You show (Table 2) RMSE for hourly data of ~0.1 ppm. Is that 'good enough' for detecting things like spatial variations or short-term emissions? What are the potential applications of methane sensors that perform to the level you have demonstrated?
Minor comments
-Line 72 cites Brandt et al as evidence for inventories missing methane leaks. My recollection is that the Brandt paper, which focuses on natural gas production, identified under estimation of methane emissions from super emitters as a major problem. So this reference may not be appropriate.
-Line 123 - The Weather Underground reference is not in the bibliography
-Is there a reference for eq 1?
-Line 219 - I assume that "sensor value" in this sentence refers to the resistance.
-Line 290 - The stated temperature range is 4-37 C. It never goes below freezing in Baltimore?
Reviewer 2
This manuscript examines the performance of a low-cost methane sensor and investigates the key environmental factors that affect sensor detection. The authors found that methane detection by the sensor is influenced significantly by temperature, absolute humidity, and CO. Sensor performance varies across seasons with better performance in winter than summer, largely due to the smaller methane concentration variation observed in summer. The measured methane concentration can be corrected by the influencing environmental factors using multi-linear regression. Overall, the manuscript is well written. However, the calibration of Figaro TGS2600 has been largely studied in the literature in the past years. Although the results presented in this study are interesting, they are mostly expected. The novelty of the study needs to be further elaborated.
1. Abstract: Please add the detection limit (after calibration) of methane sensor used in this study.
2. Lines 152-159: Would the addition of the activated carbon-impregnated cloth also reduce the signal of methane? How long can the cloth be used with good performance? Is there any effect on CO and water removal?
3. Lines 341-343: Please provide the calibration detail (curves) of the CO sensor in the supplemental material.
4. Table 2: Why was the R^2 value of the calibration data smaller than that of the validation, especially for the winter data?
5. Figure 3B: It seems that the model is not good at predicting high methane concentration. Any explanation?
6. Figure 3F: Why not separate the methane calibration into two parts (< 2.3 ppm and > 2.3 ppm)?
7. Figure 4: I got confused. Sensor calibration was performed in a laboratory chamber. Why there was a diurnal trend of methane concentration during calibration? What do you mean by the diurnal pattern in methane concentration during the calibration period?
In addition, it seems that there was a greater variation in methane concentration in the reference measurement than in the sensor measurement. Please explain.
Can the authors also comment on the greater variation of methane concentration observed during the nighttime as compared to those in the daytime, although this is beyond the scope of the current study?
8. Table 3: Please add the detection limit of the sensor in each study to the table.
9. Lines 493-497: I agree with the authors that a lower sampling height may better capture the characteristics of a local emission. However, a sampling height of 29 m is not very high. Can the authors comment on the scale of the underlying zone of influence for a sampling height of 29 m for methane emission?
10. VOC interference: Buehler et al. (AMT, 2021) examined the effect of VOC on methane sensor performance using high ethanol concentration. Although I agree that VOC may affect the performance of methane detection, can the authors comment on to what degree VOCs at ambient levels can affect the sensor results? Have the authors compared the sensor performance with and without VOC removal during the field deployment?
11. Other meteorological parameters: in addition to temperature and humidity, wind direction (linked to the source) and speed (dispersion and dilution) may also affect the prediction of methane concentrations. Have the authors considered these factors in the calibration model?
12. One-hour resolution data: In this study, the sensor data was compared to Picaro CRDS instrument at 1-hour resolution. Have the authors compared the data at a higher time resolution? How did the sensor perform at higher time resolutions?
![](https://www.rsc-cdn.org/pubs-core/2022.0.146/content/NewImages/icon-response.png)
![](https://www.rsc-cdn.org/pubs-core/2022.0.146/content/NewImages/chevron-down.png)
Dear Dr. Mohr,
We are thankful for your careful evaluation of our submission and grateful for the constructive feedback from the two referees. We have made substantial revisions to the manuscript to improve methodological clarity and study rationale. We hope you find that our responses and revisions adequately address the comments brought forth by the referees.
Looking forward to hearing from you!
Joyce
The authors want to thank the reviewers for taking the time to provide thorough comments. We have addressed the comments below.
Reviewer 1 comments:
This manuscript describes the evaluation of low-cost methane sensors. The sensors were tested in an urban environment, though presumably the lessons learned in this study could also be applied to using these sensors in rural environments (e.g., for monitoring natural gas wells). Overall, the paper is well written and relevant to the journal. Please see my comments below.
Response: Thank you for highlighting the transferability of this work. We believe the cross sensitivities identified in this study would be true in both urban and rural environments.
1. The laboratory evaluation of the sensors was used in part to quantify potential interferences from other pollutants (CO and NOx) and environmental conditions (T and RG). Supplemental Figure 3 is supposed to show that the sensors do not respond to NO or NO2. However, the absolute value of the best-fit slopes in panels C and D are larger than the slope for CH4 in panel A. Maybe the sensors do respond to these species but the typical range in the atmosphere is small enough that the response is rarely or never important? I think the explanation in this section could therefore be clearer.
Response: We are thankful for this constructive comment from this referee and agree that Supplemental Figure 3 would benefit from additional explanation. The figure caption has been updated to include a discussion of the typical CO2, NO, and NO2 concentrations observed in the study area. We added “Ambient concentrations of CO2, NO, and NO2 exist on different scales than CH4 concentrations, thus the slopes cannot be directly compared. However, median ambient concentrations of CO2, NO, and NO2 typically range from 420 – 490 ppm, 0 – 0.040 ppm, and 0 – 0.040 ppm, respectively, in the study area. Within these concentration ranges, the CH4 sensor response is effectively negligible.”
2. The sensors have major performance differences in summer versus winter. E.g., Line 321 "Given the low CH4 variance conditions observed in the summer season" - is the problem that there is less variation in methane in the summer (there are clear peaks), or that the sensors perform poorly in summer? There is evidence of similar seasonality in performance of electrochemical NO2 sensors - they perform reasonably well in winter and worse in summer when there is more potential for ozone interference.
Response: We acknowledge that there is a possibility that other environmental conditions in the summer could influence sensor performance, but we did not find any potential interferences from the other environmental covariates measured at the co-location site that could explain the sensor’s poorer performance in the summer season. The inclusion of the hour of day predictor may account for some of the unmeasured parameters that follow diurnal patterns such as ozone. While the low variability of CH4 concentrations in the summer period may not be the only contributor to poor sensor performance, low variance conditions are known to decrease sensor R2. As the referee has noted, there are still clear peaks in the summer period (up to 2.5 ppm), but these are notably lower than peaks observed in the winter period which go up to 3.2 ppm. The evaluation of sensor performance by season depends on which metric is of most importance. We suspect that the ability of the sensor to detect the large peaks in the winter contributed to the larger R2 in the winter compared to the summer. However, the mean bias is lower in the summer, possibly also related to the lower dynamic range.
3. Figure 4 focuses on the diurnal patterns using the model that includes the hour of day (HOD) as a parameter. I am a bit skeptical of this model because it could "get the right answer for the wrong reason" (e.g., because HOD is correlated with some sort of typical variation in methane concentrations). Are the diurnal patterns similar for the model that does not include HOD?
Response: We observed comparable diurnal patterns for the model without HOD, but found the model without HOD to underestimate CH4 concentrations in the early morning hours leading to worse time series agreement. We have added a discussion of the HOD predictor to the paper and provided comparisons of the diurnal trend and time series with and without this predictor in Supplemental Figure 5. After the description of the diurnal trend in the manuscript, we have added “The inclusion of the hour of day predictor in the calibration model did not change the overall trend of hourly averaged CH4 concentrations but resulted in greater agreement with the reference instrument particularly in the early hours of the morning (6 – 7 am) and the afternoon (1 – 8 pm). Importantly, this addition also resulted in a greater agreement between the sensor and reference time series in the low variance summer season (Supplemental Figure 5). This improvement suggests that the hour of day predictor could be accounting for other pollutants or conditions that were not directly measured in this study which exhibit diurnal trends, such as ozone concentration.”
4. Line 499 - it's not immediately clear that you account for VOC interference. It's not shown directly in your lab or field tests. Perhaps this is addressed in a previous paper, but it does not seem to be directly addressed here.
Response: VOC interference was accounted for in the design of the sensor setup and removed using a carbon-impregnated cloth over the sensing component. For clarity, we added to section 2.1 (Sensor Incorporation and Monitor Design) “To reduce VOC interferences in the sensing component, the sensors used in this study were covered with a layer of activated carbon-impregnated cloth (Zorflex® Double Weave) held in place with a retaining ring that was wrapped on the sides with Teflon tape (see Buehler et al., 2021). Under ambient conditions, the activated carbon in the cloth substrate does not interact with CH4 or CO and potential interactions with water vapor would be weak. The lifetime of this method is dependent on ambient VOC concentrations, but this technique was previously shown to be effective in filtering out ethanol interferences as high as 2% in a controlled laboratory setting and remained effective after continuous outdoor VOC exposure for 3 months (Buehler et al., 2021).” Additional details are provided in the paper by Buehler et al. 2021 cited in section 2.1 (Sensor Incorporation and Monitor Design).
5. Your study has the longest averaging time in Table 3. This perhaps suggests that you see nominally better performance because of signal averaging and hence noise reduction. What is the relevant timescale for monitoring urban methane? Is 1 hr sufficient, or do we need 1 minute data?
Response: We agree that the longer averaging time contributes to the sensor’s observed performance in this study but note that the relevant averaging time is dependent on the study question. In this paper, we assess the use of this sensor for long term environmental monitoring for which 1-hr averaging times can give us sufficient understanding of spatial and seasonal methane trends. Other studies which intend to use the sensor to identify CH4 plumes or to identify specific CH4 emissions sources would benefit from monitoring at a 1-minute resolution. Some of these studies and their aims are referenced in Table 3. To clarify the objective of our study in this inter-study comparison section we added that “This study was the first cross-seasonal deployment of the sensor in urban conditions and evaluated sensor performance at 1-hr resolution over 16 weeks (8 weeks in the winter and 8 weeks in the summer) to determine the suitability of the sensor for long-term deployment for the monitoring of spatial and seasonal CH4 trends.” Additionally, data from the reference Picaro CRDS reference instrument was only available to us in 1-hr resolution. This is noted in the methods portion of the manuscript (line 209) and prevents an assessment of accuracy at higher temporal resolution.
6. Methane was only evaluated up to 3 ppm. Why was that upper limit selected? The winter data in Figure 3 seem to miss a few peaks. Maybe this is a consequence of only testing up to 3 ppm?
Response: The laboratory calibration ranges were selected based on relevant ambient concentration ranges in the published literature prior to monitor development. This range was chosen for laboratory calibration prior to field deployment in our study area. However, a different unit of the same sensor that was tested in our laboratory (not shown) was calibrated between (0 – 5 ppm) and exhibited the same linear response to increasing methane concentration as the sensors tested in this paper. Given the linear responses of the sensor to increasing methane concentrations, we expect that calibration beyond 3 ppm in the laboratory would not substantially change our results.
7. I think a big question in the sensor community is simply "are these good enough for ambient methane measurement?" I would like to see the authors comment on that a bit. You show (Table 2) RMSE for hourly data of ~0.1 ppm. Is that 'good enough' for detecting things like spatial variations or short-term emissions? What are the potential applications of methane sensors that perform to the level you have demonstrated?
Response: We thank this referee for commenting on the real-life applicability of the sensor and have added a paragraph to the end of section 2.2.2 "Despite the elevated sampling inlet, our findings suggest that after adjusting for temperature and humidity dependencies, filtering for VOCs, and correcting for CO cross-sensitivity, the sensor is a valuable supplement to existing monitoring strategies to identify localized trends and CH4 hotspots. With an overall 2.8% bias from the reference and 0.08 ppm RMSE throughout the 16-week deployment, the sensor effectively captures diurnal and seasonal CH4 trends, with notably better performance (R2 = 0.65) in high CH4 variance conditions. Thus, in combination with individualized field calibration prior to deployment, the sensor is a strong candidate for deployment as part of low-cost sensing networks to identify CH4 emissions trends proximal to potential emissions sources.”
Minor comments
-Line 72 cites Brandt et al as evidence for inventories missing methane leaks. My recollection is that the Brandt paper, which focuses on natural gas production, identified under estimation of methane emissions from super emitters as a major problem. So, this reference may not be appropriate.
Response: This has been corrected.
-Line 123 - The Weather Underground reference is not in the bibliography
Response: This has been added.
-Is there a reference for eq 1?
Response: This has been added.
-Line 219 - I assume that "sensor value" in this sentence refers to the resistance.
Response: This has been clarified.
-Line 290 - The stated temperature range is 4-37 C. It never goes below freezing in Baltimore?
Response: The sensor is co-located inside an enclosed indoor monitoring space that is connected to an outdoor sampling inlet. Thus, the temperature of the room did not drop below freezing over the sampling period. We note in section 3.2.2. that the calibration may not cover the full range of expected ambient operating conditions due to the enclosed nature of the NIST measurement station and the constant internal heat generated by the CH4 sensor and other sensing components within the multipollutant monitor. We state “the instrument experienced smaller seasonal temperature and AH variations than were observed by temperature and humidity sensors located outdoors. Over the 8-month deployment, the sensor experienced AH ranging from 3 g/m3 – 24 g/m3, and temperatures from 17°C – 38°C. For comparison, the outdoor AH and temperature recorded during this time ranged from 3 g/m3 – 27 g/m3 and -7°C – 38°C, respectively (MD Weather History). As such, there is less range in the deployment data used to generate coefficients for the temperature and humidity predictors. This could lead to greater uncertainty on these predictors at high humidity and low-temperature conditions if the box were set up fully outdoors.”
Reviewer 2 comments
Comments to the Author
1. This manuscript examines the performance of a low-cost methane sensor and investigates the key environmental factors that affect sensor detection. The authors found that methane detection by the sensor is influenced significantly by temperature, absolute humidity, and CO. Sensor performance varies across seasons with better performance in winter than summer, largely due to the smaller methane concentration variation observed in summer. The measured methane concentration can be corrected by the influencing environmental factors using multi-linear regression. Overall, the manuscript is well written. However, the calibration of Figaro TGS2600 has been largely studied in the literature in the past years. Although the results presented in this study are interesting, they are mostly expected. The novelty of the study needs to be further elaborated.
Response: We thank this referee for their thoughtful comments and have now provided more explicit mentions of the novelty of the study and the advances that this knowledge brings to the low-cost sensor field. In the discussion section, we added “The unique design of the multipollutant sensor box which measured multiple other environmental conditions and pollutants at the sampling site allowed us to account for all the cross-sensitivities indicated by the manufacturer and present more accurate CO corrections than previously available.” The study objective in the abstract has been updated: “This study evaluates sensor performance across seasons with specific attention to the sensor’s understudied carbon monoxide (CO) interferences and environmental dependencies through long-term ambient co-location in an urban environment.” The results section of the abstract has also been updated: “the results highlight the utility of sensor deployment in more variable ambient CH4 conditions and demonstrate the importance of accounting for temperature and humidity dependencies as well as point of measurement CO concentrations with low-cost CH4 measurements”. Aside from the introduction of the novel point of measurement CO correction, this study was also the first cross-seasonal deployment of the sensor in urban conditions. We highlight the seasonal performance differences observed in the field and suggest ideal deployment conditions for future long-term methane monitoring such as setup for ground-level measurement. Thus, in the inter-study comparison paragraph after Table 3 we also added: “This study was the first cross-seasonal deployment of the sensor in urban conditions and evaluated sensor performance at 1-hr resolution over 16 weeks (8 weeks in the winter and 8 weeks in the summer) to determine the suitability of the sensor for long-term deployment for the monitoring of spatial and seasonal CH4 trends.”
2. Abstract: Please add the detection limit (after calibration) of methane sensor used in this study.
Response: Given the high ambient background methane concentrations observed in previous field deployments of this sensor, we did not seek to find a sub-ambient detection limit in this work. However, the paper by Buehler et al., 2021 detailing the setup of the multipollutant monitors found detection at sub-ambient concentrations and linear responses across ambient background concentrations. The following has been added to laboratory calibration results section, “The sensor’s CH4 detection limit was assessed in a previous paper detailing the setup of the multipollutant sensor box and found to be below ambient background concentrations (Buehler et al., 2021). In the laboratory calibration from this study, we observed sensor responses in CH4 concentrations as low as 0.7 ppm, which is significantly lower than ambient background CH4 concentrations reported in environmental studies which range from 1.6 - 1.9 ppm (Collier-Oxandale et al., 2018; Eugster et al., 2019; Riddick et al., 2020).”
3. Lines 152-159: Would the addition of the activated carbon-impregnated cloth also reduce the signal of methane? How long can the cloth be used with good performance? Is there any effect on CO and water removal?
Response: Under ambient conditions, the activated carbon in the cloth substrate will not interact with methane—hence its suitability here. Similarly, carbon monoxide will not sorb to the activated carbon, and interactions with water vapor would be weak at ambient conditions. More details about the sensor setup and VOC reduction from the activated carbon are available in Buehler et al. 2021 which is cited in section 2.1 (Sensor Incorporation and Monitor Design). The purpose of the cloth is removing VOCs as feasible to activated carbon and the lifetime is thus dependent on ambient VOC concentrations. Though we note that the method was effective in filtering out VOC interference resulting from ethanol concentrations as high as 2% even after continuous exposure to outdoor VOC levels for 3 months (Buehler et al). We have added to section 2.1 “Under ambient conditions, the activated carbon in the cloth substrate does not interact with CH4 or CO and potential interactions with water vapor would be weak. The lifetime of this method is dependent on ambient VOC concentrations, but this technique has proven effective in filtering out ethanol interferences as high as 2% in a controlled laboratory setting and remained effective after continuous outdoor VOC for 3 months (Buehler et al., 2021).”
4. Lines 341-343: Please provide the calibration detail (curves) of the CO sensor in the supplemental material.
Response: This has been added to the supplemental material (Supplemental Figure 4) and referenced in the main text.
5. Table 2: Why was the R^2 value of the calibration data smaller than that of the validation, especially for the winter data?
The high R2 during the wintertime validation period is largely influenced by the strong CH4 peaks that are well predicted by the sensor in the last two weeks of the winter season. When the calibration and validation periods are reversed with the calibration period spanning the last 2.5 weeks of the winter season (2/10 - 2/29), the calibration R2 improves to 0.64 (See table below added to supplement as Supplemental Table 3). Before Table 2a in the manuscript, we added: “The R2 from the validation period appears to be largely influenced by the strong CH4 peaks from 2/10 - 2/29 that are well predicted by the sensor. This led to a higher R2 (0.55) in the validation period than the calibration period (R2 = 0.43). When the calibration and validation periods are reversed with the calibration period spanning 2/10 - 2/29, the calibration R2 improves to 0.64 (Supplemental Table 3).”
Supplemental Table 3. Model 5 R2 by different calibration and validation period splits
Calibration/validation period split Calibration R2 Validation R2
1:3
First 2.5 weeks of winter and summer season for calibration, remaining data for validation 0.43 0.55
5:4
First 5 weeks of winter and summer for calibration, remaining data for validation 0.55 0.57
1:3 reverse
Last 2.5 weeks of winter and summer for calibration, remaining data for validation 0.64 0.34
5:4 reverse
Last 5 weeks of winter and summer for calibration, remaining data for validation 0.66 0.46
6. Figure 3B: It seems that the model is not good at predicting high methane concentration. Any explanation?
Response: Our calibration model accounted for all the potential sensor sensitivities listed by the manufacturer. This included corrections for temperature, absolute humidity, and CO as well as VOC removal from the sensing component using a carbon-impregnated cloth wrapped over the sensing component. While this is one of the most comprehensive lists of predictors tested for this sensor, we cannot preclude the possibility of interference from unmeasured parameters which could influence CH4 measurement at high concentrations. Low-cost sensor calibration using regression models is also known to suffer from peak underestimation since regression is a mean-reverting method (Heffernan et al. 2022). Thus, for extreme peak values, the regression prediction can be lower, resulting in underestimation. We have added a statement about regression underestimation and modified the paragraph under Table 2b to read “The calibrated sensor captured CH4 dips and peaks over the co-location period but failed to capture the full range of CH4 variability. In the winter when CH4 concentrations ranged from 1.9 ppm – 3.2 ppm, the calibrated sensor underestimated high concentrations (>2.3 ppm) and overestimated minor spikes in CH4 concentration below 2.3 ppm (Figure 3C). This is consistent with knowledge of potential peak underestimation with mean-reverting methods like regression (Heffernan et al. 2022). The regression approach was chosen despite these limitations given the importance of calibration accessibility and interpretability to a wide range of low-cost sensor users.”
7. Figure 3F: Why not separate the methane calibration into two parts (< 2.3 ppm and > 2.3 ppm)?
Response: We attempted this in our initial analysis using a knot at the median methane concentration but noted negligible improvements to model fit. We have additionally tried splitting the calibration by < 2.3 ppm and > 2.3 ppm as recommended by the referee and found no significant improvements to the model fit. Given the linear relationship between sensor response and CH4 concentration observed in the laboratory calibration, keeping the methane predictor as a simple linear variable is the most interpretable and generalizable method to calibrate the sensor.
8. Figure 4: I got confused. Sensor calibration was performed in a laboratory chamber. Why there was a diurnal trend of methane concentration during calibration? What do you mean by the diurnal pattern in methane concentration during the calibration period?
Response: The sensor was evaluated first in the laboratory (Figures 1-2) to determine important covariates for measurement in the field. The final sensor calibration model was determined using data from the field co-location (Figures 3-4) and included a predictor for the hour of day which could capture atmospheric trends or represent the influence of unmeasured pollutants or environmental conditions that follow diurnal trends. In a comparison of the calibration model with and without the hour of day predictor, accounting for the diurnal trend using the hour of day predictor improved the time series to better capture CH4 peaks and dips throughout the deployment period. After the description of the diurnal trend in the manuscript, we have added “The inclusion of the hour of day predictor in the calibration model resulted in greater agreement with the reference instrument particularly in the early hours of the morning (6 – 7 am) and the afternoon (1 – 8 pm). Importantly, this addition also resulted in a greater agreement between the sensor and reference time series in the low variance summer season (Supplemental Figure 5). This improvement suggests that the hour of day predictor could be accounting for other pollutants or conditions that were not directly measured in this study which exhibit diurnal trends such as ozone concentration.”
9. In addition, it seems that there was a greater variation in methane concentration in the reference measurement than in the sensor measurement. Please explain.
Response: The lower variation in CH4 concentration in the sensor measurement is likely related to the mean reverting nature of the regression calibration method (Heffernan et al. 2022). We have provided details about this issue in our response to comment #6 and updated the text in the discussion accordingly.
10. Can the authors also comment on the greater variation of methane concentration observed during the nighttime as compared to those in the daytime, although this is beyond the scope of the current study?
Response: The greatest variations in CH4 concentration were observed between 4 am and 10 am. This corresponds to the times with the greatest average CH4 concentration. It has been noted that large vertical gradients overnight and into the early morning could be indicative of local anthropogenic sources and respiration from the biosphere (Karion et al., 2020). Additionally, ambient CH4 concentrations are known to vary seasonally and have been observed to vary by diurnal variation in ambient temperature. It is possible that the variability is, in part, attributable to changes in ambient air temperature in the early morning. We did not add this to the discussion of observed CH4 trends as the manuscript is intended to assess the efficacy of this sensor compared to the reference instrument.
11. Table 3: Please add the detection limit of the sensor in each study to the table.
Response: Detection limits are not provided by the manufacturer and have not been reported in previous studies of the sensor. Given the high ambient background methane concentrations observed in previous field deployments of this sensor, we did not seek to find a sub-ambient detection limit in this work. However, the paper by Buehler et al., 2021 detailing the setup of the multipollutant monitors found detection at sub-ambient concentrations and linear responses across ambient background concentrations. The following has been added to laboratory calibration results section “The sensor’s CH4 detection limit was assessed in a previous paper detailing the setup of the multipollutant sensor box and found to be below ambient background concentrations (Buehler et al., 2021). In the laboratory calibration from this study, we observed sensor responses in CH4 concentrations as low as 0.7 ppm, which is significantly lower than ambient background CH4 concentrations reported in environmental studies which range from 1.6 - 1.9 ppm (Collier-Oxandale et al., 2018; Eugster et al., 2019; Riddick et al., 2020).”
12. Lines 493-497: I agree with the authors that a lower sampling height may better capture the characteristics of a local emission. However, a sampling height of 29 m is not very high. Can the authors comment on the scale of the underlying zone of influence for a sampling height of 29 m for methane emission?
Response: We agree with this referee that a 29 m sampling height is not very high compared to many sampling towers which have sampling inlets located more than 70 m above ground to monitor atmospheric concentrations. However, the sensor co-location site at the NIST Northeast corridor urban test bed, which samples from a 29 m sampling inlet, is designed to identify the spatial and temporal trends of anthropogenic greenhouse gas emissions from the entire surrounding urban area (Karion et al., 2020). While part of this effort is related to the identification of urban sources, the elevated sampling inlets were selected to strike a balance between the necessity to observe and distinguish sources and the ability to simulate observations across regional towers spread across the Northeast US using transport and dispersion models. The paragraph has been updated “A possible contributor to the limited hourly CH4 variability during this deployment is the elevated inlet on the measurement tower used for the co-located in-field testing at the NIST Northeast Corridor Urban Test Bed site. The sampling inlets for this project were selected to strike a balance between the identification of anthropogenic greenhouse gas emissions from the surrounding urban area and the ability to simulate observations through transport and dispersion models across regional towers across the Northeast (Karion et al., 2020). However, in denser urban monitoring networks, it is beneficial to measure trace gas concentrations closer to ground-level sources so that finer spatial gradients can be used to identify potential sources and emissions estimates. Since both the reference and low-cost instruments were sampling from an inlet located 29 m above ground level in suburban Baltimore, CH4 emissions were likely more diluted compared to surface level measurements, where sensors may experience larger fluctuations due to closer proximity to CH4 sources. As such, the sensor may be better suited for ground-level measurement with surface-level co-location to provide a wider range of data to train the sensor calibration models.”
13. VOC interference: Buehler et al. (AMT, 2021) examined the effect of VOC on methane sensor performance using high ethanol concentration. Although I agree that VOC may affect the performance of methane detection, can the authors comment on to what degree VOCs at ambient levels can affect the sensor results? Have the authors compared the sensor performance with and without VOC removal during the field deployment?
Response: The manufacturer’s datasheet indicates sensor sensitivity to ethanol and isobutane between 1 – 100 ppm with the strongest responses at higher concentrations (> 4 ppm). While the sensor’s response to testes VOC might be higher than methane at 5 – 10 ppm, the ambient concentrations of VOCs are typically at least 3 – 4 orders of magnitude lower at sub-ppb levels while methane concentrations are around 2 ppm or greater. The paper by Buehler et al. 2021 examined the sensor’s response to methane with and without the activated carbon filter in place (SI figure 7).
14. Other meteorological parameters: in addition to temperature and humidity, wind direction (linked to the source) and speed (dispersion and dilution) may also affect the prediction of methane concentrations. Have the authors considered these factors in the calibration model?
Response: Meteorological parameters like wind direction, speed, and dispersion may affect the measured CH4 concentrations at our sampling site but should not impact the calibration or operation of the sensor, especially since the sensor is located inside and sampling with a controlled flow rate.
15. One-hour resolution data: In this study, the sensor data was compared to Picaro CRDS instrument at 1-hour resolution. Have the authors compared the data at a higher time resolution? How did the sensor perform at higher time resolutions?
Response: Data from the reference Picaro CRDS reference instrument was only available to us in 1-hr resolution. This is noted in the methods portion (section 2.3) of the manuscript. In this paper, we assess the use of this sensor for long term environmental monitoring for which 1-hr averaging times can give us sufficient understanding of spatial and seasonal methane trends. To clarify the objective of our study in this inter-study comparison section we added that “This study was the first cross-seasonal deployment of the sensor in urban conditions and evaluated sensor performance at 1-hr resolution over 16 weeks (8 weeks in the winter and 8 weeks in the summer) to determine the suitability of the sensor for long-term deployment for the monitoring of spatial and seasonal CH4 trends.” Additionally, data from the reference Picaro CRDS reference instrument was only available to us in 1-hr resolution. This is noted in the methods portion of the manuscript (line 209) and prevents an assessment of accuracy at higher temporal resolution.
Round 2
Revised manuscript submitted on 22 12 2022![](https://www.rsc-cdn.org/pubs-core/2022.0.146/content/NewImages/icon-letter.png)
![](https://www.rsc-cdn.org/pubs-core/2022.0.146/content/NewImages/chevron-down.png)
19-Feb-2023
Dear Ms Lin:
Manuscript ID: EA-ART-08-2022-000100.R1
TITLE: Laboratory and Field Evaluation of a Low-cost Methane Sensor and Key Environmental Factors for Sensor Calibration
Thank you for submitting your revised manuscript to Environmental Science: Atmospheres. I am pleased to accept your manuscript for publication in its current form. I have copied any final comments from the reviewer(s) below.
You will shortly receive a separate email from us requesting you to submit a licence to publish for your article, so that we can proceed with the preparation and publication of your manuscript.
You can highlight your article and the work of your group on the back cover of Environmental Science: Atmospheres. If you are interested in this opportunity please contact the editorial office for more information.
Promote your research, accelerate its impact – find out more about our article promotion services here: https://rsc.li/promoteyourresearch.
We will publicise your paper on our Twitter account @EnvSciRSC – to aid our publicity of your work please fill out this form: https://form.jotform.com/211263048265047
How was your experience with us? Let us know your feedback by completing our short 5 minute survey: https://www.smartsurvey.co.uk/s/RSC-author-satisfaction-energyenvironment/
By publishing your article in Environmental Science: Atmospheres, you are supporting the Royal Society of Chemistry to help the chemical science community make the world a better place.
With best wishes,
Dr Claudia Mohr
Associate Editor, Environmental Science: Atmospheres
![](https://www.rsc-cdn.org/pubs-core/2022.0.146/content/NewImages/icon-comments.png)
![](https://www.rsc-cdn.org/pubs-core/2022.0.146/content/NewImages/chevron-down.png)
Reviewer 2
The authors have sufficiently addressed my comments. I don't have further comments. The manuscript is well written. Low-cost sensor techniques have emerged as potential advanced in atmospheric science research. The results presented herein with regard to the long-term measurement of methane using low-cost sensor are valuable to the literature. I would like to recommend its publication in Environmental Science: Atmospheres.
Reviewer 1
Overall the authors have responded to my comments from the initial round of reviews and I think this manuscript is worthy of publication. I agree wth Reviewer 2's comment that calibrations for the Figaro sensors have been previously published. This means that most of the novelty of this study relies on testing the sensors in an urban environment (previous studies were mostly in rural or natural gas production areas) and the direct inclusion of CO interferences. Nonetheless, I think this work merits publication in its current form.
One minor comment is that Fig 3A and 3E may not be colorblind friendly.
Transparent peer review
To support increased transparency, we offer authors the option to publish the peer review history alongside their article. Reviewers are anonymous unless they choose to sign their report.
We are currently unable to show comments or responses that were provided as attachments. If the peer review history indicates that attachments are available, or if you find there is review content missing, you can request the full review record from our Publishing customer services team at RSC1@rsc.org.