Remote sensing of methane point sources with the MethaneAIR airborne spectrometer

Guanter, Luis; Warren, Jack; Omara, Mark; Chulakadabba, Apisada; Roger, Javier; Sargent, Maryann; Franklin, Jonathan E.; Wofsy, Steven C.; Gautam, Ritesh

doi:https://doi.org/10.5194/egusphere-2024-3577

Preprints

https://doi.org/10.5194/egusphere-2024-3577

Preprints

02 Jan 2025

| 02 Jan 2025

Remote sensing of methane point sources with the MethaneAIR airborne spectrometer

Luis Guanter, Jack Warren, Mark Omara, Apisada Chulakadabba, Javier Roger, Maryann Sargent, Jonathan E. Franklin, Steven C. Wofsy, and Ritesh Gautam

Abstract. The MethaneAIR imaging spectrometer was originally developed as an airborne demonstrator of the MethaneSAT satellite mission. MethaneAIR enables accurate methane concentration retrievals from high spectral resolution measurements in the 1650-nm methane absorption feature at a nominal spatial sampling of 5x25 m. In this work, we present a computationally-efficient data processing chain optimized for the detection and quantification of methane plumes with MethaneAIR. It involves the retrieval of methane concentration enhancements (ΔXCH₄) with the high-precision matched-filter retrieval, which is applied to 1650-nm retrievals for the first time. Methane plumes are detected through visual inspection of the resulting ΔXCH₄ maps. Flux rates are estimated from the detected plumes using the integrated mass enhancement (IME) method. The evaluation of the proposed methods included comparisons with simulated plumes, with existing plume retrieval and quantification methods for MethaneAIR, and with controlled methane releases. We applied our processing chain to MethaneAIR at-sensor radiance datasets acquired over the Permian Basin during flights in 2021 and 2023, which resulted in the detection of hundreds of point sources above 100–200 kg/h, with a conservative detection limit around 120 kg/h. Our results show the consistency of MethaneAIR's ΔXCH₄ matched-filter retrievals, and their potential for the detection and quantification of methane point sources across large areas.

Received: 15 Nov 2024 – Discussion started: 02 Jan 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 7687 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (7687 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

14 Aug 2025

Detection and quantification of methane plumes with the MethaneAIR airborne spectrometer

Luis Guanter, Jack Warren, Mark Omara, Apisada Chulakadabba, Javier Roger, Maryann Sargent, Jonathan E. Franklin, Steven C. Wofsy, and Ritesh Gautam

Atmos. Meas. Tech., 18, 3857–3872, https://doi.org/10.5194/amt-18-3857-2025,https://doi.org/10.5194/amt-18-3857-2025, 2025

Short summary

Luis Guanter, Jack Warren, Mark Omara, Apisada Chulakadabba, Javier Roger, Maryann Sargent, Jonathan E. Franklin, Steven C. Wofsy, and Ritesh Gautam

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-3577', Anonymous Referee #1, 18 Jan 2025

The authors provide a straightforward reanalysis of MethaneAir data that has been processed with a more point source centric algorithm. The paper is clear, well written, and very interesting to see a matched filter algorithm applied to a high spectral resolution instrument like MethaneAir. I do have a few comments, most important Comment #4, as I think these results have broad implications around the importance of point sources generally. I ask the authors to clarify and provide this context before I recommend for submission.

1. Line 145. Can you confirm or clarify how the injection of WRF-LES concentrations was performed? You calculate transmission due to extra CH4 column concentration and then apply to MethaneAir radiance?
2. Line 194. Why aren't matched filter retrievals suitable for estimation of total area budgets? Because the background normalization in an MF algorithm "removes" regional gradients? Ultimately in an area flux inversion, one has to create XCH4 enhancements relative to the background for assimilation. If one plotted retrieved XCH4 enhancements derived from matched filters vs. CO2 proxy, and they correlated reasonably well, I don't see why a matched filter algorithm couldn't be used. Please explain.
3. Figure 8. Why are only 8 data points shown here, when El Abbadi et al. (2024) reports 24 controlled releases were performed for MethaneAir? Shouldn't all points be shown? Were there plumes that didn't perform well with this new algorithm applied, hence they are not shown?
4. Please include point source datasets as part of the SI
5. Related to comment #4 - I am curious about how the improvement on detection limit affects the general understanding of point vs area sources, which appears to be a central mission thrust of MethaneAir. Looking at other datasets that are available online (https://showcase.earthengine.app/view/methanesat), I count 28 point sources that were detected from RF06, while this study reports 121. Relatedly, that dataset on Google Earth Engine states that point sources make up 33,700 kg/h compared to a total flux of 91,000 kg/h (37%). How much methane total do you now quantify from point sources using this new matched filter algorithm? It appears that it would have to be higher, potentially much higher. As some bottom-up studies have leveraged MethaneAir to suggest a small contribution from emission sources above 100 kg/h (e.g., https://doi.org/10.5194/egusphere-2024-1402), it appears that the conclusions from those studies may have been an artifact of point source detection limit. Though it is out of scope for this paper to comment on those studies, it is appropriate for you to state how much total CH4 there is from the Permian scenes you processed, and how that relates to total fluxes derived from the CO2 proxy method.

Citation: https://doi.org/10.5194/egusphere-2024-3577-RC1
- AC2: 'Reply on RC2', Luis Guanter, 07 Mar 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2024-3577/egusphere-2024-3577-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2024-3577-AC2
RC2:
'Comment on egusphere-2024-3577', Anonymous Referee #2, 24 Jan 2025

The authors provide CH4 retrieval and emission quantification methods for the MethaneAIR imaging spectrometer based on a matched filter and integrated mass enhancement method. The manuscript is well written and provides interesting results. The manuscript does not have a code and data availability section. I have some comments on the methodology that should be addressed for me to recommend the paper for publication.
Introduction
L20ff: The grouping of methane imagers based on 1600 nm and 2300 nm windows is a bit arbitrary. I would argue that the main difference between AVIRIS and MAMAP-2D instruments are the difference in spatial and spectral resolution.
L40f: Please define area sources. Is a landfill already an area source?
L47ff: Maybe already explain here why CO2-proxy retrievals are less precise than matched filters.

Method
Figure 1: Instead of arbitrary spectra for CH4, CO2 and H2O, it would be nice show spectra for typical atmospheric concentrations.
L86ff: Foote et al. (2020) introduces an albedo correction term to remove systematic errors in XCH4 plumes due to deviations between the mean spectrum and the local spectrum. The systematic errors are likely to introduce systematic errors in the emission estimates. I think it is necessary to test if the albedo correction affects the results.
L95: How do you account for varying observation angles and surface elevation during data acquisition?
L101f: The small number of samples also affect the mean vector. Did you test the effect of computing the mean vector for a larger sample on your retrieval?
L114f: Kuhlmann et al. (2024, https://doi.org/10.5194/egusphere-2024-3494) identified CH4 emissions from vent stack in Romania using AVIRIS-NG that were not visible in high-resolution images. How many plumes did you reject, because they are not linked to any infrastructure, and do you see the possibility that you miss such sources in your analysis?
L113: Do you use the plume length or the square root of the detectable plume area as length scale L?
L128ff: Effective wind speed also depends on emission height and vertical mixing. Maasakkers et al. (2022) derive their empirical equation for a landfill, which I would assume, emits near the surface, while emissions from oil and gas can be elevated from vent stacks or on top of processing facilities. How do you account for this in your method?
L139f: Please provide more information about the DI method.
L143ff: Section 2.4 does not provide enough information to judge the accuracy of the end-to-end simulator. I would assume that it does not include systematic errors in the plume, which might explain why Figure 5 shows good agreement between retrieval and input. I suggest to either remove the end-to-end simulator from the manuscript or provide more details including a more detailed analysis, which should be quite interesting.
Results
Figure 2: Please add a (rough) scale to the image.
L182ff / Figure 4: I really would like to see the difference between proxy and matched filter (as in Fig. 5). Do you find systematic differences between the methods, in particular inside the plume, what might be the reason, and how would they affect your emission estimate?
L198ff: (see my previous comment on the end-to-end simulator)
L207ff: Please provide more information how the DI method has been implemented in this study. Following Chulakadabba et al. (2023), the DI method sums over all pixels along rectangular for difference from the source location to the edge of the detectable plume (except for subtracting the background, which would be about zero for the matched filter). This isn't much different from the IME method, which sums over all pixels in the detectable plume, while excluding the background (which is close to zero).
The main difference between IME and DI method seems to be the effective wind speed. What wind speed is used for the DI method and how does compare to effective wind speed used with the IME method? Would the differences vanish if the same data source (GEOS-FP or HRRR) for the wind speed is used for both methods?
L223f: It would interesting do discuss the potential for systematic errors inside the plume here (see my previous comment).
L226f: I do not see why the detection limit should not always increase with wind speed. However, detection limit can depend on other factors such as turbulence.
L238f: What is the uncertainty of the emission rates from the controlled releases?
Figure 8: Why is the RMSE measured in ppb?
L280ff: Does the wind speed vary spatially and temporally in the campaign area during data acquisition? I expect that would affect the detection limit during the campaign.
Figure 12: Do you see a dependency of flux rates on wind speed?
L292: Does the albedo change between the campaigns?
L292f: It is unclear what do you mean with "enhanced spatial variability of ΔXCH4".

Conclusions
L312f: Please specify why a computationally efficient retrieval would improve the detection limit.
L332: I suggest adding that a major advantage is minimizing the number of false positives.

Citation: https://doi.org/10.5194/egusphere-2024-3577-RC2
- AC1: 'Reply on RC2', Luis Guanter, 07 Mar 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2024-3577/egusphere-2024-3577-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2024-3577-AC1
- AC2: 'Reply on RC2', Luis Guanter, 07 Mar 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2024-3577/egusphere-2024-3577-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2024-3577-AC2

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-3577', Anonymous Referee #1, 18 Jan 2025

The authors provide a straightforward reanalysis of MethaneAir data that has been processed with a more point source centric algorithm. The paper is clear, well written, and very interesting to see a matched filter algorithm applied to a high spectral resolution instrument like MethaneAir. I do have a few comments, most important Comment #4, as I think these results have broad implications around the importance of point sources generally. I ask the authors to clarify and provide this context before I recommend for submission.

1. Line 145. Can you confirm or clarify how the injection of WRF-LES concentrations was performed? You calculate transmission due to extra CH4 column concentration and then apply to MethaneAir radiance?
2. Line 194. Why aren't matched filter retrievals suitable for estimation of total area budgets? Because the background normalization in an MF algorithm "removes" regional gradients? Ultimately in an area flux inversion, one has to create XCH4 enhancements relative to the background for assimilation. If one plotted retrieved XCH4 enhancements derived from matched filters vs. CO2 proxy, and they correlated reasonably well, I don't see why a matched filter algorithm couldn't be used. Please explain.
3. Figure 8. Why are only 8 data points shown here, when El Abbadi et al. (2024) reports 24 controlled releases were performed for MethaneAir? Shouldn't all points be shown? Were there plumes that didn't perform well with this new algorithm applied, hence they are not shown?
4. Please include point source datasets as part of the SI
5. Related to comment #4 - I am curious about how the improvement on detection limit affects the general understanding of point vs area sources, which appears to be a central mission thrust of MethaneAir. Looking at other datasets that are available online (https://showcase.earthengine.app/view/methanesat), I count 28 point sources that were detected from RF06, while this study reports 121. Relatedly, that dataset on Google Earth Engine states that point sources make up 33,700 kg/h compared to a total flux of 91,000 kg/h (37%). How much methane total do you now quantify from point sources using this new matched filter algorithm? It appears that it would have to be higher, potentially much higher. As some bottom-up studies have leveraged MethaneAir to suggest a small contribution from emission sources above 100 kg/h (e.g., https://doi.org/10.5194/egusphere-2024-1402), it appears that the conclusions from those studies may have been an artifact of point source detection limit. Though it is out of scope for this paper to comment on those studies, it is appropriate for you to state how much total CH4 there is from the Permian scenes you processed, and how that relates to total fluxes derived from the CO2 proxy method.

Citation: https://doi.org/10.5194/egusphere-2024-3577-RC1
- AC2: 'Reply on RC2', Luis Guanter, 07 Mar 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2024-3577/egusphere-2024-3577-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2024-3577-AC2
RC2:
'Comment on egusphere-2024-3577', Anonymous Referee #2, 24 Jan 2025

The authors provide CH4 retrieval and emission quantification methods for the MethaneAIR imaging spectrometer based on a matched filter and integrated mass enhancement method. The manuscript is well written and provides interesting results. The manuscript does not have a code and data availability section. I have some comments on the methodology that should be addressed for me to recommend the paper for publication.
Introduction
L20ff: The grouping of methane imagers based on 1600 nm and 2300 nm windows is a bit arbitrary. I would argue that the main difference between AVIRIS and MAMAP-2D instruments are the difference in spatial and spectral resolution.
L40f: Please define area sources. Is a landfill already an area source?
L47ff: Maybe already explain here why CO2-proxy retrievals are less precise than matched filters.

Method
Figure 1: Instead of arbitrary spectra for CH4, CO2 and H2O, it would be nice show spectra for typical atmospheric concentrations.
L86ff: Foote et al. (2020) introduces an albedo correction term to remove systematic errors in XCH4 plumes due to deviations between the mean spectrum and the local spectrum. The systematic errors are likely to introduce systematic errors in the emission estimates. I think it is necessary to test if the albedo correction affects the results.
L95: How do you account for varying observation angles and surface elevation during data acquisition?
L101f: The small number of samples also affect the mean vector. Did you test the effect of computing the mean vector for a larger sample on your retrieval?
L114f: Kuhlmann et al. (2024, https://doi.org/10.5194/egusphere-2024-3494) identified CH4 emissions from vent stack in Romania using AVIRIS-NG that were not visible in high-resolution images. How many plumes did you reject, because they are not linked to any infrastructure, and do you see the possibility that you miss such sources in your analysis?
L113: Do you use the plume length or the square root of the detectable plume area as length scale L?
L128ff: Effective wind speed also depends on emission height and vertical mixing. Maasakkers et al. (2022) derive their empirical equation for a landfill, which I would assume, emits near the surface, while emissions from oil and gas can be elevated from vent stacks or on top of processing facilities. How do you account for this in your method?
L139f: Please provide more information about the DI method.
L143ff: Section 2.4 does not provide enough information to judge the accuracy of the end-to-end simulator. I would assume that it does not include systematic errors in the plume, which might explain why Figure 5 shows good agreement between retrieval and input. I suggest to either remove the end-to-end simulator from the manuscript or provide more details including a more detailed analysis, which should be quite interesting.
Results
Figure 2: Please add a (rough) scale to the image.
L182ff / Figure 4: I really would like to see the difference between proxy and matched filter (as in Fig. 5). Do you find systematic differences between the methods, in particular inside the plume, what might be the reason, and how would they affect your emission estimate?
L198ff: (see my previous comment on the end-to-end simulator)
L207ff: Please provide more information how the DI method has been implemented in this study. Following Chulakadabba et al. (2023), the DI method sums over all pixels along rectangular for difference from the source location to the edge of the detectable plume (except for subtracting the background, which would be about zero for the matched filter). This isn't much different from the IME method, which sums over all pixels in the detectable plume, while excluding the background (which is close to zero).
The main difference between IME and DI method seems to be the effective wind speed. What wind speed is used for the DI method and how does compare to effective wind speed used with the IME method? Would the differences vanish if the same data source (GEOS-FP or HRRR) for the wind speed is used for both methods?
L223f: It would interesting do discuss the potential for systematic errors inside the plume here (see my previous comment).
L226f: I do not see why the detection limit should not always increase with wind speed. However, detection limit can depend on other factors such as turbulence.
L238f: What is the uncertainty of the emission rates from the controlled releases?
Figure 8: Why is the RMSE measured in ppb?
L280ff: Does the wind speed vary spatially and temporally in the campaign area during data acquisition? I expect that would affect the detection limit during the campaign.
Figure 12: Do you see a dependency of flux rates on wind speed?
L292: Does the albedo change between the campaigns?
L292f: It is unclear what do you mean with "enhanced spatial variability of ΔXCH4".

Conclusions
L312f: Please specify why a computationally efficient retrieval would improve the detection limit.
L332: I suggest adding that a major advantage is minimizing the number of false positives.

Citation: https://doi.org/10.5194/egusphere-2024-3577-RC2
- AC1: 'Reply on RC2', Luis Guanter, 07 Mar 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2024-3577/egusphere-2024-3577-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2024-3577-AC1
- AC2: 'Reply on RC2', Luis Guanter, 07 Mar 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2024-3577/egusphere-2024-3577-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2024-3577-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

AR by Luis Guanter on behalf of the Authors (07 Mar 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (24 Mar 2025) by Dominik Brunner

RR by Anonymous Referee #1 (03 Apr 2025)

Suggestions for revision or reasons for rejection

I appreciate the authors' responses to my comments and feel the manuscript has greatly improved. I also appreciate the inclusion of the data with the SI. I still have a few lingering comments before I can recommend for publication.

1. Responding to my original comment #3 - in an effort to be transparent, it is still important to show the controlled release results from 29 October. What do you mean by more stable winds? How are you sure that the conditions in the Permian were of sufficient stability to quantify emissions? It's useful to see controlled release results in a less ideal environment because it gives a general sense of how accuracy depends on environmental conditions out of one's control.

2. Regarding my original comment #5 on point vs area sources. Appreciate the response, and I suppose it could be argued that it is out of scope to talk area fluxes in this manuscript, though I generally disagree. That said, putting area fluxes aside, I do believe the newly proposed text is misleading. A central focus of this manuscript is that by application of a matched filter algorithm, the effective detection limit of MethaneAir has reduced. However, in the authors' response, they filter out all the new plumes they detected with this new algorithm for their comparison with the CO2_proxy algorithm - i.e., after filtering out plumes > 200 kg/h, then summing their plume dataset, they get a similar emission total. This greatly undermines what makes this study so compelling!

What I would like to see is one additional sentence or two where the authors describe that when summing all plumes detected from the CO2 proxy approach, they get 26.7 t/h, but when they sum all plumes detected with the matched filter approach, they get 35.9 t/h (I'm just summing the provided RF06 list of detections). This means that there are 9.2 t/h more emissions attributed to point sources that is purely the result of an algorithm update.

This is a fundamentally important point as observing systems are not just the result of instrument specifications but are a result of the algorithms applied to them, and for point source detection, the plume detection protocols as well. And depending on algorithm, you may come to slightly different conclusions about an area/process/etc. The authors should do more highlight this point, in my opinion.

Hide

RR by Anonymous Referee #2 (07 Apr 2025)

ED: Publish subject to minor revisions (review by editor) (23 Apr 2025) by Dominik Brunner

AR by Luis Guanter on behalf of the Authors (08 May 2025) Author's response Author's tracked changes Manuscript

ED: Publish as is (12 May 2025) by Dominik Brunner

AR by Luis Guanter on behalf of the Authors (19 May 2025)

Journal article(s) based on this preprint

14 Aug 2025

Detection and quantification of methane plumes with the MethaneAIR airborne spectrometer

Luis Guanter, Jack Warren, Mark Omara, Apisada Chulakadabba, Javier Roger, Maryann Sargent, Jonathan E. Franklin, Steven C. Wofsy, and Ritesh Gautam

Atmos. Meas. Tech., 18, 3857–3872, https://doi.org/10.5194/amt-18-3857-2025,https://doi.org/10.5194/amt-18-3857-2025, 2025

Short summary

Luis Guanter, Jack Warren, Mark Omara, Apisada Chulakadabba, Javier Roger, Maryann Sargent, Jonathan E. Franklin, Steven C. Wofsy, and Ritesh Gautam

Viewed

Total article views: 743 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
521	202	20	743	12	35

HTML: 521
PDF: 202
XML: 20
Total: 743
BibTeX: 12
EndNote: 35

Views and downloads (calculated since 02 Jan 2025)

Month	HTML	PDF	XML	Total
Jan 2025	237	62	7	306
Feb 2025	33	20	1	54
Mar 2025	43	23	2	68
Apr 2025	40	19	3	62
May 2025	43	16	2	61
Jun 2025	40	27	5	72
Jul 2025	65	24	0	89
Aug 2025	20	11	0	31

Cumulative views and downloads (calculated since 02 Jan 2025)

Month	HTML	PDF	XML	Total
Jan 2025	237	62	7	306
Feb 2025	33	20	1	54
Mar 2025	43	23	2	68
Apr 2025	40	19	3	62
May 2025	43	16	2	61
Jun 2025	40	27	5	72
Jul 2025	65	24	0	89
Aug 2025	20	11	0	31

Viewed (geographical distribution)

Total article views: 690 (including HTML, PDF, and XML) Thereof 690 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 14 Aug 2025

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (7687 KB)
Metadata XML

Short summary

This study presents a data processing scheme for the detection and quantification of methane emissions using the MethaneAIR airborne spectrometer. We show that the proposed methods enable the detection of smaller plumes than other existing methods, and that improves the potential of MethaneAIR to survey methane point sources across large regions


Total:	0
HTML:	0
PDF:	0
XML:	0