the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 31 Mar 2025
Performance Evaluation of Multi-Source Methane Emission Quantification Models Using Fixed-Point Continuous Monitoring Systems
Abstract. Quantifying methane emissions from oil and gas facilities is crucial for emissions management and accurate facility-level GHG inventory development. This paper evaluates the performance of several multi-source methane emission quantification models using the data collected by fixed-point continuous monitoring systems as part of a controlled release experiment. Two dispersion modeling approaches (Gaussian plume, Gaussian puff) and two inversion frameworks (least-squares optimization and Markov-Chain Monte-Carlo) are applied to the measurement data. In addition, a subset of experiments are selected to showcase the application of computational fluid dynamic (CFD) informed calculations for direct solution of the advection-diffusion equation. This solution utilizes a three-dimensional wind field informed by solving the momentum equation with the appropriate external forcing to match on-site wind measurements. Results show that the Puff model, driven by high-frequency wind data, significantly improves localization and reduces bias and error variance compared to the Plume model. The Markov-Chain Monte-Carlo (MCMC) based inversion framework further enhances accuracy over least-squares fitting, with the Puff MCMC approach showing the best performance. The study highlights the importance of long-term integration for accurate total mass emission estimates and the detection of anomalous patterns. The findings of this study can help improve emissions management strategies, aid in facility-level emissions risk assessment, and enhance the accuracy of greenhouse gas inventories.
- Preprint
(4026 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-1266', Anonymous Referee #1, 06 May 2025
Reviewer of “Performance Evaluation of Multi-Source Methane Emissions Quantification Models Using Fixed-Point Continuous Monitoring Systems”
May 1st, 2025
Dear Authors,
This is a super interesting paper! The authors present a comprehensive evaluation of several techniques for multi-source methane emissions detection and quantification using fixed-point continuous monitoring systems. The techniques include two dispersion models including Gaussian plume and Gaussian puff, two inversion models including least-square optimization and Markov-Chain Monte-Carlo, and a computational fluid dynamics (CFD) model. The authors find that the combination Gaussian puff and Monte Carlo simulation is the most accurate in localization and detection of methane leaks compared to the other models and is not as computationally expensive as a CFD model. The manuscript is well written and a good fit for AMT. I recommend accepting it for publication after correcting for the following comments and questions:
General:
- Greenhouse gas inventories and facility operations – Could the authors elaborate on how this technique will aid facilities in improving their greenhouse gas inventories? It is mentioned in the last line of the abstract, but I’d like to see a paragraph on the application of these methods to oil and gas facility operators, how it could be implemented, and how easy would the models be for operators to use?
- Length/brevity – The authors clearly know these models well and have robust statistical evidence for the accuracy of the models, but this paper is around 50 pages long. I would like to see the authors work on making the paper more concise. Perhaps some of the introduction can be cut or combined to make the thesis of the paper clearer. Also, maybe some parts of Section 3 Methodology can be moved to a supplement, and the detailed explanation of the summary statistics to the supplement (keeping some of the explanation in the main body obviously). Right now, the introduction and methodology read like a literature review, so I’d like to see it condensed to focus on the main point of the paper that the Puff/MCMC model works best for the CM system and the data explaining why. Another suggestion could be to add the description of the figures to the figure captions themselves instead of in the body of the paper.
- Meteorological Conditions – I’d like to see in the discussion how the meteorological conditions could affect the outcome of the various models, for example if it’s very windy would more leaks be missed? How does meteorology affect the performance of the CMS emissions models?
- Paper thesis - I’d like to see more discussion on why the Puff model is the most reasonable approach operationally. Is the point of the paper to compare different measurement techniques or to offer a solution for facilities to implement?
- Cumulative facility emissions rate – Could you elaborate on the sentence in the conclusion on the cumulative mass emissions estimate (L947)? I think this is critical for understanding total flux emissions from a facility and deserves more attention. Maybe a table comparing all the methods together and how they compare them to what METEC reported.
- Validation – could you elaborate more on validating the models against the reported METEC emissions and locations? (goes with the previous comment)
- Number of CMS instruments – How many CMS instruments would be needed to perform an accurate DLQ using the Puff/MCMC method (and other methods as well)?
Specific:
- Introduction – Add references for factual statements L16, L19, L26, L43 (AVO), L45 (OGI), L73 (‘smoke alarms’), L75 (CMS development), L859 (why Appalachia?)
- L12 – “anomalous (emissions) patterns”?
- L55 – Which one, 1-2 kg/hr or 200 kg/hr?
- The last two paragraphs of the intro are good. More of this.
- Data – Conceptually, I’m having a hard time understanding how a controlled release works, are there only 5 locations where a leak could be? Does each leak have a different release rate? Could my confusion be cleared up by a more detailed figure?
- L205 – This is good and targets the scope of the paper, move this to the intro?
- Figure 2 – Make the wind data into a wind rose? Use lat/lon for the location figure instead of x and y? Are the colors in the concentration figure for each instrument?
- L290 – How do you determine k? Is it a measurement?
- L317 – remove quotes from Pasquill, it is a noun (person).
- L345 – Could you elaborate on which ‘fewer assumptions’ are made in Gaussian Puff?
- Section 3.3 – I like the beginning of the first paragraph that explains the questions this study is attempting to answer – could you make this clearer in the introduction? Maybe move this to the intro?
- Figure 8/9 – Move to supplement?
Overall, this is an interesting paper and valuable to understanding facility level methane emissions, good job!
Citation: https://doi.org/10.5194/egusphere-2025-1266-RC1 -
AC1: 'Reply on RC1', Ali Lashgari, 30 May 2025
Dear Editor and Referees,
We would like to express their sincere gratitude for your thorough review and valuable feedback on our manuscript. We appreciate the time and effort you have invested in volunteering to evaluate our work. We believe that your insightful comments and suggestions have been instrumental in improving the quality and clarity of our manuscript.
The attached document provides details on how we have addressed each of your comments (highlighted in plum), including revisions that we made to the manuscript. All the added and eliminated/reorganized text are highlighted in green and strikethrough, respectively. We are also submitting an annotated copy of the manuscript that tracks additions for the benefit of the reviewers and editor. We hope that our responses and the revised manuscript meet your expectations and address your concerns. Once again, we appreciate your constructive feedback and thank you for considering our manuscript for publication.
Sincerely,
Ali Lashgari
-
RC2: 'Comment on egusphere-2025-1266', Anonymous Referee #2, 27 May 2025
Review of Performance Evaluation of Multi-Source Methane Emission Quantification Models Using Fixed-Point Continuous Monitoring Systems
The manuscript studies detection, localization, and quantification (DLQ) of methane leaks using two methods (Gaussian plume and puff) and two inversion techniques (least squares and MCMC). Field data is obtained from METEC with up to five release locations per experiment and sensing with a network of continuous monitors. The puff MCMC method shows advantages in terms of error (and especially in reducing false positives and negatives), but is also the most complex of the four. CFD on a subset of the experiments is also investigated. Overall the work is topical and results are of interest, however I feel the results could be fleshed out a little more, as described in #2 below,
Main
- Half the metrics are directly related to detection (TP, FP, FN, TN, N(L=5), Lbar), however how is detection actually done in the different methods?
- A lot of space spent on the models and metrics, comprising Section 3, a lot of which is already described in the literature. I think it could be shortened, if possible, to better highlight the results Section 4. A key novelty of the manuscript is the multi-source estimation estimate, along with large number of experiments with continuous monitors. Many analyses could be envisioned, in particular detection curve vs emission rate, whether any equipment groups perform better (perhaps due to prevailing wind patterns or other factors), simulating if there were fewer sensors (as mentioned might be realistic), interference (if small leaks sometimes are hidden by larger ones), effect of experiment time vs DLQ accuracy (30 minutes vs 8 hours), etc.
Minor
L43 add AVO in parentheses
Cheptonui 2024 a/b are same paper
Sec 2 – for releases at multiple equipment groups, does each release rate simply randomly belong to the overall distribution in Fig 1b?
L270 I am surprised to characterize the importance of the stability class and dispersion coefficient parameterization as minimal. Sure they may be representing the same behavior, but don’t they empirically have a significant effect?
L275 GMP -> GPM
Eq 8 (for the Gaussian puff) seems unusual – please check. Normally there is a 2*pi^(3/2) factor, and importantly, time dependence
L439 Is a no flux condition typical in these simulations? A 200 m boundary seems like it would significantly affect the dispersion behavior
Fig 5 some information is not visible on this plot (puff LSQ on L=4 and plume MCMC on L = 1)
Figure 11 suggest using different marker types, as the colors by themselves can be difficult to distinguish
L173 Inverse distance weighting is mentioned, and “sonic anemometers” (plural) here, but is unclear to me where/how many where used. Could this be added to Fig 2 or otherwise?
Citation: https://doi.org/10.5194/egusphere-2025-1266-RC2 -
AC2: 'Reply on RC2', Ali Lashgari, 30 May 2025
Dear Editor and Referees,
We would like to express their sincere gratitude for your thorough review and valuable feedback on our manuscript. We appreciate the time and effort you have invested in volunteering to evaluate our work. We believe that your insightful comments and suggestions have been instrumental in improving the quality and clarity of our manuscript.
The attached document provides details on how we have addressed each of your comments (highlighted in plum), including revisions that we made to the manuscript. All the added and eliminated/reorganized text are highlighted in green and strikethrough, respectively. We are also submitting an annotated copy of the manuscript that tracks additions for the benefit of the reviewers and editor. We hope that our responses and the revised manuscript meet your expectations and address your concerns. Once again, we appreciate your constructive feedback and thank you for considering our manuscript for publication.
Sincerely,
Ali Lashgari
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 274 | 103 | 18 | 395 | 9 | 18 |
- HTML: 274
- PDF: 103
- XML: 18
- Total: 395
- BibTeX: 9
- EndNote: 18
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1