the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 01 Nov 2023
Using metal oxide gas sensors for the estimate of methane controlled releases: reconstruction of the methane mole fraction time-series and quantification of the release rates and locations
Abstract. Fugitive methane (CH4) emission occur in the whole chain of oil and gas production, from the extraction, transportation, storage and distribution. The detection and quantification of such emissions are conducted usually from surveys as close as possible to the source location. However, these surveys are labor intensive, costly and they do not provide continuous monitoring of the emissions. The deployment of permanent networks of sensors in the vicinity of industrial facilities would overcome the limitations of surveys by providing accurate estimates thanks to continuous sampling of the plumes. High precision instruments are too costly to deploy in such networks. Low-cost sensors like Metal oxide semiconductors (MOS) are presented as a cheap alternative for such deployments due to its compact dimensions and to its sensitivity to CH4. In this study we test the ability of two types of MOS sensors from the manufacturer Figaro® (TGS 2611-C00 and TGS 2611-E00) deployed in six chambers to reconstruct an actual signal from a source in open air corresponding to a series of controlled CH4 releases and we assess the accuracy of the emission estimates computed from reconstructed CH4 mole fractions from voltages measurements of these sensors. A baseline correction of the voltage linked to background variations is presented based on an iterative comparison of neighboring observations. Two reconstruction models were compared, multilayer perceptron (MLP) and 2nd degree polynomial, providing similar performances meeting our target requirement on all the chambers when the input variable is the TGS 2611-C00 sensor. The emission estimates were then computed using an inversion approach based on the adjoint of a Gaussian dispersion model obtaining promising results with an emission rate error of 25 % and a location error of 9.5 m.
- Preprint
(36404 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on amt-2023-52', Anonymous Referee #1, 20 Nov 2023
Title: Using metal oxide gas sensors for the estimate of methane controlled releases: reconstruction of the methane mole fraction time-series and quantification of the release rates and locations
Martinez et al.
Review
This paper describes how metal oxide sensors can be used quantify the methane concentration in air. These concentrations are used to calculate an emission and the calculated emission rate compared against known rate to generate some idea of uncertainty in emission location and quantification.
General Comments
The manuscript is poorly written and difficult to read in some parts (for example, the title is quite odd). The concept isn’t very novel and several studies have presented the conversion of voltages output by Figaro MOS sensors into mixing ratios. The novelty is the methods by which the emission location and the emission rate are calculated. Details on both processes is lacking and it is currently unclear how novel the approaches used here are. Also, this work would benefit from a deeper dive into the applicability of the approach in the real world. Details are given below.
Specific Comments
Abstract
Language needs to be improved throughout.
There are no details of the tests here: flow rates, meteorological conditions, time of year, how long the test were run.
Too much extraneous detail between lines 7 and 14. Should be condensed.
Last sentence (L14 to 16) doesn’t make sense.
Introduction
Throughout the introduction - It reads like a list of barely related bullet points.
This needs to be rewritten and resubmitted for review. Currently, I find this very hard to comment on as there are too many inaccuracies and poorly written sentences. I have picked out some examples but this list is not exhaustive.
For example:
At L18: “Fossil fuel anthropogenic methane (CH4) emissions related to the production, exploitation and transport of coal, oil and natural gas, account for 35% of global anthropogenic emissions (Saunois et al., 2020).”
Isn’t all incorrect, only poorly written. It would be better written as:
A recent study suggests that in 20XX methane (CH4) emissions from the production, transportation, storage and distribution of fossil fuels (e.g. coal, oil and natural) account for 35% of global anthropogenic CH4 emissions (Saunois et al., 2020).
Then delete the next sentence.
L26: “Atmospheric measurements” is too vague. Please describe what is being measured. “The downwind measurements CH4 concentration in the air” (or similar).
L37: “limiting the impact of atmospheric transport modelling uncertainties”. This is the key issue with networks of sensors.
L38: “Chamberland and Veeravalli (2006)” this is an older study to be referencing. Also, the main issue with quantifying fugitive emissions on production sites is differentiating the vented and combustion emissions from fugitives.
L40: “dense network” another key issue with monitoring O&G productions sites. How many sensors? Typically, only four are deployed by a solution provider regardless of site size. Key question is how to optimize?
L43: “metal oxide sensing material” not strictly true. The metal oxide doesn’t sense anything, more physically changes in the presence of methane.
L46: “suited to long time deployment” Again, not true. These sensors require relatively high amounts of energy (as you state in the following sentence), work poorly in low humidity and work poorly in low temperatures (as you state in the nest sentence).
L62 – L71 Suggest you read:
Riddick, S. N., Ancona, R., Cheptonui, F., Bell, C. S., Duggan, A., Bennett, K. E. and Zimmerle, D. J. (2022) A cautionary report of calculating methane emissions using low-cost fence-line sensors. Elementa: Science of the Anthropocene 10(1). https://doi.org/10.1525/elementa.2022.00021
L78 – L109 Much of this is methods and should be condensed.
Methods
L112 – L128: Please condense this section to a description of the site. Please remove any extra detail superfluous to this study.
L129: Which heights? What is the ATEX zone?
L131: I would like the information here instead of having to find a different paper.
L136: I have concerns over air drawn through 100m of tubing at 6 lpm. I assume transit time will be significant as will gas mixing in the tubing.
Tables 1 & 2 aren’t adding much. Could go to SI
L150 – L158: How were the supply voltage controlled? Also, sensitivity would be affected by the ADC used, can you comment on that?
Section 2.3 – Please filter this for unnecessary information. Why mention data that were not used?
L173 – Not being able to detect < 4 ppm is concerning.
L 177 – Why 1-minute averages? Seems a long time. Was the sonic near anything else?
Table 3 -Why were these emission rates chosen? 1.4 to 18 kg h-1 are quite high emissions, what would be typical emission rates on a production site? What have other people measured as fugitives?
Section 2.4 – This is particularly badly written. I am finding it difficult to follow. It also has too much information that could be put in the SI. Essentially TGS data were fitted to the methane analyzer data.
L197 – Can you described the strengths and shortcomings of the two approaches. The main issue with approach 2 is the TGS sensors give erroneous output in low humidity conditions. This could strongly affect your linear interpolation if properly considered. At the very least, I would have thought the data were filtered for low RH.
Section 2.6 – Again, should be condensed.
Section 2.7 – This needs to be unpacked and explained clearly. Is there a figure that could help the reader? From what is written here, you are using the mole fraction (as calculated by a TGS) to infer four unknown variables (x, y, z and emission rate). Surely you need more than one measurement to do this, how many? What assumptions does this approach use? How realistic is it to well pad emissions? Again, this is impenetrable and needs to be written succinctly. A figure also helps the reader.
L282: What data were used in the Gaussian plume model? Please be explicit. How do you overcome the GP’s inability to work at distances less than 100 m?
L290 – L293: I do not know what this means. Please rewrite.
Results
The current information presented here could be condensed.
I would really like to know how applicable this method would be in the real world. For example, a well pad does not only have emissions from a single fugitive point which remain active until someone fixes it. Need to consider and comment on:
- There could be multiple emissions at the same time: vented, combustion and fugitive. How would this approach distinguish between types?
- Emissions (all types) can be intermittent. How long does an emission have to run to be detected?
- Gas coming out could be very hot, how does this affect your approach?
- Are any of the Gaussian Plume assumption violated?
- How does the GP approach work with distances < 100m?
- Density of sensors – is it realistic to have 16 sensor locations on a site? What is the optimum for you approach and what is realistic?
- Computational time. How long does it take to calculate the RSS matrix? Is this realistic?
Technical comments
Currently, the manuscript is difficult to follow and I strongly suggest a comprehensive rewrite.
Citation: https://doi.org/10.5194/amt-2023-52-RC1 - AC1: 'Reply on RC1', Rodrigo Andres Rivera Martinez, 19 Apr 2024
-
RC2: 'Comment on amt-2023-52', Anonymous Referee #3, 24 Nov 2023
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2023-52/amt-2023-52-RC2-supplement.pdf
- AC2: 'Reply on RC2', Rodrigo Andres Rivera Martinez, 19 Apr 2024
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 318 | 86 | 23 | 427 | 23 | 17 |
- HTML: 318
- PDF: 86
- XML: 23
- Total: 427
- BibTeX: 23
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
