the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 Sep 2023
A Modified Gaussian Plume Model for Mobile in situ Greenhouse Gas Measurements
Abstract. Atmospheric methane measurements are important for evaluating high resolution methane inventories and monitoring emissions reductions. Despite recent international efforts to harmonize measurement methodologies and techniques, currently there are no standardized or internationally accepted techniques for estimating emissions from mobile in situ concentration measurements. We present measurements from two different mobile in situ methane laboratories, and compare emission rates calculated from four Gaussian plume Bayesian optimal estimation strategies and a statistical algorithm. For mobile transects from the slower flow-rate instrument, we find a significant asymmetric smoothing artifact. The effect of this asymmetry is most significant for short transects of small (0–50 kg CH4 day−1), nearby methane sources, where the plume crossing time is comparable to the mean residence time of the instrument. We develop a model of this effect, demonstrate how this model can be applied to Gaussian plume inversions, and describe its limitations. We use these results to compute emissions rate estimates for two methane sources from Toronto’s wastewater management system to demonstrate the use and limitations of Gaussian plume inversions to quantify methane emissions in an urban environment. Overall, we highlight the importance of using observed plume enhancement areas rather than the more commonly used enhancement heights for determining comparable emissions estimates between different mobile laboratories.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(9132 KB)
-
Supplement
(1877 KB)
-
This preprint has been withdrawn.
- Preprint
(9132 KB) - Metadata XML
-
Supplement
(1877 KB) - BibTeX
- EndNote
Interactive discussion
Status: closed
-
RC1: 'Comment on amt-2023-193', Anonymous Referee #1, 27 Oct 2023
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2023-193/amt-2023-193-RC1-supplement.pdf
-
RC2: 'Comment on amt-2023-193', Anonymous Referee #2, 31 Oct 2023
This manuscript presents an analysis of mobile CH4 measurements on two platforms. A vehicle uses a Picarro CH4analyzer with 1-2 second time response. A bicycle uses an LGR instrument with much slower time response, resulting in lags and skewed peak shapes.
The authors use the instrument response functions to correct the peak shapes before Gaussian inversion, using a small controlled release experiment. They conclude that even after this correction, peak area is preferred over peak enhancement height and use the instrument response functions.
The conclusion that the instrument response factor for very slow instruments needs to be considered is self-evident. The follow-on conclusion about the use of area vs peak height needs more support, as it is unclear to me whether this applies only to the specific formulation of the Gaussian method used by the authors.
Specific additional gaps in this paper are:
- Discussion of source height. Just like stability class, source height can have a major impact on simulated emissions. Which source heights were chosen for the WWTP? How does this impact the error?
- Discussion of timescales of stability classes. The Pasquill-Gifford stability classes were developed for plumes measured on stationary sensors for 15min-1hr. These mobile measurements have much shorter transect times, and so it is expected that the plumes will appear narrower (more stable – D stability classes) even in urban settings. How do the measured plume widths compare to simulated gaussian plume widths? Is this why the area estimate performs better?
- The treatment of the uncertainties determined from the staged release requires more rigor. What metric is used to get the ± error bar on the slope? ± error bars should be defined based on a degree of confidence. 40% error is extremely low for any Gaussian inversion method. Other studies have shown that factor errors 1.5 – 3 at 95% confidence are more typical, and those errors are asymmetrical.
- Linear fits of estimated vs controlled release: In the figures, markers are hard to see, slopes could be listed on the graph along with R2. Consider also looking at the ratio of estimated/controlled, which should center around 1, and does not overly weight the result from the high emitter.
- Figures are difficult to interpret. At the minimum, they should have more extensive legends to show all shaded areas.
- The use of the Bayesian statistics needs more motivation/explanation. What is the underlying distribution that the model samples from, and how does this relate to the dataset? Is the model given a lognormal distribution shape as we might expect from the real distribution of emitters?
Citation: https://doi.org/10.5194/amt-2023-193-RC2 -
EC1: 'Comment on amt-2023-193', Huilin Chen, 09 Nov 2023
Dear authors,
First, thanks for submitting your work to the AMT. Your manuscript has been fully reviewed.
Unfortunately, both reviewers suggested rejection. I went through their comments and do agree with their evaluations.
Therefore, my recommendation is not to submit your responses and a revised version. I understand that you can withdraw your submission.
Best regards,
Huilin
Citation: https://doi.org/10.5194/amt-2023-193-EC1
Interactive discussion
Status: closed
-
RC1: 'Comment on amt-2023-193', Anonymous Referee #1, 27 Oct 2023
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2023-193/amt-2023-193-RC1-supplement.pdf
-
RC2: 'Comment on amt-2023-193', Anonymous Referee #2, 31 Oct 2023
This manuscript presents an analysis of mobile CH4 measurements on two platforms. A vehicle uses a Picarro CH4analyzer with 1-2 second time response. A bicycle uses an LGR instrument with much slower time response, resulting in lags and skewed peak shapes.
The authors use the instrument response functions to correct the peak shapes before Gaussian inversion, using a small controlled release experiment. They conclude that even after this correction, peak area is preferred over peak enhancement height and use the instrument response functions.
The conclusion that the instrument response factor for very slow instruments needs to be considered is self-evident. The follow-on conclusion about the use of area vs peak height needs more support, as it is unclear to me whether this applies only to the specific formulation of the Gaussian method used by the authors.
Specific additional gaps in this paper are:
- Discussion of source height. Just like stability class, source height can have a major impact on simulated emissions. Which source heights were chosen for the WWTP? How does this impact the error?
- Discussion of timescales of stability classes. The Pasquill-Gifford stability classes were developed for plumes measured on stationary sensors for 15min-1hr. These mobile measurements have much shorter transect times, and so it is expected that the plumes will appear narrower (more stable – D stability classes) even in urban settings. How do the measured plume widths compare to simulated gaussian plume widths? Is this why the area estimate performs better?
- The treatment of the uncertainties determined from the staged release requires more rigor. What metric is used to get the ± error bar on the slope? ± error bars should be defined based on a degree of confidence. 40% error is extremely low for any Gaussian inversion method. Other studies have shown that factor errors 1.5 – 3 at 95% confidence are more typical, and those errors are asymmetrical.
- Linear fits of estimated vs controlled release: In the figures, markers are hard to see, slopes could be listed on the graph along with R2. Consider also looking at the ratio of estimated/controlled, which should center around 1, and does not overly weight the result from the high emitter.
- Figures are difficult to interpret. At the minimum, they should have more extensive legends to show all shaded areas.
- The use of the Bayesian statistics needs more motivation/explanation. What is the underlying distribution that the model samples from, and how does this relate to the dataset? Is the model given a lognormal distribution shape as we might expect from the real distribution of emitters?
Citation: https://doi.org/10.5194/amt-2023-193-RC2 -
EC1: 'Comment on amt-2023-193', Huilin Chen, 09 Nov 2023
Dear authors,
First, thanks for submitting your work to the AMT. Your manuscript has been fully reviewed.
Unfortunately, both reviewers suggested rejection. I went through their comments and do agree with their evaluations.
Therefore, my recommendation is not to submit your responses and a revised version. I understand that you can withdraw your submission.
Best regards,
Huilin
Citation: https://doi.org/10.5194/amt-2023-193-EC1
Data sets
GTA Bike Surveys - Summer 2018 - Calibrated data Debra Wunch, Colin Arrowsmith, Sébastien Ars, Emily Knuckey, Nasrin Mostafavi Pak, Jaden L. Phillips https://doi.org/10.5683/SP2/U5CVFZ
GTA Bike Surveys - Summer 2019 - Uncalibrated data Sébastien Ars, Juliette Lavoie, Rica Cruz, Cameron Macdonald, Genevieve Beauregard, Liz Cunningham, Debra Wunch https://doi.org/10.5683/SP2/SBIZ1F
GTA Bike Surveys - Summer 2020 - Calibrated data Lawson Gillespie, Sébastien Ars, Tianjie Feng, Nasrin Mostafavi Pak, and Debra Wunch https://doi.org/10.5683/SP3/JEIZIF
GTA Bike Surveys - Summer 2021 - Calibrated data Lawson Gillespie, Sébastien Ars, Michael Raczkowski, Nasrin Mostafavi Pak, and Debra Wunch https://doi.org/10.5683/SP3/ZGMAI7
GTA Bike Surveys - Summer 2022 - Calibrated data Lawson Gillespie, Sébastien Ars, Mishaal Kandapath, Stephanie Gu, Amy Mann, Nasrin Mostafavi Pak, and Debra Wunch https://doi.org/10.5683/SP3/PGAIV7
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 632 | 228 | 62 | 922 | 75 | 35 | 40 |
- HTML: 632
- PDF: 228
- XML: 62
- Total: 922
- Supplement: 75
- BibTeX: 35
- EndNote: 40
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Lawson David Gillespie
Sébastien Ars
James Phillip Williams
Louise Klotz
Tianjie Feng
Stephanie Gu
Mishaal Kandapath
Amy Mann
Michael Raczkowski
Mary Kang
Felix Vogel
Debra Wunch
This preprint has been withdrawn.
- Preprint
(9132 KB) - Metadata XML
-
Supplement
(1877 KB) - BibTeX
- EndNote