the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Jan 2024
Development and deployment of a mid-cost CO2 sensor monitoring network to support atmospheric inverse modeling for quantifying urban CO2 emissions in Paris
Abstract. To effectively monitor the highly heterogeneous urban CO2 emissions using atmospheric observations, there is a need to deploy cost-effective CO2 sensors at multiple locations within the city with sufficient accuracy to capture the concentration gradients in urban environments. Its measurements could be used as input of an atmospheric inversion system for the quantification of emissions at the sub-city scale or separate specific sectors. Such quantification would offer valuable insights into the efficacy of local initiatives and could also identify unknown emission hotspots that require attention. Here we present the development and evaluation of a mid-cost CO2 instrument designed for continuous monitoring of atmospheric CO2 concentrations with a target accuracy of 1 ppm on hourly mean measurement. We assess the sensor sensitivity in relation to environmental factors such as humidity, pressure, temperature and CO2 signal, which leads to the development of an effective calibration algorithm. Since July 2020, eight mid-cost instruments have been installed within the city of Paris and its vicinity to provide continuous CO2 measurements, complementing the seven high-precision Cavity Ring-Down Spectroscopy (CRDS) stations that have been in operation since 2016. A data processing system, called CO2calqual, has been implemented to automatically handle data quality control, calibration and storage, which enables the management of extensive real-time CO2 measurements from the monitoring network. Colocation assessments with the high-precision instrument show that the accuracies of the eight mid-cost instruments are within the range of 1.0 to 2.4 ppm for hourly afternoon (12–17 UTC) measurements. The long-term stability issues require manual data checks and instrument maintenance. The analyses show that CO2 measurements can provide evidence for underestimations of CO2 emissions in the Paris region and a lack of several emission point sources in the emission inventory. Our study demonstrates promising prospects in integrating mid-cost measurements along with high precision data into the subsequent atmospheric inverse modeling to improve the accuracy of quantifying the fine-scale CO2 emissions in the Paris metropolitan area.
-
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
(2908 KB)
-
Supplement
(2551 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2908 KB) - Metadata XML
-
Supplement
(2551 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-125', Anonymous Referee #1, 04 Mar 2024
The manuscript titled “Development and deployment of a mid-cost CO2 sensor monitoring network to support atmospheric inverse modeling for quantifying urban CO2 emissions in Paris” is generally well written and scientifically justified. We recommend it to be published after addressing the following points.
It would be helpful to have additional discussion of the spatial density of the network. The reader can deduce this from Figure 1, but average distance to the nearest site would be helpful to include in Section 2.4. How does the spatial density of the HPP network compare to the density of the Picarro network? How does the HPP spatial density compare to the previous studies of similar moderate-cost sensor networks (i.e. Wu et al., 2016; Turner et al., 2016)?
P and p are used seemingly interchangeably for pressure. (Eq 1 uses P, but p is used frequently in the text and other figures).
Figure 1 and Figure 2 emphasize the hardware of the sensor and the data infrastructure. These are important to include but might be better to be included in the SI than the main text as the figures contain a lot of details that are not necessarily central to the main points of the paper.
Figure 6 is too small to read comfortably, especially the text.
Figure 8: Add some break / distinction between the Picarros and HPPs
Figure 9: Please consider flipping the direction of the difference (site - JUS), so that lower than JUS values are negative. Currently this figure gives the impression of elevated concentrations outside of the urban center. Alternatively, clarify in the figure caption that the difference is JUS minus other sites. Same note for figure S9.
Page 9, line 9: Could the author(s) comment or speculate on the variable sensor performance? Is this a difference in sensor hardware or experimental condition?
Page 1, line 14: ambiguous pronoun reference (Its)
Page 1, line 16: Missing word (should be “to separate”)
Page 1, line 28: should be “prospects for”
Page 1, line 33: “spatial and temporal variations” of what? Emissions?
Page 3, line 15: “in dimensions” is redundant
Page 3, line 21: a SHT75 -> an SHT75
Page 12, line 33 “on-going” -> “ongoing”
Citation: https://doi.org/10.5194/egusphere-2024-125-RC1 - AC1: 'Reply on RC1', Jinghui Lian, 12 Jun 2024
-
RC2: 'Comment on egusphere-2024-125', Anonymous Referee #2, 29 Apr 2024
The manuscript, “Development and deployment of a mid-cost CO2 sensor monitoring network to support atmospheric inverse modeling for quantifying urban CO2 emissions in Paris,” provides a thorough description of the development and evaluation of a new urban mid-cost sensor monitoring network. This is a well-written document that is useful for those working on greenhouse gas quantification with urban monitoring networks. The authors should address the following (minor) comments:
- Page 4, lines 1-2: The wording is a bit unclear - Is the flushing pump installed or is it not? The text says it “could be” installed, but Figure 1a shows that it is installed.
- Page 5: In lines 35-37, the authors describe the target gas calibration using only 2 minutes to allow for flushing, but the calibration in parallel with the CRDS instrument requires 7 minutes to flush as described in lines 16-17. Could the authors provide justification for the shorter flushing time?
- Page 7, lines 15-16: It is unclear why the authors chose to apply one-minute averaging to the data before calibration.
- Page 11, lines 18-19: I do not agree with the authors when they state with certainty that the small differences in CO2 mole fraction between these sites and JUS can be attributed solely to their proximity to the site when OBS, the site closest to JUS, has a substantial difference in mole fraction. I would expect the city center to have substantial heterogeneity in CO2 mole fraction.
- Figure 9 and paragraph beginning page 11, line 31: I recommend the authors specify the sign convention of the gradient.
Technical comments:
- Page 1, line 14: “Its” should be “Their” if the authors are referring to the mid-cost sensors.
- Figure 4: The tables in each figure should have units.
- Page 8, line 34: The authors are missing an “and” here, and I have a suspicion that they meant to use HPP5 rather than HPP7 as an example for improvement from the p correction.
- Figure 6: The text is hard to read because of the small font size and light colors.
- Figures 7 and 9: The wind roses for OBS (both figures) and JUS (Figure 7) are too small to interpret.
- Page 11, line 32: “from July 2020 and December 2022” should be “from July 2020 to December 2022.”
- Figure S5: A time series of 2 years of hourly data in a figure is hard to interpret. It could be helpful to include some averaging to make it easier to read, maybe in an additional figure.
Citation: https://doi.org/10.5194/egusphere-2024-125-RC2 - AC2: 'Reply on RC2', Jinghui Lian, 12 Jun 2024
-
RC3: 'Comment on egusphere-2024-125', Anonymous Referee #3, 30 Apr 2024
Lian et al., Development and deployment of a mid-cost CO2 sensor monitoring network to support atmospheric inverse modeling for quantification of urban CO2 emissions in Paris, is a well-written and thorough piece of science that will be a useful resource for others working in this burgeoning field.
I have only two substantive comments that the authors might consider addressing briefly within the text.
Firstly, on page 5, line 17, the authors write: "measuring dry air from two target cylinders with known CO2 mole fractions. … for a duration of 10 minutes, utilizing only the last three minutes of data", when discussing the calibration. Later in that paragraph, they say, "The target gas is injected … for a duration of 3 minutes and only the last-minute data are used" to deal with sensor drift. It would be helpful if the authors commented on the sufficiency of the three minutes of target gas measurement. Is the measurement stable after two minutes? Did the authors run the target gas for longer periods and determine this was optimal for eliminating sensor drift while minimizing gas consumption?
Secondly, the authors have invested a huge effort in developing, testing and deploying the HPP sensors. Much of that effort will be 'banked' e.g. the data handling investment, but there remain significant recalibration efforts when replacing the target tanks every 4-5 months. A short comment on the relative cost saving over the lifetime of the HPP sensors relative to investment in a higher-precision instrument such as a Picarro would be helpful to the audience.
I make some minor recommendations to improve accessibility of the text.
P2, line 27: use 'the ninth' rather than 'a ninth',
There is some inconsistency in the text about the use of p and P to denote pressure. Please stick to one or the other.
p6, line 34. Change to "For a list of internal flags for some important physical parameters, refer to Table S1”
p8. line 33-34.
“The p correction substantially reduces the RMSEs of ΔCO2 to 1.6ppm (HPP4) to 49.7 ppm (HPP7)” is slightly confusing. Suggest revision to: The p correction generally substantially reduces the RMSEs of ΔCO2. For instance, in HPP4, the p correction reduces the RMSE of ΔCO2 to 1.6ppm (an improvement of 88% relative to the Raw, H2O and T corrected RMSE).”
While it is commendable that the authors also cite HPP7, the relative improvement in that case was only 1%, significantly lower than the other seven sensors, which causes some confusion for the reader taking in the whole dataset.
p22. Figure 6 is very dense. Perhaps panel b) with the modeling data could be omitted and the aspect ratio adjusted for an improved reader experience.
p.24 Figure 8 is also very complex and hard to read. Possibly this information could be moved to the supplementary material and a smaller sub-set (possibly only winter and summer and/or every second site by distance from JUS) displayed in Fig 8, to improve readers experience.
Citation: https://doi.org/10.5194/egusphere-2024-125-RC3 - AC3: 'Reply on RC3', Jinghui Lian, 12 Jun 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-125', Anonymous Referee #1, 04 Mar 2024
The manuscript titled “Development and deployment of a mid-cost CO2 sensor monitoring network to support atmospheric inverse modeling for quantifying urban CO2 emissions in Paris” is generally well written and scientifically justified. We recommend it to be published after addressing the following points.
It would be helpful to have additional discussion of the spatial density of the network. The reader can deduce this from Figure 1, but average distance to the nearest site would be helpful to include in Section 2.4. How does the spatial density of the HPP network compare to the density of the Picarro network? How does the HPP spatial density compare to the previous studies of similar moderate-cost sensor networks (i.e. Wu et al., 2016; Turner et al., 2016)?
P and p are used seemingly interchangeably for pressure. (Eq 1 uses P, but p is used frequently in the text and other figures).
Figure 1 and Figure 2 emphasize the hardware of the sensor and the data infrastructure. These are important to include but might be better to be included in the SI than the main text as the figures contain a lot of details that are not necessarily central to the main points of the paper.
Figure 6 is too small to read comfortably, especially the text.
Figure 8: Add some break / distinction between the Picarros and HPPs
Figure 9: Please consider flipping the direction of the difference (site - JUS), so that lower than JUS values are negative. Currently this figure gives the impression of elevated concentrations outside of the urban center. Alternatively, clarify in the figure caption that the difference is JUS minus other sites. Same note for figure S9.
Page 9, line 9: Could the author(s) comment or speculate on the variable sensor performance? Is this a difference in sensor hardware or experimental condition?
Page 1, line 14: ambiguous pronoun reference (Its)
Page 1, line 16: Missing word (should be “to separate”)
Page 1, line 28: should be “prospects for”
Page 1, line 33: “spatial and temporal variations” of what? Emissions?
Page 3, line 15: “in dimensions” is redundant
Page 3, line 21: a SHT75 -> an SHT75
Page 12, line 33 “on-going” -> “ongoing”
Citation: https://doi.org/10.5194/egusphere-2024-125-RC1 - AC1: 'Reply on RC1', Jinghui Lian, 12 Jun 2024
-
RC2: 'Comment on egusphere-2024-125', Anonymous Referee #2, 29 Apr 2024
The manuscript, “Development and deployment of a mid-cost CO2 sensor monitoring network to support atmospheric inverse modeling for quantifying urban CO2 emissions in Paris,” provides a thorough description of the development and evaluation of a new urban mid-cost sensor monitoring network. This is a well-written document that is useful for those working on greenhouse gas quantification with urban monitoring networks. The authors should address the following (minor) comments:
- Page 4, lines 1-2: The wording is a bit unclear - Is the flushing pump installed or is it not? The text says it “could be” installed, but Figure 1a shows that it is installed.
- Page 5: In lines 35-37, the authors describe the target gas calibration using only 2 minutes to allow for flushing, but the calibration in parallel with the CRDS instrument requires 7 minutes to flush as described in lines 16-17. Could the authors provide justification for the shorter flushing time?
- Page 7, lines 15-16: It is unclear why the authors chose to apply one-minute averaging to the data before calibration.
- Page 11, lines 18-19: I do not agree with the authors when they state with certainty that the small differences in CO2 mole fraction between these sites and JUS can be attributed solely to their proximity to the site when OBS, the site closest to JUS, has a substantial difference in mole fraction. I would expect the city center to have substantial heterogeneity in CO2 mole fraction.
- Figure 9 and paragraph beginning page 11, line 31: I recommend the authors specify the sign convention of the gradient.
Technical comments:
- Page 1, line 14: “Its” should be “Their” if the authors are referring to the mid-cost sensors.
- Figure 4: The tables in each figure should have units.
- Page 8, line 34: The authors are missing an “and” here, and I have a suspicion that they meant to use HPP5 rather than HPP7 as an example for improvement from the p correction.
- Figure 6: The text is hard to read because of the small font size and light colors.
- Figures 7 and 9: The wind roses for OBS (both figures) and JUS (Figure 7) are too small to interpret.
- Page 11, line 32: “from July 2020 and December 2022” should be “from July 2020 to December 2022.”
- Figure S5: A time series of 2 years of hourly data in a figure is hard to interpret. It could be helpful to include some averaging to make it easier to read, maybe in an additional figure.
Citation: https://doi.org/10.5194/egusphere-2024-125-RC2 - AC2: 'Reply on RC2', Jinghui Lian, 12 Jun 2024
-
RC3: 'Comment on egusphere-2024-125', Anonymous Referee #3, 30 Apr 2024
Lian et al., Development and deployment of a mid-cost CO2 sensor monitoring network to support atmospheric inverse modeling for quantification of urban CO2 emissions in Paris, is a well-written and thorough piece of science that will be a useful resource for others working in this burgeoning field.
I have only two substantive comments that the authors might consider addressing briefly within the text.
Firstly, on page 5, line 17, the authors write: "measuring dry air from two target cylinders with known CO2 mole fractions. … for a duration of 10 minutes, utilizing only the last three minutes of data", when discussing the calibration. Later in that paragraph, they say, "The target gas is injected … for a duration of 3 minutes and only the last-minute data are used" to deal with sensor drift. It would be helpful if the authors commented on the sufficiency of the three minutes of target gas measurement. Is the measurement stable after two minutes? Did the authors run the target gas for longer periods and determine this was optimal for eliminating sensor drift while minimizing gas consumption?
Secondly, the authors have invested a huge effort in developing, testing and deploying the HPP sensors. Much of that effort will be 'banked' e.g. the data handling investment, but there remain significant recalibration efforts when replacing the target tanks every 4-5 months. A short comment on the relative cost saving over the lifetime of the HPP sensors relative to investment in a higher-precision instrument such as a Picarro would be helpful to the audience.
I make some minor recommendations to improve accessibility of the text.
P2, line 27: use 'the ninth' rather than 'a ninth',
There is some inconsistency in the text about the use of p and P to denote pressure. Please stick to one or the other.
p6, line 34. Change to "For a list of internal flags for some important physical parameters, refer to Table S1”
p8. line 33-34.
“The p correction substantially reduces the RMSEs of ΔCO2 to 1.6ppm (HPP4) to 49.7 ppm (HPP7)” is slightly confusing. Suggest revision to: The p correction generally substantially reduces the RMSEs of ΔCO2. For instance, in HPP4, the p correction reduces the RMSE of ΔCO2 to 1.6ppm (an improvement of 88% relative to the Raw, H2O and T corrected RMSE).”
While it is commendable that the authors also cite HPP7, the relative improvement in that case was only 1%, significantly lower than the other seven sensors, which causes some confusion for the reader taking in the whole dataset.
p22. Figure 6 is very dense. Perhaps panel b) with the modeling data could be omitted and the aspect ratio adjusted for an improved reader experience.
p.24 Figure 8 is also very complex and hard to read. Possibly this information could be moved to the supplementary material and a smaller sub-set (possibly only winter and summer and/or every second site by distance from JUS) displayed in Fig 8, to improve readers experience.
Citation: https://doi.org/10.5194/egusphere-2024-125-RC3 - AC3: 'Reply on RC3', Jinghui Lian, 12 Jun 2024
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 633 | 213 | 39 | 885 | 63 | 20 | 20 |
- HTML: 633
- PDF: 213
- XML: 39
- Total: 885
- Supplement: 63
- BibTeX: 20
- EndNote: 20
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Olivier Laurent
Mali Chariot
Luc Lienhardt
Michel Ramonet
Hervé Utard
Thomas Lauvaux
François-Marie Bréon
Grégoire Broquet
Karina Cucchi
Laurent Millair
Philippe Ciais
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2908 KB) - Metadata XML
-
Supplement
(2551 KB) - BibTeX
- EndNote
- Final revised paper