the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Sep 2024
The monitoring network of greenhouse gas (CO2, CH4) in the Paris' region
Abstract. There is a growing interest for the study of greenhouse gas emissions over urban areas. In this context, a network for measuring greenhouse gas concentrations was set up in Paris in 2015. Since then, seven stations located in and around Paris and equipped with cavity ring-down spectrometers (CRDS) have been monitoring gas concentrations of different species (CO2, CH4, CO) on a continuous basis. Procedures for maintenance, calibration, data processing have been adapted to ensure that the network is operational, providing good quality data in near-real time (NRT) and with high availability. The CO2 and CH4 concentrations show a growth rate of the baseline (linear trend of the averaged concentrations over the Paris region) concentrations (+2.34 ppm CO2/year and +11.1 ppb CH4/year) between 2015 and 2022 consistent with that observed at remote observatories such as Jungfraujoch (Switzerland). The amplitude of the CO2 seasonal cycle is around 20 ppm (i.e. around 5 %) while that of CH4 is 40 ppb (i.e. around 2 %). The concentration gradients calculated as the differences between up- and down-wind concentrations, can be used to infer the emissions for long lived species. We use the measurements from the 100 m Saclay tower, outside of Paris, as a regional background during suitable wind direction conditions. The largest differences with these background measurements are observed in the two stations that are located within Paris (JUS and CDS), with urban offset up to ~ 50 ppm CO2 and ~ 100 ppb CH4 in winter and ~ 10 ppm CO2 and ~ 50 ppb CH4 in summer. In winter, the co-variability of CO and CO2 hourly measurements (correlation coefficient r ~ 0.9 in all stations) indicates that the concentration variability is driven by anthropogenic emissions. Conversely, in summer, lower correlations between these two gases concentration (r ~ 0.3 in peri-urban stations and r ~ 0.6 at CDS and JUS) shows the more dominant role of vegetation fluxes.
- Preprint
(3036 KB) - Metadata XML
-
Supplement
(157 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2024-2826', Anonymous Referee #1, 07 Oct 2024
-
CC1: 'Comment on egusphere-2024-2826', Maria de Fatima Andrade, 06 Mar 2025
This is an important manuscript that provides a detailed description of the GHG network in Paris, established in 2015. The authors present data from seven measurement sites. I consider it is valuable for publication. There is a lack of data concerning urban measurements of GHG. Some suggestions to the manuscript.
- The manuscript should clarify where the data can be accessed.
- One key point for discussion is the impact of different sampling heights on CO₂ concentrations. The authors should address how they analyze the data and draw conclusions about transport processes, given that the stations are located at different altitudes.
- Regarding calibration, the manuscript does not explain how the target gases are used to assess bias and repeatability. Additionally, the one-point calibration with the reference gas requires further clarification. What corrections are applied for water vapor?
- The manuscript includes many figures. Some, such as those related to calibration procedures (e.g., Figure 2), could be moved to an appendix for clarity.
- In lines 280-285, the authors link CO levels to road traffic and residential heating. However, another potential source that should be discussed is long-range transport of biomass burning plumes.
- The CH₄ peaks are attributed to air mass transport. Are these air masses from rural areas? Could they also be linked to natural gas usage or waste treatment? A clearer discussion of CH₄ sources is needed.
- The variation of CO₂ and CO could be analyzed on an hourly basis to better evaluate the influence of anthropogenic emissions and vegetation-related sources and sinks.
- The english could be improved to make the text more fluid.
- Some figures are with bad quality and difficult to reas (Figure 4 for instance)
Citation: https://doi.org/10.5194/egusphere-2024-2826-CC1 -
RC2: 'Comment on egusphere-2024-2826', Anonymous Referee #2, 10 May 2025
Comments for egusphere-2024-2826
This manuscript provides a comprehensive overview of the greenhouse gas (GHG) monitoring network in the Paris region, including site descriptions, calibration strategies, data quality assurance, and observed trends from 2015–2022. The long-term, high-frequency dataset is a valuable contribution to urban GHG monitoring and modeling efforts. The manuscript is detailed, but the analysis and presentation could benefit from further clarification and refinement. I recommend publication after attention to the items below.
Major Comments
Several papers describing aspects of this network have been published previously. A more comprehensive summary of what is already in the literature and what is new here would be helpful. Also, references to other networks should be updated to include the most recent work on inversions and network setup.
The authors should also consider whether their assumption that CH4 and CO do not have the same source is justified by the data. I agree that inventories widely assume that the two have different sources in cities. However, these results suggest methane emissions are proportional to CO and therefore leaks are proportional to consumption.
Identification of Plumes:
Both of the filtering methods need to be explained more clearly. While the authors refer to a “statistical filter” derived from high-frequency variability (citing El Yazidi et al., 2018 and Cristofanelli et al., 2023), the manuscript does not provide a self-contained summary of how this filter works. Since this filtering step is critical to all subsequent analyses including diurnal cycles, covariability, and gradients, it should be briefly described in the main text. Specific suggestions:
The paper should provide a brief description of the statistical method (e.g., moving median, standard deviation thresholding, percentile ranges). What thresholds are applied? 2) Clarify which types of anomalies the filter is designed to catch (e.g., define short spikes clearly). How does the filter determine the baseline concentration levels? 3) Is the same filter applied uniformly to all species and stations, or are there species-specific thresholds or site-specific tuning? 4) Was the output of the statistical filter visually checked, or compared to known events (e.g., maintenance logs or known pollution spikes)?
The wind sector filter approach to identifying and removing local contamination at GNS and OVS is interesting but currently underdeveloped. The manuscript should provide enough information for someone to repeat the analysis and arrive at the same result. The description states that filter thresholds for wind direction, wind speed, and standard deviation were "determined empirically," but no clear procedure or rationale is provided. Were these thresholds optimized? Consider describing the steps taken to identify thresholds and validating them (e.g., correlation drops, concentration anomalies, footprint analysis). The authors mention that 5–13% of hourly data are flagged as contaminated using either of data filters but didn’t show any quantitative analysis results, which leads to the next question: how does the wind sector filter compare with the high-frequency spike filter used network-wide? Are there overlapping selections? A brief comparison of data removed by each filter and whether one dominates would improve clarity.
Role of SAC100 as Background Site:
The manuscript uses SAC100 as a background reference for gradient calculations but also notes its altitude-dependent decoupling under stable conditions. Can the authors quantify how often SAC100 is truly representative of background air, maybe via PBLH diagnostics or footprint modeling? Presenting uncertainty estimates for concentration gradients associated with this would improve the robustness of inversion-readiness claims.
Rationale for Analyzing Differences between the Measurement and the Temporal Average:
The authors calculate residuals by subtracting a 3-month rolling mean from the hourly time series and then examine the correlation between these residuals for CO₂ vs. CO and CH₄ vs. CO. However, this approach raises several concerns: 1) Since the main goal of this part of analysis is to understand emission-related co-variability, removing the seasonal cycle and long-term trends could obscure the very patterns of interest, particularly for species with strong source-seasonality like CO₂. It’s unclear why this step is necessary. 2) More directly relevant would be to analyze co-variability between enhancements (e.g., ΔCO₂ or ΔCH₄ relative to a background site like SAC100), especially during downwind conditions. This would isolate the urban emission signal rather than mixing local and regional variability into a residual. This also aligns with how inversion systems interpret urban plume signals. 3) Arbitrary smoothing window? The use of a 3-month rolling average is not clearly justified. How sensitive are the correlations and slopes to this choice? Would the use of daily, weekly, or climatological baselines yield different results? I recommend that the authors either (i) shift the co-variability analysis to focus on signal–background enhancements, or (ii) more clearly justify and validate their use of detrended residuals.
Minor Comments
Line 20: Specify how the reported baseline growth rates of CO2 (+2.34 ppm/year) and CH4 (+11.1 ppb/year) were computed (e.g., linear regression, seasonal detrending).
Line 61-69: The legend showing “Emissions totales” in Figure 1 was not explained in the main text or in the caption. Also, that information was not used in the analysis.
Line 128-134: A brief description of ICOS protocols and classification is needed for readers to understand what is going on here.
Line 155: Please clarify if CO data at OVS was used in the analysis due to the unsolved problem in its calibration.
Line 198-200: What does “the impact of spike filter” mean? Are you referring to the data filtered out or the data selected to use in the analysis?
Fig. 7: It’s very hard to read. Consider changing it into a year-long monthly distribution plot.
L335-336: “outliers not detected by the filters” but removed as outliers here, which is ambiguous and confusing. More explanation is needed.
L350–355: Consider briefly explaining why “… amplitudes in CH4 diurnal cycles remain fairly stable across seasons and is likely driven by vertical mixing rather than emissions…” given that the residential sector showing strong seasonality (45% vs. 7%, Line 301-305).
Line 496: “Bio-fuels may be responsible…” — please clarify if this is speculation or supported by inventory evidence/previous study.
Citation: https://doi.org/10.5194/egusphere-2024-2826-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 806 | 201 | 258 | 1,265 | 52 | 31 | 54 |
- HTML: 806
- PDF: 201
- XML: 258
- Total: 1,265
- Supplement: 52
- BibTeX: 31
- EndNote: 54
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of "The monitoring network of greenhouse gas (CO2,
CH4) in the Paris’ region" by Doc et al., for Atmospheric Chemistry and Physics (ACP).
This manuscript describes the CO2, CH4, and CO observations in the Paris urban GHG network. (although the title only mentions CO2 and CH4, CO measurements are described so perhaps the title should be adjusted to reflect this?). Overall the analysis is sound, although it is fairly simple, it is valuable. However, this same group has published several modeling analyses that use the CO2 data from this network in a much more sophisticated analysis to derive emissions. This paper should be put into context of that other work and presented for what it is, which is a description of the underlying observations and the methods used to collect them along with introducing the CH4 and CO observations and looking at relationships between the species. While the gradient and tracer/ratio analyses were useful, I still would categorize this as a "data" paper, i.e., perhaps not suitable for ACP but rather ESSD. As such, it should also present the data in a public repository in order to be published. I also think that more space could be devoted to the details of the calibration systems at the sites, as it is now it was a bit cursory. I am not sure I recommend publication in ACP but rather would advise the authors re-work this into an ESSD paper and make the data public alongside. While the addition of CH4 and CO separates this from the Lian et al papers, much of the analysis in this work is about CO2, making some broad conclusions about gradients and seasonality (e.g. that co2 goes down in summer due to photosynthesis!) that are redundant to the other work. If the authors would like to keep this in ACP, perhaps additional analysis of the non-CO2 gases would be more useful.
Overall comments:
1- There is no data availability statement at all in this work, is that data being made public somewhere? [Even if not put in ESSD, there should be some statement for data access].
2- ppm and ppb should be defined along with what exactly these measurements are (presumably dry-air mole fractions)? (ppm = micromoles per mole, or parts per million, etc)
3- Could the authors put this work into context with the Xueref-Remy 2018 ACP work? Looking at the station names and map of that paper, none are the same. Is this the case - has the network changed so dramatically since 2011 that none of the stations are the same, or is that effort still ongoing separately? Clearly, that paper did analysis on less than one full year of data, while here multiple years are analyzed, in fact there is no temporal overlap with that work. It would still be helpful to understand the history here.
4- Some grammar and many awkward wording issues - I recommend a read-through for English language. Many units are sometimes abbreviated and sometimes spelled out, these should be just abbreviated according to conventions (m, km, etc.).
L48: A list of references for this network is provided here, but (as noted above with regards to the Xueref-Remy 2018 paper), some context would be useful here as to how this work advances understanding over the previously cited works that use this same data. It seems a little unusual to present basic observational analysis after full model inversions have been conducted, for example as in Lian 2023. What is being added here? Two additional gases, but much of the analysis is about CO2 here and fairly simplistic.
No mention of drying the air or water correction. Can the authors please comment on whether they are reporting dry air mole fractions, and if so, how they are obtained?
figure 1, caption should indicate the source of the emissions data for the colors, and the units. (are these annual? what year?).
L85 north-est should be north-east.
L 106 - repetitive, wind measurements at 100 m were already mentioned earlier.
L105 - is this the only station with weather measurements?
Calibration section - This was a bit confusing and should be clarified. Do all sites have target run once a day but not a daily ref? that's only at JUS? A little more detailed explanation would be nice here - I do not believe these details have been published before.
L188-190 can the statistical (not the wind filter described later) filter be briefly described here without only the references? (i.e. do you filter out minutes whose standard deviations exceed a threshold? If so, what is that threshold?).
L199 - previously it was stated that these would be refered to as "statistical" fitlering and "wind" filtering, but here this paragraph refers to spike filtering - does this just mean both filter together?
L243-249 I would think that the statistical filter would not be able to sufficiently filter out the GNS spikes since they originate farther away, but the wind filter would. Seems backward from what is being said here, perhaps the authors can clarify?
L250 Here it is stated that this analysis uses the data without the data that is filtered. Is this data removed entirely from the series that is made available to the public? Or is it retained with a flag indicated that there is some local contamination, for users that might be interested in analyzing these? What about the data that is filtered at all the stations by the statistical filter, is this treated differently from those isolated by the wind filter? Were these filters used in Lian et al., 2023?
L268- is there a reference for the Jungfraujoch data?
L270 re: CH4, can these trends and seasonal cycle magnitudes be compared with those at MLO or Jungfraujoch as the CO2 values were? After all, a trend different than the global mean trend may mean something about local or regional emission trends.
It is disappointing after all the discussion of the wind filter to determine that it is not filtering the data as the authors wanted for their analysis - it seems a different, more restrictive filter should be implemented, if not in general, then at least for this analysis? Or perhaps then there should be no filter and there could be an interprentation of the impact of the nearby emission on the analysis? Having a half-filter seems like the worst of both worlds.
L383 - whole day meaning 24-hours?
Gradients usually means a change over a distance, and here they are just differences, so I would advocate they be referred to as differences between stations.
Section 5.1 title has poor grammar? Also, it seems that differences with SAC are presented not just when the SAC station is strictly upwind, but over many wind directions, so this should not be called upwind and downwind. I would re-name this section "Annual differences between each station and SAC100"
Fig 12 and analysis of residuals (detrended and deseasonalized), I would recommend using a different nomenclature for these detrended values, rather than the capital Delta which earlier in this paper represented the difference in concentration from the SAC100 data, to avoid confusion.
Fig 12 caption, define r and s? Is this a york or ODS slope or ordinary least squares?
In the captions, "doted" should be "dotted", but actually more accurately should say "dashed" as the regression lines seem to be the gray dashed lines, and the red dotted lines are the slope =1 lines (which should be mentioned in the caption).
L379 "int" should be "in"
L495, Waste (both wastewater and landfills) are dominated sources in cities.
L516 made available to whom?
L521 "around" should be "approximately"
L523 "highlights the increasing efficiency of residential ..."
References:
typos, esp. in first reference. ("dby"?)
Ref 25 Lian et al 2024 on mid-cost network - I would like to look this up but the reference is incomplete here. (no source listed).
32 and 33 are the same
Reference list has numbers while citations do not.