Review of updated version of “Carbon emission reduction requires attention to the contribution of natural gas use: Combustion and leakage”.
First and foremost, I want to congratulate the authors on working very hard and addressing each and every question I have raised. I believe that their effort has led to a more technically valid paper. I also believe that the paper now has stronger conclusions and deeper scientific merit. I have a few remaining questions, mostly about new points raised in their responses. I have structured some of my points to help the authors to write their “Caveats and Limitations” section of the conclusion (https://www.atmospheric-chemistry-and-physics.net/policies/guidelines_for_authors.html). I believe that with another round of much less intensive revisions, that the paper will be ready to contribute substantially to the ACP and wider academic community.
Specific points:
(1) The author responds: “However, few studies have evaluated the impact of using different combination of time window and quantile on background value calculation. Yet, the choice of both the time window length and the quantile does indeed affect the final calculated background concentration. Here, using mobile measurement results near the gas storage tank in summer as an example, we evaluated the impact of different window-quantile combinations on background value calculation. This part has been added to supplementary,”.
The idea of using a background calculation is inherently complex because the observations of both the background and the signal contain uncertainty. Even if you had access to the data of Shangdianzi, it would still contain uncertainty. For this reason, there are at least three recent papers which have worked to advance the idea of how to account for the join uncertainty’s impact on emissions estimation, although none of them used observations from flux towers [Lu F. et al., 2025; Lu L. et al., 2025; Zheng et al., 2025]. I am not sure if you have the ability to repeat their approaches, or if you should think about how to write this new type of approach into the part of the conclusion that deals with caveats and limitations.
(2) The author responds: “Replace identified outliers with linearly interpolated values from adjacent points. Consecutive outliers ≤ 3 are treated as a single outlier; consecutive outliers ≥ 4 are considered local trends and excluded from outlier classification.”
I would think that you would need to carefully consider the footprint during these times before they are classified as either an outlier or a trend. The impact would be vastly different if the footprint during these times is similar to the normal conditions that you have demonstrated, or looked different from the normal conditions you have demonstrated. It also is possible that you are removing an actual emissions event, and replacing it with an interpolated value which is smaller, in which case you have just underestimated the actual emissions. Perhaps this is reasonable, but perhaps not, especially given the uncertainty in the observations themselves, as mentioned above in point 1.
(3) The author responds: “It can be seen that the source area covers the most urban area of Beijing. It basically covers the entire Fifth Ring Road area of Beijing but does not extend to other provinces, thus excluding long-range transport from other provinces.”.
This map shows a footprint out to the 90th percentile. It would be interesting to see how far the remaining 10% of the footprint looks, especially if the fluxes computed during that remaining 10% fall outside of the range of the fluxes occurring within this 90%. Some simple graphs of the statistics of the computed fluxes (i.e., PDFs) occurring at within each ring, as well as outside of the outer ring should help us address this issue.
(4) The author responds: “Second, the explanation lies in errors associated with the turbulent flux measurement system. This uncertainty is difficult to quantify because the sources of error are diverse, such as signal loss due to frequency attenuation in closed-path systems, the occurrence of negative values when real fluxes approach zero caused by the instrument's low signal-to-noise ratio, and the failure of the steady-state assumption underlying the eddy covariance method under conditions of weak turbulence. Unfortunately, no study can fully quantify the causes of negative values in flux observations currently, particularly over highly heterogeneous urban surfaces, where quantifying these uncertainties becomes especially challenging. Due to weaker turbulence development at night, flux measurement uncertainty increases, and the probability of observing negative fluxes is higher. Fluxes frequently fluctuate around zero during these periods.”.
This is not the interpretation that I would make. I would follow your comment that the fluxes around zero during these periods are actually approximately zero, plus some amount of white noise from your instrument. Therefore, instead of considering both the positive and negative fluxes, instead you should assume all are a function of white noise, and ignore all of the both very small negative and very small positive fluxes, since white noise is just as equally positive as negative. This will then reduce your overall number of valid points, and likely lead to an increase in the overall emissions, although this will need to be carefully considered. This is a response similar to what the papers talked about in response to point (1). This may be too difficult to work out for this paper, but it should at least be pointed out as being important for future study and/or as a limitation of the current approach.
(5) The author responds: “Another a essential point is that CO has a long lifespan in the atmosphere, and it takes several tens of days to decay into CO2 (Drummond et al., 2009; Weinstock et al., 1969), so the impact of long-distance transmission of CO is relatively small.”
I have checked these references and they both refer to modeling studies or global average values. Due to CO’s very large variation in concentration, and that its lifetime is related to OH, which also varies by orders of magnitude, local lifetimes may also vary substantially. Recent observational papers using satellites and light-physical models in tandem, have demonstrated that in highly emitting regions that the lifetime of CO in the actual atmosphere is far shorter[Lin et al., 2020; Wang et al., 2021; Wang et al., 2025]. One such paper has specifically demonstrated that the production of CO2 from CO is not insignificant in Shanxi, which is directly upwind of Beijing at least some fraction of the time[Li et al., 2025]. Again, this additional work may be too much of an extension, but it could be mentioned as a limitation.
(6) The author responds: “Unfortunately, due to the lack of basic data from other cities or provinces, we are unable to measure its specific value accurately, a rough method was applied to estimated China's overall natural gas leakage rate based on existing reports and literature as follows, we have modified the Section 4.2 according to updated national leakage rate of natural gas.”.
I would be very careful with this approach. For example, I have personally observed a very large amount of piped gas in use in Shanxi, much more so than you would realize based on the population, or other statistics used in your approach. Simple scaling approaches have been demonstrated to miss these substantial sources[Qin et al. 2023; Hu et al., 2024]. Again, such work would not likely be scalable based on your techniques used, but at least such a mention should be put into the “Caveats and Limitations” section of the conclusion.
References:
Hu, W., Qin, K., Lu, F. et al. Merging TROPOMI and eddy covariance observations to quantify 5-years of daily CH4 emissions over coal-mine dominated region. Int J Coal Sci Technol 11, 56 (2024). https://doi.org/10.1007/s40789-024-00700-1
Li, X., Cohen, J.B., Tiwari, P. et al. Space-based inversion reveals underestimated carbon monoxide emissions over Shanxi. Commun Earth Environ 6, 357 (2025). https://doi.org/10.1038/s43247-025-02301-5
Lin, C., Cohen, J.B., Wang, S., Lan, R., Deng, W. A new perspective on the spatial, temporal, and vertical distribution of biomass burning: quantifying a significant increase in CO emissions. 2020 Environ. Res. Lett. 15 104091 DOI 10.1088/1748-9326/abaa7a
Lu, F., Qin, K., Cohen, J. B., He, Q., Tiwari, P., Hu, W., Ye, C., Shan, Y., Xu, Q., Wang, S., and Tu, Q.: Surface-observation-constrained high-frequency coal mine methane emissions in Shanxi, China, reveal more emissions than inventories, consistent with satellite inversion, Atmos. Chem. Phys., 25, 5837–5856, https://doi.org/10.5194/acp-25-5837-2025, 2025.
Qin, K., Hu, W., He, Q., Lu, F., and Cohen, J. B.: Individual coal mine methane emissions constrained by eddy covariance measurements: low bias and missing sources, Atmos. Chem. Phys., 24, 3009–3028, https://doi.org/10.5194/acp-24-3009-2024, 2024.
Wang, S., Cohen, J. B., Deng, W., Qin, K., & Guo, J. (2021). Using a new top-down constrained emissions inventory to attribute the previously unknown source of extreme aerosol loadings observed annually in the Monsoon Asia free troposphere. Earth's Future, 9, e2021EF002167. https://doi.org/10.1029/2021EF002167
Wang, S., Cohen, J.B., Guan, L. et al. Observationally constrained global NOx and CO emissions variability reveals sources which contribute significantly to CO2 emissions. npj Clim Atmos Sci 8, 87 (2025). https://doi.org/10.1038/s41612-025-00977-2
Zheng, B., Cohen, J. B., Lu, L., Hu, W., Tiwari, P., Lolli, S., Garzelli, A., Su, H., and Qin, K.: How can we trust TROPOMI based Methane Emissions Estimation: Calculating Emissions over Unidentified Source Regions, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2025-1446, 2025. |
Review of “Carbon reduction requires attention to the contribution of natural gas use: Combustion and leakage”.
The manuscript presents the results of a 73 day long observational campaign of methane (CH4) and carbon dioxide (CO2) fluxes made on a tall tower in Beijing. In addition, surface mobile measurements over different time periods were also made to address some specific geospatial regions within the domain covered by the flux tower. The methods to prepare and analyze the data are very standard. The findings include that the emissions of CH4 are likely anthropogenic in nature due to their similarity in time and direction to those of CO2. Comparisons were made with other previous campaigns in earlier years. They then draw some conclusions about the changes in CO2 and CH4 over time and relate these to various different policies.
The authors clearly have demonstrated that their basic measurements of flux are reasonable and representative. There should be no doubt about this point, and hence the fundamental data underlying the project looks sound. However, there are many issues. One such issue about the data is that the individual half-hour averaged flux time series over the entire time studied is not available anywhere. However, the details in the figures of the entire-campaign averaged hour-by-hour data clearly demonstrates that the hour-to-hour and day-to-day variability are both important. They also demonstrate that there are issues likely occuring at the half-hour scale, but they cannot be analyzed or discussed based on the current figures and data provided. Therefore, analyzing the data or evaluating analysis done cannot be validates, and the potential strong impacts of these 30-minute scale variations cannot be analyzed or presented. This weakens the paper.
Furthermore, there are many technical issues with the paper, as outlined in detail below. Perhaps it is just writing style or many small and unintentional mistakes. However, the net of all of these small mistakes and mis-communications in tandem lead to serious doubts as to the overall findings of the work.
Another issue is that the subsequent analysis of the data does not present any new perspectives or science. Advances in analysis by the community over the past few years should be followed. The base 30-minute and vehicle-obtained data both need to be processed at higher temporal and spatial frequency, not a mere “summer” comparison. More rigorous techniques than simple linear correlation between concentration enhancements and single variables (the other species concentration enhancement or temperature) also need to be performed. Uncertainty in the models, in the data, and in the assumptions need to be considered. Analysis of variance and of multi-species need to be performed in tandem. Specific details are presented in more detailed comments below.
Additionally, the use of background subtraction may lead to substantial errors. First, there are the issues of the observational uncertainty in the background value. Second, there are more modern papers demonstrating that background subtraction is not needed. Third, in those cases in which long-range transport is present, background subtraction is flawed in connection with the flux tower computational assumptions. This is because the equations underlying the flux calculation assume that the upper air is clean and that the emissions come from the local surface. Recent papers have demonstrated that there is in fact long-range transport into Beijing from upwind industrial sources in central China, and therefore any such events would need to be excluded from the data before analysis is performed. I raise this point since in analysis done both by my group as well as others, the time period studied in this work contains at least one such long-range transport event. Analysis of the 30-minute time series may help identify this event, and possibly others as well. In addition, this paper introduces the use of a 5-minute window to identify background values. However, given the size of the domain, this is not consistent. The observed wind speed will take more than 5 minutes to go from the edge of the domain to the tower location, and hence the length of the averaging period must be at least this long. This will change from day-to-day and hour-to-hour. The time likely needs to be longer, to account for any atmospheric recycling occurring within the domain.
For all of these reasons, I recommend that the work undergo major revisions before it be considered further. However, due to the strong people on the team, I do believe that with a considerable amount of hard work and time, that they can raise the level of the paper to such that it will make a good ultimate contribution to ACP. I am happy to continue to work with any future revisions which are brought forward.
Specific Issues:
Of course, there are newer techniques such as published in ACP in 2025 this year based on a study of CH4 in central China which completely does away with background subtraction and enhancement calculation. You could consider this new approach as well and completely avoid the issues of enhancement and background subtraction. Or you can work hard to justify why your background subtraction is valid and how it contributes to overall uncertainties in the conclusions.