the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 13 Jun 2023
High-resolution U.S. methane emissions inferred from an inversion of 2019 TROPOMI satellite data: contributions from individual states, urban areas, and landfills
Abstract. We quantify 2019 methane emissions in the contiguous U.S. (CONUS) at 0.25° × 0.3125° resolution by inverse analysis of atmospheric methane columns measured by the Tropospheric Monitoring Instrument (TROPOMI). A gridded version of the U.S. Environmental Protection Agency (EPA) Greenhouse Gas Emissions Inventory (GHGI) serves as the basis for the prior estimate for the inversion. We optimize emissions and quantify observing system information content for an eight-member inversion ensemble through analytical minimization of a Bayesian cost function. We achieve high resolution with a reduced-rank characterization of the observing system that optimally preserves information content. Our optimal (posterior) estimate of anthropogenic emissions in CONUS is 30.9 (30.0–31.8) Tg a-1, where the values in parentheses give the spread of the ensemble. This is a 13 % increase from the 2023 GHGI estimate for CONUS in 2019. We find livestock emissions of 10.4 (10.0–10.7) Tg a-1, oil and gas of 10.4 (10.1–10.7) Tg a-1, coal of 1.5 (1.2–1.9) Tg a-1, landfills of 6.9 (6.4–7.5) Tg a-1, wastewater of 0.6 (0.5–0.7), and other anthropogenic sources of 1.1 (1.0–1.2) Tg a-1. The largest increase relative to the GHGI occurs for landfills (51 %), with smaller increases for oil and gas (12 %) and livestock (11 %). These three sectors are responsible for 89 % of posterior anthropogenic emissions in CONUS. The largest decrease (28 %) is for coal. We exploit the high resolution of our inversion to quantify emissions from 73 individual landfills, where we find emissions are on median 77 % larger than the values reported to the EPA’s Greenhouse Gas Reporting Program (GHGRP), a key data source for the GHGI. We attribute this underestimate to overestimated recovery efficiencies at landfill gas facilities and to under-accounting of site-specific operational changes and leaks. We also quantify emissions for the 48 individual states in CONUS, which we compare to the GHGI’s new state-level inventories and to independent state-produced inventories. Our posterior emissions are on average 34 % larger than the 2022 GHGI in the largest 10 methane-producing states, with the biggest upward adjustments in states with large oil and gas emissions, including Texas, New Mexico, Louisiana, and Oklahoma. We also calculate emissions for 95 geographically diverse urban areas in CONUS. Emissions for these urban areas total 6.0 (5.4–6.7) Tg a-1 and are on average 39 (27–52) % larger than a gridded version of the 2023 GHGI, which we attribute to underestimated landfill and gas distribution emissions.
- Preprint
(18911 KB) - Metadata XML
-
Supplement
(659 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2023-946', Anonymous Referee #1, 21 Jul 2023
Nesser et al., present detailed results for methane emissions in the contiguous U.S. as derived via inverse modelling using satellite observations of methane columns. They present new and very interesting results appropriate for EGUsphere and the paper is very well written. I strongly recommend publication in EGUsphere after the issues listed below have been carefully addressed by the authors.
Most of my comments relate to minor issues (see below) but there is one major aspect:
Whereas an ensemble of inversions has been used to obtain reliable results (which is very good) only one satellite data product has been used. Taking into account the high (also political) relevance of the results - the authors present very detailed emission results for various sectors and down to quite high spatial resolution (state, urban and even facility level) - I wonder if this is really appropriate. I have doubts in particular because an initial (i.e., quite old) satellite data product has been used which has known bias issues. Several of these bias issues are listed in the manuscript (line 209 following). It is written in the paper that because of the bias issues the authors could not use the data as they are but had to correct and filter them. For example, they had to remove data with too low methane values (e.g., data below 1700 ppb as these data were considered unrealistic). Bias corrections done by users are not uncommon but should be avoided if better data exist. On line 355 the authors write that they are aware of better data: „We also compare the TROPOMI v14 data used here to the most recent data (v19), which has improved bias corrections and performance compared to GOSAT in North America (Balasus et al., 2023). We find no correlation (R2 = 0.03) between our posterior scaling factors and the mean (v14 - v19) difference, suggesting that biases in the v14 data do not influence our posterior emissions.“. While this is good to know, this is by far not sufficient. The high spatial resolution methane emission information is contained in small scale methane gradients and even small differences of the methane data can have a large impact on the derived emissions. I therefore strongly recommend repeating the analysis using the latest (v19 or higher) product.
If resources permit, I recommend to also use an ensemble approach for the satellite data (even if only 2-3 products can be used) given the high relevance of the results and that these satellite data are the key input data for the inversion. This would give additional strong confidence that the quite large reported differences compared to the EPA inventory are a reliable finding and do not significantly depend on the peculiarities of a given satellite data product. In this context I would expect that the ensemble-based results will have larger (and likely more realistic) error bars than the (surprisingly small) error bars as shown in, e.g., Fig. 4. In this case I suggest to include also the TROPOMI WFMD v1.8 XCH4 data product described in Schneising et al., 2023. In that publication several comparisons with the operational TROPOMI product are shown (note that this operational RemoTeC algorithm product is very similar as the scientific RemoTeC product used here) indicating higher accuracy of the WFMD product due to much reduced bias (e.g., much reduced “striping errors” etc.). Such a satellite data ensemble approach would strengthen the manuscript even further, which is relevant here given the importance of the paper and the level of detail reported.
Minor aspects:
Line 61: Not sure if the “3% success rate” is valid for the real data used here as the cited (quite old) document (Hasekamp et al., 2019) covers algorithm related theoretical aspects not based on multi-year statistics.
Sect. 2: The time dependence of the emissions is prescribed and not optimized, or? If yes, then this needs to be explicitly mentioned somewhere.
Line 205: Statement “We use only high-quality retrievals …”. This is misleading. If data are flagged “good” than this does not necessarily mean they are “good”. Please rephrase to describe what actually has been done (We use only data with quality flag “good” (or equivalent)).
Eq. (6): If I understand correctly than emission delta_x is the methane emission with unit mass per time per area whereas all other emissions in the manuscript are reported as mass per time. If this is true than please add this information or add the division by area on the right hand side of the equation explicitly.
Line 338 following: “We isolate anthropogenic emissions by removing contributions from wetlands and other natural sources following Sect. 2.8.”. Perfect knowledge of wetland emissions compared to anthropogenic emissions is a very strong assumption given the distribution and magnitude of the wetland emissions as shown in Fig. 1. Is this considered somehow for the (surprisingly small) reported error bars (as shown in, e.g., Fig. 4)?
Line 364: Please check “27.6 (22.6 – 23.9)” as mean value outside range.
Line 444: Please check “2.8 (2.8 – 2.9)”.
Line 446: Please consider citing also Veefkind et al., 2023, studying also Permian emissions.
References
Schneising, O., Buchwitz, M., Hachmeister, J., Vanselow, S., Reuter, M., Buschmann, M., Bovensmann, H., and Burrows, J. P.: Advances in retrieving XCH4 and XCO from Sentinel-5 Precursor: improvements in the scientific TROPOMI/WFMD algorithm, Atmos. Meas. Tech., 16, 669–694, https://doi.org/10.5194/amt-16-669-2023, 2023.
Veefkind, J. P., Serrano-Calvo, R., de Gouw, J., Dix, B., Schneising, O., Buchwitz, M., Barré, J., van der A, R. J., Liu, M., Levelt, P. F., Widespread frequent methane emissions from the oil and gas industry in the Permian basin, Journal of Geophysical Research: Atmospheres, 128, e2022JD037479, pp. 13, https://doi.org/10.1029/2022JD037479, 2023.
Citation: https://doi.org/10.5194/egusphere-2023-946-RC1 -
RC2: 'Comment on egusphere-2023-946', Anonymous Referee #2, 08 Dec 2023
Overview:
Nesser et al. present a study using space-based observations of CH4 in conjunction with an atmospheric transport model and prior emissions estimates to solve for an optimal set of methane emissions for the 48 contiguous US states. With this analysis they then breakdown the optimized emissions by sector, state, and urban region to further refine the atmospheric constraint on US methane emissions. The topic is important, timely, and well-suited for ACP. The paper is well-written. The method and analysis looks overall sound. Overall this paper helps advance our knowledge of US methane emissions. I have only a couple concerns and suggestions I outline below. Once the authors address these (which shouldn’t be too onerous) I would enthusiastically support publication.
Comments:
Temporal resolution/seasonality: I couldn’t find any explicit discussion of the temporal resolution of the analysis. My understanding is that a 3 hour time-step is used in the geos-chem simulations, and that the inversion is conducted for 1 full calendar year. That is to say, the analysis assumes all fluxes are static over the year and all observations, from different days with sensitivity to the same surface regions put equal (depending on winds of course) weighting on the posterior flux solution. Is that correct? It would be helpful to explicitly discuss the temporal element of the inversion. It further than would be helpful to discuss the assumptions underlying this and the implications/impacts of those assumptions. Some sources are known to have large intermittency (oil&gas) – would you expect to over/under sample those or get a reasonable average picture? Do you have equal measurements across seasons? How do you treat emissions seasonal variance in the prior/posterior? Some sources, such as wetlands, have large seasonality – how do you address that? How do you consider seasonality in your landfill analysis? If you are biased toward summer observations, you might conclude high emissions as landfills have a non-insignificant temperature relationship with emissions. Is there an uncertainty term that should be added for addressing seasonal challenges?
Region of study: You should be explicit that this paper analyzes emissions from onshore the 48 contiguous states – you are not constraining offshore oil/gas emission nor AK/HI. Just a detail important for precision. Also relevant in discussing the oil&gas sector for the entire US, as both offshore and AK are non-negligible portions of that sector.
Uncertainty: I may have missed this, but how did you construct the confidence interval for the total posterior flux from the US (is that from the 8 member ensemble?)? Please clarify.
I also didn’t understand how you determined uncertainty when you get to the posterior sectoral breakdown – please expand.
Source attribution: How does the assumption on prior source distribution impact the results? If you iterate and correct %’s based on your results, would it change the sectoral? Is this type of uncertainty conveyed in the sectoral uncertainty as presented?
Transport error: How was transport error handled/accounted for or is it not represented in uncertainty? What additional uncertainty may this add? Regional studies often find >10% uncertainty from transport error alone.
Posterior oil and gas: not necessary, but might be nice to show what your basin numbers are compared to airborne/tower studies that have been done in a bunch of the basins, even in the most recent airborne studies were ~5 years earlier. Examples include: Denver-Julesberg, Uintah, Bakken, Marcellus, among some others..
Line 50 “These top-down emission estimates are most useful if they achieve high spatial resolution and maximize the information content of the observation- model system”. Saying “most useful” is ill defined and misleading, as depending on question of interest high spatial resolution may not lead to most optimal result. Suggest rephrasing
Final sentence: you should add something like “that may be attributable to decreasing emissions between study periods” for context so the LA comparison isn’t misinterpreted.
Data availability: It would be good to include data files that have all the states and all the urban region prior/p
Citation: https://doi.org/10.5194/egusphere-2023-946-RC2 -
AC1: 'Comment on egusphere-2023-946', Hannah Nesser, 20 Jan 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-946/egusphere-2023-946-AC1-supplement.pdf
Status: closed
-
RC1: 'Comment on egusphere-2023-946', Anonymous Referee #1, 21 Jul 2023
Nesser et al., present detailed results for methane emissions in the contiguous U.S. as derived via inverse modelling using satellite observations of methane columns. They present new and very interesting results appropriate for EGUsphere and the paper is very well written. I strongly recommend publication in EGUsphere after the issues listed below have been carefully addressed by the authors.
Most of my comments relate to minor issues (see below) but there is one major aspect:
Whereas an ensemble of inversions has been used to obtain reliable results (which is very good) only one satellite data product has been used. Taking into account the high (also political) relevance of the results - the authors present very detailed emission results for various sectors and down to quite high spatial resolution (state, urban and even facility level) - I wonder if this is really appropriate. I have doubts in particular because an initial (i.e., quite old) satellite data product has been used which has known bias issues. Several of these bias issues are listed in the manuscript (line 209 following). It is written in the paper that because of the bias issues the authors could not use the data as they are but had to correct and filter them. For example, they had to remove data with too low methane values (e.g., data below 1700 ppb as these data were considered unrealistic). Bias corrections done by users are not uncommon but should be avoided if better data exist. On line 355 the authors write that they are aware of better data: „We also compare the TROPOMI v14 data used here to the most recent data (v19), which has improved bias corrections and performance compared to GOSAT in North America (Balasus et al., 2023). We find no correlation (R2 = 0.03) between our posterior scaling factors and the mean (v14 - v19) difference, suggesting that biases in the v14 data do not influence our posterior emissions.“. While this is good to know, this is by far not sufficient. The high spatial resolution methane emission information is contained in small scale methane gradients and even small differences of the methane data can have a large impact on the derived emissions. I therefore strongly recommend repeating the analysis using the latest (v19 or higher) product.
If resources permit, I recommend to also use an ensemble approach for the satellite data (even if only 2-3 products can be used) given the high relevance of the results and that these satellite data are the key input data for the inversion. This would give additional strong confidence that the quite large reported differences compared to the EPA inventory are a reliable finding and do not significantly depend on the peculiarities of a given satellite data product. In this context I would expect that the ensemble-based results will have larger (and likely more realistic) error bars than the (surprisingly small) error bars as shown in, e.g., Fig. 4. In this case I suggest to include also the TROPOMI WFMD v1.8 XCH4 data product described in Schneising et al., 2023. In that publication several comparisons with the operational TROPOMI product are shown (note that this operational RemoTeC algorithm product is very similar as the scientific RemoTeC product used here) indicating higher accuracy of the WFMD product due to much reduced bias (e.g., much reduced “striping errors” etc.). Such a satellite data ensemble approach would strengthen the manuscript even further, which is relevant here given the importance of the paper and the level of detail reported.
Minor aspects:
Line 61: Not sure if the “3% success rate” is valid for the real data used here as the cited (quite old) document (Hasekamp et al., 2019) covers algorithm related theoretical aspects not based on multi-year statistics.
Sect. 2: The time dependence of the emissions is prescribed and not optimized, or? If yes, then this needs to be explicitly mentioned somewhere.
Line 205: Statement “We use only high-quality retrievals …”. This is misleading. If data are flagged “good” than this does not necessarily mean they are “good”. Please rephrase to describe what actually has been done (We use only data with quality flag “good” (or equivalent)).
Eq. (6): If I understand correctly than emission delta_x is the methane emission with unit mass per time per area whereas all other emissions in the manuscript are reported as mass per time. If this is true than please add this information or add the division by area on the right hand side of the equation explicitly.
Line 338 following: “We isolate anthropogenic emissions by removing contributions from wetlands and other natural sources following Sect. 2.8.”. Perfect knowledge of wetland emissions compared to anthropogenic emissions is a very strong assumption given the distribution and magnitude of the wetland emissions as shown in Fig. 1. Is this considered somehow for the (surprisingly small) reported error bars (as shown in, e.g., Fig. 4)?
Line 364: Please check “27.6 (22.6 – 23.9)” as mean value outside range.
Line 444: Please check “2.8 (2.8 – 2.9)”.
Line 446: Please consider citing also Veefkind et al., 2023, studying also Permian emissions.
References
Schneising, O., Buchwitz, M., Hachmeister, J., Vanselow, S., Reuter, M., Buschmann, M., Bovensmann, H., and Burrows, J. P.: Advances in retrieving XCH4 and XCO from Sentinel-5 Precursor: improvements in the scientific TROPOMI/WFMD algorithm, Atmos. Meas. Tech., 16, 669–694, https://doi.org/10.5194/amt-16-669-2023, 2023.
Veefkind, J. P., Serrano-Calvo, R., de Gouw, J., Dix, B., Schneising, O., Buchwitz, M., Barré, J., van der A, R. J., Liu, M., Levelt, P. F., Widespread frequent methane emissions from the oil and gas industry in the Permian basin, Journal of Geophysical Research: Atmospheres, 128, e2022JD037479, pp. 13, https://doi.org/10.1029/2022JD037479, 2023.
Citation: https://doi.org/10.5194/egusphere-2023-946-RC1 -
RC2: 'Comment on egusphere-2023-946', Anonymous Referee #2, 08 Dec 2023
Overview:
Nesser et al. present a study using space-based observations of CH4 in conjunction with an atmospheric transport model and prior emissions estimates to solve for an optimal set of methane emissions for the 48 contiguous US states. With this analysis they then breakdown the optimized emissions by sector, state, and urban region to further refine the atmospheric constraint on US methane emissions. The topic is important, timely, and well-suited for ACP. The paper is well-written. The method and analysis looks overall sound. Overall this paper helps advance our knowledge of US methane emissions. I have only a couple concerns and suggestions I outline below. Once the authors address these (which shouldn’t be too onerous) I would enthusiastically support publication.
Comments:
Temporal resolution/seasonality: I couldn’t find any explicit discussion of the temporal resolution of the analysis. My understanding is that a 3 hour time-step is used in the geos-chem simulations, and that the inversion is conducted for 1 full calendar year. That is to say, the analysis assumes all fluxes are static over the year and all observations, from different days with sensitivity to the same surface regions put equal (depending on winds of course) weighting on the posterior flux solution. Is that correct? It would be helpful to explicitly discuss the temporal element of the inversion. It further than would be helpful to discuss the assumptions underlying this and the implications/impacts of those assumptions. Some sources are known to have large intermittency (oil&gas) – would you expect to over/under sample those or get a reasonable average picture? Do you have equal measurements across seasons? How do you treat emissions seasonal variance in the prior/posterior? Some sources, such as wetlands, have large seasonality – how do you address that? How do you consider seasonality in your landfill analysis? If you are biased toward summer observations, you might conclude high emissions as landfills have a non-insignificant temperature relationship with emissions. Is there an uncertainty term that should be added for addressing seasonal challenges?
Region of study: You should be explicit that this paper analyzes emissions from onshore the 48 contiguous states – you are not constraining offshore oil/gas emission nor AK/HI. Just a detail important for precision. Also relevant in discussing the oil&gas sector for the entire US, as both offshore and AK are non-negligible portions of that sector.
Uncertainty: I may have missed this, but how did you construct the confidence interval for the total posterior flux from the US (is that from the 8 member ensemble?)? Please clarify.
I also didn’t understand how you determined uncertainty when you get to the posterior sectoral breakdown – please expand.
Source attribution: How does the assumption on prior source distribution impact the results? If you iterate and correct %’s based on your results, would it change the sectoral? Is this type of uncertainty conveyed in the sectoral uncertainty as presented?
Transport error: How was transport error handled/accounted for or is it not represented in uncertainty? What additional uncertainty may this add? Regional studies often find >10% uncertainty from transport error alone.
Posterior oil and gas: not necessary, but might be nice to show what your basin numbers are compared to airborne/tower studies that have been done in a bunch of the basins, even in the most recent airborne studies were ~5 years earlier. Examples include: Denver-Julesberg, Uintah, Bakken, Marcellus, among some others..
Line 50 “These top-down emission estimates are most useful if they achieve high spatial resolution and maximize the information content of the observation- model system”. Saying “most useful” is ill defined and misleading, as depending on question of interest high spatial resolution may not lead to most optimal result. Suggest rephrasing
Final sentence: you should add something like “that may be attributable to decreasing emissions between study periods” for context so the LA comparison isn’t misinterpreted.
Data availability: It would be good to include data files that have all the states and all the urban region prior/p
Citation: https://doi.org/10.5194/egusphere-2023-946-RC2 -
AC1: 'Comment on egusphere-2023-946', Hannah Nesser, 20 Jan 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-946/egusphere-2023-946-AC1-supplement.pdf
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,436 | 687 | 42 | 2,165 | 149 | 33 | 38 |
- HTML: 1,436
- PDF: 687
- XML: 42
- Total: 2,165
- Supplement: 149
- BibTeX: 33
- EndNote: 38
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
