the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Jun 2023
Two years of satellite-based carbon dioxide emission quantification at the world’s largest coal-fired power plants
Andrew Thorpe
Charles Miller
Alana Ayasse
Ralph Jiorle
Riley Duren
Ray Nassar
Jon-Paul Mastrogiacomo
Robert Nelson
Abstract. Carbon dioxide (CO2) emissions from combustion sources are uncertain in many places across the globe. Satellites have the ability to detect and quantify emissions from large CO2 point sources, including coal-fired power plants. In this study, we tasked the PRecursore IperSpettrale della Missione Applicativa (PRISMA) satellite imaging spectrometer and the Orbiting Carbon Observatory-3 (OCO-3) instrument onboard the International Space Station at over 30 coal-fired power plants routinely between 2021–2022. CO2 plumes were detected in 50 % of acquired PRISMA scenes, which is consistent with the combined influence of viewing parameters on detection (solar illumination, surface reflectance) and unknown factors (like daily operational status). We compare satellite-derived emission rates to in situ stack emission observations and find average agreement to within 27 % for PRISMA and 30 % for OCO-3, though more observations are needed to robustly characterize the error. We highlight two examples of fusing PRISMA with OCO-2 and OCO-3 observations in South Africa and India. For India, we acquired PRISMA and OCO-3 observations on the same day and use the high spatial resolution capability of PRISMA (30 m spatial/pixel resolution) to partition relative contributions of two distinct emitting power plants to the net emission. Though an encouraging start, two years of tasking these satellites did not produce sufficient observations to estimate annual average emission rates within low (<15 %) uncertainties. However, as the constellation of CO2-observing satellites is poised to significantly improve in the coming decade, this study offers an approach to leverage multiple observation platforms to better understand large anthropogenic emission sources.
-
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
(1472 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1472 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Daniel Cusworth et al.
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-1408', Anonymous Referee #1, 18 Jul 2023
The authors have presented CO2 emissions from power plants detected from PRISMA and OCO-3. PRISMA observations with high spatial resolution detect the plume of individual power plants. This is very impressive. The topic is very interesting. However, some major issues are needed to be addressed before publication.
Major comments:
The structure of the paper still needs to be improved. Sometimes it is not easy to follow. The study shows the results of PRISMA, but the main advantages of using PRISMA and IME are not very clearly emphasized.
The authors have written a nice introduction of studies using OCO2 and OCO3 but the introduction related to PRISMA is only one line. In section 2.2 the PRISMA instrument is introduced here. Detail information (spatial resolution, overpass time, etc.) of the instrument is missing. Please add more background information about PRISMA and studies using PRISMA.
Please check the method section. There is no section 2.1. The numbering of subsection started directly from section 2.2. Please provide a subsection to introduce some datasets used in the study, such as the GEM dataset, CEMS dataset.
L185, the calculation of L is assuming that the plume is square. But in reality, not many plumes are square. I assume that the L is underestimated here. This will also affect the uncertainties of emissions.
I am wondering if there are any validation of the xco2 retrievals of PRISMA. Is there any bias of the xco2 retrievals? How much is the uncertainty? For comparison with OCO2 and OCO3, are there any systematic biases among the retrieval products?
When compare IME and the Gaussian Plume method for OCO-3, we clearly see that the estimates from IME clustered together (Figure 5). This means that the sensitivity of IME for OCO-3 is lower than the Gaussian Plume method. Lower resolution of OCO-3 limits the number of plume pixels. This point should be added when you described the discrepancy between IME and the Gaussian Plume method.
Please add DOI number in references
Minor comments:
L31 ‘better understand’ than what? Compare to which part? Please make it more specific.
L112, ‘non-overlapping’. Non-overlapping with what? Please specify it here.
About table 1, please add some explain in the manuscript about why the plume detections are lower than the clear-sky observations. There are many cases that having clear-sky observation and plume detections, but there shows ‘NA’. The explanation is not very clear in the text.
L167 Please specify that the calculation of background is described in section 3.2
L169 ‘one-pixel dilation filter’, do you have reference for this filter?
L186-187: The Ueff calculated from Varon et al. (2018) is for methane plume observations with GHGsat instrument. Is the relationship also suitable for CO2?
L283, ‘210 scenes, 104 were determined.’ Please explain more clearly why there are fewer determined.
L 306, it would be better to mention one or two sentence of the ‘logistic regression classification function’ cited from (Fan et al., 2008).
L446-448: Please check the sentence and rewrite it. You mentioned the Adani plant two times.
Lin 448-451: The uncertainties of PRISMA look very low here: only 16%. You explain that the discrepancy is caused by the uncertainty. The main discrepancy of emission rates between PRISMA and OCO3 is that there are more emission sources mixing in the detected plume of OCO3. We see this in Figure 7b that the XCO2 enhancement around red point is among the CO2 plumes of all power plants and industrial area in Mundra. Because OCO3 has coarse spatial resolution, and this is an important reason for the difference. This should be mentioned here.
L 451-452 Validation won’t reduce the uncertainties but only to better quantify the uncertainties.
Figure 7 Please add the coordinate in each sub-figure to help readers to see the scale of each map.
Citation: https://doi.org/10.5194/egusphere-2023-1408-RC1 -
RC2: 'Comment on egusphere-2023-1408', Anonymous Referee #2, 21 Jul 2023
The authors have observed a selection of major coal-fired power plants using the PRISMA and OCO-3 satellite instruments to make estimates of CO2 emissions. The results show how difficult this objective is to achieve with the current available instruments. Even though the authors have obtained emission estimates using varied techniques of using the data from these instruments, the analysis resulted in too little sampling to obtain emission estimates with acceptably low uncertainties. Estimates can only be obtained from these instruments under exactly the right conditions in terms of cloud amount, solar zenith angle, and surface reflectivity. The authors have considered all of these effects and had to use only small fractions of the observed scenes that had been obtained from the satellites (only about 1/3 of the observed scenes from PRISMA over 2 years were usable, and about 10% of OCO-3 scenes were usable). The manuscript, for the most part, adequately describes the analysis techniques that were used and sufficiently describes the results. Some specific improvements are suggested below.
Detailed comments:
lines 16-19 of abstract and other locations in the text: I know that the terminology among the users of these satellite instruments frequently includes the term "tasked" to indicate what the satellite is instructed to do. However, in a manuscript which may be read by others outside the satellite community, another term would be more understandable. Here in this sentence, I would recommend revising to "In this study the PR......Station observed over 30......2021-2022." Rewording is recommended in several places were "task" or "tasked" are used.
lines 35-36: Give some relative numbers for the importance of power plants and other sources such as motor vehicles for CO2 emissions.
lines 93-94: "observing" and "observed" instead of "tasking' and "tasked"
line 188: If I understand correctly, Ueff would be the wind speed at plume level. I don't understand why this wind speed would be significantly smaller than U10. Please provide some explanation.
line 446: the second time "Adani" appears in this sentence, it should be changed to "Tatu Mundra".
line 465: We observed a global....
Citation: https://doi.org/10.5194/egusphere-2023-1408-RC2 - AC1: 'Comment on egusphere-2023-1408', Daniel Cusworth, 29 Aug 2023
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-1408', Anonymous Referee #1, 18 Jul 2023
The authors have presented CO2 emissions from power plants detected from PRISMA and OCO-3. PRISMA observations with high spatial resolution detect the plume of individual power plants. This is very impressive. The topic is very interesting. However, some major issues are needed to be addressed before publication.
Major comments:
The structure of the paper still needs to be improved. Sometimes it is not easy to follow. The study shows the results of PRISMA, but the main advantages of using PRISMA and IME are not very clearly emphasized.
The authors have written a nice introduction of studies using OCO2 and OCO3 but the introduction related to PRISMA is only one line. In section 2.2 the PRISMA instrument is introduced here. Detail information (spatial resolution, overpass time, etc.) of the instrument is missing. Please add more background information about PRISMA and studies using PRISMA.
Please check the method section. There is no section 2.1. The numbering of subsection started directly from section 2.2. Please provide a subsection to introduce some datasets used in the study, such as the GEM dataset, CEMS dataset.
L185, the calculation of L is assuming that the plume is square. But in reality, not many plumes are square. I assume that the L is underestimated here. This will also affect the uncertainties of emissions.
I am wondering if there are any validation of the xco2 retrievals of PRISMA. Is there any bias of the xco2 retrievals? How much is the uncertainty? For comparison with OCO2 and OCO3, are there any systematic biases among the retrieval products?
When compare IME and the Gaussian Plume method for OCO-3, we clearly see that the estimates from IME clustered together (Figure 5). This means that the sensitivity of IME for OCO-3 is lower than the Gaussian Plume method. Lower resolution of OCO-3 limits the number of plume pixels. This point should be added when you described the discrepancy between IME and the Gaussian Plume method.
Please add DOI number in references
Minor comments:
L31 ‘better understand’ than what? Compare to which part? Please make it more specific.
L112, ‘non-overlapping’. Non-overlapping with what? Please specify it here.
About table 1, please add some explain in the manuscript about why the plume detections are lower than the clear-sky observations. There are many cases that having clear-sky observation and plume detections, but there shows ‘NA’. The explanation is not very clear in the text.
L167 Please specify that the calculation of background is described in section 3.2
L169 ‘one-pixel dilation filter’, do you have reference for this filter?
L186-187: The Ueff calculated from Varon et al. (2018) is for methane plume observations with GHGsat instrument. Is the relationship also suitable for CO2?
L283, ‘210 scenes, 104 were determined.’ Please explain more clearly why there are fewer determined.
L 306, it would be better to mention one or two sentence of the ‘logistic regression classification function’ cited from (Fan et al., 2008).
L446-448: Please check the sentence and rewrite it. You mentioned the Adani plant two times.
Lin 448-451: The uncertainties of PRISMA look very low here: only 16%. You explain that the discrepancy is caused by the uncertainty. The main discrepancy of emission rates between PRISMA and OCO3 is that there are more emission sources mixing in the detected plume of OCO3. We see this in Figure 7b that the XCO2 enhancement around red point is among the CO2 plumes of all power plants and industrial area in Mundra. Because OCO3 has coarse spatial resolution, and this is an important reason for the difference. This should be mentioned here.
L 451-452 Validation won’t reduce the uncertainties but only to better quantify the uncertainties.
Figure 7 Please add the coordinate in each sub-figure to help readers to see the scale of each map.
Citation: https://doi.org/10.5194/egusphere-2023-1408-RC1 -
RC2: 'Comment on egusphere-2023-1408', Anonymous Referee #2, 21 Jul 2023
The authors have observed a selection of major coal-fired power plants using the PRISMA and OCO-3 satellite instruments to make estimates of CO2 emissions. The results show how difficult this objective is to achieve with the current available instruments. Even though the authors have obtained emission estimates using varied techniques of using the data from these instruments, the analysis resulted in too little sampling to obtain emission estimates with acceptably low uncertainties. Estimates can only be obtained from these instruments under exactly the right conditions in terms of cloud amount, solar zenith angle, and surface reflectivity. The authors have considered all of these effects and had to use only small fractions of the observed scenes that had been obtained from the satellites (only about 1/3 of the observed scenes from PRISMA over 2 years were usable, and about 10% of OCO-3 scenes were usable). The manuscript, for the most part, adequately describes the analysis techniques that were used and sufficiently describes the results. Some specific improvements are suggested below.
Detailed comments:
lines 16-19 of abstract and other locations in the text: I know that the terminology among the users of these satellite instruments frequently includes the term "tasked" to indicate what the satellite is instructed to do. However, in a manuscript which may be read by others outside the satellite community, another term would be more understandable. Here in this sentence, I would recommend revising to "In this study the PR......Station observed over 30......2021-2022." Rewording is recommended in several places were "task" or "tasked" are used.
lines 35-36: Give some relative numbers for the importance of power plants and other sources such as motor vehicles for CO2 emissions.
lines 93-94: "observing" and "observed" instead of "tasking' and "tasked"
line 188: If I understand correctly, Ueff would be the wind speed at plume level. I don't understand why this wind speed would be significantly smaller than U10. Please provide some explanation.
line 446: the second time "Adani" appears in this sentence, it should be changed to "Tatu Mundra".
line 465: We observed a global....
Citation: https://doi.org/10.5194/egusphere-2023-1408-RC2 - AC1: 'Comment on egusphere-2023-1408', Daniel Cusworth, 29 Aug 2023
Peer review completion
Journal article(s) based on this preprint
Daniel Cusworth et al.
Daniel Cusworth et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 458 | 246 | 20 | 724 | 12 | 14 |
- HTML: 458
- PDF: 246
- XML: 20
- Total: 724
- BibTeX: 12
- EndNote: 14
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1472 KB) - Metadata XML