Benchmarking data-driven inversion methods for the estimation of local CO<sub>2</sub> emissions from synthetic satellite images of XCO<sub>2</sub> and NO<sub>2</sub>
<p>The largest anthropogenic emissions of carbon dioxide (CO<span class="inline-formula"><sub>2</sub></span>) come from local sources, such as cities and power plants. The upcoming Copernicus CO<span class="inline-formula"><sub>2</sub&...
Saved in:
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2025-01-01
|
| Series: | Atmospheric Measurement Techniques |
| Online Access: | https://amt.copernicus.org/articles/18/211/2025/amt-18-211-2025.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841527647273746432 |
|---|---|
| author | D. Santaren J. Hakkarainen G. Kuhlmann E. Koene F. Chevallier I. Ialongo H. Lindqvist J. Nurmela J. Tamminen L. Amorós D. Brunner G. Broquet |
| author_facet | D. Santaren J. Hakkarainen G. Kuhlmann E. Koene F. Chevallier I. Ialongo H. Lindqvist J. Nurmela J. Tamminen L. Amorós D. Brunner G. Broquet |
| author_sort | D. Santaren |
| collection | DOAJ |
| description | <p>The largest anthropogenic emissions of carbon dioxide (CO<span class="inline-formula"><sub>2</sub></span>) come from local sources, such as cities and power plants. The upcoming Copernicus CO<span class="inline-formula"><sub>2</sub></span> Monitoring (CO2M) mission will provide satellite images of the CO<span class="inline-formula"><sub>2</sub></span> and NO<span class="inline-formula"><sub>2</sub></span> plumes associated with these sources at a resolution of 2 km <span class="inline-formula">×</span> 2 km and with a swath of 250 km. These images could be exploited using atmospheric-plume inversion methods to estimate local CO<span class="inline-formula"><sub>2</sub></span> emissions at the time of the satellite overpass and their corresponding uncertainties. To support the development of the operational processing of satellite imagery of the column-averaged CO<span class="inline-formula"><sub>2</sub></span> dry-air mole fraction (XCO<span class="inline-formula"><sub>2</sub></span>) and tropospheric-column NO<span class="inline-formula"><sub>2</sub></span>, this study evaluates <i>data-driven inversion methods</i>, i.e., computationally light inversion methods that directly process information from satellite images, local winds, and meteorological data, without resorting to computationally expensive dynamical atmospheric transport models. We designed an objective benchmarking exercise to analyze and compare the performance of five different data-driven inversion methods: two implementations with different complexities for the cross-sectional flux approach (CSF and LCSF), as well as one implementation each for the integrated mass enhancement (IME), divergence (Div), and Gaussian plume (GP) model inversion approaches. This exercise is based on pseudo-data experiments with simulations of synthetic <i>true</i> emissions, meteorological and concentration fields, and CO2M observations across a domain of 750 km <span class="inline-formula">×</span> 650 km, centered on eastern Germany, over 1 year. The performance of the methods is quantified in terms of the accuracy of single-image emission estimates (from individual images) or annual-average emission estimates (from the full series of images), as well as in terms of the number of instant estimates for the city of Berlin and 15 power plants within this domain. Several ensembles of estimations are conducted using different scenarios for the available synthetic datasets. These ensembles are used to analyze the sensitivity of performance to (1) data loss due to cloud cover, (2) uncertainty in the wind, or (3) the added value of simultaneous NO<span class="inline-formula"><sub>2</sub></span> images. The GP and LCSF methods generate the most accurate estimates from individual images. The deviations between the emission estimates and the true emissions from these two methods have similar interquartile ranges (IQRs), ranging from <span class="inline-formula">∼</span> 20 % to <span class="inline-formula">∼</span> 60 % depending on the scenario. When taking cloud cover into account, these methods produce 274 and 318 instant estimates, respectively, from the <span class="inline-formula">∼</span> 500 daily images, which cover significant portions of the plumes from the sources. Filtering the results based on the associated uncertainty estimates can improve the statistics of the IME and CSF methods but does so at the cost of a large decrease in the number of estimates. Due to a reliable estimation of uncertainty and, thus, a suitable<span id="page212"/> selection of estimates, the CSF method achieves similar, if not better, accuracy statistics for instant estimates compared to the GP and LCSF methods after filtering. In general, the performance of retrieving single-image estimates improves when, in addition to XCO<span class="inline-formula"><sub>2</sub></span> data, collocated NO<span class="inline-formula"><sub>2</sub></span> data are used to characterize the structure of plumes. With respect to the estimates of annual emissions, the root mean square errors (RMSEs) for the most realistic benchmarking scenario are 20 % (GP), 27 % (CSF), 31 % (LCSF), 55 % (IME), and 79 % (Div). This study suggests that the Gaussian plume and/or cross-sectional approaches are currently the most efficient tools for providing estimates of CO<span class="inline-formula"><sub>2</sub></span> emissions from satellite images, and their relatively light computational cost will enable the analysis of the massive amount of data to be provided by future satellite XCO<span class="inline-formula"><sub>2</sub></span> imagery missions.</p> |
| format | Article |
| id | doaj-art-eba9c3632fa44ec7b484ce9b925f4f45 |
| institution | Kabale University |
| issn | 1867-1381 1867-8548 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | Atmospheric Measurement Techniques |
| spelling | doaj-art-eba9c3632fa44ec7b484ce9b925f4f452025-01-15T11:51:19ZengCopernicus PublicationsAtmospheric Measurement Techniques1867-13811867-85482025-01-011821123910.5194/amt-18-211-2025Benchmarking data-driven inversion methods for the estimation of local CO<sub>2</sub> emissions from synthetic satellite images of XCO<sub>2</sub> and NO<sub>2</sub>D. Santaren0J. Hakkarainen1G. Kuhlmann2E. Koene3F. Chevallier4I. Ialongo5H. Lindqvist6J. Nurmela7J. Tamminen8L. Amorós9D. Brunner10G. Broquet11Laboratoire des Sciences du Climat et de l'Environnement, LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay, Gif-sur-Yvette, FranceEarth Observation Centre, Finnish Meteorological Institute, Helsinki, FinlandLaboratory for Air Pollution/Environmental Technology, Swiss Federal Laboratories for Materials Science and Technology (Empa), Dübendorf, SwitzerlandLaboratory for Air Pollution/Environmental Technology, Swiss Federal Laboratories for Materials Science and Technology (Empa), Dübendorf, SwitzerlandLaboratoire des Sciences du Climat et de l'Environnement, LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay, Gif-sur-Yvette, FranceEarth Observation Centre, Finnish Meteorological Institute, Helsinki, FinlandEarth Observation Centre, Finnish Meteorological Institute, Helsinki, FinlandEarth Observation Centre, Finnish Meteorological Institute, Helsinki, FinlandEarth Observation Centre, Finnish Meteorological Institute, Helsinki, FinlandEarth Observation Centre, Finnish Meteorological Institute, Helsinki, FinlandLaboratory for Air Pollution/Environmental Technology, Swiss Federal Laboratories for Materials Science and Technology (Empa), Dübendorf, SwitzerlandLaboratoire des Sciences du Climat et de l'Environnement, LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay, Gif-sur-Yvette, France<p>The largest anthropogenic emissions of carbon dioxide (CO<span class="inline-formula"><sub>2</sub></span>) come from local sources, such as cities and power plants. The upcoming Copernicus CO<span class="inline-formula"><sub>2</sub></span> Monitoring (CO2M) mission will provide satellite images of the CO<span class="inline-formula"><sub>2</sub></span> and NO<span class="inline-formula"><sub>2</sub></span> plumes associated with these sources at a resolution of 2 km <span class="inline-formula">×</span> 2 km and with a swath of 250 km. These images could be exploited using atmospheric-plume inversion methods to estimate local CO<span class="inline-formula"><sub>2</sub></span> emissions at the time of the satellite overpass and their corresponding uncertainties. To support the development of the operational processing of satellite imagery of the column-averaged CO<span class="inline-formula"><sub>2</sub></span> dry-air mole fraction (XCO<span class="inline-formula"><sub>2</sub></span>) and tropospheric-column NO<span class="inline-formula"><sub>2</sub></span>, this study evaluates <i>data-driven inversion methods</i>, i.e., computationally light inversion methods that directly process information from satellite images, local winds, and meteorological data, without resorting to computationally expensive dynamical atmospheric transport models. We designed an objective benchmarking exercise to analyze and compare the performance of five different data-driven inversion methods: two implementations with different complexities for the cross-sectional flux approach (CSF and LCSF), as well as one implementation each for the integrated mass enhancement (IME), divergence (Div), and Gaussian plume (GP) model inversion approaches. This exercise is based on pseudo-data experiments with simulations of synthetic <i>true</i> emissions, meteorological and concentration fields, and CO2M observations across a domain of 750 km <span class="inline-formula">×</span> 650 km, centered on eastern Germany, over 1 year. The performance of the methods is quantified in terms of the accuracy of single-image emission estimates (from individual images) or annual-average emission estimates (from the full series of images), as well as in terms of the number of instant estimates for the city of Berlin and 15 power plants within this domain. Several ensembles of estimations are conducted using different scenarios for the available synthetic datasets. These ensembles are used to analyze the sensitivity of performance to (1) data loss due to cloud cover, (2) uncertainty in the wind, or (3) the added value of simultaneous NO<span class="inline-formula"><sub>2</sub></span> images. The GP and LCSF methods generate the most accurate estimates from individual images. The deviations between the emission estimates and the true emissions from these two methods have similar interquartile ranges (IQRs), ranging from <span class="inline-formula">∼</span> 20 % to <span class="inline-formula">∼</span> 60 % depending on the scenario. When taking cloud cover into account, these methods produce 274 and 318 instant estimates, respectively, from the <span class="inline-formula">∼</span> 500 daily images, which cover significant portions of the plumes from the sources. Filtering the results based on the associated uncertainty estimates can improve the statistics of the IME and CSF methods but does so at the cost of a large decrease in the number of estimates. Due to a reliable estimation of uncertainty and, thus, a suitable<span id="page212"/> selection of estimates, the CSF method achieves similar, if not better, accuracy statistics for instant estimates compared to the GP and LCSF methods after filtering. In general, the performance of retrieving single-image estimates improves when, in addition to XCO<span class="inline-formula"><sub>2</sub></span> data, collocated NO<span class="inline-formula"><sub>2</sub></span> data are used to characterize the structure of plumes. With respect to the estimates of annual emissions, the root mean square errors (RMSEs) for the most realistic benchmarking scenario are 20 % (GP), 27 % (CSF), 31 % (LCSF), 55 % (IME), and 79 % (Div). This study suggests that the Gaussian plume and/or cross-sectional approaches are currently the most efficient tools for providing estimates of CO<span class="inline-formula"><sub>2</sub></span> emissions from satellite images, and their relatively light computational cost will enable the analysis of the massive amount of data to be provided by future satellite XCO<span class="inline-formula"><sub>2</sub></span> imagery missions.</p>https://amt.copernicus.org/articles/18/211/2025/amt-18-211-2025.pdf |
| spellingShingle | D. Santaren J. Hakkarainen G. Kuhlmann E. Koene F. Chevallier I. Ialongo H. Lindqvist J. Nurmela J. Tamminen L. Amorós D. Brunner G. Broquet Benchmarking data-driven inversion methods for the estimation of local CO<sub>2</sub> emissions from synthetic satellite images of XCO<sub>2</sub> and NO<sub>2</sub> Atmospheric Measurement Techniques |
| title | Benchmarking data-driven inversion methods for the estimation of local CO<sub>2</sub> emissions from synthetic satellite images of XCO<sub>2</sub> and NO<sub>2</sub> |
| title_full | Benchmarking data-driven inversion methods for the estimation of local CO<sub>2</sub> emissions from synthetic satellite images of XCO<sub>2</sub> and NO<sub>2</sub> |
| title_fullStr | Benchmarking data-driven inversion methods for the estimation of local CO<sub>2</sub> emissions from synthetic satellite images of XCO<sub>2</sub> and NO<sub>2</sub> |
| title_full_unstemmed | Benchmarking data-driven inversion methods for the estimation of local CO<sub>2</sub> emissions from synthetic satellite images of XCO<sub>2</sub> and NO<sub>2</sub> |
| title_short | Benchmarking data-driven inversion methods for the estimation of local CO<sub>2</sub> emissions from synthetic satellite images of XCO<sub>2</sub> and NO<sub>2</sub> |
| title_sort | benchmarking data driven inversion methods for the estimation of local co sub 2 sub emissions from synthetic satellite images of xco sub 2 sub and no sub 2 sub |
| url | https://amt.copernicus.org/articles/18/211/2025/amt-18-211-2025.pdf |
| work_keys_str_mv | AT dsantaren benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT jhakkarainen benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT gkuhlmann benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT ekoene benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT fchevallier benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT iialongo benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT hlindqvist benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT jnurmela benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT jtamminen benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT lamoros benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT dbrunner benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub AT gbroquet benchmarkingdatadriveninversionmethodsfortheestimationoflocalcosub2subemissionsfromsyntheticsatelliteimagesofxcosub2subandnosub2sub |