the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Jun 2026
Impact of satellite observations on assimilation inversion of high-resolution urban-scale carbon dioxide fluxes
Abstract. Cities are major sources of global carbon. Accurately quantifying urban-scale carbon dioxide (CO2) fluxes is essential for supporting targeted emission reduction policies and effective monitoring. To address the limitations in the accuracy of current urban carbon emission estimates, we developed FEISSO, an urban-scale CO2 flux inversion system that integrates a Lagrangian atmospheric transport model with a Bayesian assimilation framework. FEISSO was used to systematically explore the feasibility of retrieving high-resolution flux distributions from satellite-based XCO2 observations. Sensitivity experiments were conducted in Weifang, Chengdu, and Xining (China) to identify key influencing factors in CO2 flux inversion. Results show that the resolution of meteorological drivers substantially affects the accuracy of simulated transport trajectories, with higher resolution (0.25°) improving the spatial fidelity of flux retrieval. Sensitivity analysis indicates that the column-averaged CO2 observation error and the total error for the inversion domain are the dominant factors affecting the total flux estimates, and they induce a "seesaw" effect in the spatial distribution of emissions. In contrast, prior flux error and spatial correlation length for land have limited influence on the total emissions but primarily affect the spatial pattern of weak emission regions and the smoothness of flux fields, respectively. Differences in topography and meteorological conditions across cities govern the temporal response of flux estimates to observations. With optimized parameter settings, the system successfully retrieved 10-days of total CO2 emissions for Weifang, Chengdu, and Xining, showing overall consistency with EDGAR and local inventory data. The retrieved emissions correspond to 85.8 %, 190.42 %, and 86.4 % of the EDGAR estimates for the three cities, respectively, while the relative differences from local inventories are 2.3 % for Weifang and 10.9 % for Xining. The results from this study demonstrate the applicability and scalability of the FEISSO system for urban CO2 flux estimation. In this study, the frequency of urban-scale inversions was limited by the current orbital coverage of the OCO-2 satellite. With future improvements in satellite observation capabilities, particularly in spatial resolution and revisit frequency, FEISSO is expected to play a pivotal role in global urban carbon emission monitoring and in the evaluation of emission reduction policy.
- Preprint
(11254 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 13 Aug 2026)
-
CEC1: 'Comment on egusphere-2026-2619 - No compliance with the policy of the journal', Juan Antonio Añel, 26 Jun 2026
reply
-
AC1: 'Reply on CEC1', Xingyu Yao, 29 Jun 2026
reply
Dear Dr. Añel,
Thank you very much for your valuable comments regarding the compliance with the GMD Code and Data Policy.
We apologize for the oversight in our previous submission. We misunderstood the requirements regarding the Code and Data Availability information and did not provide a complete description of the code repository and data sources in the manuscript.
Following your suggestion, we have now uploaded the FEISSO demonstration dataset, including the relevant code and example files, to Zenodo:
https://doi.org/10.5281/zenodo.21027418 (Yao et al., 2026).
Due to the current submission stage, we are unable to directly modify the manuscript at this moment. However, we will update the Code and Data Availability section in the next version of the manuscript. Lines 689-691 in the original manuscript will be revised as follows:
"Code and Data availability
The FEISSO framework, including the source code and demonstration dataset used in this study, is publicly available on Zenodo at https://doi.org/10.5281/zenodo.21027418 (Yao et al., 2026). The satellite observations used in this study were obtained from the NASA OCO-2 Level 2 Standard XCO2 products (https://oco2.gesdisc.eosdis.nasa.gov/data/OCO2_DATA/OCO2_L2_Standard.11r/). The anthropogenic emission inventory data used to construct the prior emissions were obtained from the EDGAR greenhouse gas emission inventory (https://edgar.jrc.ec.europa.eu/dataset_ghg2024) and the ODIAC emission inventory (https://db.cger.nies.go.jp/dataset/ODIAC/DL_odiac2025.html)."We appreciate your guidance in improving the transparency and reproducibility of our work. We hope that the updated repository and the revised Code and Data Availability statement will now satisfy the requirements of the GMD Code and Data Policy.
Sincerely,
Xingyu YaoCitation: https://doi.org/10.5194/egusphere-2026-2619-AC1
-
AC1: 'Reply on CEC1', Xingyu Yao, 29 Jun 2026
reply
-
RC1: 'Comment on egusphere-2026-2619', Anonymous Referee #1, 06 Jul 2026
reply
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-2619/egusphere-2026-2619-RC1-supplement.pdf
-
RC2: 'Comment on egusphere-2026-2619', Anonymous Referee #2, 14 Jul 2026
reply
The manuscript presents FEISSO-Carbon, a Lagrangian–Bayesian inversion framework for estimating high-resolution urban CO₂ emissions using OCO-2 XCO₂ observations. The development of a kilometer-scale inversion framework, together with the sensitivity analysis of key inversion parameters, is potentially valuable for satellite-based urban carbon monitoring. The manuscript is generally well organized, and the discussion of parameter sensitivity provides useful guidance for future applications. However, I have several major concerns regarding the validation strategy, methodological description, and interpretation of the inversion results. In its current form, I do not believe the manuscript provides sufficient evidence to support several of its major conclusions.
1. The current title emphasizes the "impact of satellite observations", which suggests that the manuscript evaluates whether satellite observations improve inversion performance. However, the main contributions are actually development of the FEISSO framework, sensitivity analysis of inversion parameters, and recommended parameter settings for urban inversions. The title should better reflect these methodological contributions.
2. The posterior emissions are primarily evaluated against EDGAR and local emission inventories. However, EDGAR (combined with ODIAC spatial allocation) is also used to construct the prior emissions. Consequently, agreement between the posterior estimates and EDGAR cannot be regarded as an independent validation. More importantly, the Chengdu posterior emissions are approximately 190% of the EDGAR estimate, which differs substantially from the prior inventory, while the abstract still concludes that the inversion shows "overall consistency". This statement appears overstated, particularly under complex terrain and sparse observational coverage. I strongly recommend including independent validation datasets, such as ground-based CO₂ or XCO₂ observations, urban CO₂ monitoring stations, or at least cross-validation using withheld OCO-2 observations. Without independent evaluation, it remains difficult to assess whether the inversion genuinely improves emission estimates.
3. The study is based on very limited OCO-2 observations. Within the selected 10-day assimilation window. Weifang uses three satellite overpasses. Xining uses three overpasses. Chengdu relies on only one OCO-2 track. A 10-day inversion based on such limited observations cannot necessarily represent monthly, seasonal, or annual urban emissions. The limitations associated with observational coverage should be discussed more thoroughly. Therefore, conclusions regarding the scalability of FEISSO and its applicability for policy evaluation appear insufficiently supported.
4. Background concentrations are derived from GEOS-Chem and corrected using AirCore observations, with three pressure levels (510, 139, and 48 hPa) used to define the background. However, the manuscript does not explain why these pressure levels are universally appropriate for different cities, seasons, and topographic conditions. No sensitivity analysis is performed to quantify how background uncertainty propagates into the posterior emissions. Since background uncertainty can be comparable to the urban enhancement signal in XCO₂ inversions, this represents one of the dominant uncertainty sources and deserves much more detailed discussion. 5. The manuscript identifies a turning point around ObsErr ≈ 1.5 ppm and GlobErr ≈ 8 Mt yr⁻¹, but does not explain the physical mechanisms responsible for this behavior.
6. The opposite responses observed in Weifang and Xining require further explanation. Section 3.2 states that Weifang and Xining exhibit opposite responses to changes in ObsErr and GlobErr, yet the underlying physical reasons are not discussed. A mechanistic explanation, considering emission patterns, atmospheric transport, and inversion constraints, would improve the interpretation.
7. The proposed "seesaw effect" is interesting but insufficiently explained. The authors should further explain why the seesaw effect occurs, under what atmospheric or inversion conditions it appears, and whether it is expected to be generally applicable or only specific to the selected case studies.
8. Important FLEXPART configuration details are missing, including the number of released particles, backward integration time, and the vertical release heights. Since these settings directly affect the source-receptor relationships, they should be explicitly documented to ensure reproducibility.
9. The manuscript concludes that high-resolution meteorological fields are critical for urban CO₂ source attribution. However, FLEXPART is still driven by meteorological fields at 0.25° resolution. Interpolating these fields onto a 0.01° inversion grid does not fundamentally increase the effective meteorological resolution. Therefore, the statement should be moderated, particularly for kilometer-scale urban inversions where sub-kilometer meteorological variability is important.
10. The manuscript states that "A super observation is defined as the median of all observations within one standard deviation in each 1 km × 1 km grid cell." However, the native OCO-2 sounding footprint (~2.25 km × 1.3 km) is larger than the aggregation grid. Please clarify how individual soundings are assigned to the 1 km grid cells.
11. The manuscript concludes that Flux Err has relatively limited influence. However, the tested range of 1–30% appears relatively narrow for high-resolution urban inventories, especially considering the large uncertainties in spatial allocation. Additional experiments using larger prior uncertainties (e.g., 50–100%), would provide stronger support for this conclusion.
12. As currently presented, Table 2 gives the impression that only six experiments were conducted while simultaneously varying all five parameters. However, the manuscript later indicates that each parameter was perturbed independently while all remaining parameters were fixed. I recommend revising the table layout or adding an explicit note to avoid misunderstanding.
13. The manuscript attributes hotspot displacement primarily to ODIAC's nighttime-light-based spatial allocation.In fact, ODIAC does not allocate all emissions using nighttime lights. Large power plants are represented using dedicated global power plant databases (e.g., CARMA), where locations are generally well constrained. Hotspot displacement originates from inaccuracies in the spatial structure of the prior inventory, while nighttime-light allocation is only one contributing factor. In addition, Bayesian inversion mainly adjusts emission magnitudes rather than relocating emission hotspots, allowing prior spatial biases to propagate into posterior estimates.
Citation: https://doi.org/10.5194/egusphere-2026-2619-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 8 | 54 | 1 | 63 | 0 | 0 |
- HTML: 8
- PDF: 54
- XML: 1
- Total: 63
- BibTeX: 0
- EndNote: 0
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear authors,
Unfortunately, after checking your manuscript, it has come to our attention that it does not comply with our "Code and Data Policy".
https://www.geoscientific-model-development.net/policies/code_and_data_policy.html
Your manuscript does not contain a Code Avaliability section. You must provide one with links to repositories that comply with the requirements exposed in the policy and containing all the code used to perform the work presented here. Also, in your Data Availability section you have not provided all the data used in the work presented in your manuscript. You must provide all the datasets used for training or running models and for validation of results. Due to this, your manuscript should have not been accepted for Discussions and review in the journal.
The GMD review and publication process depends on reviewers and community commentators being able to access, during the discussion phase, the code and data on which a manuscript depends, and on ensuring the provenance of replicability of the published papers for years after their publication. Please, therefore, publish your code and data in one of the appropriate repositories and reply to this comment with the relevant information (link and a permanent identifier for it (e.g. DOI)) as soon as possible. We cannot have manuscripts under discussion that do not comply with our policy.
Later, if the Topical Editor decides to continue with the review or publication process of your manuscript and you are requested to upload a new version of it, then The 'Code and Data Availability’ section of your manuscript must also be modified to cite the new repository locations, and corresponding references added to the bibliography.
I must note that if you do not fix this problem, we cannot continue with the peer-review process or accept your manuscript for publication in GMD.
Juan A. Añel
Geosci. Model Dev. Executive Editor