the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 13 Jul 2026
Anthropogenic CO2 Emissions in China Constrained by OCO-2/3 XCO2 Observations
Abstract. Accurately quantifying anthropogenic CO2 emissions is essential for evaluating carbon budget and mitigation strategies. However, traditional "bottom-up" emission inventories suffer from substantial uncertainties and update time lags, urgently requiring top-down constraints from atmospheric observations while accounting for confounding terrestrial biogenic interferences. In this study, we extended RegGCAS, a regional carbon assimilation system based on the WRF-CMAQ atmospheric chemical transport model and the Ensemble Kalman Filter algorithm. By assimilating column-averaged dry-air CO2 mole fractions (XCO2) from OCO-2/3 satellite observations, we inverted anthropogenic CO2 emissions over mainland China during winter 2022–2023. The results revealed that the total national anthropogenic CO2 emissions amounted to 2808.3 ± 157.0 Tg, 16.1 % higher than the MEIC inventory. For key emission regions, emissions increased by 13.1 % in the Beijing-Tianjin-Hebei region, whereas they decreased by 10.4 % in the Yangtze River Delta. The system captured distinct urban-suburban emission adjustment differences in key regions, with reductions in city centers and increases in surrounding areas. It also reflected short-term emission fluctuations related to anthropogenic activity changes, such as the Spring Festival work stoppages. Evaluation demonstrates that assimilation effectively reduces prior emission errors by 68.0 %. Validation shows that posterior simulation RMSE decrease by 5.8 % against the assimilated OCO-2/3 XCO2, and by 15.3 %, 7.7 %, and 25.2 % against independent TCCON, ObsPack, and urban site observations, respectively, confirming the enhanced accuracy of the posterior emission estimates. This study provides a reliable inversion framework for tracking regional carbon dynamics and refining bottom-up emission inventories.
- Preprint
(2025 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3241', Anonymous Referee #1, 07 Sep 2026
-
RC2: 'Comment on egusphere-2026-3241', Anonymous Referee #2, 18 Sep 2026
General Comments:
This manuscript extends the RegGCAS regional carbon assimilation system to assimilate OCO-2 and OCO-3 XCO2 retrievals and estimate anthropogenic CO2 emissions over mainland China during the winter of 2022-2023. The national-scale application, the use of direct satellite XCO2 constraints, and the combination of an OSSE, ground-based evaluation data, a NOx-proxy comparison, and sensitivity experiments are potentially valuable. The reported regional differences between the Beijing-Tianjin-Hebei region (BTH) and the Yangtze River Delta (YRD), as well as the inferred urban-suburban adjustments and short-term changes around the Spring Festival, are also of broad interest.
However, several methodological and interpretive issues currently prevent a robust assessment of the main conclusions. In particular, the performance of WRF-CMAQ for CO2 transport up to 100 Pa was not well evaluated; the state-vector design and its temporal evolution are insufficiently described; correlated and systematic errors in the OCO retrievals are not adequately addressed; the OSSE is close to a perfect-model twin experiment; atmospheric transport errors are not evaluated; and the reported uncertainty of 5.6% is not statistically well justified. In addition, some of the fine-scale spatial and causal interpretations appear stronger than the evidence presented. I therefore recommend major revision. Addressing the points below would substantially improve the study's reliability, reproducibility, and interpretation.
Major Comments:
1. Validation of CO2 transport and stratosphere-troposphere exchange in WRF-CMAQ: The manuscript assumes that WRF-CMAQ can accurately transport CO2 throughout the atmospheric column using only 25 layers between the surface and 100 Pa. However, CMAQ has been developed and evaluated primarily for tropospheric air pollutants, and the applicability of this particular CMAQ v5.0.2 configuration to full-column CO2 transport is not established in the manuscript. Previous CMAQ studies of stratosphere-troposphere exchange have generally focused on ozone and have used dedicated treatments and extensive evaluation against ozonesonde, aircraft, and satellite observations. Such studies do not by themselves validate passive CO2 transport in the present configuration.
This issue is important because errors in upper-tropospheric and stratospheric transport, vertical mixing, or the model-top boundary condition may produce XCO2 biases that the assimilation subsequently compensates for by adjusting surface anthropogenic emissions. The relatively large sensitivity of the posterior emissions to the prescribed background field further demonstrates this potential ambiguity.
Please provide appropriate references demonstrating that the selected WRF-CMAQ configuration has previously been validated for full-column CO2 transport, including stratosphere-troposphere exchange. If no such validation exists, the manuscript should provide a dedicated evaluation. At a minimum, the authors should report the complete vertical-layer structure, the number of layers around and above the tropopause, any layer collapsing performed by MCIP, the upper-boundary treatment for CO2, and the conservation properties of the vertical remapping. A sensitivity experiment with enhanced vertical resolution, particularly in the upper troposphere and lower stratosphere, should quantify the resulting changes in modeled XCO2 and posterior emissions. Comparison of simulated vertical CO2 profiles with aircraft, AirCore, or other available profile observations, or with an independently validated high-vertical-resolution transport model, would substantially strengthen the study.
2. Treatment of the atmosphere above the model top: CMAQ has a model top of 100 Pa, whereas the OCO pressure grid and pressure-weighted column extend to the top of the atmosphere. The manuscript does not explain how the CO2 profile above the CMAQ model top is constructed when applying Eq. (4). Although this region represents only approximately 0.1% of the atmospheric column, an incorrect treatment could introduce an XCO2 bias of up to approximately 0.4 ppm, which is comparable to the residual biases discussed in the manuscript. Please clarify whether the OCO prior profile, the uppermost CMAQ value, or another climatology is used above the model top, and whether the pressure weights are renormalized. A simple sensitivity test using alternative upper-atmospheric treatments should be reported. If the OCO prior is used above 100 Pa, this should be stated explicitly; in that case, the expected effect is likely small because the above-model contribution to the averaging-kernel correction would be zero.
3. The meaning of 2808.3 ± 157.0 Tg CO2 and the 5.6% uncertainty should be reconsidered: The manuscript does not clearly define how the ±157.0 Tg CO2 uncertainty was calculated. It appears to correspond to the 5.6% spread across the sensitivity experiments. However, these experiments are a small set of deterministic alternative configurations rather than independent samples from a defined probability distribution. Their standard deviation therefore cannot automatically be interpreted as a posterior standard uncertainty or confidence interval.
Moreover, changing the background field and biogenic-flux product changes the national posterior estimate by -10.3% and -11.7%, respectively, both larger than 5.6%. Please provide the exact uncertainty calculation, the included error sources, any weighting scheme, and the method used to obtain the regional uncertainty values. It would be preferable to distinguish the posterior ensemble variance from structural uncertainty across model configurations and from systematic uncertainty associated with observations, transport, boundary conditions, and biogenic fluxes. In its current form, the evidence does not support statements such as “high reliability” or “reliable, operational framework.”
Specific and minor comments
1. Inconsistent regional fractions: Section 3.1 states that BTH and YRD account for 10.8% and 16.0% of the national posterior total, respectively. However, the reported values give 294.2 / 2808.3 = 10.5% for BTH and 345.9 / 2808.3 = 12.3% for YRD. In particular, the reported YRD fraction of 16.0% is inconsistent with the stated totals. Please verify the calculation and clarify whether different denominators were used.
2. Figure 4 panel references: The text refers to the TCCON results as Fig. 4h-i (line 447), whereas the caption lists only panels (a)-(g). Please correct the panel labels and all related cross-references.
Citation: https://doi.org/10.5194/egusphere-2026-3241-RC2
Data sets
Anthropogenic CO2 Emissions in China Constrained by OCO-2/3 XCO2 Observations Shenpeng Qiu and Shuzhuang Feng https://doi.org/10.5281/zenodo.20605680
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 191 | 105 | 34 | 330 | 34 | 26 |
- HTML: 191
- PDF: 105
- XML: 34
- Total: 330
- BibTeX: 34
- EndNote: 26
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript extends RegGCAS to assimilate OCO-2/3 XCO2 observations and invert anthropogenic CO2 emissions over mainland China during winter 2022–2023. National posterior emissions are reported to be 16.1% higher than the 2019 MEIC prior, with opposite adjustments in BTH and YRD. Results are evaluated through an OSSE, independent observations (TCCON, ObsPack, urban site), and sensitivity experiments. The study addresses an important topic and is worthy of publication after addressing the comments below.
General comments
Specific comments
Line 36–39 and Line 584–586: The improvement against assimilated OCO-2/3 observations is an internal consistency diagnostic, not an independent validation. The Methods section correctly distinguishes between the two, but the Abstract and Conclusions list the OCO-2/3 fit alongside TCCON, ObsPack, and the urban site as if they were equivalent validations. Please correct this.
Line 104–105: The text states that winter biogenic fluxes in the YRD account for only 0.2% of anthropogenic emissions, yet the BIO_CASA sensitivity experiment (Table 2) shows an 11.7% change in posterior anthropogenic emissions when switching the biogenic flux product. Please explain why biogenic fluxes that are small in absolute magnitude have such a large impact on the inversion.
Line 256–278: Please describe the spatiotemporal collocation methods used to compare model output with TCCON, ObsPack, and the ZJ urban site.
Line 379-381: I do not agree that the two methods are consistent in BTH based on Fig. 2. Reporting regional totals and also time series may help a reader. In addition, how interference of biogenic fluxes affects the inference on anthropogenic emissions. This is a key question that is not adequately discussed.
Line 382–394: Posterior emissions show a marked drop during the Spring Festival (Fig. 1f), while the prior has only monthly-mean temporal resolution. Could part of this signal arise from meteorological variations during the holiday period, rather than from emission changes alone?
Line 609–610: The statement "the inherent structural biases identified within the bottom-up inventories are likely applicable throughout the year" is not sufficiently supported by winter-only results. Please soften this claim.
Fig. 4g: Just a comment: I appreciate that the authors evaluate the posterior simulation against the ZJ site located within the source region. This is valuable.
Technical corrections
Line 36–37: decrease -> decreases
Line 244: "(Taylor et al., 2023) reported" -> "Taylor et al. (2023) reported".
Line 314: Which months are included as winter?
Line 342: "overestimated" -> "overestimation".
Line 364: A little bit of ambiguity on “the proxy-based method”, as you have just mentioned “proxy variables” in the last paragraph. Better to describe the method with clearer language (maybe “anthropogenic CO2 emissions from the NO2-proxy method”).
Line 447: "Fig. 4h-i" is referenced, but the Fig. 4 caption only describes panels (a)–(g). Please check and correct the panel labels.
Line 812: "GenevaWMO-No. 1368". Change to "Geneva, WMO-No. 1368".
References
Baker, D. F., Bösch, H., Doney, S. C., O'Brien, D., and Schimel, D. S.: Carbon source/sink information provided by column CO2 measurements from the Orbiting Carbon Observatory, Atmos. Chem. Phys., 10, 4145–4165, https://doi.org/10.5194/acp-10-4145-2010, 2010.
Friedlingstein, P., et al.: Global Carbon Budget 2025, Earth Syst. Sci. Data, 18, 3211–3288, https://doi.org/10.5194/essd-18-3211-2026, 2026.