the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Jul 2026
First comparison of XCO2 products from DQ-1 ACDL and passive optical satellites
Abstract. Spaceborne active CO2 IPDA lidar provides an active remote-sensing approach for XCO2 observations that differs from passive NIR/SWIR remote sensing. However, comparisons between DQ-1 ACDL and passive optical satellite XCO2 products remain limited. Here we compare DQ-1 ACDL with OCO-2, OCO-3, GOSAT, and GOSAT-2 XCO2 products from June 2022 to December 2024 using a unified framework that combines CAMS-based spatiotemporal coherence assessment, satellite sampling availability, daily 2° aggregation, spatiotemporal collocation, and XCO2 column-definition correction. The results show that DQ-1 is broadly consistent with OCO-2 and OCO-3 over coherent and well-sampled regions, with mean differences mostly below 0.5 ppm and similar spatial distributions, meridional structures, and regional monthly variations. Nighttime DQ-1 XCO2 is also consistent with CAMS-derived XCO2, indicating stable performance relative to an external model reference. Independent validation against TCCON shows that daytime DQ-1 XCO2 has a correlation coefficient of 0.92, a mean bias of 0.21 ppm, and an RMSE of 1.49 ppm, with statistics comparable to GOSAT. Regional results further indicate that DQ-1 does not simply duplicate existing passive satellite observations, but provides complementary XCO2 observational coverage under high aerosol loading, at high latitudes, and in regions where passive observations are limited. Overall, DQ-1 ACDL is comparable to passive satellite XCO2 products under the unified framework, and can serve as an important complement to existing passive satellite XCO2 observations, supporting future active–passive joint XCO2 constraints, regional carbon flux inversions, and emission monitoring.
- Preprint
(8664 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3864', Artem Feofilov, 07 Aug 2026
-
RC2: 'Comment on egusphere-2026-3864', Anonymous Referee #2, 25 Aug 2026
This manuscript showed the validation results of XCO2 from the DQ-1 satellite instrument, compared with other satellite XCO2 and reanalysis data. All analyses look sincerely conducted, results are well organized, and writing quality seems fine. However, there are some questions and comments about suggested results as below. Final decision for this manuscript will be executed after checking authors' response and revised manuscript.
1) Readers often read abstract first, sometimes only to see the main lesson of that paper. Thus the completeness of abstract is very important. In that context, the abstract of this manuscript should be improved more. What is DQ-1 ACDL? How does it relate to the IPDA? Why do we need to evaluate the DQ-1 ACDL data? Overall, the research motivation was not well transferred from this abstract.
2) In Figure 1, the number of available GOSAT-2 data number looks larger than that of GOSAT-1 data, which is OK because the GOSAT-2 is recent one. Then why the data number of OCO-3 is smaller than that of OCO-2? Additionally, the number of available OCO-3 data decreases over the tropical Pacific, but increases over the 40-50 °N in the North America and Asia. How to explain this? I did not see meaningful authors' statements about this issue. This seems important in terms of sampling bias factor related to provided validation results.
3) Overall, GOSAT-2 XCO2 overestimation looks obvious. What is the reason? I know that the main topic of this manuscript is DQ-1. Once the GOSAT-2 overestimation issue was suggested in this manuscript, however, authors need to explain its reason. I only found a single statement about this issue in Line 397-398 as next "This systematic positive bias may be related to the fact that the GOSAT-2 SWIR L2 product used here has not been empirically bias corrected", however, I could not figure out the meaning of this statement.
4) In Figure 5, the bias looks large over the land of Australia and tropical Pacific. What is the reason to explain this? There was authors' statement in Line 425 "This pattern suggests that the residual differences are more regional and sampling dependent than globally systematic". Again, I could not figure out the meaning of this statement. This statement is consistent with results and discussions in next paragraphs?
5) Figure 6 (XCO2 comparison between DQ-1 and OCO-2) looks similar Figure 7 (XCO2 comparison between DQ-1 and OCO-2). In tropical Pacific, the validation data number in OCO-3 (Fig. 7) is smaller than those in OCO-2 (Fig. 6), probably due to the data number difference (Fig. 1). Based on Figure 2, I can notice that the data number difference between OCO-2 and OCO-3 suddenly becomes larger after August 2023 for a year. What happen in this period? This malfunction of OCO-3 can make the data difference in tropical Pacific between OCO-2 and OCO-3? (Over Australia, RMSE data points look similar between OCO-2 and OCO-3). Is there any special issue of OCO-3 for only ocean surface monitoring?
6) In Line 447, there is a statement as next "These patterns suggest that the spatial characteristics of DQ-1–OCO-2/OCO-3 bias and RMSE cannot be explained by surface reflectance alone, and may also be related to ocean glint observation conditions and the distribution of valid collocated samples". How the ocean glint differs? Is there any clue that can be applied to high RMSE in tropical Pacific ocean? Since the bias is especially large over this tropical Pacific (different from other ocean area), authors need to add more detailed statement about this part, at least based on previous study results.
7) Please cite figure number appropriately after important statements (especially in chapter 3.4). I cannot make good connection between written statements and suggested figures.
8) Figure 8 (DQ-1 vs. CAMS-derived DQ-1) is interesting. The RMSE is also high over Australia, Sahara desert, and tropical Pacific, similar to FIgure 6 and 7. Then my question is how about the comparison results between OCO and CAMS-derived XCO2? If they are closer, then DQ-1 research team can achieve some tips about bias factors.
9) Please add the site name in the caption of Figure 10 (there are many abbreviation in the legend, but I do now know what those are).
Citation: https://doi.org/10.5194/egusphere-2026-3864-RC2 -
RC3: 'Comment on egusphere-2026-3864', Anonymous Referee #3, 28 Aug 2026
General comments:
This manuscript presents an intercomparison of DQ-1/ACDL XCO₂ with OCO-2, OCO-3, GOSAT, and GOSAT-2. The topic is relevant, and the comparison of active and passive XCO₂ observations is useful. My main concern is that several choices in sample selection and collocation are not sufficiently justified, which makes it difficult to assess the robustness of the reported agreement. I believe the paper makes a valuable contribution and is suitable for publication after revision.
Special comments:
- The analysis is restricted to the top 10% of grid cells based on the coherence and sampling score. Why was the 90th-percentile threshold chosen? Because this selection favors relatively homogeneous, well-sampled regions, it would be useful to show whether the reported bias and RMSE change when a less restrictive threshold is used.
- The comparison uses 2° × 2° daily averages with a maximum time difference of 4 h. Please clarify the rationale for these choices, especially the 4 h threshold. A simple test using a smaller spatial or temporal window would help show how sensitive the results are to the collocation criteria.
- CAMS is involved in the region screening, column-definition correction, and nighttime DQ-1 comparison. How large is the column-definition correction in practice? Showing the comparison statistics before and after this correction would make its influence on the results clearer.
- GOSAT-2 shows a large positive bias relative to the other products, but the manuscript also notes that this product has not been empirically bias corrected. This makes the direct comparison with bias-corrected OCO products somewhat difficult to interpret. The discussion should clarify this distinction, or include bias-corrected GOSAT-2 results if available.
- All satellite products are further restricted to 380–480 ppm after the product-specific quality screening. What is the reason for applying this additional range filter? Please also report the number or fraction of observations removed for each product.
- The TCCON comparison provides an important independent evaluation, but more information on the collocations is needed. Please report the sample size for each satellite and indicate, if possible, whether a small number of stations dominates the pooled statistics. Some estimate of uncertainty for the reported bias and RMSE would also be useful.
- Please clarify whether the daily grid-cell XCO₂ values are simple arithmetic means or uncertainty-weighted means.
- The reduced OCO-3 coverage during the stowage/reconfiguration period should be separated from limitations caused by clouds, aerosols, or solar illumination.
- The title “First comparison” seems somewhat strong, since previous studies cited in the manuscript have already compared or validated DQ-1 against other XCO₂ datasets. The main novelty appears to be the unified multi-satellite comparison framework.
- Please check the manuscript for typographical errors.
-
CC1: 'Comment on egusphere-2026-3864', Kai Qin, 29 Aug 2026
This study presents a systematic comparison of DQ-1 ACDL XCO2 with passive satellite products from OCO-2, OCO-3, GOSAT, and GOSAT-2. Wang et al. develop a unified comparison framework that combines CAMS-based assessment of background-field spatiotemporal coherence, multi-satellite sampling availability, daily 2° spatial aggregation, spatiotemporal collocation, and correction for differences in XCO2 column definitions among the satellite products. The study further investigates the remaining inter-product differences through surface-reflectance analysis, validation against TCCON observations, nighttime comparisons with an external model reference, and regional sampling analyses.
The datasets used in this study are comprehensive, and the proposed framework provides a useful basis for assessing the comparability and potential joint use of active and passive satellite XCO2 observations. The manuscript is generally well structured and relevant to the scope of Atmospheric Chemistry and Physics. However, several methodological details and some of the principal conclusions require further clarification and quantitative support. I recommend publication after the following comments have been adequately addressed.
Main comments
- Although both DQ-1 ACDL and passive satellite products provide XCO2 retrievals, the two approaches differ in their retrieval formulations and effective vertical weighting. Please add a brief comparison in Sects. 2.1.1–2.1.2, noting that passive XCO2 retrievals depend on prior CO2 profiles and are characterized by column averaging kernels, whereas ACDL derives a weighted column from differential absorption measurements and its weighting functions. This clarification would help readers understand that the original active and passive XCO2 products should not be interpreted as directly equivalent without accounting for differences in their effective column definitions, and would provide the necessary context for the column-definition correction in Sect. 2.2.3.
- The novelty of the manuscript could be stated more clearly relative to previous DQ-1 validation studies and intercomparisons among passive satellite products. Please clarify in the final part of the Introduction, and briefly reinforce in the Conclusions, that the principal contribution of this study is the systematic comparison of DQ-1 with OCO-2, OCO-3, GOSAT, and GOSAT-2 within a common framework that jointly accounts for background-field coherence, multi-sensor sampling availability, daily gridded collocation, observation-time differences, and differences in XCO2 column definitions among the satellite products. The authors should also explain briefly how this framework extends beyond previous validation of individual DQ-1 products and conventional intercomparisons among passive satellite products.
- The study contains several interconnected processing steps, including comparison-region selection, satellite-data aggregation and collocation, column-definition correction, inter-satellite comparisons, TCCON validation, nighttime DQ-1 assessment, surface-reflectance analysis, and regional sampling evaluation. However, the overall methodological structure is difficult to follow from the text alone. I suggest adding a workflow figure showing the input datasets, principal processing steps, and relationships among the different analyses. In particular, the different roles of CAMS in comparison-region selection, column-definition correction, and nighttime DQ-1 assessment should be clearly distinguished.
- The CAMS-based treatment of XCO2 column-definition differences is a central component of the proposed framework, but its magnitude and influence on the final results are not sufficiently demonstrated. The authors should present the spatial distributions of the column-definition differences between DQ-1 and OCO-2, OCO-3, GOSAT, and GOSAT-2 for the collocated samples or selected comparison regions, and briefly discuss where these differences are relatively pronounced. In addition, the correlation coefficient, mean difference, and RMSE should be reported before and after column-definition correction for each satellite pair. This comparison would quantify the actual influence of the correction and show whether and to what extent the correction affects the inter-product statistics and conclusions.
- The manuscript reports that GOSAT-2 is systematically higher than the other satellite products and TCCON and attributes this mainly to the absence of empirical bias correction. However, differences in the bias-correction status of the products may confound differences related to sensors and retrieval algorithms with those caused by product post-processing. The official NIES release notes for the GOSAT-2 SWIR L2 Full-Physics V02.20 product provide land- and ocean-specific empirical bias-correction formulas and coefficients based on TCCON GGG2020 data. The authors should apply the corresponding correction to the pixel-level GOSAT-2 XCO2 retrievals and update the inter-satellite and TCCON comparisons. At minimum, results before and after empirical correction should be presented as a sensitivity analysis. The product version used during the study period and the order in which empirical bias correction and column-definition correction are applied should also be stated explicitly.
Minor comments
- 1. Notation throughout the manuscript: Please check the entire manuscript, including the title, Abstract, main text, equations, tables, figure captions, and axis labels, and consistently replace “CO₂” and “XCO₂” containing the Unicode subscript character “₂” with “CO2” and “XCO2” produced using standard subscript formatting.
- 2. P3, L66: The word “characteristics” is repeated in the phrase “global XCO2 observation characteristics characteristics”. Please remove the duplicated word.
- 3. Table 1: The column heading “Native resolution/scale” combines temporal resolution, spatial resolution, and sampling mode, which are not directly comparable across the listed datasets. Please either separate this information into “Temporal resolution” and “Spatial resolution/sampling” columns or revise the heading and entries so that the dataset characteristics are described consistently.
- 4. (1)–(16): Please format all displayed equations consistently following the ACP/Copernicus manuscript template. The equations should be left-aligned, with the equation numbers aligned on the right and numbered consecutively in order of appearance.
- 5. P7, L180 and L189: “Where” should be changed to lowercase “where”, as the text following the equation continues the preceding sentence. In addition, please remove the extra space before the comma after “d” on L180.
- 6. P8, Eq. (6): The text following the equation states that represents “the overall sampling sufficiency of grid cell ”. The latter expression should be replaced by “grid cell ”.
- 7. The colorbar in Figure 1 needs to be adjusted. The present colorbar cannot clearly distinguish regions with no valid observations from those where the number of valid observations is lower than 20. The color corresponding to zero valid observations could be changed to white or another color clearly distinct from dark blue.
- 8. Figure 2 and the associated text: The term “valid-observation cell-days” does not accurately describe the reported quantity and should be replaced with “number of valid observations”. Please define the counting unit explicitly at its first occurrence and revise the terminology consistently throughout the relevant methods, results, figure captions, and axis labels. The manuscript should also state explicitly that DQ-1 Asc+Desc is calculated as the direct sum of the ascending and descending observation counts.
- P14, L366: “Final score” appears within a sentence and should be changed to lowercase “final score” for consistency with sentence-style capitalization.
- 10. Figure 4 and Table A2: Please clearly specify the sign convention used for the mean bias or mean difference, for example, “the product shown on the y axis minus the product shown on the x axis”. The regression method used to derive the slope and intercept should also be specified in the figure caption or table note.
- 11. Table A1: The caption refers to the “valid observation periods” of the TCCON sites, but the table does not appear to include a corresponding observation-period column. Please add this information or revise the caption accordingly. The column headings could also be standardized as “Latitude (°N)”, “Longitude (°E)”, and “Altitude (m)”.
- 12. The manuscript uses TCCON data, but the corresponding data providers are not acknowledged. Please add an appropriate acknowledgement and include the relevant dataset citations or DOIs for the TCCON sites listed in Table A1.
- Author contributions: “Author contributions.” appears twice at the beginning of this section. Please remove the duplicated text.
- References: The bibliographic information for Han et al. (2017, 2018) appears incomplete. Please provide the journal name, volume, page range or article number, and DOI, following the same format used for the other references.
Citation: https://doi.org/10.5194/egusphere-2026-3864-CC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 167 | 65 | 22 | 254 | 19 | 23 |
- HTML: 167
- PDF: 65
- XML: 22
- Total: 254
- BibTeX: 19
- EndNote: 23
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The reviewed manuscript discusses the XCO2 product retrieved from the DQ-1 ACDL space-borne lidar and compares it both with other space-borne XCO2 measurements and with the reference ground-based stations. The need for accurate CO2 measurements is obvious, its mean value characterizes global changes, whereas its sources and sinks tell us about carbon cycle, emissions, and transport effects. Compared to existing and past space-borne passive measurements (SCIAMACHY, GOSATs, OCOs, TanSat), the space-borne lidar has certain advantages. It uses a controlled source of probing light whereas the passive sounders depend on the incident solar radiance that is not pointed the same way throughout the globe. In particular, this changes the observability of the nighttime part of the globe that is completely inaccessible for the passive XCO2 sounders relying on reflected sunlight (OCO-2, GOSAT, TanSat). The thermal-infrared emission sounders like AIRS, IASI, or CrIS do observe CO2 emissions, but their channels are characterized by altitude-dependent contribution functions related to CO2 concentrations, whereas the differential absorption method probes the whole atmospheric column down to the ground. For this method, the absence of solar photons contaminating the signals makes the nighttime observations preferable to daytime ones. From this point of view, ACDL might serve either as a reference for other sounders or as a strong supplement to existing XCO2 data.
The manuscript introduces the problem, discusses an extensive list of comparison datasets that includes not only spaceborne passive remote sensors like OCOs or GOSATs, but also a network of ground-based Fourier-spectrometers TCCON that retrieve XCO2 from solar near-infrared absorption spectra and serve as a reference for space-borne observations.
Throughout the manuscript, the authors show different cross-comparisons between the space-borne passive, space-borne active, and ground-based observations, and conclude that the overall agreement between the compared instruments is astoundingly good, with bias less than 0.3 ppmv and RMSE less than 1.5ppmv for all instruments except GOSAT-2. The correlation coefficients between the datasets are on the order of 0.91-0.95. The manuscript is topical, it fits well the scope of the journal, is well organized and well-written.
I believe it can be published in the journal with minor revisions, which I propose for the discussion below.
Minor comments
The abstract relies heavily on acronyms (XCO2, IPDA, TCCON, and the various instrument names) without expanding any of them at first use. This is fine for a specialist audience already familiar with the history of spaceborne XCO2 observations, but makes the abstract nearly unreadable for a general reader outside this subfield. Since the abstract is often the only part of the paper read by non-specialists, could the authors, please, either expand all acronyms at first use or rewrite the abstract in plain language?
Lacking elements of the method
Lines 55-72 and/or lines 102-113: it makes sense to add some information about the optical path of ACDL to explain why the vertically resolved lidar signal does not carry the information about the CO2 and only the column value is retrievable. The references given in this section explain this, but adding a couple of sentences regarding the method will simplify the flow.
Neither of these paragraphs discusses the differences of daytime and nighttime observations. Since the online and offline channels share nearly the same additive solar background but different signal levels, an imperfectly subtracted background biases the online/offline ratio toward unity and, therefore, biases XCO2 low – an effect that should vanish at night. Could the authors show a daytime-vs-nighttime comparison over similar scenes to bound this residual bias, if not already addressed in the cited references? Maybe it makes sense to provide a theoretical estimate for exactly the same atmospheric column to exclude any effects caused by diurnal change.
A related point: the differential absorption method requires a valid hard target (surface or opaque cloud top) to close the two-way path. Could the authors please note the fraction of shots lost to a missing hard target and whether this loss is geographically or seasonally structured?
The structure and the order of presentation
The validation with TCCON observation appears only after the diagnosis of residuals in passive spaceborne instrument comparison, whereas I’d say that TCCON observations should have more weight and therefore precede them. See the other comment regarding the order of presentation and TCCON measurements below.
Section 3.1 is dedicated to sampling and takes a lot of space – I am not sure that it belongs to the Results. In principle, it could have been reduced or moved to Appendix without losing the main message of the section.
Throughout the Results section, some of the differences between the datasets are explained whereas the other ones are just reported. Maybe it makes sense to give the priorities or weights to these differences: for example, one chooses TCCON as a reference as in Section 3.7 and discusses the differences only with respect to these observations. Again, this will require a slight reorganization of the order of sections. If the authors believe that TCCON comparisons are of lesser importance, I’d like to understand, why – for me, they are a good candidate for the reference ones because they depend on smaller number of unknowns.
For the figures showing geographical distribution of XCO2 differences, did the authors try to subset the datasets over the oceans using the surface winds values? Does the smoother sea produce better results?
Technical corrections
Fig. 1: the panels share the same color bar and OX/OY axes. In this case it would be logical to merge them and keep only 2xOX axis labels, 3xOY axis labels, and 1 color bar.