the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Feb 2026
Satellite observations reveal heterogeneous atmospheric composition responses to rapid emission changes
Abstract. We developed a unified machine learning framework to retrieve daily, 1 km resolution, gap-free concentrations of six major atmospheric pollutants across China, providing a consistent basis for quantifying atmospheric composition responses to rapid emission perturbations. Our results reveal pronounced spatiotemporal variability across pollutant species, with recovery times ranging from two to eight weeks following abrupt emission reductions. Most air pollutants, such as particulate matter (PM) and NO2, exhibited rapid declines and subsequent rebounds, consistent with changes in anthropogenic emissions, whereas O3 showed the opposite response, reflecting nonlinear photochemical processes under reduced NOx conditions. In contrast, SO2 and CO displayed more sustained decreases, indicating longer-term structural changes in combustion-related sources. By integrating explainable artificial intelligence with atmospheric predictors, we disentangle meteorological and emission-driven contributions to the variability of secondary pollutants across spatial scales. In Wuhan, reduced anthropogenic emissions contributed to a 22 % decrease in PM2.5 during the emission-reduction period, whereas enhanced atmospheric oxidation associated with meteorological variability led to a 40 % increase in O3. During the subsequent recovery phase, meteorological factors dominated interannual variability, driving a 16 % rebound in PM2.5 but a 5 % decline in O3. These findings elucidate the chemical and physical mechanisms governing atmospheric composition under rapid perturbations in emissions and underscore the nonlinear coupling among primary emissions, secondary formation, and meteorological processes.
- Preprint
(30189 KB) - Metadata XML
-
Supplement
(56801 KB) - BibTeX
- EndNote
Status: open (until 01 Apr 2026)
- RC1: 'Comment on egusphere-2026-891', Anonymous Referee #1, 10 Mar 2026 reply
-
RC2: 'Comment on egusphere-2026-891', Anonymous Referee #2, 10 Mar 2026
reply
The authors developed a unified deep learning framework to generate a high–spatial-resolution, full-coverage dataset of six major air pollutants across China for the period 2019–2022, which provides a systematic view of how different air pollutants, and their combined index, AAQI, responded during periods of rapid emission reductions and the recovery phase. I think this work is meaningful, as it offers valuable insights into underlying recovery patterns among pollutants, as well as the key drivers behind the transformations in PM2.5 and O3 levels. These results can provide important scientific evidence and data support for future air quality management and policy design.
However, I have several major concerns, particularly regarding the consistency and reliability of the reconstructed historical predictions, which are critical to the robustness of the study. Additional minor comments are also provided below.
Major comments:
- The input variables used for SHAP originate from multiple datasets (Line265-270: CHAP, CAMS reanalysis, ERA5). Do these datasets have different spatial resolutions? If so, I am concerned that directly applying them without harmonization could introduce substantial bias into the 3.5 section results. Were all inputs resampled to a common resolution before analysis? If yes, please detail the preprocessing steps in the Methods section. Otherwise, it may be necessary to re-run the experiments with consistent spatial resolution to ensure robustness.
- The author repeatedly refers to population-weighted concentrations (e.g., Lines 568, 581), yet neither the main text nor the supplementary materials provide the corresponding formula or weighting methodology. The explicitly define the calculation approach are needed, and ensure that all derived metrics are clearly documented. In addition, specify the population dataset used, including its reference year. Given the known uncertainties in population data, please clarify whether any harmonization or bias correction was applied?
- The discussion emphasizes the case of Wuhan, however, the validation appears to be conducted primarily at the national scale, which remains unclear whether the dataset maintains its reliability at finer spatial scales, particularly for urban areas such as Wuhan. I believe a dedicated validation specifically for Wuhan is necessary to demonstrate the dataset’s capability to support high-precision, city-scale air quality management. Similarly, since the manuscript discusses differences between northern and southern regions, region-specific validation should also be conducted to confirm the dataset’s applicability across these contrasting environments.
Minor comments:
- There are several instances of awkward phrasing and incomplete expressions, like in Line 311, “… can effectively estimate?”. It is recommended that the manuscript undergo professional language editing by a native English speaker or editing service.
- There are inconsistencies between American and British spelling, for example, “modeling” in Section 2.1 versus “modelling” in Table S1. Please standardize spelling throughout the manuscript.
- The subscripts are incorrectly formatted (e.g., Line 571: “NOx”; Line 664: “NO2”). Please ensure proper chemical notation throughout the manuscript. Similar issues are present in Figure S9 (e.g., “5 PM10” in the caption)
- In Table S1 of the Supplement, PM10is reported in μg/m3, whereas the main text uses μg m⁻3. Please standardize all units according to journal guidelines.
Citation: https://doi.org/10.5194/egusphere-2026-891-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 148 | 121 | 14 | 283 | 46 | 11 | 14 |
- HTML: 148
- PDF: 121
- XML: 14
- Total: 283
- Supplement: 46
- BibTeX: 11
- EndNote: 14
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This study develops a machine-learning framework to retrieve gap-free, 1-km2 daily concentrations of six major pollutants across China and uses these datasets to investigate atmospheric responses to rapid emission perturbations. The results highlight strong pollutant-specific responses and nonlinear interactions among emissions, meteorology, and atmospheric chemistry during emission reduction and recovery periods. Overall, this work is well-conducted and suitable for publication after addressing the comments listed below.
Major comments:
Minor comments: