the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Jul 2025 | Subsequently updated
NOx emissions changes from 2019 to 2021 in Eastern China as estimated through variational inversions and TROPOMI satellite data
Abstract. China is one of the largest emitters of nitrogen oxides NOx (= NO + NO2) worldwide, and up-to-date estimates are crucial as the country faces rising pressure to curb emissions. We estimate NOx emissions over Eastern China (101.75–132.25° E; 17.75–50.25° N) from 2019 to 2021, focusing on the impacts of COVID-19 and the Chinese Lunar New Year (LNY). Using high-resolution NO2 observations from TROPOMI, onboard the Sentinel-5 Precursor satellite, our estimates are at the regional, national and provincial scales. They are produced using the Community Inversion Framework (CIF), coupled to the CHIMERE regional chemistry transport model at 0.5° resolution.
Our results show a sharp drop in NOx emissions by −40% in February 2020, as compared to 2019, driven mostly by lockdown-related mobility restrictions, and partially due to LNY festivities. Provincial reductions in February 2020 include −38% in Shanghai, −29% in Qinghai, −31% in Jiangsu, −36% in Hubei, −24% in Henan, and −16% in Beijing. Total NOx emissions (anthropogenic + biogenic) over Eastern China fell by 0.2 TgNO2/year in 2020 vs. 2019, but rose again in 2021, exceeding 2019 levels by +4% (16.7 TgNO2 in 2021 vs 16.0 TgNO2 in 2019).
Our estimates of recent past years offer insights to guide future strategies and policies to reduce NOx emissions in China and its provinces. These results highlight the advantages of combining high dimensional variational inversion methods with high-resolution satellite data, to strengthen air quality monitoring and support more effective regulations.
- Preprint
(3342 KB) - Metadata XML
-
Supplement
(4676 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
Version 2 | 24 Oct 2025
RC2: 'Comment on egusphere-2025-3329', Anonymous Referee #2, 24 Dec 2025-
AC2:
'Reply on RC2', Rimal Abeed, 13 Feb 2026
- Answer 1: The number of grid cells for the two cities that were displayed in Figure 5 and discussed in section 3.1 was 8 for Beijing and 2 for Shanghai (Figure R2.1) (please refer to RC2.pdf for the figure). Following this comment, we acknowledge that Shanghai may not have been sufficiently resolved in space (raising uncertainties in the diagnostics for such a city), therefore, we replaced it (in Figure 5) by the province of Henan, covered by 64 grid cells. However, in order to cover such a large area (101.75-132.25°E; 17.75-50.25°N) with a resolution equal to that of the chemistry transport model (CHIMERE, 0.25°), and using a variational inversion system, it is currently difficult to achieve resolutions lower than 0.5°. Moreover, one does not find in the literature equivalent examples lower this scale and resolution. As illustrated by the surface area of Beijing, this resolution is nonetheless sufficient to analyze the budget over major cities and provinces. It is accepted in the literature to show emissions per province at a resolution of 0.5°, for instance Hui Li et al. (2024) show emissions per province in China, using a Mass-Balance method, at a resolution of 0.5°×0.625°.
- Answer 2: Separating the anthropogenic and the biogenic emissions will not change much in our current interpretations. In the supplementary material (Figure S7, shown below), we show the biogenic and anthropogenic emissions separated. As we can see from these figures, anthropogenic emissions are highly dominant, with the biogenic emissions varying between 0.2 and 4% only in this domain, and therefore total emissions are largely representative of the anthropogenic emissions rather than the biogenic. However, the inverse modeling framework that we are using separates the anthropogenic and the biogenic emissions. It also assumes that biogenic emissions are negligible throughout the year compared to the anthropogenic ones (Figure S7). (please refer to RC2.pdf for the figure). Our set-up of the prior estimate of the biogenic emissions, and of the prior relative uncertainty in this estimate, both limit our ability to increase it by a very large amount, while, based on the results we obtain, we assume that the actual biogenic emissions are significant in spring summer and that a large part of these emissions is erroneously aggregated to the anthropogenic emissions by the inverse modeling system during these seasons. This is mainly the point of our discussion in section 3.3 and in Figure 9. We thus have two options: 1) to analyze in section 3 (Results) the retrieved anthropogenic emissions before discussing our assumption, in section 3.3, that it actually includes erroneously a significant part of biogenic emission in spring-summer, or 2) to state from the beginning that the separation between the anthropogenic and biogenic emissions is not robust enough, and that we should restrain ourselves to analyze anthropogenic + biogenic emissions before discussing what can be said about the inter-annual variations of the anthropogenic emissions, based on the temporal and spatial patterns of the resulting fluxes. Hence, we prefer to stick to the second option, which is the one that is presented in the current version of the article. With this said, we did consider your comment about separating the anthropogenic and the biogenic emissions in the analysis, and we added to Figure 9 a second panel showing the anthropogenic emissions (Figure 9a). We also edited the discussion of this Figure to better explain the peak in summer from the anthropogenic emissions. We show Figure 9 (a-b) below and the discussion that follows. (please refer to RC2.pdf for the figure). Moreover, extending the analysis and discussion to more years, and studying the effect of the meteorological factors on the biogenic emissions is out of the scope for this study; that aims to look at the changes of NOx emissions duing COVID-19 and the Lunar New Year in Eastern China, as derived from variational inversions using the CIF inversion model (coupled to CHIMERE chemistry Transport model)
-
AC2:
'Reply on RC2', Rimal Abeed, 13 Feb 2026
-
RC1:
'Comment on egusphere-2025-3329', Anonymous Referee #1, 11 Dec 2025
This study quantifies changes in NOx emissions over eastern China from 2019 to 2021, with particular attention to the reduction associated with the COVID-19 lockdowns and the Chinese Lunar New Year. A noteworthy aspect is that the authors derive total NOx emissions, encompassing both anthropogenic and biogenic contributions. Given the substantial reduction from 2019 to 2020, which is likely driven by government interventions associated with COVID-19 restrictions and the Lunar New Year slowdown, and considering the clear rebound in 2021, the interannual pattern reported in the study appears reasonable. The manuscript provides useful insights and is, overall, clearly presented and well-structured. However, certain critical details require clarification. I support the publication of this manuscript, provided that the following major comments can be addressed.
Major comments:
In this study (line 214), the NOx emission estimates are reported in teragrams of nitrogen dioxide equivalent (Tg NO2) for both the prior and posterior datasets. I am not fully familiar with this convention and would appreciate clarification on the rationale for adopting NO2-equivalent units, particularly in comparison with more conventional representations such as NOx expressed directly in tons. Given that this study relies on TROPOMI observations with high spatial resolution and daily temporal coverage, rather than annual or monthly averaged emission inventories, it is not immediately clear whether expressing emissions in Tg NO2 offers advantages or disadvantages when dealing with such pronounced temporal and spatial heterogeneity.This study combines several emission inventories to represent different sectors and applies multiple downscaling approaches to convert monthly emission estimates into daily values. A key issue arises in Figure 2 (line 217) and Figure S1 to S3, where prior emission values below 0.3 kt NO2 in February 2019 are set to zero when computing the relative difference in order to reduce noise in the figure. It is well understood that extremely small values in prior fields often reflect numerical noise rather than actual emissions. However, the manuscript only discusses this adjustment in the context of figure presentation and does not clearly explain how grids with prior values below 0.3 kt NO2 are treated during the inversion itself. Specifically, it remains unclear whether these low-value regions are assimilated as noisy priors, masked out, or handled with any constraint in the retrieval process.
Figure 9 is discussed only briefly, yet it shows a monthly trend that differs substantially from the posterior estimates. A deeper analysis is recommended to clarify the potential drivers of this discrepancy and to explain the underlying causes of the observed emission pattern.
Although the study claims advantages in high temporal resolution, the daily emission variations are not clearly demonstrated. Most results remain at monthly or seasonal scales, and even in Figure 11, the time series which is about two months for Henan is shown with a 14-day moving average, which overly smooths the data. This presentation does not effectively showcase the value of daily-scale emission information.
As shown in Figures 13 and 14, the year 2020 marks not only the onset of the COVID-19 lockdowns but also the final year of China’s 13th Five-Year Plan. Although many studies have reported substantial emission reductions during this period, attributing these changes solely to the lockdown measures is not appropriate, especially given the large spatial and socio-economic heterogeneity of the cities examined. Broader policy-driven structural adjustments and long-term emission control measures likely contributed as well, and should therefore be considered in the interpretation. To avoid potential over-attribution, the authors are encouraged to either (1) clarify how the analysis distinguishes the lockdown-induced reductions from broader economic influences, or (2) revise the discussion to reflect that the observed changes may result from both the lockdown measures and underlying economic conditions.
Minor comments:
The title of Section 3.3, “Seasonal cycle of NOx total emissions in Eastern China” appears to reflect only the content associated with Figure 9. However, Figures 10 and 11 focus on the NOx emission changes around the Lunar New Year period, which does not fall under the concept of a seasonal cycle. To avoid confusion and more accurately represent the scope of this section, it would be clearer to revise the title of Section 3.3 to better encompass both the seasonal analysis and the LNY-related emission variations.Line 477 contains only a single sentence presented as an independent paragraph. It would be clearer and more consistent with academic writing conventions either to merge this sentence with the preceding or following paragraph.
In Figure 10 and 11, the comparison of different weeks following the Lunar New Year provides useful insight into the holiday-related impact on NOx emissions in China. However, when comparing week-to-week trends across different years, is the underlying daily grid spatial coverage consistent among the three years. If the number of valid emission grids in 2020 differs from (maybe much smaller than) those in 2019 and 2021, the conclusion regarding the pandemic-related disruption would not be fully robust.
In Figure 10, connecting the data points for different provinces with lines adds little interpretive value. The key information lies in whether the relative changes are above or below zero, similar to the Figure 13. It is recommended to remove these lines and consider alternative ways to visually highlight the direction and magnitude of the changes, which would improve the clarity and readability of the figure.
The title of Section 3.4, “Impact of the COVID-19 lockdowns in Eastern China”, suggests that the analysis only focus on the effects of the lockdowns. However, as indicated in lines 525 to 530, part of the observed emission reductions may also stem from concurrent economic downturns or other many factors which are not related with lockdown. It is not sufficiently precise to use the current title and framing of Section 3.4 when other factors, such as economic slowdown, also contribute to the observed emission reductions.
Citation: https://doi.org/10.5194/egusphere-2025-3329-RC1 -
AC1:
'Reply on RC1', Rimal Abeed, 13 Feb 2026
Major comments:
- Answer 1: We do not report the NOx emissions via the mass of actual NO2 molecules emitted, we report the total NOx emissions as the mass of equivalent amount of NO2 if all NOx molecules that have been emitted were emitted as NO2 (as often done, in particular when the mass is given in “Tg” or “tons” without more details); as a result, our unit choice cannot have any impact on the spatial and temporal variations of the emissions. We rechecked a wide range of scientific articles and found that all of them use either NO21–4 or N-equivalent5–8 units. In some inventories, the NOx emissions are expressed as NO such as CAMS-GLOB-ANT9. Converting the total NOx emissions between NO2, NO, and N‑equivalent units is straightforward, so we routinely convert our emissions into all conventional units to enable comparison with other inventories; however, to maintain a consistent presentation in the article, we stick to a single unit. Lastly, the reason we do not show the emissions in tons instead of Tg is to avoid stating large numbers, since NOx emissions in China are among the highest globally (for example, 16 Tg of annual NOx emissions in 2019 corresponds to 16000000 tons).
- Answer 2: Values below 0.3 ktNO2 are only filtered (set to zero) for the computation and display of relative differences; they are not filtered during the inversion itself. In relative terms, such small values behave like noise and produce large apparent changes over low‑emission regions, which are not the focus of the analysis. To obtain a more meaningful graphical representation, all values lower than 0.3 ktNO2 are therefore masked in the relative difference plots. We clarified this in the caption of Figure 2, now it reads as follows (we added the text in Bold and Italic): “[…] Values below 0.3 kt NO2 in the prior estimate for February 2019 are set to zero when calculating the relative difference (d) only for display purposes; these are not filtered in the inversion itself.”
- Answer 3: We believe that the monthly variability of the posterior estimates is different from that of the prior emissions due to an underestimation of biogenic emissions in the MEGAN inventory, and therefore an underestimation of biogenic emissions in our prior estimates. We briefly state this in the discussion of Figure 2 (lines 467 – 476). We tested this hypothesis by using CAMS-GLOB-SOIL instead of MEGAN in several sensitivity tests of the forward prior simulations (not shown in this work). However, performing a full additional analysis and including it in this scientific article would further divert the focus and make an already dense manuscript even longer, which is beyond the scope of the present work. That is why, in this article we choose to keep the explanation brief.
- Answer 4: We show daily TVCDs (prior, posterior and TROPOMI TVCDs) in the supplementary materials, for the period 2019 – 2021, for Eastern China (Figures S4, S5 and S6). We added daily emissions in the supplementary materials for provinces (S13 to S40) and we show below one of them: (please refer to RC1.pdf for the figure).
- Answer 5: We do not attribute the changes in emissions during 2020 solely to the lockdown measures, and we state this in section 3.3 (lines 493 to 495): “Miyazaki et al. (2020a) showed that nearly 80% of the emission reductions, during the period 23 Jan – 29 Feb 2020, are due to the lockdown measures in each province”. Our analysis is therefore in alignment with the second suggestion. In section 3.4, we equally discuss the changes in import and export during March-April-May 2020 (lines 525 to 530): “During March-April-May of 2020, emissions decreased by -20% along the China-Mongolia-Russia Economic Corridor (Figure 12a), one of the main economic belts connecting China through Inner Mongolia, then Mongolia and Russia (World Bank, 2019). This reduction in NOx emissions may be a result of the decrease in the Chinese exports in April and May 2020 by -15% (GACC, 2020), due to the lockdown measures applied by the Chinese government and to the decrease in Russian exports to China , from ~ $60 billion in 2019 to ~$50 billion in 2020 (GZERO, 2022; OEC, 2022).” However, it is also important to note that most of these economic changes are in fact linked to the pandemic and the measures that were applied.
Minor comments:
- Answer 1: We changed the title of section 3.3 to “Seasonal cycle of NOx and the Lunar New Year contribution”
- Answer 2: The sentence is now merged with the paragraph preceding it.
- Answer 3: The number of valid emission grids in 2020 is the same as in 2019 and 2021, for all the provinces, during the first and second weeks of analysis. Here we show an example of two provinces. The title of each subplot includes the week studied (w1, or w2), the year, and the number of pixels (px = available emission grids per province, per week). Providing that the number of grids in 2020 is equivalent to those in 2019 and 2021, we believe that our conclusion regarding the pandemic-related disruption is robust. (please refer to RC1.pdf for the figure).
- Answer 4: we removed the line connecting the data points, and now the figure looks like that: (please refer to RC1.pdf for the figure).
- Answer 5: we changed the title of Section 3.4 and now it reads as follows: “Impact of the COVID-19 lockdowns and economic changes on NOx emissions”.
-
AC1:
'Reply on RC1', Rimal Abeed, 13 Feb 2026
-
Version 1 | 17 Jul 2025
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 763 | 245 | 43 | 1,051 | 204 | 28 | 41 |
- HTML: 763
- PDF: 245
- XML: 43
- Total: 1,051
- Supplement: 204
- BibTeX: 28
- EndNote: 41
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Audrey Fortems-Cheiney
Grégoire Broquet
Robin Plauchu
Isabelle Pison
Antoine Berchet
Elise Potier
Gaëlle Dufour
Adriana Coman
Dilek Savas
Guillaume Siour
Henk Eskes
Beatriz Revilla-Romero
Antony Delavois
Philippe Ciais
- V1 , 17 Jul 2025
This article presents analysis on NOx emission changes from Eastern China during 2019 to 2021, highlighting the drop of emissions due to the COVID lockdown and the Chinese New Year. The article is well organized but unfortunately I couldn’t see much new features or points in this work compared to existing works. Substantial revisions should be undertaken before the reconsidering of this paper for publication on ACP. Here are my specific comments: