Spatial-temporal variations of atmospheric NH<sub>3</sub> concentration and its dry deposition across China based on one decade of satellite and ground-based observations

Sun, Fan; Cui, Yu; Su, Jiayin; Zhang, Yifan; Shi, Xuejing; Zhang, Junqing; Liu, Huili; Xiao, Qitao; Lu, Xiao; Zeng, Zhao-Cheng; Griffis, Timothy J.; Hu, Cheng

doi:https://doi.org/10.5194/egusphere-2025-3090

Preprints

https://doi.org/10.5194/egusphere-2025-3090

Preprints

10 Jul 2025

| 10 Jul 2025

Spatial-temporal variations of atmospheric NH₃ concentration and its dry deposition across China based on one decade of satellite and ground-based observations

Fan Sun, Yu Cui, Jiayin Su, Yifan Zhang, Xuejing Shi, Junqing Zhang, Huili Liu, Qitao Xiao, Xiao Lu, Zhao-Cheng Zeng, Timothy J. Griffis, and Cheng Hu

Abstract. Ammonia (NH₃), a key alkaline gas in the atmosphere, significantly influences ecosystem nitrogen cycling and the formation of fine particulate matter (PM_2.5). However, limited ground-based monitoring hinders understanding of NH₃’s spatial and temporal dynamics and its dry deposition across China, which is ranked as one of global largest NH₃ emission hotspots. This study integrated 2013–2023 satellite-derived NH₃ column concentrations from the Cross-track Infrared Sounder (CrIS) with ground in-situ observations. We used the GEOS-Chem transport model and a random forest algorithm to simulate NH₃ dry deposition fluxes and explore the driving forces behind observed trends. Our results show that NH₃ concentrations were the highest in the North China Plain (>10 ppb), with notable annual and seasonal increases. NH₃ concentration in 2023 were 14–31 % higher than in 2013. CrIS retrievals aligned well with in-situ data, though were generally about twice as high. Dry deposition fluxes exhibited a clear east-west gradient, with maxima in the North China Plain and Sichuan Basin. Increases in NH₃ concentrations and deposition were most pronounced in urban, cropland, and forest regions, with urban areas experiencing the fastest growth and grasslands the highest total deposition. The national mean NH₃ concentration and dry deposition flux were 4.98 ppb and 0.51 g m^⁻2 yr^⁻1, respectively. Anthropogenic emissions explained 77 % of the variability in NH₃ concentration trend, while meteorological factors accounted for the remainder. 70 %–80 % of deposition trend was governed by atmospheric NH₃ concentration changes. This study highlights growing ammonia pollution and informs nitrogen management strategies in China.

Received: 30 Jun 2025 – Discussion started: 10 Jul 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 2549 KB)

Supplement (3579 KB)

Download & links

Fan Sun, Yu Cui, Jiayin Su, Yifan Zhang, Xuejing Shi, Junqing Zhang, Huili Liu, Qitao Xiao, Xiao Lu, Zhao-Cheng Zeng, Timothy J. Griffis, and Cheng Hu

Status: final response (author comments only)

RC1:
'Comment on egusphere-2025-3090', Anonymous Referee #1, 07 Sep 2025
General comments
This manuscript addresses an important topic, spatial–temporal variability of atmospheric NH3 and its dry deposition across China, which has not been jointly studied before, especially for using CrIS in China, as far as I know. However, the main problem is that the logical connection between derived surface-level/near-surface NH3 concentrations and the derived NH3 dry-deposition fluxes is not sufficiently explained. In its current form, it is difficult to follow the storyline between concentrations and dry depositions and contains several conceptual and presentation problems in figures and tables that must be resolved before publication. The title (“one decade of satellite and ground-based observations”) is misleading because the text does not make clear which data sources (RF-derived GEOS-Chem simulations, satellite, or ground obs) dominate the results and how they are linked. Suggest alternatives, something like, “Decadal changes in atmospheric ammonia and dry deposition across China inferred from space-ground measurements, and model simulations". The manuscript frequently mixes satellite, ground, reanalysis, and inventory products without a clear, reproducible workflow.
Clarify the satellite and ground linkage and what is actually shown in Figs. 3–4
The satellite product is described as a “near-surface column average at ~900 m” while ground sites measure at ~1 m. The rationale for using a regression to “correct” or calibrate the satellite is not justified, and Fig. 4b shows that the R2 does not improve after correction. If regression does not raise R2, explain why the regression is still preferred (e.g., reduces bias, corrects seasonal bias, etc.). If the vertical gradient between ~900 m and 1 m is relatively constant, justify why a simple multiplicative (or additive) conversion factor was not used instead of a regression.

Explicitly state what the satellite product represents (column, layer height, vertical averaging kernel). If you intend to present surface-level NH3, then produce maps and time series of the surface concentration (satellite-derived and corrected by sites) in Sect. 3.1–3.2. If you still keep the near-surface average, explain plainly at the beginning of the results to describe the retrieval layer.

Emission inventories: document, justify choices, and correct low-level mistakes
The manuscript references “six inventories”, but it is unclear why different inventories were used for SO2, NOx, and NH3, and Text S3/Table S2 contains errors (institution names, versions, resolutions).

Add a table listing all inventories used with: name, publisher/institution, version/year, spatial resolution, temporal resolution, main purpose, and how each inventory was used in your study.

Explain why different inventories were selected for different species. If possible, use a consistent set of inventories for cross-species comparison, or present a justification for why species-specific choices are necessary.

State whether biomass burning emissions were included in the simulations and, if so, which dataset was used. If biomass burning is excluded, provide justification.

Random Forest (RF) applications and predictor consistency
The RF is used for two distinct purposes: (A) to extend/estimate dry deposition velocity (Vd) across 2013–2023 from 2015 simulations, and (B) to identify key drivers of atmospheric NH3. The methods (Sect. 2.4.2) only describe the prior usage incompletely, and there are inconsistent predictor sources (ERA5 used for RF, MERRA-2 used for GEOS-Chem). Fig. 10 and its description are confusing (panel a vs b; emissions vs deposition drivers).

Revise Fig. 10 and its caption. Make it explicit what each panel displays. If panels show different metrics (contribution to concentration vs contribution to deposition), label and discuss them separately.

If SO2 and NOx are included as predictors, present their individual contributions (don’t lump them into “anthropogenic emissions” only).

For RF validation: show diagnostics (train/test split) and present performance metrics separately for validation. For spatial maps (e.g., Fig. 5b for RF-predicted Vd in 2015) indicate whether values shown include both training and validation pixels; better: show a validation map or a scatter of observed vs predicted Vd for the validation set.

Trend analysis and how representative the 24 sites + satellite decade are
The trend analysis relies on 24 ground sites and 11 years of satellite data. Few ground sites have >10 years of continuous records (as mentioned in the Introduction); this could bias trend estimates.

Provide a clear description of the temporal coverage at each of the 24 sites, or justify the site selection procedure

Where inventories disagree with inferred trends (e.g., fertilizer usage trends vs EDGAR/MEIC vs policy implementation), explicitly discuss the discrepancy. Possible reasons: (1) differences between bottom-up inventories and top-down estimates, (2) regional heterogeneity in fertilizer use, (3) post-2015 changes not captured in inventory updates, (4) changes in emission factors or agricultural practices.

The authors should cross-check with additional inventories or top-down emission estimates. Reword the manuscript to avoid implying firm causal attribution unless supported by consistent evidence (inventory trends, policy timing, and observational trends).

Quantitative comparison of changes in NH3, SO2 and NOx and their role in deposition
The manuscript claims the increase in dry deposition flux is driven by NH3 concentration increases arising from declining SO2/NOx. However, Fig. S13 indicates NH3 emissions also decline ~20% over the decade, which contradicts the claim. There is no quantitative comparison of the rates of change of NH3 vs SO2/NOx emissions or concentrations.

Provide table or plots that show trends for emissions and concentrations of NH3, SO2, and NOx over the study period. Quantitatively compare declining speeds for NH3, SO2, and NOx emissions/concentrations. If NH3 emissions themselves decreased, explain how a concurrent increase in observed NH3 concentrations could arise. Show analyses that reconcile emissions and observed concentrations.

Specific comments
Line 67, Paulot et al. (2014) provides emissions for 2005–2008 but does not compare emissions from India. Please update these numbers using more recent emission estimates:
Cropland emissions → Zhan 2020 (already cited in line 443), Xu, P., Li, G., Zheng, Y. et al. Fertilizer management for global ammonia emission reduction. Nature 626, 792–798 (2024). https://doi.org/10.1038/s41586-024-07020-z

top-down estimates -> Luo, Z., Zhang, Y., Chen, W., Van Damme, M., Coheur, P.-F., and Clarisse, L.: Estimating global ammonia (NH3) emissions based on IASI observations from 2008 to 2018, Atmos. Chem. Phys., 22, 10375–10388, https://doi.org/10.5194/acp-22-10375-2022

bottom up -> from global inventories such as EDGAR, CEDS

Line 76-78; 106-107; 581-583, Add appropriate references.

Line 119-121; 137-139, unclear —> please rewrite for clarity.

Line 143-149, it is mentioned that "long-term studies remain scarce, and the drivers of spatiotemporal variation in NH3 concentrations and dry deposition ..." but this does not clearly connect with the previous sentence on high NH3 concentrations in China. Also, a decade-long study may not fully address the gap in long-term observations.

Satellite-based atmospheric NH3 concentration section: you mentioned two overpass times and two satellite missions. Were both times and missions used for the entire study period? Were all NH3 data over 73°–136°E and 3°–54°N included, or only those over China?

Line 197/572, Is “land use types” the same as “land cover types” mentioned in lines 208/575?

GEOS-Chem simulation: You simulate Vd at 0.5° × 0.625° over China. Why not: (1) directly use F from simulations, and (2) analyze at this resolution instead of downscaling to 0.25°?

Line 257, Clarify what “two approaches” refers to.

Line 305, "soil moisture" is not a meteorological variable -> land-surface/hydrological variable.

Eq.3 and Eq.4, derived from Eq.2, why is it dlnC/dlnF instead of dlnF/dlnC?

Line 354-357, Higher accuracy is typically associated with higher thermal contrast; conversely, lower thermal contrast would lead to higher uncertainties in NH3 retrievals.

Line 369-372, it is mentioned "lowering NH3 emissions from pastoral sources". Please specify the exact sources.

Line 465, Explain how the “7 % per year growth rate” was calculated and provide the national average growth rate.

Line 557-558, Does this mean Vd is decreasing in these regions? Clarify.

Line 645, it is mentioned "the continuous expansion of urban areas from 2013 to 2023", Consider adding a supplementary figure showing this expansion.

Line 700, I don't see any introduction about the "logarithmic differential method" in the main text or SI

Line 718-727, Acid rain does not appear directly related to your NH3 concentration results; consider removing or tightening this section.

Line 742-750, Move to Methods section or remove if not part of the main results.

Line 782-783, if you also used ground obs from this NNDMN, why are there differences?

Line 784-785, urban and rural regions?

Line 835, specify what "atmospheric dynamics" based on your conclusions

Table 1, Suggest removing from the main text -> information is already in Fig. 1.

Table 2/3 and line 424/454, Add relative annual growth rates (percentage) to check if the increase in summer/Huang–Huai–Hai Plain is still the largest.

Table 4, Consider combining the left and right parts into a single table—current format is confusing.

Table 5, The comparison with global results and by land cover may not be necessary; consider simplifying.

Fig. 1b, Provide explanation for percentage values

Fig. 2a-j, specify which subplots show trends and which show concentrations.

Fig. S8 and S9, ubplot order is inconsistent -> please standardize.

Fig. 7d, Clarify whether this shows interannual variability; define the term in the text and specify it in the subplot y-axis label.
Citation: https://doi.org/10.5194/egusphere-2025-3090-RC1
- AC2: 'Reply on RC1', Cheng Hu, 16 Oct 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-3090/egusphere-2025-3090-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-3090-AC2
RC2:
'Comment on egusphere-2025-3090', Anonymous Referee #2, 09 Sep 2025

The manuscript attempts to investigate spatial-temporal variations of atmospheric NH3 concentration and its dry deposition across China in 2013-20123 by combining satellite-based, ground-level observational data in publica domain and 3-D modeling results. The analysis sounds scientific, but it needs a substantial revision on potential uncertainty and modeling accuracy. The major comments are listed as below:
1) Key scientific questions are too general to be valuable by considering the uncertainties associated and previous studies published in the literature. The authors are encouraged to deeply think the issues.

2) Figure 4, the size of data is too small by considering one decade observations, what happens?

3) Modleing results always suffer from the errors. However, most of air quality modeling results in China well reproduce PM2.5 in approximately 1/3 days in each year. However, it is not case in other times because of the poor prediction of one or several meteorological conditions. The authors should select the 1/3 days with good prediction performance for machine learning.

4) Line 399-400, “Elevated temperatures further enhance volatilization from manure and urban waste, intensifying atmospheric NH3 levels.”. The reviewer has much concern on the statement, i.e., the authors might not know what exactly happen for agriculture emissions of NH3 in China? With a large population moving from the country land to the city in the last decade, the sources are negligible.

Minor comments:
1)The effective number through the manuscript are total off and needs to be corrected. Principally, it should be consistent with the analytic error, i.e., 5-10% analytic errors correspond two effective numbers.

2)Abstract, lines 37, “our results”, what does it means? Modeling results? Observations from CrIS ,AMoN-China or NNDMN?

3)Abstract, line 40-42, “Dry deposition fluxes exhibited a clear east-west gradient, with maxima in the North China Plain and Sichuan Basin. “ The sentence is problematic. Sichuan Basin should be located in southwestern China, correct?

Citation: https://doi.org/10.5194/egusphere-2025-3090-RC2
- AC1: 'Reply on RC2', Cheng Hu, 16 Oct 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-3090/egusphere-2025-3090-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-3090-AC1

Fan Sun, Yu Cui, Jiayin Su, Yifan Zhang, Xuejing Shi, Junqing Zhang, Huili Liu, Qitao Xiao, Xiao Lu, Zhao-Cheng Zeng, Timothy J. Griffis, and Cheng Hu

Supplement

https://doi.org/10.5194/egusphere-2025-3090-supplement

Fan Sun, Yu Cui, Jiayin Su, Yifan Zhang, Xuejing Shi, Junqing Zhang, Huili Liu, Qitao Xiao, Xiao Lu, Zhao-Cheng Zeng, Timothy J. Griffis, and Cheng Hu

Viewed

Total article views: 823 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
718	88	17	823	66	7	13

HTML: 718
PDF: 88
XML: 17
Total: 823
Supplement: 66
BibTeX: 7
EndNote: 13

Views and downloads (calculated since 10 Jul 2025)

Month	HTML	PDF	XML	Total
Jul 2025	141	34	7	182
Aug 2025	127	13	2	142
Sep 2025	406	22	3	431
Oct 2025	44	19	5	68

Cumulative views and downloads (calculated since 10 Jul 2025)

Month	HTML	PDF	XML	Total
Jul 2025	141	34	7	182
Aug 2025	127	13	2	142
Sep 2025	406	22	3	431
Oct 2025	44	19	5	68

Viewed (geographical distribution)

Total article views: 794 (including HTML, PDF, and XML) Thereof 794 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 21 Oct 2025

Short summary

This study used satellite data and models to track ammonia concentration and dry deposition across China from 2013 to 2023. Ammonia levels rose sharply, especially in urban and farming regions, with the North China Plain showing the highest values. Human activity was the main driver of change. These findings highlight growing environmental risks and provide key insights for managing air quality and nitrogen pollution in one of the world’s major emission hotspots.


Total:	0
HTML:	0
PDF:	0
XML:	0

Spatial-temporal variations of atmospheric NH3 concentration and its dry deposition across China based on one decade of satellite and ground-based observations

Supplement

Viewed

Viewed (geographical distribution)

Spatial-temporal variations of atmospheric NH₃ concentration and its dry deposition across China based on one decade of satellite and ground-based observations