General comments:

This study investigates fertilizer-induced soil NO_x emissions and their contributions to atmospheric NO₂, as well as quantifies the impacts on regional air quality during March over North China. Unlike previous studies that focused on summertime, this work examines the early spring fertilizer application season, providing new insights into the significance of soil NO_x on regional particulate matter (PM) and ozone concentrations. These insights are particularly important as fossil fuel combustion-related NO_x emissions decline, making other sources, such as soil emissions, increasingly important.

The authors first analyze two decades of satellite-retrieved atmospheric NO₂ data over North China and identify recurring sub-peaks in March. They link these sub-peaks to fertilizer application activities and validate this hypothesis through air quality model simulations using the BDSNP mechanism for NO emission estimations. The study further assesses the impacts of fertilizer-induced NO_x emissions on PM and ozone levels, highlighting the importance of this often-overlooked source in the context of air quality management.

Overall, this paper provides evidence to support its conclusions and presents a relatively comprehensive analysis of the influence of soil NO_x emissions on air quality. The manuscript is well-organized and clear. However, there are some concerns regarding the uncertainties associated with BDSNP mechanisms in the WRF-Chem model, which may introduce some biases into the analysis. These uncertainties are not sufficiently discussed. Additionally, some details are missing in the method part, and certain discussions are insufficient, along with several technical issues that need to be addressed.

I recommend accepting this paper once these concerns have been addressed.

Specific comments:

1. Line 77: I recommend defining the study area as “North China Plain (NCP)” rather than “North China” for accuracy and consistency with the geographical locations shown in Figure 1 and other similar studies.
2. Please cite recent key studies on soil NO_x emissions and their air quality impacts in the NCP or China in the Introduction, such as

• Lu, Xiao, et al. “The underappreciated role of agricultural soil nitrogen oxide emissions in ozone pollution regulation in North China.” Nature Communications 12.1 (2021): 5021.
• Huang, Ling, et al. “Insights into soil NO emissions and the contribution to surface ozone formation in China.” Atmospheric Chemistry and Physics23 (2023): 14919-14932.

1. Line 130: Please specify the unit for the variables in the formula.
2. Line 159-164: Please add details of the OMI-NO₂ and IASI-NH₃ products, such as the hosting satellites, product versions, orbit types, and local overpass times.
3. Line 164: Please clarify the interpolation method to map the IASI and OMI data to your study region and resolutions. For example, is it area-weighted or error-weighted?
4. Line 169: Please include a map showing the spatial distribution of the 141 observation sites.
5. Please clarify how you calculate the seven-day means. Is it a moving averaging?
6. I recommend adding a figure (possibly in the Supplement) to show the full annual cycle of NO₂ columns, to better illustrate the seasonal variation and highlight the sub-peak in March compared to other months, rather than showing only March in the main text and June and October in the Supplement.
7. Please clarify whether the NO₂ column density refers to the total column density or near-surface levels.
8. Please revise the title of Figure 2 to better reflect its content, which includes NO₂ columns in March and annual emission inventories.
9. Line 207: Please clarify how HTAP calculates the soil NO_x emissions and how this differs from BDSNP mechanism.
10. Figure S3 & S4: Please spell out “VCD”.
11. Line 241: Missing citations?
12. Line 261: I recommend adding a line on Figure 3b to show the differences between the simulations with and without soil NO_x emissions to emphasize their impact on the sub-peaks of atmospheric NO₂.
13. Line 292: Please show the locations of these observation sites. Are they near agricultural areas?
14. Although the authors attempt to evaluate the model performance in predicting soil NO_x emissions, the lack of direct comparisons against flux measurements remains a limitation. Please discuss the uncertainties that may be introduced with the BDSNP scheme.
15. Line 322. The referenced information does not appear in Figure S1. Please correct the citation or clarify.
16. Regarding the O3 diurnal cycle, please clarify whether the BDSNP mechanism in your WRF-Chem simulates diurnal variation in soil NOx emissions or first performs monthly predictions with fixed scaling factors to determine the diurnal changes. If not, discuss how this may affect the interpretation of diurnal O₃ patterns.
17. Line 348: The italicization of via is unnecessary.
18. Line 381: Please verify whether the y-axis label should be ΔO₃ instead of the ΔNO₂.
19. Line 359: The difference between r = 0.997 and r = 0.994 is minimal and likely not significant. Consider tempering this statement.
20. Line 371: Soil is also an important HONO source, a precursor of OH radicals. Does your model include soil HONO emissions? If not, please discuss how this omission might affect your conclusions on OH and atmospheric oxidation capacity.
21. Line 394-405: In this paragraph, the authors compare their findings with other studies to highlight the different impacts of soil NOx on ozone formation, showing suppression in springtime in this study versus enhancement in summertime in other studies. However, the comparison is incomplete. Several recent studies focusing on soil NO_x and ozone formation in North China are not mentioned, while studies from California are cited instead, despite potentially different background conditions and atmospheric environments. I recommend including more regionally relevant studies to support the comparison, considering the nonlinear responses of ozone to its precursors.

• Lu, Xiao, et al. “The underappreciated role of agricultural soil nitrogen oxide emissions in ozone pollution regulation in North China.” Nature Communications 12.1 (2021): 5021.
• Huang, Ling, et al. “Insights into soil NO emissions and the contribution to surface ozone formation in China.” Atmospheric Chemistry and Physics23 (2023): 14919-14932.
• Shen, Y., Xiao, Z., Wang, Y., Xiao, W., Yao, L., & Zhou, C. (2023). Impacts of agricultural soil NOx emissions on O3 over Mainland China. Journal of Geophysical Research: Atmospheres, 128(4), e2022JD037986.
• Tan, W., Wang, H., Su, J., Sun, R., He, C., Lu, X., ... & Fan, S. (2023). Soil emissions of reactive nitrogen accelerate summertime surface ozone increases in the North China Plain. Environmental Science & Technology, 57(34), 12782-12793.

Additionally, the authors attribute the seasonal differences in ozone responses to sunlight intensity driving ozone formation regime shifts. However, this explanation is not robust, as no ozone sensitivity indicators (such as empirical metrics or modeled VOC-/NOx-limited regimes) are provided to support this claim. Please consider expanding this section with additional localized studies and include more concrete evidence to justify your conclusions.

1. Line 416 Please consider adding OH changes to Figure 8 as you describe the changes in OH when mentioning the atmospheric oxidizing capacity.
2. Line 425: Please clarify whether this statement about PM and NO₂ sensitivity refers specifically to NO₂ from soil sources.
3. Line 433 to 438: Please add supporting references, such as FAO reports on fertilizer trends and studies linking global warming with soil NO emissions.
4. Please check the accessibility of the real-time air pollution data website http://beijingair.sinaapp.com in the Code/Data availability part.

Technical corrections:

1. The subscript formatting of NO_x is inconsistent throughout the manuscript, for example, between Line 47 and Line 99. Please ensure the notation of NO_x is consistent across the text, figures, and tables.
2. Citation formatting is inconsistent. In some sections, numbered citations are used, while in others, author–year formats appear. Please standardize the citation style according to the journal's guidelines and ensure consistency throughout the manuscript.

The authors use a chemical transport model, satellite retrievals of NO2 and NH3, and surface station observations to evaluate fertilizer-induced NO emissions and their impacts on atmospheric composition in a region of China.  An analysis of OMI tropospheric NO2 columns reveals a putative peak around March each year, roughly corresponding to the timing of a fertilizer application at a single agricultural research station in the region, and to peaks of total column NH3 from IASI.  They implement the BDSNP soil NOx emission scheme into WRF-Chem, and conduct simulations for spring 2020, finding that the inclusion of soil emissions produces a better match to both satellite and station observations of NO2.  The simulations exhibit an increase in ozone concentrations when soil emissions are included, which the authors attribute to NOx titration during nighttime and to consumption of OH during daytime.  The simulations also exhibit an increase in nitrate and ammonium aerosols, but a decrease in sulfate aerosols, which they attribute to reduced SO4 production resulting from the consumption of OH.

Assessment:

I like what the authors are trying to do here, and I think they have laid the groundwork for a compelling paper highlighting the role of agricultural emissions in determining atmospheric composition in North China during spring.  I do think the manuscript requires substantial revision prior to publication. 

General comments: 

1. When using OMI data across the entire measurement period, the OMI row anomaly needs to be acknowledged—it needs to be mentioned in the text and either the decision not to remove pixels affected by the row anomaly needs to be justified, or the approach used to compensate for the row anomaly needs to be described. If seasonality is the only temporal variability being analyzed, it might be fine to use all the data from the retrieval. But once a multi-year time series is used, it may be more appropriate to exclude any pixels affected by the row anomaly from all years of retrievals. (Also, in particular because of the reduced coverage provided by OMI in 2020 as a result of the row anomaly, TROPOMI would probably provide an improved benchmark for comparisons to WRF-Chem. I don’t know that it’s necessary for publication, though)

2. Additional analysis incorporating land use masks (e.g., using MODIS) would strengthen the attribution of the changes in OMI NO2 (and IASI NH3) columns to agriculture.

3. I fully expect the March sub-peaks to be related to fertilizer, but it seems necessary to exclude the possibility that these could be pulses related to soil thaw. Soil freeze/thaw transitions are sources of huge pulses of N2O in higher latitude agricultural ecosystems. There’s at best mixed evidence as to whether there are also substantial NO pulses, but there’s some suggestion (presented at the most recent American Geophysical Union meeting) that the N2O pulses may result from nitrification rather than denitrification, which could suggest a role for NO.

I think all that needs to be established is that soils in the region thaw well in advance of the initiation of the March pulse.  This could be done with SMAP, which has a global freeze/thaw flag – it is an official product for latitutdes above 50N, but it is available for all pixels.  Reanalysis soil temperature could also be used, but note that soil temperature freezing points may vary and are not 0C.  (But if temperatures are consistently above 0C in February, that seems like it reasonably eliminates the possibility).

4. It appears that there is an NH3 module added to WRF-Chem that is not described.

5. I think it’s very important to frame the WRF-Chem simulations as a case study: there are only 3 months of simulation as the basis for all of the insights into emissions, atmospheric chemistry, and atmospheric composition. The limited duration of the simulation is a big limitation on the generalizability of the model results, and the presentation of results needs to reflect this limitation.

6. Due to an error in formatting references, it is not possible to determine what fertilizer and manure datasets were used (and I believe the reference for these datasets are not included in the reference list). The manuscript does need more information on how fertilizer was handled in WRF-Chem. If the fertilizer and manure datasets are the ones used by Hudman et al. 2012, they are datasets for the year 2000, with fertilizer application timing determined by MODIS EVI for 2001-2004.  So the fertilization application rates and timing may be quite different from practices in 2020, the year simulated in this study.

In addition,  It is not clear how fertilization dates and application amounts in the simulation were determined. If using the fertilizer files from Hudman et al. 2012, they include an assumption that 75% of fertilizer is added at the green-up day as determined by MODIS EVI, with the remaining 25% applied constantly throughout the rest of the season, which was probably an error (the reverse is a common assumption in crop models).  It is also likely quite a different temporal distribution of fertilizer applications from the fertilization practices at the Fengqiu station, and so should be discussed. In addition, we need information on how fertilizer emissions were scaled in WRF-Chem (in the original BDSNP, fertilizer NO emissions were scaled to Stehfest & Bouwman 2006’s estimate of 1.8 Tg N yr-1; DOI 10.1007/s10705-006-9000-7, but there are quite a range of estimates). Understanding and discussing the fertilizer datasets used—especially the fertilization rates--is absolutely necessary to understand the atmospheric composition implications—it will make a big difference if these are 100 kg N ha-1 rates vs. 600 kg N ha-1 rates.

More generally, the manuscript would benefit from more information on agricultural and fertilizer management in the region—there is almost nothing presented.  North China has been notorious for having high fertilization rates (https://doi.org/10.1126%2Fscience.1170261), which may be central to understanding the magnitude of the emissions impact and its effects on chemistry. When comparing to other studies (e.g., Oikawa 2015, Almaraz 2018), it is also important to understand what the cropping systems are. But the manuscript provides no information on fertilization rates or practices in the region or used in the simulations. 

Minor comments:

• The reference list is incomplete, and includes a number of references not cited in the manuscript, and which do not seem related to the manuscript (e.g., multiple references from Li, G. et al on organic aerosols, HONO, and garbage burning)

• All figures need to include information on the time period (I think always Feb-April 2020?) and the domain represented (I think almost always the domain described in Figure 1a + Table 1).

• It could be very interesting to compare the diurnal WRF-Chem analyses to GEMS NO2 and ozone.

• In general, the discussion does little to contextualize the findings within what is already known. This is also true of the introduction: there is very limited background on the rather large literature on NO emissions from agriculture, including in China.

Line comments:

Line 24: use ‘nitrogen dioxide (NO2)’ at first occurrence in abstract

Line 44: need to define NO and NO2 as nitric oxide and nitrogen oxide

Line 49: Huber is not an appropriate reference here. I did not check the other references.

Line 53: To be more specific, it is particularly after the rewetting of dry soils.

Line 55-58: the references here are not appropriate, and do not support the global budget.  The Oikawa, Huber, and Almarez references are all regional studies within the US.  The Tang reference is a technical paper with 10 days of measurements from a single site.

Line 62: again, the Almarez reference is only applicable to southern California and cannot support this statement.  Much better references: DOI: 10.1111/gcb.16864, DOI: 10.1111/gcb.16193, DOI 10.1007/s10705-006-9000-7, doi:10.1029/2001GB001811, etc. etc.

And generally soils emit NO, not NOx. 

Line 65: Note that both references are for California.  For other world regions see doi:10.1016/j.agrformet.2010.10.008,  doi: 10.1111/gcb.13644, and DOI 10.1007/s11104-005-4894-4, http://dx.doi.org/10.1016/j.atmosenv.2014.11.052, among many other studies.

Line 68: as noted in my comment on line 65, there is a substantial body of literature reporting in situ measurements of NO fluxes in agricultural systems going back at least to the early 1990s.

Lines 68-73: Please also see additional studies explicitly focused on the effects of fertilizer-induced soil NO emissions on air quality, including one conducted at a continental scale: doi: 10.1111/gcb.13644 and 10.1016/j.atmosenv.2018.02.040

Line 84: again, because Oikawa is focused on California’s Imperial Valley, it is not an appropriate reference for a discussion of the timing of fertilizer applications in North China.  The introduction does need a paragraph on cropping systems and agricultural practices (including fertilizer use) in the region.

Line 86: I’m not convinced that a pulse is ‘unexpected’, but it is interesting and surprising that it is not in the HTAP inventory. I think it’s worth looking to see if this is true for other inventories.

Line 99: BDSNP does not include a parameterization for NH3 emissions (or, more appropriately, bi-directional NH3 flux).  It exclusively simulates NO emissions.  Also, define NH3 as ammonia at this first occurrence.

Figure 1: I’m afraid I only see lighter orange contours, and then dark orange pixels.  Might be useful to indicate how the agriculture and urban locations were defined.

Table 1: Please include a description in the text of the simulation experiments conducted, in addition to the Table.  3 months of simulation limits generalizability; multiple years would provide more compelling evidence.

Line 121-156: The reference formatting used in these lines is incorrect (making it impossible to know what papers are being referred to).

Line 126: I don’t believe this is accurate: in BDSNP there are dynamic N pools from fertilizer and deposition, but there is not a separate natural N pool.  Natural soil emissions are captured partly by the representation of N deposition in Navail, and partly by A’biome, which is a fixed value.

Line 160-168: There is not enough detail here.  Were L2 or L3 data used? If L2, how was regridding conducted?  Also, it’s fairly standard to include more information on the instruments—overpass time, launch date, orbit, etc.  And, critically, with OMI, the row anomaly needs to be discussed (see general comments).

Line 169: please include the frequency of measurements—are the Picarro measurements conducted at the same frequency as the other species?

Figure 2: please add the IASI NH3 time series to Figure 2a, perhaps in blue.

Line 231/232 and line 237/238: Quantitative data on agricultural practices would be very helpful here.  With respect to the timing of planting, this can vary interannually, typically responding to temperature or precipitation patterns, and in a season, will vary among farms within a region.  It is clear in Figure 2a that the timing of fertilizer applications at Fengqiu station varies considerably from year to year.  And in general, research station practices are not always reflective of regional practice: it would strengthen the argument to have additional data on planting or fertilization dates from the region.

Line 241: the reference is not included (probably Tang 2020).  In any event, it has been well known for decades that fertilization is a significant source of NO (there are, after all, national inventories of fertilizer-induced NOx), and there’s a large literature that can be cited; citing just this one reference is not an effective way to make this argument.   

Line 247: the NH3 module (presumably one that can accommodate bi-directional fluxes) has not been described.

Line 249: I do not understand what this sentence is intending to express: “Soil NOx emission rate calculated by the model gradually increases while adding the soil NOx emission mechanism related to agricultural fertilization."

Line 254-259: the manuscript relies very heavily on Oikawa 2015, Tang 2020, and a few other papers—expanding the use of the literature would improve the manuscript.  Here, Hudman 2010 (https://acp.copernicus.org/articles/10/9943/2010/) may be particularly relevant—it presents work with OMI and GEOS-Chem looking at fertilizer NO emissions over the US, among other topics.

Figure 3: How are the fertilization periods determined in Figure 3a?  With respect to 3b, the match between WRF-Chem and OMI is remarkable, given that BDSNP emission rates are scaled globally. 

Line 281-290 and Figure 4: Be specific about what this comparison is –is this one surface site and one grid cell being compared?

Line 286, 287: spell out IOA and MB

Line 338: Please provide more detail on the difference between NH3 and NO2 in Figure 7.  If these are fertilizer emissions, and they are being produced in WRF-Chem (with, presumably, soil N being the primary driver of emission rate), why and how would the “spatial distribution of NH3 emission rates” explain the difference?

Line 343: Minor comment, but I don’t understand why different units are used for NO2 and NH3 here

Line 353: the decrease in ozone concentrations in agricultural areas is interesting and unexpected—one thinks of ozone being NOx limited in rural areas.  Repeating a concern from a major comment, are the fertilizer rates used in WRF-Chem very high (e.g., 300 kg N ha-1), and may that contribute to the differences with other world regions? Can you include ozone isopleths? Since soils emit NOx in the form of NO, perhaps the equilibrium reactions for NO + NO2 contributes to the reduction in ozone:

NO + O₃ -> NO₂ + O₂                                                                            (1)

NO₂ + hν (+O₂) -> NO + O₃                                                                   (2)

Line 397: two other papers showing increased O3 in response to fertilizer NOx: 10.1111/gcb.13644, 10.1016/j.atmosenv.2018.02.040

Line 432-435: this argument could be strengthened if it quantified the emissions from the urban areas and compared them to the agricultural emissions.

Line 435: the connection to global warming/climate change is not clear. I would remove it. maybe spend time on the unexpected response of ozone instead.

Lines 450-453: I’m not sure the authors are fully representing what we know about agricultural NO emissions. I’d like to see the literature much better represented here, and in the discussion in general.

Line 455: I’d change “long-term fertilization record” to a “2 decade record of fertilization events at a research station.” (I’d really love to see additional information on fertilizer management in the region: research stations are not necessarily representative of local or regional practices).  In addition, the model simulations are decidedly not long-term.

Line 456: change “provide sufficient evidence to illustrate their” to “provide evidence consistent with a”

Line 461 change “leads” to “lead”

Line 461-470: These results are really quite limited: they are for a single 3-month period in 2020.  I think in this conclusion section in particular, it really needs to be emphasized that these quantitative estimates are from a limited case study; we have no idea if the responses are similar in other years with different meteorology.

Line 469: specify that this increase occurs when? March 2020?

Line 469-471: There is an enormous literature on efforts to reduce N losses from fertilizer—this statement doesn’t really reflect that effort, or the fact that historically, fertilizer has been over-applied in North China and could be used much more efficiently.  Additional discussion on how the results may depend on the fertilizer inputs used would be valuable.

Line 473-475: Temperature response has not been a topic in this manuscript, and there’s no sense of how much changes in temperature will affect emissions in this manuscript. I would remove this.