This manuscript presents different ammonia emission scenarios’ impact on PM2.5 mitigation through both a chemical-transport model and a dispersion model. A major revision is recommended before acceptance.

Major comments:

1. Please add a paragraph to summarize ammonia emission sectors in the UK, e.g., the percentage of NH3 emissions from housed dairy, pigs, and poultry. Are there any noticeable industrial and vehicle NH3 emissions near the selected farms?
2. The changes in emissions between different uptake scenarios are relatively small - low2030 is 8.5% and medium2030 is 9.3%. It is redundant to have 3 scenarios with such small emission differences.
3. What is the temporal resolution of the model? Ammonia emissions in winter are much lower than in summer. Wintertime NH3 control is likely to be more effective than summertime. Please comment on the seasonality of ammonia emissions and PM2.5 compositions instead of using the annual mean. 
4. What is the interannual variability between 2019 and 2030? Are there any warming trends during the studied time period? If so, please comment on the impact on ammonia volatilization. Also, please comment on the contribution from transport and deposition changes.
5. Are NOx emissions underestimated? Any sensitivity test with NOx emissions? 
6. Please clarify how meteorological differences between 2019 and 2030 are taken into consideration for the decay distance comparison. Additionally, I suggest including a figure to display the concentration maps generated by the dispersion model for Farm 3.

Minor comments:

Tabel S1: no SO42-?

The area of research encompassed by the title of this paper is of global interest and importance. PM2.5 is the air pollutant with the greatest human health burden, and ammonia emissions contribute to a significant proportion of PM2.5 composition, but are proving hard to abate. It is of interest to scientists and policy-makers to determine the extent to which NH3 emissions – from agriculture in particular – might be feasibly reduced, and the effect of these emissions reductions on PM2.5 concentrations.

Unfortunately, however, despite the important research context, this paper ultimately contains little useful new data or depth of insight. The data also appear to be subject to much uncertainty.

The paper is also not helped by writing that, in my opinion, is not of sufficiently good quality for a journal paper. There is much poor or confusing grammar, and long-winded ways of saying things. Scientifically, I found it hard to follow a lot of the description of the methodology, and I also found it very difficult to extract the points being made from a lot of the text in Section 3.2.

There is also a question of prior publication of a lot of the work. Scientifically, the paper reports independent applications of an atmospheric chemistry transport model (CMAQ) and of a local dispersion model (ADMS), to simulate, respectively, annual mean PM2.5 over the UK and rate of PM2.5 concentration fall-off from 5 individual farms. In both cases, model runs are performed using baseline emissions (for 2030) and scenarios corresponding to low, medium and high NH3 emission mitigation measures being applied to NH3 sources on farms nationally, or to the 5 individual farms specifically. The most novel part of the work overall is the methodology used to derive information about potential extent of uptake of various possible NH3 emission mitigation methods and the conversion of this qualitative scenario data into quantitative changes in actual NH3 emissions from farms. (According to what is written in the present paper, that part of the study involved questionnaires and discussions with farmers and other stakeholders, and use of emissions modelling tools.) However, all the scenario and emissions development is contained in other publications: Jenkin & Wiltshire, and Leonard & Wiltshire. The current paper uses these previously derived emissions and emission factors directly.

As well as the emissions methodology and data being published elsewhere it appears that the CMAQ modelling results have also essentially been published elsewhere. There is another recent paper by Pommier et al. with title “The Impact of Farming Mitigation Measures on Ammonia Concentrations and Nitrogen Deposition in the UK” published in Atmosphere 2025, 16(4), 353; https://doi.org/10.3390/atmos16040353. This other paper already describes (i) the development of the same potential mitigation measures as described in this paper, (ii) the results from the CMAQ modelling of these mitigation measures on particulate NH4+ and PM2.5 concentrations across the UK, and (iii) an explanation for the rather low reductions in the latter two entities simulated by the modelling. Whilst it could be argued that the current paper presents a little more data of the authors’ CMAQ modelling of PM2.5, the descriptions of the model limitations and of the authors’ conclusions on the impact of their NH3 mitigation scenarios on UK PM2.5 concentrations is essentially the same.  

In relation to the CMAQ modelling part of this work, Section 3.1.1 of the present paper presents rather poor model simulation of present day (2019) annual mean PM2.5 at rural PM2.5 monitoring sites. The model values underestimate measurements by a factor of two on average, and the spatial correlation across 48 measurement sites is only 0.58, which implies an explanation of variation of only 36%, i.e. well under half of the measured spatial variation is captured by the model. The authors do not describe further diagnostic model-measurement comparisons. I checked the previous Pommier et al. paper referred to above, and found that its Supplementary Material section does contain some other model-measurement comparisons, for NH4+ and for NH3; but the model-measurement comparisons for both these species essentially have no correlation at all. It is surely a concern for a study whose focus is on the link between NH3 emissions and the effect on NH4+ and PM2.5 that the model simulations are so poor. The author state that their model performance statistics “are not fully satisfactory” but then carry on with presentation of the results from this “not fully satisfactory” modelling. The reader requires far more convincing from the authors that data from their CMAQ modelling yields reliable insight.

The one aspect of the current paper that doesn’t appear to have been published elsewhere is the ADMS dispersion modelling of fall-off of NH3 and PM2.5 concentrations in a few km distance from individual farms. As one would expect, the concentrations fall-off rapidly, down to 10% of max concentration within 1000 meters or shorter. However, these results are caveated by the authors with a long list of sources of uncertainty or difficulty with the input emissions data (that are published elsewhere) so again I was left wondering about the reliability of these results. Regardless of the accuracy or otherwise of these dispersion modelling results I was left wondering what would catch the international reader’s attention from the fact that emissions from a point source dilute quite rapidly. Instead, the main message appears to be that more data and research are needed.  

Some specific comments:

L46: Sentence starting “Resulting of” is not grammatically correct. It is also not clear to me what the point is that this sentence is making.

L59: Sentence starting “Results confirmed” is not grammatically correct.

L63: The statement here that Ge et al. (2022) suggests NH3 reduction only has minor improvement on PM2.5 in the UK contradicts what the sentence starting in L59 says.

L123: There is reference here and in Appendix A to the investigation of 19 mitigation measures to reduce NH3 emissions, yet Table A1 lists 38 different measures that it is stated were used in each of the three scenarios. To confuse the reader further, Table B1 lists 20 mitigation measures. How many, and which, mitigation measures were actually incorporated in the model emission scenarios?

L162: What is the word “This” at the start of this sentence referring to?

L215, and Figure 2: The explanation of why CO emissions decrease in the agricultural NH3 emission mitigation scenarios is not at all clear. Surely the NH3 emission reductions measures have essentially no impact on CO emissions, or if they do have some impact – such as account being taken of how combustion emissions may be reduced through the process of enacting the NH3 mitigation measures – then surely there would be reductions in NOx emissions (and possibly also in primary PM emissions) alongside the reductions in CO emissions? If there is some form of bias in the modelling of CO emissions, as the text seems to imply, then why weren’t the CO emissions for the mitigation scenarios set to be the same as for the base2030 run?

L231: Simply stating that “filling” was used where data capture was poor is not helpful: what quantitative approach was used for data filling? Also, start a new sentence at “Where”.

L234: This sentence mentions “Each farm…” but nothing has been mentioned about individual farms so far in this paper. What farms? There are positions of some farms marked on a map in Fig 2b but this map is never referred to in the text.

L275-L297: it is very difficult to follow what nature of temporality in emission profiles was actually developed and applied in the ADMS model for which particular sources on each farm.

L275-297: It appears that no seasonality in NH3 emissions were applied to most, or all, of the outdoor sources of NH3, yet surely outdoor NH3 emissions are very temperature, i.e. season, dependent.

L290: The sentence starting “Loafing areas” cannot be understood.

L315: Presumably the authors mean that the value 0.58 they quote here is the ratio between PM2.5 and PM10 concentrations, but they write it the other way around.

L340-L344: The fact that increasing NH3 emissions by 50% doesn’t increase PM2.5 doesn’t automatically imply that present-day NH3 emissions are not limiting to PM2.5. To address that question requires simulating with lower NH3 emissions than present-day NH3 emissions.

L357: I’m sure the authors quote a MRE value here that is too small by a factor of 100. As their NMB value is -51%, I presume there MRE value should read ~-50% here, rather than ~-0.5%. It would not be possible to have NMB and MRE values for the same dataset that differ by a factor 100.

L563: There is an error here, the mean relative error does not have unit of concentration. If expressed as a ratio then it is unitless.