Modelling changes in secondary inorganic aerosol formation and nitrogen deposition in Europe from 2005 to 2030

Abstract. Secondary inorganic PM2.5 particles are formed from SOx, NOx and ammonia emissions, through the formation of either ammonium sulphate or ammonium nitrate. EU limits and WHO guidelines for PM2.5 levels are frequently exceeded in Europe, in particular in the winter months. In addition the critical loads for eutrophication are exceeded in most of the European continent. Further reductions in ammonia emissions and other PM precursors beyond the 2030 requirements could alleviate some of the health burden from fine particles, and also reduce the deposition of nitrogen to vulnerable ecosystems. Using the regional scale EMEP/MSC-W model, we have studied the effects of year 2030 ammonia emissions on PM2.5 concentrations and depositions of nitrogen in Europe in the light of present (2017) and past (2005) conditions. Our calculations show that in Europe the formation of PM2.5 from ammonia to a large extent is limited by the ratio between the emissions of ammonia on one hand, and SOx plus NOx, on the other hand. As the ratio of ammonia to SOx and NOx is increasing, the potential to further curb PM2.5 levels through reductions in ammonia emissions is decreasing. Here we show that per gram of ammonia emissions mitigated, the resulting reductions in PM2.5 levels simulated using 2030 emissions are about a factor of 2.6 lower than when 2005 emissions are used. However, this ratio is lower in winter, thus further reductions in the ammonia emissions in winter may have similar potentials as SOx and NOx in curbing PM2.5 levels in this season. Following the expected reductions of ammonia emission, depositions of reduced nitrogen should also decrease in Europe. However, as the reductions in NOx emission are larger than for ammonia, the fraction of total nitrogen (reduced plus oxidised nitrogen) deposited as reduced nitrogen is increasing and may exceed 60 % in most of Europe by 2030. Thus the potential for future reductions in the exceedances of critical loads for eutrophication in Europe will mainly rely on the ability to reduce ammonia emissions.



Introduction
Concentrations of particles with a diameter of less than 2.5 µm (PM 2.5 ) have been decreasing in most of Europe since the turn of the century as a combined result of reductions in anthropogenic emissions of primary particles and gaseous PM 2.5 precursors. Emissions of NH 3 play a central role in the secondary particle formation and are also major contributors to the exceedances of critical loads for eutrophication (Tsyro et al., 2020). In most parts of Eu-rope emissions of in particular SO x (emitted predominantly as SO 2 but also as SO 2− 4 ), and NO x (NO + NO 2 ), have been steadily decreasing in the past decades. At the same time, emissions of NH 3 have changed much less, decreasing in some European countries and increasing in others (see Sect. 3 and Appendix B in EMEP Status Report 1/2020, 2020). Further reductions of SO x , NO x , and NH 3 emissions are required by the year 2030 according to the EU NEC2030 directive (https://www.eea.europa.eu/themes/air/ air-pollution-sources-1/national-emission-ceilings, last ac-cess: 14 December 2021), but the projected percentage reductions in NH 3 emissions in NEC2030 are smaller than for SO x and NO x . In the atmosphere SO 2 is oxidised to  and NO x to HNO 3 . Contrary to SO x and NO x , more than 90 % of the NH 3 emissions are from agriculture, with only minor contributions from industry and traffic (IIASA, 2020). As a result these emissions are in general not co-located with the SO x and NO x emissions. In addition the temporal distribution of the emissions differ, with NH 3 emissions peaking in spring and summer, whereas anthropogenic SO x and NO x emissions in general peak in winter. In the fine mode, ammonium sulfate ((NH 4 ) 2 SO 4 ) particles are first formed from NH 3 and SO 2− 4 . Any excess NH 3 can then form ammonium nitrate (NH 4 NO 3 ) in thermodynamic equilibrium with HNO 3 (see e.g. Simpson et al., 2012). When NH 3 is in excess relative to both SO 2− 4 concentrations and the equilibrium with HNO 3 , the formation of NH + 4 salts will slow down at some point (when there is less acid available to react with NH 3 ), and free NH 3 (NH 3 in excess of SO 2− 4 ) will be present. With NH 3 emissions greatly exceeding SO x and NO x emissions already before 2005, one could question the effects of small or moderate reductions in NH 3 emissions on secondary inorganic aerosols (SIA25), a major component in PM 2.5 .
Using emissions as described in Jiang et al. (2020), Aksoyoglu et al. (2020) showed that the fraction of NH + 4 in SIA25 was similar when calculated with 1990 versus 2030 emissions. With 1990 versus 2030 emissions, the fraction of SO 2− 4 in SIA25 dropped significantly, whereas the NO − 3 fraction increased, compensating for the reduction in SO 2− 4 . In many air pollution episodes in Europe involving PM 2.5 , NH 4 NO 3 has accounted for a large portion of the aerosol mass (Petit et al., 2017;Vieno et al., 2016). With a large surplus of NH 3 relative to HNO 3 it could be that NH 4 NO 3 formation will be virtually unaffected by changes in NH 3 emissions. Both NO x and NH 3 are relatively short-lived, with a lifetime in the atmosphere of about 1 d (Seinfeld and Pandis, 2016). Given the difference in both spatial and temporal distribution in the sources of NH 3 , NO x , and SO x , substantial local variability in the ratio between NH 3 on one hand and SO 2− 4 and/or HNO 3 on the other hand can be expected. Thus, locally the formation of SIA25 may be limited by the availability of either NH 3 or HNO 3 and SO 2− 4 due to the lack of co-location in both space and time of the sources of these species.
Here we apply the EMEP MSC-W model to investigate how PM 2.5 concentrations, and deposition of reduced nitrogen (NH 3 + NH + 4 ), have changed from 2005 to 2017. But the main focus is on model calculations for 2030, assuming that the NEC2030 requirements will be met. Given that NH 3 concentrations in Europe are generally in substantial excess of HNO 3 concentrations, we explore to what extent additional reductions in NH 3 emissions will contribute to further reductions in NH + 4 and subsequently to reductions in PM 2.5 levels, and to what extent the response to further NH 3 emissions is linear. We try to answer this with a sensitivity study for PM 2.5 for post NEC2030, applying additional step-wise NH 3 emission reductions on top of the NEC2030 requirements holding all other emissions constant. At the same time we also investigate to what extent reductions in NH 3 emissions may affect deposition of reduced nitrogen and the exceedance of critical loads for nitrogen deposition.

Model description
The  Simpson et al. (2012), with later model updates described in Simpson et al. (2020) and references therein. In the EMEP model the composition of the metastable aqueous aerosols of the inorganic system containing NO − 3 and NH + 4 and water and the system containing NH 3 and HNO 3 in the gas phase is calculated using the MARS equilibrium model (Binkowski and Shankar, 1995). In Tsyro and Metzgert (2019) the EMEP model results using the MARS model are compared to model calculations with EQUSAM4clim (Metzger et al., 2016(Metzger et al., , 2018, giving very similar results. The EMEP model is available as open-source code (see code availability) and is under continuous development, receiving feedback from a host of users. It is regularly evaluated against measurements (see Gauss et al. (2017Gauss et al. ( , 2019Gauss et al. ( , 2020) for the most recent evaluations. Scatter plots of model versus measurements for the concentrations of several key species, as well as for the wet depositions of reduced and oxidised nitrogen, are shown in Appendix A. The model performance is comparable for both 2005 and 2017, even though the selection of measurement sites differs for the two years. Measurements are also available for subsets of common sites for the two years, in general showing comparable biases between the model and measurements for the concentrations of key species for the years 2005 and 2017. The EMEP model has also participated in model intercomparisons and model evaluations in a number of peer-reviewed publications (Karl et al., 2019;Colette et al., 2011Colette et al., , 2012Jonson et al., 2018). In Vivanco et al. (2018) depositions of sulfur and nitrogen species in Europe have been calculated by 14 regional models and compared to measurements, and in Theobald et al. (2019) the modelcalculated trends in the wet deposition of SO 2− 4 as well as reduced and oxidised nitrogen from six models, including the EMEP model, are compared to measurements from 1990 to 2010. These two studies showed good results for the EMEP model. Out of the 14 models included in the study by Vivanco et al. (2018), the EMEP model was one of very few with low fractional biases compared to measurements for the wet depositions of reduced nitrogen (−0.01), oxidised nitrogen (−0.05), and SO 2− 4 (−0.11). For the trend studies presented in Theobald et al. (2019), the fractional bias for the years 1990 to 2010 was −0.18, −0.02, and 0.22 for the wet deposition of reduced nitrogen, oxidised nitrogen, and SO 2− 4 respectively, but the overall overestimation of SO 2− 4 was mainly caused by an overestimation in the first years of the period. Running the EMEP model in global mode, Ge et al. (2021) showed that the model captures the overall spatial and seasonal variations well for the major inorganic pollutants NH 3 , NO 2 , SO 2 , HNO 3 , NH + 4 , NO − 3 , and SO 2− 4 and wet depositions in East Asia, Southeast Asia, Europe, and North America.

Definition of the critical load for eutrophication
A critical load (CL) is defined as "a quantitative estimate of an exposure to one or more pollutants below which significant harmful effects on specified sensitive elements of the environment do not occur according to present knowledge" (Nilsson and Grennfelt, 1988). CLs are calculated for different receptors (e.g. terrestrial ecosystems, aquatic ecosystems), and a sensitive element can be any part (or the whole) of an ecosystem or ecosystem process. CLs have been derived for several pollutants and different negative effects. Here we restrict ourselves to CL defined to avoid the eutrophying effects of nitrogen deposition (CLeutN). Like sulfur, nitrogen can also have acidifying impacts in ecosystems, but the areas affected by acidification are strongly decreasing in Europe compared to earlier decades, and therefore the focus of this paper is on the eutrophying effects (Slootweg et al., 2015;EEA, 2014;Hettelingh et al., 2017).
The CLeutN for a site is either empirically derived or calculated from steady-state simple mass balance (SMB) equations. In the SMB method, non-harmful nitrogen-fixing processes are described mathematically and combined with a chemical criterion (e.g. an acceptable N concentration in the soil solution). This summation is then compared to the corresponding deposition value. Methods to compute CLs are summarised in the Mapping Manual of the ICP Modelling and Mapping CLRTAP (2017) (see also De Vries et al., 2015), which is used within the Convention on Long-range Transboundary Air Pollution (https://unece.org/ 40-years-clean-air, last access: 14 December 2021).
If the deposition of the pollutant under consideration is greater than the CL at a site, the CL is designated as exceeded. Such site-specific exceedances can be summarised for different spatial entities (e.g. grid cells, countries). This method is called average accumulated exceedance and is defined as the weighted average of exceedances for all ecosystems within the selected region, where the weights are the respective ecosystem areas .
The CL exceedances presented here were calculated using the current CL database, which is described in Hettelingh et al. (2017) and stored by the current Coordination Cen-tre for Effects (CCE) at the German Federal Environmental Agency. The calculation is based on an extensive set of input data and equations. A detailed description is included in the Mapping Manual of the ICP Modelling and Mapping. (CLRTAP, 2017, Sect. 5). This dataset is also used, among other things, to support European assessments and negotiations on emission reductions Reis et al., 2012;EEA, 2014).

Model runs
The EMEP model runs have been performed with 2017 meteorological conditions. In these model runs, emissions esti- . In this study we use ECLIPSE version 6a (hereafter referred to as "ECLIPSEv6a"), which is a global emission dataset widely used by the scientific community. Some of the methods used in ECLIPSEv6a are described in the recent publication of Höglund-Isaksson et al. (2020). NH 3 and NO x emissions from all EU28 countries and selected non-EU European countries are listed in Table 1. For year 2005 non-EU28 EMEP and ECLIPSEv6a, emissions are very similar for NO x , but for NH 3 the EMEP emissions are in general higher.
In order to explore the effects that further emission reductions of NH 3 in 2030 may have on PM 2.5 concentrations and nitrogen depositions, additional model sensitivity runs have been made. The 2030 NH 3 emissions have been reduced by up to 50 % in steps of 10 %. In addition the 2030 emissions of SO x and NO x have been reduced by 10 % and the 2005 NH 3 emissions by 10 %. All model runs are listed in Table 2.  Fig. 2, substantial reductions in PM 2.5 concentrations are expected from 2005 to 2030, caused by reductions in NH 3 emissions, and even larger reductions in SO x and NO x emissions, in Europe. Even so, in 2030 elevated PM 2.5 concentrations still persist in some areas, notably in the Po Valley in Italy and the Benelux countries (Belgium, the Netherlands, and Luxembourg). In these areas, anthropogenic primary PM 2.5 and PM 2.5 precursor emissions are expected to also remain high in 2030. As a result, the limit values for PM 2.5 recommended by WHO (WHO, 2005) are expected to also be exceeded in these locations in 2030. Figure 1b shows the concentrations of SIA25 averaged over the EU28 countries in 2005, 2017, and 2030 split by component (SO 2− 4 , NO − 3 , and NH + 4 ). Even if the reductions in NH 3 emissions in the EU28 are much smaller than the corresponding reductions in SO x and NO x , the calculated percentage contributions to SIA25 from NH + 4 are virtually unchanged between 2005, 2017, and 2030, confirming the findings in Aksoyoglu et al. (2020) for the Payerne measurement site in Switzerland. The lack of change in the fraction of NH + 4 is not surprising, as NH + 4 is associated either with (NH 4 ) 2 SO 4 or with NH 4 NO 3 . As the molecular weight is 18 g mol −1 for NH 4 , 96 g mol −1 for SO 2− 4 , and 62 g mol −1 for NO 3 , the resulting percentage contribution by weight from NH 4 for both NH 4 NO 3 and (NH 4 ) 2 SO 4 is roughly 25 %, consistent with the contributions shown in Fig. 1b. Between 2005 and 2017 the percentage reductions in SO x emissions in the EU28 were more than twice as large as the reductions in NO x , resulting in an increase in the nitrate fraction in SIA25. From 2017 to 2030 the EU28 reductions in NO x are expected to be larger than for SO x , resulting in a slight decrease in the fraction of NO − 3 and a corresponding increase in the SO 2− 4 fraction in SIA25.  Table 1. In most central European high-emitting countries less reduced nitrogen is deposited than is emitted. Several countries facing the sea, with very few upwind sources, exemplified by Ireland and Portugal, receive far less deposition than they emit. At the same time the Nordic countries (Norway, Sweden, and Finland) and the Baltic countries (Estonia, Latvia, and Lithuania), located downwind of central Europe, receive more deposition of reduced nitrogen than they emit. For the European Union as a whole, the fraction of deposited over emitted reduced nitrogen is between 0.7 and 0.8 for all three emission years considered. The remaining 0.2-0.3 is deposited either at sea or in non-EU countries. About 15 % of the NH 3 emitted within the model domain is advected out of the model domain, but much of this is coming from non-EU countries close the eastern model boundaries.

Deposition of reduced nitrogen
As a result of the lower ambitions for reductions in NH 3 emissions compared to NO x emissions, a larger portion of  Switzerland  280  497  257  343  57  186  454  171  373  69  189  446  123  302  71  Iceland  88  26  39  26  40  70  43  43  37  46  17  33  28  25  47  Norway  660  179  486  247  34  496  275  408  307  43  281  192  274  214  44  Albania  76  138  123  86  41  76  199  92  124  57  72  197  78  103  57  Turkey  2042  2355  2382  1626  41  2389  6092  2139  3822  64  2421  3992  2074  2555  55  Bosnia H.  100  130  286  225  44  94  175  198  240  55  103  205  147  215  59  N. Macedonia  113  65  121  69  36  73  84  80  76  49  73  58  70  65  48  Serbia  508  467  502  389  44  450  535  359  432  55  213  250  198  306  61  Montenegro  24  19  55  39  41  43  17  42  35  45  13  16  33  37  53 the total nitrogen deposition is expected to come from NH 3 . This is illustrated in Fig. 4, which shows the fraction of reduced nitrogen in the total nitrogen deposition calculated with 2005, 2017, and 2030 emissions. The figure shows that this fraction increases significantly from 2005 to 2017, with a further increase expected from 2017 to 2030. By 2030 the percentage of the total nitrogen deposition resulting from NH 3 emissions is expected to exceed 60 % in large parts of Europe. The percentage contributions are also listed as an average for EU28 and as averages for individual European countries in Table 1. This underpins the findings from IIASA (2018) that the potential for further reductions of the exceedances of CL for eutrophication mainly depends on our ability to control future NH 3 emissions. As shown in Fig. 5, the calculated CL for eutrophication is exceeded for all three years (2005, 2017, and 2030). Even though the level of exceedance has been substantially reduced from 2005 to 2017 and large reductions in depositions are also expected from 2017 to 2030, the total area in Europe where the CL is exceeded remains high for all three years. The percentage of the area where the CL for eutrophication is exceeded is listed in Table 3 for individual European countries.   respectively. Compared to 2005, the absolute effects of 10 % further emission reductions in 2030 are smaller. Partially, this is because the percentage emission reductions in 2005 give a larger reduction in absolute numbers compared to percentage emission reductions based on the lower 2030 emissions.

Effects of NH 3 emission controls
As an example, 10 % of the emissions from the EU in 2005 (3574 Gg) will give a smaller reduction than 10 % reductions in 2030 (2900 Gg). However, these absolute changes in NH 3 emissions are not large enough to explain a decrease in the magnitude seen in Fig. 2c versus 2d. As seen in Fig. 6, there is more free NH 3 , shown as NH 3 over total reduced nitrogen, in 2030 relative to 2005. A larger portion of free NH 3 could partially explain the 2 %-4 % annual increase in NH 3 observed by satellites between 2008 and 2018 in countries like Belgium, the Netherlands, France, Germany, Poland, Italy, and Spain (Damme et al., 2020). A 10 % reduction in NH 3 emissions will gradually a make smaller impact on the formation of NH + 4 in future years. This is exemplified by the EU28 countries in Table 4. For 2005, we find that as an annual average 10 % reductions of NH 3 emissions were about 4 times more efficient than 10 % reductions in NO x and almost twice as efficient as SO x in reducing PM 2.5 per gigagram emitted. For 2030, we find that as an annual average the efficiency of mitigating PM 2.5 concentrations by reduc-  ing NH 3 emissions by 10 % has been reduced from 0.61 to 0.22 ng N m −3 per Gg NH 3 emitted, a reduction of a factor of about 2.6 from 2005. Over the same time span, the efficiency of a further 10 % reduction in NO x emissions has gone up by about a factor of 1.8 (from 0.15 to 0.27) and by about a factor of 1.6 (0.37 to 0.58) for a 10 % further reduction in SO x emissions.
The dry deposition of NH 3 is faster than that of NH + 4 . As the fraction of NH 3 in total reduced nitrogen increases from 2005 to 2030 (as discussed in Sect. 4.1), reduced nitrogen may be deposited closer to its sources and potentially in-creasingly more in the same country as it is emitted. A trend in deposition versus emissions for the individual countries (deposition divided by emissions in Table 1) is not readily seen based on the model calculations. The geographical extent of the countries in Europe is relatively small, and there is considerable variability in the emission trends for NH 3 between the individual EU28 countries, also affecting the trends in the depositions in neighbouring countries.
As a large portion of the emitted reduced nitrogen is deposited close to its sources, changes in emissions close to the outer EU28 geographical borders should affect this fraction  more for EU28 as a whole than emission changes in central parts. NH 3 emissions in large EU28 countries such as Germany and France have increased between 2005 and 2017, whereas emissions in several countries close to the eastern and southeastern geographical EU28 borders, such as Bulgaria, Romania, and Greece, have decreased. For the EU28 countries as a whole the fraction of deposited over emitted reduced nitrogen is between 0.7 and 0.8 for all three years considered (2005, 2017, and 2030). It would be possible to inves-tigate the hypothesis of a possible decrease in the transport distance of reduced nitrogen by looking at so-called source receptor matrices for the different years (e.g. studying how the contribution from the country to itself have changed over the years). Such experiments are planned as a follow-up of this paper. More than 90 % of the NH 3 emissions are from agriculture (IIASA, 2020), with low emissions in winter and a maximum in spring, as opposed to both NO x and SO x emissions peaking in winter. As a result, there is more SO 2− 4 and HNO 3 relative to NH 3 in winter than in other seasons. Also, the condensation process forming NH 4 NO 3 aerosols is favoured by low temperatures. As a result Fig. 7 shows that for PM 2.5 by far the largest effects of further reductions of NH 3 emissions are modelled for the winter months. Notably, most of PM 2.5 pollution episodes, including exceedances of the EU limits or WHO AQ guidelines for daily mean PM 2.5 concentrations, are most frequent in large parts of Europe during the winter period (see Tsyro et al., 2019). The smallest effects are calculated for the summer months, when both SO x and NO x emissions are at a minimum. Thus in summer, reductions are mainly confined to the southwestern parts of the North Sea, where ship emissions of NO x are large. This seasonal behaviour is also seen in the measurements at Preila in Lithuania, with low NH + 4 concentrations in summer and higher concentrations in the cold season (Davuliene et al., 2021). Furthermore, they found that the relative abundance of NH 4 NO 3 has increased at the expense of (NH 4 ) 2 SO 4 as a result of particularly large reductions in SO x emissions in the last decades.
The seasonal behaviour of PM 2.5 formation from NH 3 is also demonstrated for EU28 in Table 4, showing that the PM 2.5 reductions that can be achieved by reducing NH 3 emissions are largest in winter and are almost constant (and low) for each 10 % increment in emission reduction in summer. With a large surplus of free NH 3 in summer, the impact of further emission reductions is small. In winter the NH 3 surplus relative to HNO 3 and SO 2− 4 is much smaller (or nonexistent), and additional NH 3 emission reductions will have larger impacts on PM 2.5 levels. Figure 8 compares the efficiency of NH 3 emissions reductions on top of the NEC2030 requirements for PM 2.5 concentrations and reduced nitrogen depositions. Starting from the expected emission levels in 2030, the maps compare the effects of the first 10 % reductions (Base-10 %) in NH 3 emissions to the effects of further reductions in NH 3 emissions from 40 %-50 % relative to Base. If linear, the effects of these 10 % increments in emissions should be equal. However, as shown in Fig. 8a, the reductions in PM 2.5 are larger for the 50 %-40 % emission reductions compared to Base-10 % reductions almost everywhere. This is further demonstrated in Table 4, listing the reductions in annual and seasonal PM 2.5 concentrations as an average over the EU28 countries in steps of 10 % relative to the 2030 NEC emissions. Both as an annual average and for each season, the reductions in PM 2.5 in- crease for each 10 % increment. The reductions in PM 2.5 increase from 0.23 ng m −3 per gigagram of NH 3 emitted for the first 10 % additional reductions to 0.35 ng m −3 per gigagram emitted for the 50 %-40 % reductions. The increase in efficiency is a result of a shift in the ratio in NH 3 versus SO x and NO x emissions. In Table 4 we also show that with the much higher SO x and NO x emissions versus NH 3 emissions, the potential of 10 % additional reductions in NH 3 emissions in curbing PM 2.5 levels was substantially higher in 2005, even when compared to 40 %-50 % reductions in 2030.

Sensitivity tests with additional emission controls
In two additional model runs we separately reduce the 2030 emissions of SO x and NO x by 10 %. As an annual average we find that 10 % reductions in both SO x and NO x emissions will lead to larger reductions in PM 2.5 levels in the EU28 than the corresponding 10 % reductions in NH 3 emissions. However, as discussed in Sect. 4.3.1, the seasonal variation is large, and in winter reductions of PM 2.5 per gigagram emitted could still be larger for NH 3 than for NO x .
For depositions of reduced nitrogen the situation is reversed. As shown in Fig. 8b the reductions in depositions achieved with NH 3 reduced between 50 % and 40 % compared to the first 10 % reductions are (marginally) smaller in the vicinity of the source regions. This can be explained by a slightly larger portion of the emitted NH 3 being converted to NH + 4 aerosols and having a slower dry deposition rate than NH 3 . As a result, the higher deposition seen in the source areas is compensated for by a much smaller but more widespread decrease elsewhere. As discussed in Sect. 4.2, only a small portion of the reduced nitrogen is advected out of the central parts of the model domain.

Discussion and conclusions
Focusing on the effects of NH 3 emissions, we have investigated how PM 2.5 concentrations and depositions of reduced nitrogen will change from 2005 to 2030, assuming that the NEC2030 emission targets will be met. In addition, we have made a sensitivity study for PM 2.5 for post NEC2030, assum-ing additional emission reductions on top of the NEC2030 requirements.
Emissions of SO x and NO x have decreased in Europe from the year 2005 to present, and further emissions reductions are expected by the year 2030. However, NH 3 emissions have so far remained high, and projected NEC2030 emission reductions of NH 3 are much smaller than for SO x and NO x . Our model calculations show that these differences in emission trends lead to a smaller fraction of the emitted NH 3 being converted to NH + 4 and an increasingly larger portion of free NH 3 versus NH + 4 in the atmosphere in Europe. Based on 10 % emission reductions of NH 3 , NO x , and SO x , we calculate that the potential for PM 2.5 formation per gigagram of NH 3 emitted is expected to drop by a factor of about 2.6 as an annual average between 2005 and 2030. Over the same time span the potential for forming PM 2.5 from NO x per gigagram emitted has increased by a factor of 1.8 and from SO x by a factor of 1.6 per gigagram emitted.
In winter, with low NH 3 emissions and relatively higher NO x and SO x emissions, the ratio of NH 3 to HNO 3 and SO 2− 4 is higher, and a larger portion of the emitted NH 3 will form particulate NH + 4 . Also, the formation of NH 4 NO 3 in equilibrium with HNO 3 and NH 3 is favoured by low temperatures. As a result we find that in winter the effects of further reductions in NH 3 emissions are larger than in other seasons and comparable to additional reductions in SO x and NO x emissions. This is in agreement with the findings in Backes et al. (2016), pointing out that even though the NH 3 emissions are highest in spring and summer due to the application of manure on the fields, emission reductions in winter have a stronger impact on the formation of secondary aerosols than in any other season. Furthermore they stated that the potential of reducing NH 3 emissions in winter is highest through the reduction of animal farming, as this source accounts for about 80 % of the NH 3 emissions in the autumn and winter months.
Following the emission reductions of NH 3 , deposition of reduced nitrogen is decreasing in Europe. However, the reductions in NO x emissions are much larger than for NH 3 , resulting in a much faster decline in oxidised nitrogen deposition compared to reduced nitrogen. Thus the fraction of reduced over total deposition of nitrogen is increasing and is expected to reach more than 60 % in large parts of Europe by the year 2030. Our calculations show that with the existing emission projections the CL for nitrogen will also be exceeded in large parts of Europe in 2030. There are also indications that reduced nitrogen inputs are more effective in decreasing biodiversity than oxidised nitrogen, as reduced nitrogen is more readily available, stimulating growth of specific plants at the expense others (see van den Berg et al., 2008, andErisman et al., 2007, andreferences therein). Furthermore they also suggest that increased levels of NH + 4 can be toxic to plants (see also Esteban et al., 2016).
Reducing, and preferably removing, these exceedances will require larger reduction in nitrogen emissions than currently projected. Given that reduced nitrogen is responsible for the major fraction of nitrogen depositions, the largest cuts should be made in the NH 3 emissions.
As discussed in Nenes et al. (2020Nenes et al. ( , 2021, gas-aerosol partitioning of total reduced and oxidised nitrogen is affected by aerosol pH level and water content, so that low (high) pH is favourable for NH + 4 (NO − 3 ) formation. The increase in the aerosol fraction in total reduced and oxidised nitrogen would lead to changes in their dry deposition and subsequently their residence times and transport distances. This effect has not been accounted for in the EMEP model. Thus some limited local effects might have been missed in our model simulations. For instance, based on the Nenes et al. (2021) results, there may be additional NO − 3 formation in areas with low acidity, such as coastal or dusty regions. Potentially this may reduce the deposition of total nitrate near these local sources, somewhat enhancing the accumulation of particles. Furthermore, as future emissions of SO x and NO x are expected to decrease, the pH of the particles is likely to increase, potentially favouring NO − 3 formation and thus decreasing dry deposition and increasing the transport distances of oxidised and thereby total nitrogen in some regions. On the other hand, our results show that overall, the fraction of reduced nitrogen in the total nitrogen has been increasing, and this increase is expected to continue until 2030. Assuming that the deposition rates for total nitrogen are mostly driven by those of reduced nitrogen (following Nenes et al., 2021), the local effects of NO − 3 formation bursts would probably not play a major role across the regions in different present and future chemical regimes. Therefore we believe that overall the main conclusions presented in our paper remain valid.
For many countries the latest source-oriented legislation may potentially reduce the emissions of SO x and NO x below their emission reduction requirements, and as a result the EU28 as a whole could be on track to overshoot the reduction requirements for these species by 2030. But for NH 3 further efforts are needed in order to meet the 2030 commitments for many countries in Europe (IIASA, 2018). Cost-effective measures to further reduce NH 3 emissions differ among various parts of Europe. According to IIASA (2020) the damage cost estimate of EUR 17.50 per kilogram of NH 3 emitted is much higher than the average abatement costs. Also, Giannakis et al. (2019) find that much more ambitious commitments for NH 3 emission reductions could be applied by Table 4. First column listing annual and seasonal concentrations of PM 2.5 as an average for the EU28 countries. Columns 2-5 list the EU28 average reductions calculated in steps of 10 % reductions in NH 3 emissions. For PM 2.5 the reductions in ng N m −3 per gigagram of reduction of NH 3 emissions are shown in brackets. The corresponding effects of 10 and 20 % reductions of NH 3 emissions in 2005 are also shown. The effects of 10 % reductions of SO x and NO x emissions in 2030 are also listed. The reductions in PM 2.5 in ng N m −3 per gigagram emitted are given in brackets with NH 3 counted as NH 3 with molecular weight 17, NO x counted as NO 2 with molecular weight 46, and SO x counted as SO 2 with molecular weight 64.

Season
Conc    Table A1. Site positions are listed in Table A2. Measurements were downloaded from http://ebas.nilu.no (last access: 14 December 2021).   Table A1. Site positions are listed in Table A2. Measurements were downloaded from http://ebas.nilu.no (last access: 14 December 2021).  Table A1. Site positions are listed in Table A2. Measurements were downloaded from http://ebas.nilu.no (last access: 14 December 2021).   Author contributions. JEJ made the model calculations and wrote most of the paper. HF assisted in designing the model scenarios and in writing the paper. TS calculated the exceedances of critical loads for eutrophication and contributed in the writing of the paper.

Competing interests.
The contact author has declared that neither they nor their co-authors have any competing interests.
Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Acknowledgements.
Computer time for EMEP model runs was supported by the Research Council of Norway through the NOTUR project EMEP (NN2890K) for CPU and the NorStore project European Monitoring and Evaluation Programme (NS9005K) for storage of data.
Financial support. This work has been partially funded by EMEP under the United Nations Economic Commission for Europe (UN ECE).
Review statement. This paper was edited by Maria Kanakidou and reviewed by two anonymous referees.