Atmospheric aerosol particles are a major constituent of ambient air and have a large impact on atmospheric chemistry, clouds, radiative transfer and climate and also induce adverse human health effects that contribute to mortality . Particulate matter (PM) with an aerodynamic diameter smaller than 2.5 µm (PM2.5) contributes to air pollution through intricate interactions between emissions of primary particles and gaseous precursors, photochemical transformation pathways and meteorological processes that control transport and deposition.

As shown by and , agricultural emissions play a leading role in the formation of PM2.5 in various regions of the world, for example in central and eastern Europe. Agricultural emissions are mostly related to animal husbandry and fertilizer use and to a lesser extent also to the burning of crop residues : around 10 % of worldwide biomass burning emissions can be ascribed to agricultural activities . The general importance of agricultural emissions for air quality was also previously identified by a number of studies e.g., and recognized through environmental policies, (e.g., the establishment of ceilings for national emissions for ammonia by the European Union Clean Air Program). The dominant trace gas emitted by agricultural activities is ammonia (NH3). Around 80–90 % of the atmospheric NH3 emissions in industrialized regions are from the agricultural sector . NH3 is formed and released during the decomposition of manure and organic matter, mostly from animal farming and the associated manure storage and field application, with an additional contribution from (synthetic) nitrogen fertilizer use. NH3 is a toxic gas at very high concentrations, with a pungent smell that irritates the eyes and respiratory system. NH3 is also a major alkaline gas in the atmosphere and plays an important role in neutralizing acids in the aerosol and cloud liquid phase, forming ammonium sulfate and ammonium nitrate (ammonium salts) . Therefore NH3 contributes to secondary aerosol formation and the overall particulate matter burden, and decreases the acidity of the aerosols, which in turn increases the solubility of weak acids (e.g., HCOOH, SO2). The aerosol pH plays an important role in the reactive uptake and release of gases, which can affect ozone chemistry, particle properties such as hygroscopic growth and scattering efficiency of sunlight and deposition processes .

showed that a 50 % reduction of NH3 emissions would lead to a 4 and 9 % decrease in PM2.5 over the eastern USA in July and January, respectively. The reduction of NH3 emissions was found to be the most effective PM2.5 control measure for the winter period over the eastern USA compared to similar reductions of SO2, NOx and VOC emissions . and found that over Europe the reduction of NH3 emissions is the most effective control strategy used to mitigate PM2.5 in both summer and winter, mainly due to a significant decrease of ammonium nitrate. Further, , showed that reducing the NH3 emissions from agriculture by 50 % could result in a decrease of PM2.5 concentrations up to 2.4 µgm-3 over the Po Valley region (Italy). This confirms the finding of , who showed that for short-lived species like NOx and NH3, short-term fluctuations of the emissions play an important role in the formation of nitrate aerosol. According to , NH3 emissions contribute 8–11 % to PM2.5 concentrations in eastern China, which is comparable to the contributions of SO2 (9–11 %) and NOx (5–11 %) emissions. However, the air quality benefits of controlling NH3 emissions could be offset by the potential enhancement of aerosol acidity. showed that, despite the large investments in sulfur dioxide emission reductions, the acid/base gas particle system in the southeastern USA is buffered by the partitioning of semivolatile NH3, making the pH insensitive to SO2 controls. Several studies have been performed on the impact of NH3 on aerosol nitrate and sulfate , mostly with a regional rather than a global view.

As PM2.5 has been clearly associated with many health impacts, including acute lower respiratory infections (ALRI), cerebrovascular disease (CEV), ischaemic heart disease (IHD), chronic obstructive pulmonary disease (COPD) and lung cancer (LC) . Due to its strong contribution to the PM2.5 mass, control strategies in NH3 emissions could possibly reduce the mortality attributable to air pollution, and air quality policy in Europe does indeed include ceilings for NH3 emissions . Studies on PM2.5 reduction due to NH3 control have been performed regionally both for Europe and the USA , while a detailed analysis on the global scale was performed by , who showed the importance of ammonia as a contributor to mortality attributable to air pollution. Nevertheless, assumed an ammonia reduction of 10 %, and the health effects were linearized around the present-day concentrations. As the exposure-response functions, linking PM2.5 to mortality attributable to air pollution, are highly nonlinear at relatively low concentrations, the mortality reduction estimation could change drastically for strong reductions of ammonia emissions. Therefore, in this work, more aggressive reductions are studied (see Sect. ).

Furthermore, there is a need to not only investigate the impact of NH3 emission reductions on PM2.5 concentrations, but also account for particle acidity and aerosol composition. The goal of this work is to understand the impact of global agricultural emissions on model-simulated PM2.5 concentrations, the effects on aerosol pH and the potential consequences for human health, with a focus on four continental regions where air quality limits and guidelines for PM2.5 are often exceeded, i.e., North America, Europe, South and East Asia. North America is defined as the region that encompasses the USA and Canada; Europe is represented by the European continent (including Turkey) excluding Russia; South Asia includes India; Sri Lanka, Pakistan, Bangladesh, Nepal and Buthan; while the East Asia region includes China, North and South Korea and Japan (see Fig. ).

This work may also support policy actions aimed at controlling ammonia emissions, e.g., formulated in the European Union Clean Air Program (http://www.consilium.europa.eu/en/policies/clean-air/), which sets ceilings for national emissions for sulfur dioxide, nitrogen oxides, volatile organic compounds, fine-particulate matter and ammonia.

In this study the EMAC (ECHAM5/MESSy Atmospheric Chemistry) model version 1.9 was used. EMAC is a combination of the general circulation model ECHAM5 version 5.3.01 and the Modular Earth Submodel System MESSy, version 1.9. Extensive evaluation of the model can be found in , , and . ECHAM5 has been used at the T106L31 resolution, corresponding to a horizontal resolution of ∼1.1×1.1∘ of the quadratic Gaussian grid and with 31 vertical levels up to 10 hPa in the lower stratosphere. The model setup is the same as that presented by and is briefly summarized here. The anthropogenic emissions are for the year 2010 from the EDGAR-CIRCE Emission Database for Global Atmospheric Research database, distributed vertically to account for different injection altitudes . Bulk natural aerosol emissions (i.e., desert dust and sea spray), are treated using offline monthly emissions files based on AEROCOM and hence are independent of the meteorological conditions. The atmospheric chemistry is simulated with the MECCA (Module Efficiently Calculating the Chemistry of the Atmosphere) submodel by , and the aerosol microphysics and gas-aerosol partitioning are calculated by the Global Modal-aerosol eXtension (GMXe) aerosol module . Gas/aerosol partitioning is calculated using the ISORROPIA-II model . Following the approach of , the year 2010 is used as the reference year, the feedback between chemistry and dynamics was switched off, and therefore all simulations described here are based on the same (binary identical) dynamics and consequent transport of tracers.

Although evaluated the model for the same configuration and emissions database, the emissions referred to the year 2005, while here the emissions for the year 2010 are used. Therefore the model is re-evaluated for the species of interest (i.e., SO42-, NO3-, NH4+ and PM2.5). The model results of this study have been evaluated against satellite based PM2.5 estimates ; the results are shown in Fig.  and are summarized in Table , also focusing on the four regions focus of this study (i.e., North America, Europe, South and East Asia). Compared to global satellite-derived PM2.5 concentrations this model version, with prescribed dust emissions, consistently overestimates PM2.5 over desert areas (see Fig. ). However, the average concentration of PM2.5 at the surface in the regions of interest is within 30 % of the observations. For Europe and South Asia, 95 % of the simulated PM2.5 concentrations are within a factor of 2 of the observations, while for North America and East Asia this is about 80 %.

Further, SO42-, NO3- and NH4+ have been compared with station observations from different databases, such as from EPA (United States Environmental Protection Agency), EMEP (European Monitoring and Evaluation Programme) and EANET (Acid Deposition Monitoring Network in East Asia) for the year 2010. The results are shown in Fig.  and summarized in Table .

While sulfate is well reproduced, with more than ∼85 % of the model results within a factor of 2 compared to the observations, nitrate is overestimated in North America and Europe by ∼50 %, although nitric acid is reproduced accurately by the model (based on comparison with observations from ; see ). As the nitrate concentrations seem to be on the high end of the observations, it must be acknowledged that the effect of reducing ammonia emissions from agriculture could be overestimated. On the other hand, nitrate predictions are in good agreement with the measurements over East Asia. Further, ammonium concentrations are well captured by the model, with more than 75 % of the modeled results being within a factor of 2 compared to the observations. For ammonium, the annual averages estimated from model results compare well with the observations (see Table ). Further, as shown by , simulated seasonal cycle of ammonium concentrations compares well with the observed one, both for Europe and Asia (with temporal correlations between model results and observations above 0.7 and 0.5, respectively). However, this is not the case on the east coast of the USA, where the correlation is below 0.2. As suggested by , this is due to the wrong seasonality of the ammonia emissions, driven by an underestimation of the livestock emissions, which have a maximum in summer and should account for 80 % of the annual NH3 emissions in the region . The agricultural emissions of ammonia in this region in the model reproduce mostly the fertilizer application as described by and therefore the real seasonality of the ammonia emissions is missing . The seasonal results over the USA should hence be taken with caution. Further evaluation can be found in and .

In the current analysis four simulations with the EMAC model have been performed to study the impacts on PM2.5 components: the evaluated reference simulation (REF) and three sensitivity calculations in which the agricultural emissions have been reduced by different percentages, 50 % in simulation REF_50, 75 % in simulation REF_75 and 100 % (i.e., removing all agricultural emissions) in simulation REF_100.

The NOx emissions from agriculture are 0.7 Tg(N)yr-1, i.e., only ∼ 1.7 % of the total NOx emissions. Most importantly, 34.3 Tg(N)yr-1 of NH3 are emitted by agricultural activities, such as livestock manure and N mineral fertilizers, accounting for ∼80 % of the anthropogenic and ∼67 % of the total global ammonia emissions. Agricultural waste burning is responsible for the emissions of 0.1 Tg(S)yr-1 of SO2 (less than 1 % of the total SO2 emissions) and 23.2 Tg(C)yr-1 of CO (∼ 5 % of the total CO emissions), as well as 0.4 and 1.9 Tg(C)yr-1 of black carbon (BC) and organic carbon (OC), respectively, representing in both cases ∼ 5 % of their total emissions.

Considering these emission magnitudes, the main effects of agricultural emissions on PM2.5 are expected from NH3 through gas-particle partitioning. Therefore, the ammonia emissions used in this work have been compared to other used databases, such as EDGARv4.3.1 Emission Database for Global Atmospheric Research, and RCP85 Representative Concentration Pathways. These data sets differ globally by ∼15 % (40.26, 47.49 and 40.62 Tgyr-1 for EDGAR-CIRCE, EDGARv4.3.1 and RCP85). This reflects the uncertainties in the emission estimates of ammonia, which could be up to 50 % on a local scale . The implementation of bidirectional exchange of ammonia between the soil and atmosphere may improve the emissions from livestock, although this approach is still associated with underestimates of emissions . Further, ammonia emitted from traffic is included (∼1 % of total ammonia emissions), although toward the lower end of what has been estimated by .

As shown by , and , a sustainable reduction of ammonia emissions between 20 to 90 % could be achieved, depending on the emission process and the methodology applied (e.g., slurry acidification, adjustment in slurry application, under-floor drying of broiler manure in buildings, replacing urea with ammonium nitrate). As the efficiencies of the abatement processes are not well established , fixed relative reductions have been applied here to all agricultural emissions. showed that for the United Kingdom a moderate reduction in ammonia emission is easily affordable, while the costs are likely to increase exponentially for reductions above 25 %. The same control measures would be even more difficult to apply in countries in which livestock production is projected to largely increase (such as Asia; ), where they should be adopted on a large scale.

showed that in North America emission controls of SO2 and NOx are likely to be very costly and probably less efficient than decreasing agricultural emissions. Therefore, the regulation of ammonia emissions from agricultural activities offers the possibility of relatively cost-effective control policies for PM2.5. Our model simulations indicate that a 50 % decrease of ammonia emissions could reduce the annual, geographical average near-surface PM2.5 concentrations up to ∼ 1.0 (11 %), 0.3 (8 %), 1.6 (5 %) and 0.6 (2 %) µgm-3 in Europe, North America, East Asia and South Asia, respectively. The reduction can even be larger in winter (up to ∼ 1.3 (11 %), 0.6 (15 %), 2.2 (5 %) and 1.0 (3 %) µgm-3, respectively) when particulate ammonium nitrate concentrations are typically higher than in summer.

Our model simulations underscore the strong nonlinearity that plays a role in the sulfate-nitrate-ammonia system, which affects the efficiency of PM2.5 controls, especially in summer when the sulfate / nitrate ratio is high. A strong reduction of PM2.5 in response to NH3 emission regulation is expected once the ammonia-limited regime is reached. As a result, the possible PM2.5 reduction could be as large as ∼ 34 and ∼ 17 % in Europe and North America, respectively. Our results also suggest that ammonia emission controls could reduce the particle pH up to 1.5 pH units in East Asia in winter and more than 1.7 pH units in South Asia, theoretically assuming complete agricultural emission removal, which could have repercussions for the reactive uptake of gases from the gas phase and the outgassing of relative weak acids.

Furthermore, the global mortality attributable to PM2.5 could be reduced by ∼250000 (95 % CI: 148–290) people yr-1 worldwide worldwide by decreasing agricultural emissions by 50 %, with a gain of 16 000 (30 %), 52 000 (19 %), 25 000 (3 %) and 105 000 (8 %) people yr-1 in North America, Europe, South and East Asia, respectively. A total phase-out of agricultural emissions would even reduce the mortality attributable to air pollution worldwide by about 801 000 people yr-1 (25 %), in Europe by about 222 000 people yr-1 (80 %), in North America by about 40 000 people yr-1 (74 %), in South Asia by about 82 000 people yr-1 (10 %) and in East Asia by about 343 000 people yr-1 (25 %). These strong impacts are related to the nonlinear responses in both the sulfate-nitrate-ammonia system and the exposure-response functions at relatively low PM2.5 concentrations.

Therefore, emission control policies, especially in North America and Europe, should involve strong ammonia emission decreases to optimally reduce PM2.5 concentrations as well as further reductions in sulfur and nitrogen oxides emissions to avoid strong acidification of particles.