Pan-European rural atmospheric monitoring network shows dominance of NH3 gas and NH4NO3 aerosol in inorganic pollution load

Abstract. A comprehensive European dataset on monthly atmospheric NH3, acid gases (HNO3, SO2, HCl) and aerosols (NH4+, NO3-, SO42-, Cl-, Na+, Ca2+, Mg2+) is presented and analyzed. Speciated measurements were made with a low-volume denuder and filter pack method (DELTA®) as part of the EU NitroEurope (NEU) integrated project. Altogether, there were 64 sites in 20 countries (2006–2010), coordinated between 7 European laboratories. Bulk wet deposition measurements were carried out at 16 co-located sites (2008–2010). Inter-comparisons of chemical analysis and DELTA® measurements allowed an assessment of comparability between laboratories. The form and concentrations of the different gas and aerosol components measured varied between individual sites and grouped sites according to country, European regions and 4 main ecosystem types (crops, grassland, forests and semi-natural). Smallest concentrations (with the exception of SO42- and Na+) were in Northern Europe (Scandinavia), with broad elevations of all components across other regions. SO2 concentrations were highest in Central and Eastern Europe with larger SO2 emissions, but particulate SO42- concentrations were more homogeneous between regions. Gas-phase NH3 was the most abundant single measured component at the majority of sites, with the largest variability in concentrations across the network. The largest concentrations of NH3, NH4+ and NO3- were at cropland sites in intensively managed agricultural areas (e.g. Borgo Cioffi in Italy), and smallest at remote semi-natural and forest sites (e.g. Lompolojänkkä, Finland), highlighting the potential for NH3 to drive the formation of both NH4+ and NO3- aerosol. In the aerosol phase, NH4+ was highly correlated with both NO3- and SO42-, with a near 1 : 1 relationship between the equivalent concentrations of NH4+ and sum (NO3- + SO42-), of which around 60 % was as NH4NO3. Distinct seasonality were also observed in the data, influenced by changes in emissions, chemical interactions and the influence of meteorology on partitioning between the main inorganic gases and aerosol species. Springtime maxima in NH3 were attributed to the main period of manure spreading, while the peak in summer and trough in winter were linked to the influence of temperature and rainfall on emissions, deposition and gas-aerosol phase equilibrium. Seasonality in SO2 were mainly driven by emissions (combustion), with concentrations peaking in winter, except in Southern Europe where the peak occurred in summer. Particulate SO42- showed large peaks in concentrations in summer in Southern and Eastern Europe, contrasting with much smaller peaks occurring in early spring in other regions. The peaks in particulate SO42- coincided with peaks in NH3 concentrations, attributed to the formation of the stable (NH4)2SO4. HNO3 concentrations were more complex, related to traffic and industrial emissions, photochemistry and HNO3 : NH4NO3 partitioning. While HNO3 concentrations were seen to peak in the summer in Eastern and Southern Europe (increased photochemistry), the absence of a spring peak in HNO3 in all regions may be explained by the depletion of HNO3 through reaction with surplus NH3 to form the semi-volatile aerosol NH4NO3. Cooler, wetter conditions in early spring favour the formation and persistence of NH4NO3 in the aerosol phase, consistent with the higher springtime concentrations of NH4+ and NO3-. The seasonal profile of NO3- was mirrored by NH4+, illustrating the influence of gas : aerosol partitioning of NH4NO3 in the seasonality of these components. Gas-phase NH3 and aerosol NH4NO3 were the dominant species in the total inorganic gas and aerosol species measured in the NEU network. With the current and projected trends in SO2, NOx and NH3 emissions, concentrations of NH3 and NH4NO3 can be expected to continue to dominate the inorganic pollution load over the next decades, especially NH3 which is linked to substantial exceedances of ecological thresholds across Europe. The shift from (NH4)2SO4 to an atmosphere more abundant in NH4NO3 is expected to maintain a larger fraction of reactive N in the gas phase by partitioning to NH3 and HNO3 in warm weather, while NH4NO3 continues to contribute to exceedances of air quality limits for PM2.5.


The negative effects of these pollutants on sensitive ecosystems are mainly through acidification (excess acidity) and eutrophication (excess nutrient N) processes that can lead to a loss of key species and decline in biodiversity (e.g. Hallsworth 10 et al., 2010;Stevens et al., 2010). They are also implicated in radiative forcing, and influence climate change through inputs of nitrogen that can alter the carbon cycle (Reis et al., 2012;Sutton et al., 2013;Zaehle & Dalmonech, 2011). Reductions in NOx emissions have been more modest, at 45 % over the same period, with emissions in 2017 of 8563 kt NOx exceeding those of SO2. By contrast, the reductions in NH3 emissions (of which over 90% come from agriculture) have been more modest, decreasing by only 18 %. Here, the decrease was largely driven by reductions in fertiliser use and livestock 20 numbers, in particular from eastern European countries, rather than through implementation of any abatement or mitigation measures. More worryingly, the decreasing trend has reversed in recent years, with emissions increasing by 5 % since 2010, to 4788 kt NH3 in 2017 (EEA, 2019).
In recent assessments, critical loads of acidity were exceeded in about 5 % of the ecosystem area across Europe in 2017 (EMEP, 25 2018). While the substantial decline in SO2 emissions has allowed the recovery of ecosystems from acid rain, NH3 from agriculture and NOx from transport are increasingly contributing to a larger fraction of the acidity load. Although NH3 is not an acid gas, nitrification of NH3 and ammonium (NH4 + ) releases hydrogen ions (H + ) that acidify soils and freshwater. The deposition of reactive N (Nr, including oxidised N: NOx, HNO3, NO3and reduced N: NH3, NH4 + ) and their contribution to eutrophication effects have also been identified by the EEA as the most important impact of air pollutants on ecosystems and 30 biodiversity (EEA, 2019). The deposition of Nr throughout Europe remains substantially larger than the level needed to protect ecosystems, with critical loads thresholds for eutrophication from N exceeded in around 62 % of the EU-28 ecosystem area and in almost all countries in Europe in 2017 (EMEP, 2018).
Following emission, atmospheric transport and fate of the gases are controlled by the following processes: short range 35 dispersion and deposition, chemical reaction and formation of NH4 + aerosols, and the long-range transport and deposition of the aerosols (Sutton et al., 1998;ROTAP, 2012). Atmospheric S and Nr inputs from the atmosphere to the biosphere occur though i) dry deposition of gases and aerosols, ii) wet deposition in rain, and iii) occult deposition in fog and cloud (Smith et al., 2000;ROTAP, 2012). The deposition processes contribute very different fractions of the total S or Nr input and different chemical forms of the pollutants at different spatial scales. NH3 is a highly reactive, water-soluble gas and deposits much faster 40 than NOx (which is not very water soluble and has low deposition velocity). Dry N deposition by NH3 therefore contributes a significant fraction of the total N deposition to receptors close to source areas and will often exert the larger ecological impacts, compared with other N pollutants (Cape et al., 2004;Sutton et al., 1998Sutton et al., , 2007. Numerous studies have shown that Nr https://doi.org /10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License. deposition in the vicinity of NH3 sources is dominated by dry NH3-N deposition (e.g. Pitcairn et al., 1998;Sheppard et al., 2011), with removal of NH3 close to a source controlled by physical, chemical and ecophysiological processes (Flechard et al., 2011;Sutton et al., 2007Sutton et al., , 2013. Unlike NOx, HNO3 (from oxidation of NOx) is very water-soluble, while NO3particles can act as cloud condensation nuclei (CCN) so that they are both scavenged quickly and removed efficiently by precipitation.
Since NOx is inefficiently removed by precipitation, wet deposition of NOx near a source is small and only becomes important 5 after NOx has been converted to HNO3 and NO3 -.
Because of the large numbers of atmospheric N species and their complex atmospheric chemistry, quantifying the deposition of Nr is hugely complex and is a key source of uncertainty for ecosystems effects assessment (Bobbink et al., 2010;Schrader et al., 2018;Sutton et al., 2007). Input by dry deposition can be estimated using a combination of measured 10 and/or modelled concentration fields with high-resolution inferential models (e.g. Smith et al., 2000;Flechard et al., 2011), or by making direct flux measurements (e.g. Fowler et al., 2001;Nemitz et al., 2008). Although it is possible to measure Nr deposition directly (e.g. Skiba et al., 2009), the flux measurement techniques are complex and resource intensive, unsuited to routine measurements at a large number of sites. The 'inferential' modelling approach provides a direct estimation of deposition from Nr measurements by applying a land-use dependent deposition velocity (Vd) to measured concentrations (Dore 15 et al., 2015;Flechard et al., 2011;Simpson et al., 2006;Smith et al., 2000).
At present, there are limited atmospheric measurements that speciate the gas and aerosol phase components at multiple sites over several years. On a European scale, atmospheric measurements of sulfur (SO2, particulate SO4 2-) and nitrogen (NH3, HNO3, particulate NH4 + , NO3 -) have been made by a daily filter pack method across the European Monitoring and Evaluation 20 Program (EMEP) networks since 1985, providing data for evaluating wet and dry deposition models (EMEP, 2016;Torseth et al., 2012). The method, however, does not distinguish between the gas and aerosol phase N species. Consequently, these data are reported as total inorganic ammonium (TIA = sum of NH3 and NH4 + ) and total inorganic nitrate (TIN = sum HNO3 and NO3 -), limiting the usefulness of the data. Speciated measurements by an expensive and labour-intensive daily annular denuder method are also made (Torseth et al., 2012), but are necessarily restricted to a small number of sites, due to the high costs 25 associated with this type of measurement. There are also networks with a focus on specific N components, for example, the national NH3 monitoring networks in the Netherlands (LML, van Zanten et al., 2017) and in the UK (National Ammonia Monitoring Network, NAMN; Tang et al., 2018a), or compliance monitoring across Europe in the case of SO2 and NOx. The UK is unique in having an extensive set of speciated gas and aerosol monitoring data from the Acid Gas and Aerosol Network (AGANet), with measurements from 1999 to the present (Tang et al., 2018b). 30 In this context, there is an ongoing need for cost-effective, easy-to-operate, time-integrated atmospheric measurement for the respective gas and aerosol phases at sufficient spatial scales. Such data would help to, 1) improve estimates of N deposition, 2) contribute to development and validation of long-range transport models, e.g. EMEP (Simpson et al., 2006) and EMEP4UK (Vieno et al., 2014(Vieno et al., , 2016, 3) interpret interactions between the gas and aerosol phases, and 4) interpret ecological responses 35 to nitrogen (e.g. ecosystem biodiversity or net carbon exchange). To contribute to this goal, a '3-level' measurement strategy in the EU Framework Programme 6 Integrated Project "NitroEurope" (NEU, www,nitroeurope.eu) between 2006 and 2010 delivered a comprehensive integrated assessment of the nitrogen cycle, budgets and fluxes for a range of European terrestrial ecosystems (Sutton et al., 2007;Skiba et al., 2009). At the most intensive level (Level 3), state-of-the-art instrumentation for high resolution, continuous measurements at a small number of 13 'flux super sites' provided detailed understanding on 40 atmospheric and chemical processes (Skiba et al., 2009). By contrast, manual methods with a low temporal frequency (monthly) at the basic level (Level 1) provided measurements of Nr components at a large number of sites (> 50 sites) in a https://doi.org /10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License. cost-efficient way in a pan-European network (Tang et al., 2009). Key species of interest included NH3, HNO3 and ammonium aerosols ((NH4)2SO4, NH4NO3).
In this paper, we present and discuss four years of monthly reactive gas (NH3, HNO3, HCl) and aerosol (NH4 + , NO3 -, SO4 2-, Cl -, Na + , Ca 2+ , Mg 2+ ) measurements from the Level 1 network set up under the NEU integrated project, complemented by two 5 years of bulk wet deposition data made at a subset of the network sites ( Figure 1). A harmonised measurement approach with a simple, cost-efficient time-integrated method, applied with high spatial coverage allowed a comprehensive assessment across Europe. Measurements across the network were coordinated between multiple European laboratories. The measurement approach and the operations of the networks, including the implementation of annual inter-comparisons to assess comparability between the laboratories, are described. The data are discussed in terms of spatial and temporal variation in concentrations, 10 relative contribution of the inorganic nitrogen and sulfur components to the inorganic pollution load, and changes in atmospheric concentrations of acid gases and their interactions with NH3 gas and NH4 + aerosol.

NEU Level 1 DELTA ® network
The NitroEurope (NEU) Level 1 network was operated between November 2006 and December 2010 to deliver the core measurements of reactive nitrogen gases (NH3, HNO3) and aerosols (NH4 + , NO3 -) for the project (Figure 1). A low-volume denuder-filter pack method, the 'DEnuder for Long-Term Atmospheric sampling' system (DELTA ® , Sutton et al., 2001a;Tang et al., 2009Tang et al., , 2018b with time-integrated monthly sampling was used, which made implementation at a large number of sites 20 possible. Other acid gases (SO2, HCl) and aerosols (SO4 2-, Cl -, Na + , Ca 2+ , Mg 2+ ) were also collected at the same time and measured by the DELTA ® method. DELTA ® measurements were co-located with all NEU Level 3 sites with advanced flux measurements (Skiba et al., 2009), and with the network of main CarboEurope-IP CO2 flux monitoring sites (www.carboeurope.eu) (Flechard et al., 2011(Flechard et al., , 2020. Two of the UK sites in the NEU DELTA ® network are existing UK NAMN (Tang et al., 2018a) and AGANet sites (Tang et al., 2018b). These are Auchencorth Moss (UK-Amo) and Bush  EBu) located in Southern Scotland. Monthly gas and aerosol data at the two sites, made as part of the UK national networks, were included in the NEU network. NEU network Nr data were used, together with a range of dry deposition models, to model dry deposition fluxes (Flechard et al., 2011) and to assess the influence of Nr on the C cycle, potential C sequestration and the greenhouse gas balance of ecosystems using CO2 exchange data from the co-located CarboEurope sites (Flechard et al., 2020).
Other measurements made at the Level 1 sites included estimation of wet deposition fluxes (Sect. 2.3) and also soil and plant 30 bioassays (Schaufler et al., 2010).
Altogether, the DELTA ® network covered a wide distribution of sites across 20 countries and 4 major ecosystem types: crops, grassland, semi-natural and forests. These sites can be described as 'rural', and were chosen to provide a regionally representative estimate of air composition. The network site map is shown in Figure 2, with site details given in Supp. Table  35 S1. Further information on the network sites are also provided in Flechard et al. (2011). Network establishment started in November 2006, with 57 sites operational from March 2007 onwards. Over the course of the network, some sites closed or were relocated due to infrastructure changes and new sites were also added. A total of 64 sites provided measurements at the end of the project, with 45 of the sites operational the entire time. In addition, replicated DELTA ® measurements were made at 4 sites: 40 https://doi.org /10.5194/acp-2020-275 Preprint.  A team of seven European laboratories shared responsibility for running the network. Measurement was on a monthly 10 timescale, with each laboratory preparing and analysing the monthly samples with documented analytical methods for between 5 and 16 DELTA sites ( Figure 2). The use of a harmonised DELTA ® methodology, coupled to defined quality protocols (Tang et al., 2009) ensured comparability of data between the laboratories (see later in Sect. 3.1 and Sect. 3.2). A network of local site operators representing the science teams of each site performed the monthly sample changes and posted the exposed samples back to their designated laboratories for analysis. Air concentration data were submitted by the laboratories for their 15 respective sites in a standard reporting template to UKCEH. Following data checks against defined quality protocols (Tang et al., 2009), the finalised dataset was uploaded to the NEU database (www.nitroeupe.eu). Establishment of the network, including the first year of measurement results on Nr components are reported in Tang et al. (2009). Information on co-located measurements and agricultural activities at each of the sites were also collected and are accessible from the NEU website (www.nitroeurope.eu). 20

DELTA ® methodology
The DELTA ® method used in the NEU Level 1 network is based on the system developed for the UK Acid Gas and Aerosol monitoring network (AGANet, Tang et al., 2018b). Full details of the DELTA ® method and air concentration calculations in the NEU network are provided by Tang et al. (2009Tang et al. ( , 2018b. The method uses a small 6 V air pump to deliver low air sampling 25 rates of between 0.2 to 0.4 L min -1 , a high sensitivity gas meter to record the typically monthly volume of air collected and a DELTA ® denuder-filter pack sample train to collect separately the gas and aerosol phase components. The sample train is made up of two pairs of base and acid impregnated denuders (15 cm and 10 cm long) to collect acid gases and NH3, respectively, under laminar conditions. A 2-stage filter pack with base and acid coated cellulose filters collects the aerosol components downstream of the denuders. The base coating used was K2CO3/glycerol which is effective for the simultaneous 30 collection of HNO3, SO2 and HCl (Ferm, 1986), while the acid coating was either citric acid for temperate climates or phosphorous acid for Mediterranean climates (Allegrini et al., 1987;Ferm, 1979;Perrino et al., 1999;Fitz, 2002). In this way, artefacts between gas and aerosol phase concentrations are minimized (Ferm et al., 1979;Sutton et al., 2001a). The DELTA ® air inlet has a particle cut-off of ~ 4.5 µm which means fine mode aerosols in the PM2.5 fraction and some of the coarse mode aerosols < PM4.5 will be collected (Tang et al., 2015). 35 A low voltage version of the AGANet DELTA ® system was built centrally by UKCEH and sent to each of the European sites where they were installed by local site contacts. These systems operated on either 6 V (off mains power with a transformer) or 12 V from batteries (wind and solar powered). Air sampling was direct from the atmosphere without any inlet lines or filters to avoid potential loss of components, in particular HNO3 that is very "sticky", to surfaces. Sampling height was 1.5 m above 40 ground/vegetation in open areas. In forested areas, the DELTA ® equipment was set up either in large clearings, or on towers at 2 -3 m above the canopy (see Flechard et al., 2011). https://doi.org /10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.

Calculation of gas and aerosol concentrations
Atmospheric gas and aerosol concentrations in the DELTA ® method are calculated from the amount of inorganic ions (NH4 + , NO3 -, SO4 2-, Cl -, and base cations) in the denuder/aerosol aqueous extracts and the volume of air sampled (from gas meter readings), which is typically 15 m 3 for a monthly sample. The volume of deionised water used to extract acid coated denuders and aerosols filters are 3 mL and 4 mL, respectively. For the base coated denuders and aerosol filters, the extract volume in 5 both cases is 5 mL An example is shown here for calculating the atmospheric concentrations of NH3 (gas) (Equation 1) and NH4 + (aerosol) (Equation 2) from the aqueous extracts, based on an air volume of 15 m 3 collected in a typical month. Pairs of base and acid coated denuders are used to collect the acid gases and alkaline NH3 gas, respectively. This allows denuder collection efficiency of, for example, NH3 (Equation 3) to be assessed as part of the data quality assessment process.
An imperfect acid coating on the denuders for example can lead to lower capture efficiencies (Sutton et al., 2001a;Tang et al., 2003). 15 Denuder collection efficiency, NH 3 (%) = 100 x NH 3 (Denuder 1) NH 3 (Denuder 1+Denuder 2) [3] A correction, based on the collection efficiency, is applied to provide a corrected air concentration (a (corrected), Equation 4) (Sutton et al., 2001a;Tang et al., 2018aTang et al., , 2018b. With a collection efficiency of 95 %, the correction amounts to 0.3 % of 20 the corrected air concentration. For an efficiency below 60 %, the correction amounts to more than 50 % and is not applied. The air concentration of (a) of NH3 is then determined as the sum of NH3 in denuders 1 and 2 (Tang et al., 2018a). By applying the infinite series correction, the assumption is that any NH3 (and other gases) that is not captured by the denuders will be collected on the downstream aerosol filter. To avoid double counting, the estimated amount of 'NH3 breakthrough' is subtracted from the NH4 + concentrations on the aerosol filter. 25

Artefact in HNO3 determination
Results from the first DELTA ® inter-comparison in the NEU network (Tang et al., 2009) (see also Sect. 2.5) and further work by Tang et al. (2015Tang et al. ( , 2018b have shown that HNO3 concentrations may be overestimated on the carbonate coated denuders used, due to co-collection of other oxidized nitrogen components, most likely from nitrous acid (HONO). In the UK AGANet, HNO3 data are corrected with an empirical factor of 0.45 derived by Tang et al. (2015). Since the correction factor for HNO3 5 is uncertain (estimated to be ± 30 %) and derived for UK conditions, no attempt has been made to correct the HNO3 data from the NEU network. The DELTA ® method remained unchanged throughout the entire network operation and provided a consistent set of measurements by the same protocol. The caveat is that the HNO3 data presented in this paper also includes an unknown fraction of oxidized N, most probably HONO, and therefore represents an upper limit in the determination of HNO3. Contribution from NO2 is likely to be small, since this is collected with a low efficiency on carbonate coated denuders 10 (Bai et al., 2003;Tang et al., 2015) and the network sites are rural, where NOx concentrations are expected to be in the low ppbs. At the French Fougéres parallel site (FR-FgsP), NaCl coated denuders were used to measure HNO3, to compare with results from K2CO3/glycerol coated denuders at the main site (FR-Fgs) (see Sect. 2.1).

NEU Bulk wet deposition network 15
The NEU bulk wet deposition network (Figure 3, Supp. Table S2) was established to provide wet deposition data on NH4 + and NO3 -. It was set up two years after the establishment of the NEU DELTA ® network, with sites located at a subset of DELTA ® sites that did not already have on-site wet deposition measurements. Sampling commenced at some sites in January 2008, with 14 sites operational from March 2008. Site changes also occurred during the operation of this network, again with some site closures and new site additions over time. In total, 12 sites provided 2 years of monthly data, with a further 6 sites providing 20 1 year of monthly data between 2008 to October 2010 when measurements ended.

<INSERT FIGURE 3>
The type of bulk precipitation collector used was a Rotenkamp sampler (Dämmgen et al., 2005), mounted 1.5 m above ground, 25 or in the case of forest sites, either in clearings or above the canopy. Each unit has two collectors providing replicated samples, comprising of a pyrex glass funnel (aperture area = 84.9 cm 2 ) with vertical sides, connected directly to a 3 L collection bottle (material = low density polyethylene) which was changed monthly. Thymol (5-methyl-2-(1-methylethyl)phenol) (150 mg) was added as a biocide (Cape et al., 2012) to a clean, dry pre-weighed bottle at the start of each collection period. This provided a minimum thymol concentration of 50 mg L -1 for a full bottle to preserve the sample against biological degradation of labile 30 nitrogen compounds during the month-long sampling.
Three European laboratories shared management and chemical analysis for the network (Figure 3). The laboratories were CEAM (all 3 Spanish sites), INRAE (French Renon site) and SHMU, designated the main laboratory responsible for all other sites. A full suite of precipitation chemistry analyses were carried out that included: pH, conductivity, NH4 + , NO3 -, SO4 2-, PO4 3-35 , Cl -, Na + , K + , Ca 2+ and Mg 2+ . Rain volumes and precipitation chemistry data were submitted in a standard template to UKCEH for checking and then uploaded to the NEU database (www.nitroeupe.eu). Samples with high P (> 1 µg L -1 PO4 3-), high K + and/or NH4 + values that are indicative of bird contamination were rejected. Annual wet deposition (e.g. kg N ha -1 yr -1 ) were estimated from the product of the species concentrations and rain volume. Determinations of organic N were also carried out on some of the rain samples in a separate investigation reported by Cape et al. (2012). 40 https://doi.org /10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.

Laboratory inter-comparisons: chemical analysis
All laboratories in the DELTA ® and bulk wet deposition networks participated in water chemistry proficiency testing (PT) schemes in their own countries, as well as the EMEP (once annual, http://www.emep.int) and/or WMO-GAW (twice annual, http://www.qasac-americas.org/lab_ic.html) laboratory inter-comparison schemes. PT samples for analysis are synthetic precipitation samples for determination of pH, conductivity and all the major inorganic ions at trace levels. In addition, UKCEH 5 also organised an annual PT scheme for the duration of the project (NEU-PT) to compare laboratory performance in the analysis of inorganic ions at higher concentrations relevant for DELTA ® measurements. This comprised the distribution of reference solutions containing known concentrations of ions that were analysed by the laboratories as part of their routine analytical procedures.  Table 1. At each test site, DELTA ® systems were randomly assigned to each of the participating laboratories. All laboratories provided DELTA ® sampling trains for each of the inter-comparison sites and carried out chemical analysis on the returned exposed samples. Measurement results were returned in a standard template to UKCEH, the central coordinating laboratory for collation and analysis.

25
<INSERT in each country was derived by dividing the 4-year averaged total emissions by the land area (km 2 ).
Gridded emissions data: Gridded emissions data (at 0.1º x 0.1º resolution) for SO2, NOx and NH3 are available from the EMEP emissions database (EMEP, 2020). The 0.1º x 0.1º gridded data for the period 2007 to 2010 were downloaded and were used to: 1. Estimate national total emissions (sum of all grid squares in each country) and 4-year averaged emission densities (t 35 km -2 yr -1 ) for Russia and Ukraine. As a check, total emissions for the other 18 countries were also calculated by this method and were the same as the national emission totals reported by the EEA (EEA, 2019).
2. Extract gas emissions for individual grids (0.1º x 0.1º) that contains a NEU DELTA ® site.

2.7
National air quality network data from the Netherlands and UK

Dutch LML network data
Atmospheric NH3 has been monitored at 8 sites in the Dutch national air quality monitoring network (LML, Landelijk Meetnet Luchtkwaliteitl) since 1993 (van Zanten et al., 2017). The low density, high time-resolution LML network is complemented by a high density monthly diffusion tube network, the Measuring Ammonia in Nature (MAN) network (http://man.rivm.nl) 5 (Lolkema et al., 2015). The MAN network has 136 monitoring locations sited within nature reserves that includes 60 Natura 2000 sites, with concentrations ranging between 1.0 and 14 μg m -3 (Lolkema et al., 2015). The focus of the MAN network is to provide site-based NH3 concentrations for the nature conservation sites, rather than a representative spatial concentration field for the country. Hourly NH3 and SO2 data which were also available from the 8 sites in the LML network were downloaded from the RIVM website (http://www.lml.rivm.nl/gevalideerd/index.php). The 4-year averaged NH3 and SO2 10 concentrations for the period 2007 to 2010 were calculated and used to complement measurement data from the 4 Dutch sites in the NEU DELTA ® network.

UK NAMN and AGANet network data
Atmospheric NH3, acid gases and aerosols are measured in the UK NAMN (since 1996) and AGANet (since 1999) (Tang et 15 al., 2018a(Tang et 15 al., , 2018b. The UK approach is a high density network with low time-resolution (monthly) measurements, combining an implementation of the DELTA ® method used in the present NEU DELTA ® network and a passive ALPHA ® method  to increase network coverage in NH3 measurements (Sutton et al., 2001b;Tang et al., 2018a). Monthly and annual data for the overlapping period of the project were extracted from the UK-AIR website (https://uk-air.defra.gov.uk/) and nested with the NEU network data for analysis in this paper.

<INSERT FIGURE 4>
Altogether, results from the combined PT schemes produced >100 observations for each reported chemical component over 30 the 4 year period. The performances of laboratories in Figure 4 can be summarised in terms of the percentage of reported results agreeing within 10 % of the true values (see summary table below Figure 4), where the true values represent the nominal concentrations in the aqueous test solutions. The best agreements was for SO4 2and NO3 -, with an average of 92 % and 87 % of all reported results agreeing within 10 % of the true value across the concentration range covered in the PT schemes. In the case of NH4 + , while an average of 90 % of reported results were within 10% of the reference at 1 mg L -1 NH4 + , laboratory 35 performance was poorer (68 % agreeing within 10 %) at lower concentrations (0.1 -0.9 mg L -1 ). Poorer performance at the low concentrations was largely due to two laboratories (CEAM and SHMU) with > 50 % of their results reading high. For Na + and Cl -, the percentages of results agreeing within 10 % of the reference were 81 % and 86 %, respectively, across the full range of PT concentrations. At concentrations above 1 mg L -1 , the agreement improved and increased to 89 % for Na + and 96% for Cl -. A larger spread around the reference values were provided for the base cations Ca 2+ and Mg 2+ at low concentrations 40 https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.
(< 1 mg L -1 ). The percentage of results passing at low concentrations below 1 mg L -1 was 36 % (Ca 2+ ) and 59 % (Mg 2+ ), increasing to 80 % (Ca 2+ ) and 90 % (Mg 2+ ) above 1 mg L -1 . The larger scatter at low concentrations is likely due to uncertainty in the chemical analysis at or close to the method limit of detection, and reflects challenges of measuring base cations, in particular Ca 2+ as this is very 'sticky' and adsorbs/desorbs from surfaces leading to analytical artefacts.

5
To show what the PT reference solution concentrations would correspond to if they were a denuder and/or aerosol extract, equivalent gas (Equation 1) and/or aerosol concentrations (Equation 2) (Sect. 2.2.1) are calculated for each of the ions and provided in the summary table in Figure 4. A 0.5 mg L -1 NH4 + solution, for example, is equivalent to an atmospheric concentration of 0.09 µg NH3 m -3 (gas), or 0.13 µg NH4 + m -3 (aerosol) for a monthly sample. In Figure 5 With the exception of a small number of outliers, most data points are close to the 1:1 line with laboratory results agreeing within ± 0.05 µg m -3 in equivalent gas and/or aerosol concentrations. These are low ambient concentrations and show that the measurement uncertainty in the analysis of very low concentrations in the PT schemes will be small for the majority of sites in the network, where concentrations were found to be much higher (see Figure 6). 15

Laboratory inter-comparison results: DELTA ® measurements
Results from 4 years of annual DELTA ® field inter-comparisons (2006 -2009), for all field sites, are combined and 20 summarised in Figure 6. The gas and aerosol concentrations measured and reported by each of the laboratories are compared with the median estimate of all laboratories in each of the scatter plots, with the colour of the symbols identifying the laboratory providing the measurements. Regression results (slope and R 2 ) in the table below the plots provide the main features of the inter-comparison. The slope is equivalent to the mean ratio of each laboratory against the median value, where values close to unity indicate closer agreement to the median value. Overall, the scatter plots show good agreement between the laboratories, 25 with some laboratories showing very close agreement to the median estimates, and more scatter observed from the others.

<INSERT FIGURE 6>
The occurrence of outliers in some of the individual monthly values indicates that caution needs to be exercised in the 30 interpretation of these data points in the inter-comparison. To average out the influence of a few individual outliers, the mean concentrations from each of the seven laboratories for each of the four field sites were calculated and compared with averaged median estimates of all laboratories for each site. A summary of the mean concentrations and the percentage difference from median is presented in mean concentrations between laboratories are broadly comparable. Each of the laboratories were also able to resolve the main differences in mean concentrations at the four field sites, ranging from the smallest concentrations at Auchencorth (e.g. median = 1.4 µg NH3 m -3 ) to higher concentrations representing a more polluted site at Paterna (e.g. median = 5.2 µg NH3 m -3 ) for the test periods (Table 2). Larger differences for HCl, Ca 2+ and Mg 2+ are due to clear outliers from one or two laboratories at the very low concentrations of these species encountered and may be related to measurement uncertainties at the low air 40 concentrations. The comparability between laboratories for each of the components is next considered in turn. The best agreement between laboratories was for the Nr gases (NH3, HNO3) and aerosol species (NH4 + , NO3 -), with slopes within ± 10 % of the median values and R 2 > 0.9 in the regression analysis from five of the laboratories ( Figure 6, Table 2). 5 This is important since Nr species were the primary focus for the NEU DELTA ® network. Slightly poorer agreement for NH3 and NH4 + were provided by CEAM and MHSC laboratories, with data points both above and below the 1:1 line ( Figure 6).
While this seems to suggest that the performance of MHSC for NH3 improved following the first inter-comparison exercise, 10 the regression slope for aerosol NH4 + increased instead from a slope of 1.26 (R 2 = 0.83, n = 41) to 1.48 (R 2 = 0.93, n = 10), suggesting an over-estimation of NH4 + concentrations (Supp. Figure S1). A possible cause may be the quality and/or variability in the aerosol filter blank values for NH4 + , as laboratory blanks are subtracted from exposed samples to estimate aerosol NH4 + concentrations. Laboratory blank results were however not reported to allow this assessment. Another possibility is a breakthrough of NH3 from the acid coated denuders onto the aerosol filters. The denuder collection efficiency of NH3 gas 15  . Table S3). This is comparable with the mean collection efficiencies of all laboratories (91 and 90 %) (Supp . Table S3), which makes NH3 breakthrough an unlikely explanation for the higher readings. The assessment of NH4 + is however more uncertain from the reduced number of data points (n = 10).

20
For the CEAM laboratory, reported NH3 concentrations were on average 16 % lower (n = 41) than the median, with a slope of 0.89 (R 2 = 0.87) and particulate NH4 + were on average 13 % lower (n = 41) than the median, with a slope of 0.42 (R 2 = 0.22) ( Figure 6). A need to improve the NH4 + analysis (Indophenol colorimetric assay) in the acid coated denuders and aerosol filters by the CEAM laboratory was identified from the 2006 inter-comparison (Tang et al., 2009). The Indophenol method for aqueous NH4 + determination is pH sensitive. Calibration solutions and quality control checks for the colorimetric assays are 25 made up in deionised water (pH 7), whereas the aqueous extracts from the DELTA ® acid coated denuders and cellulose filters are acidic (pH ~3). Determination of NH4 + in the denuder extracts may therefore be under-estimated if the pH of the indophenol reaction has not been adjusted for the increased acidity in the sample extracts. When the 2006 data are excluded from the regression analysis, the slopes for NH3 and NH4 + increased to 1.02 (R 2 = 0.94, n = 12) and 0.98 (R 2 = 0.51, n = 12), respectively (Supp. Figure S1). The improved agreement with other laboratories after the 2006 inter-comparison suggests that the method 30 under-read was largely resolved, reflected in an improvement in the slope. Despite some uncertainties in the NH3/NH4 + measurements, the laboratories were able to clearly resolve the main differences in mean concentrations at the four different field sites in all years ( Table 2). The results presented here for CEAM and MHSC highlight the importance of the initial intercomparison exercise in identifying and resolving sampling and analytical issues at the start of the project.

3.2.2
Inter-comparisons: SO2, SO4 2-Six laboratories provided slopes within 12 % of the median values in the regression analysis for SO2 ( Figure 6). The smaller R 2 values were from two laboratories (CEAM and SHMU, R 2 < 0.7), with data points both above and below the 1:1 line. For INRAE, the larger slope of 1.6 (R 2 = 9) was due to a single high SO2 reading reported for Auchencorth of 2.0 µg SO2 m -3 , compared with the median of 1.4 µg SO2 m -3 . When the mean SO2 concentrations measured by INRAE are compared with the 40 median, the difference was on average 13 %, providing acceptable agreement, which suggests that the high reading may just be an outlier. There was more scatter in the inter-comparison for SO4 2-, although the majority of points are still close to the 1:1 https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License. line ( Figure 6). Six laboratories provided slopes within 12 % of the median values in the regression analysis also for SO4 2-.
The regression slope from CEAM for SO4 2was 1.2 (R 2 = 0.9) which is still within 20% of the median. The SO2 and SO4 2measurements were broadly comparable between the laboratories, with mean concentrations agreeing on average within 6 % of the median (Table 2).

3.2.3
Inter-comparisons: HCl, Cl -5 The HCl inter-comparison show clear outliers from the CEAM laboratory, with concentrations that were on average up to 2 times higher than other laboratories (slope = 1.8). For example, a mean concentration of 1.8 µg HCl m -3 was reported by CEAM for Paterna, compared with a median of 0.7 µg HCl m -3 . Apart from CEAM, the mean concentrations of HCl reported by the other laboratories were generally comparable ( Table 2). The larger % differences between the measured mean and median at each site reflect the challenges of measuring the very low concentrations of HCl at these sites of < 0.5 µg HCl m -3 10 (slightly higher at Paterna). HCl results were reported by NILU for the 2008 inter-comparison exercise only, limiting the number of measurements (n = 4) available for comparison.
The comparison for Clshowed better agreement of the CEAM laboratory results with other laboratories, in both the intercomparison of individual monthly values ( Figure 6) and the mean concentrations (Table 2). Like HCl, larger % differences 15 between the measured concentrations and median at each site may be attributed to higher measurement uncertainties at the low concentrations of Cl -. For NILU, there were only 2 data points for Clfrom the Auchencorth site in the 2008 inter-comparison.
Overall, the inter-comparison for HCl and Clshowed that the laboratories were able to resolve the main differences in mean concentrations at the different sites even at the low concentrations encountered.

3.2.4
Inter-comparisons: Base cations (Na + , Ca 2+ , Mg 2+ ) Measurements of Ca 2+ and Mg 2+ were the most uncertain, with the largest scatter in the inter-comparisons ( Figure 6). Despite the trace levels of these base cations at all field sites, 4 laboratories (INRAE, UKCEH, SHMU, VTI) provided data close to the 1:1 line, demonstrating close agreement between these laboratories. The clear outliers above the 1:1 line are from CEAM, MHSC and NILU, with slopes > 2. While MHSC over-read Ca 2+ and Mg 2+ , their results for Na + were in better agreement with 25 other laboratories, with a slope of 0.9 (R 2 = 0.5) ( Figure 6). There was a lot of scatter in the data however, with outlier points both above and below the 1:1 line, suggesting measurement uncertainties in their base cation measurements. For NILU, the only base cation results reported by the laboratory were for the 2008 DELTA ® inter-comparisons at Auchencorth and Braunschweig. This accounts for the low number of data points (n = 4) from the NILU laboratory. The median concentrations of Ca 2+ and Mg 2+ at both field sites were very low (< 0.1 µg m -3 ), which makes comparison with the few data reported from 30 NILU highly uncertain. Like NILU, CEAM also did not report base cations results for all of the DELTA ® inter-comparison.

Comparisons according to ecosystem types
Annual averaged concentrations of gases and aerosols measured in the NEU DELTA ® network are presented in Figure 7, with sites grouped according to each of four major ecosystem types: crops, grassland, forests and semi-natural. These are the classifications used in dry deposition models, where ecosystem-specific deposition velocities (Vd) are combined with measurement data to produce estimates of Nr dry deposition (Flechard et al., 2011). In some models such as the Concentration 40 https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.

<INSERT FIGURE 7> 5
A total of 64 sites from 20 different countries, including replicated measurements at 4 of the sites, are compared in Figure 7.
Not all of the sites were however operational all of the time or at the same time. Changes in the numbers and locations of sites occurred over the duration of the network, for example, due to site closures, relocations and/or new site additions. The annual averaged concentrations plotted for each site are the mean of all available annual means. Where the annual averaged 10 concentration is derived from less than 4 full years of data, the number of years providing the mean is shown, in brackets, next to the site data in the graph. To avoid bias in the calculation of annual means, due to seasonality in the data (see later in Sect. Sect. 2.2.3). The disadvantage of a NaCl coating, however, is that it can only collect HNO3 and not the other acid gases. A third carbonate denuder is necessary in the sample train to collect and measure SO2, since SO2 is only partially captured and HCl cannot be measured on NaCl denuders (Tang et al., 2015(Tang et al., , 2018b. This explains the smaller SO2 concentrations reported by the FR-FgsP site, with break-through of SO2 (inefficiently captured by NaCl denuders) onto the aerosol filters resulting in larger particulate SO4 2concentrations than the Fr-Fgs site. For the SK04 site, measurement reproducibility for the 4 years of 30 parallel data for N and S component was good, with agreement ranging from 0.4 % (NH4 + ) to 15 % (SO4 2-). HCl and Na + and determinations were however more uncertain with differences of 21 and 28%, respectively. It has to be noted, however, that the concentrations of the two components were very low, at < 0.2 µg HCl m -3 and < 0.4 µg Na + m -3 . The differences in concentrations are therefore actually within ± 0.1 µg m -3 for HCl and within ± 0.2 µg m -3

3.5), years with incomplete
.for Na + . 35 A key feature in Figure 7 is the dominance of N over S species at most sites, when expressed as µg m -3 of the element. The mean percentage contribution of sum Nr (NH3-N, HNO3-N, NH4 + -N, NO3 --N) concentrations to the total mass of gas and aerosol species measured is 52 % (range = 24 -80%), twice as much as from sum S (SO2-S and SO4 2--S; mean = 23 %, range = 7 -53%) ( Figure 8). This is consistent with more substantial reductions in SO2 emissions (−72%) than achieved with NOx Secondary NH4 + particles are mainly in the 'fine' mode with diameters of less than 2.5 µm (PM2.5) and estimated to contribute 15 between 10 to 50 % of ambient PM2.5 mass concentration in some parts of Europe (Putaud et al., 2010, Schwartz et al., 2016).
An assessment by Hendriks et al. (2013) found that secondary NH4 + contributed 10 -20% of the PM2.5 mass in densely populated areas in Europe and even higher contributions in areas with intensive livestock farming. Concentrations of PM2.5 continue to exceed the EU limit values of 25 μg m -3 annual mean in large parts of Europe in 2017 (EEA, 2019). Particulate NH4 + data presented from the DELTA ® network therefore highlights the potential contribution of NH3 of agricultural origin to 20 fine NH4 + aerosols in PM2.5. The formation and transport of these secondary aerosols poses a serious risk to human health, since PM2.5 are linked with increased mortality from respiratory and cardiopulmonary diseases (AQEG, 2012).
A considerable fraction of the aerosol components measured was made up of sea salt (Na + and Cl -), with contributions from sum (Na + and Cl -) ranging from 4 % of the total aerosol loading at the inland Höglwald site in Germany (DE-Hog) to 43 % at 25 Dripsey (IE-Dri), a coastal site in Ireland (Figure 7). With the reduction in European emissions and concentrations of the gases SO2, NOx and NH3 for formation of NH4 + -containing aerosols, sea salt is therefore assuming a proportionate increase of the aerosol composition, consistent with observations from a recent European assessment of composition and trends in long-term EMEP measurements (EMEP, 2016). The concentrations of Ca 2+ and Mg 2+ were very low across the network, with values (mean of all sites = < 0.1 µg m -3 ) that were at or below method limit of detection (LOD = ~ 0.1 µg m -3 ). These data are also 30 considered to be under-estimated due to the DELTA particle sampling cut-off (~ PM4.5) and they were excluded from further assessment in this paper.

Comparisons with national gas emissions
In Figure 9, the annual averaged gas and aerosol concentrations of grouped sites from each country are plotted with the 35 corresponding national emission densities derived for NH3, NOx and SO2. The emissions data in the graphs are the 4-year averages for the period 2007 to 2010, expressed as emissions per unit area of the country per year (t km -2 yr -1 ) (see Sect. 2.6) and ranked in order of increasing emission densities. The error bars, where shown, is the range (min and max) of annual averaged concentrations of sites in each country. Where error bars are not visible, this indicates either that the country has measurement from just one site, or the range of concentrations measured are very close to the average. From the visual 40 comparisons, national mean measured concentrations in each country appear to scale reasonably well with the ranked emission densities. This is supported by further regression analyses which showed significant correlation between annual averaged https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.

<INSERT FIGURE 9>
<INSERT FIGURE 10> <INSERT  Figure 10B1) and HNO3 vs NOx (p < 0.05, Figure 10B2), but not for SO2 15 ( Figure 10B3, Supp. Figure S3). Some interesting features also emerged in the NH3 comparisons, with clustering of data according to ecosystem types ( Figure 10B1). The cropland sites have highest NH3 concentrations compared with gridded emissions (slope = 0.03, R 2 = 0.34, p = 0.08, n = 10), followed by grassland sites (slope = 0.01, R 2 = 0.87, p < 0.001, n = 10) ( Fig. 10B1, Supp. Figure S3). Forest (slope = 0.007, R 2 = 0.87, p < 0.001, n = 35) and semi-natural sites (slope = 0.004, R 2 = 0.25, p = 0.11, n = 11) are similar, with smaller NH3 concentrations compared with their gridded emissions. Since NH3 is 20 spatially heterogeneous even at a local sub-grid scale (e.g. Dragosits et al., 2002), the smaller concentrations at semi-natural and forest sites in grids with large emissions indicates these sites may be located further away from sources in the grid (Tang et al., 2018a;van Zanten et al., 2017). Dry deposition of NH3 is also largest to forests and semi-natural areas (larger Vd than to crops/grass ecosystem types, e.g. Smith et al., 2000;Flechard et al., 2011), which could also contribute to the smaller concentrations at higher emissions. Relationship between emissions and concentrations in the atmosphere is however complex, 25 influenced by other factors such as chemical interactions, variations in meteorological conditions and long-range transboundary import.
The lack of correlation between SO2 concentrations and gridded emissions ( Figure 10B3) suggests that a 0.1° x 0.1° grid may be too local a spatial scale for an emission-concentration comparison for SO2, as SO2 is likely to be highly localised with 30 emissions occurring from a smaller number of large point sources at an elevated height. Indeed, emissions in neighbouring grids surrounding each site are highly variable. For example, the 4-year averaged SO2 emissions in the 4 EMEP grids around the Italian San Rossore site (IT-SRo) varied between 0.47 to 610 kt SO2 yr -1 . Further analysis was also carried out comparing site mean concentrations against the averaged emissions of an extended number of EMEP grids (4 x grids) (Supp. Figure S4).
Since the analysis provided similar results to the comparisons with individual gridded emissions, they are not included for 35 further discussions in this paper. All regression plots and summary statistics for both comparisons (gridded emissions from single grids or from average of 4 grids) are provided in Supp. Figures S3 and S4.

Spatial variability across geographical regions
The form and concentrations of the different gas and aerosol components measured also varied according to geographic regions 40 across Europe (Figure 11). Smallest concentrations (with the exception of SO4 2and Na + ) were in Northern Europe https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.
(Scandinavia), with broad elevations across other regions. Gas-phase NH3 and particulate NH4 + were the dominant species in all regions ( Figure 11). NH3 showed the widest range of concentrations, with largest concentrations in Western Europe (mean = 2.4 NH3 m -3 , range = 0.2 -7.1 µg NH3 m -3 , n = 26 in 4 countries). By contrast, HNO3 and SO2 concentrations were largest in high NOx and SO2 emitting countries in Central and Eastern Europe (Sect. 3.3.3). Particulate SO4 2concentrations were however more homogeneous between regions, which may be attributed to atmospheric dispersion and long-range 5 transboundary transport of this stable aerosol between countries in Europe (Szigeti et al., 2015;Schwarz et al., 2016). In the aerosol components, the spatial correlations between NO3 -, NH4 + and NH3 illustrates the potential for NH3 emissions to drive the formation and thus regional variations in NH4 + and NO3aerosol. Particulate SO4 2concentrations in Northern Europe (Scandinavia) were similar to other countries, despite having the smallest SO2 and NH3 emissions and concentrations ( Figure   9). By comparison, the smaller particulate NH4 + and NO3concentrations in Northern Europe are consistent with smallest 10 emissions (NH3 and NOx) and concentrations of NH3 and HNO3 ( Figure 9). As discussed later in Sect. 3.4, the larger SO4 2concentrations reported in Northern Europe were flagged up as anomalous from ion balance checks (ratio of NH4 + :sum anions).

Comparisons by grouped components
In the following sections, variations in concentrations of the different gas and aerosol components according to ecosystem types (crops, grassland, forests and semi-natural) and in relation to emissions (NH3, NOx and SO2) are further discussed. For ease of interpretation, components are grouped as follows: reduced N (NH3, NH4 + ), oxidised N (HNO3, NO3 -), S (SO2, SO4 2-), HCl, Na + and Cl -. 20

Reduced N (NH3 and NH4 + )
Broad differences in NH3 concentrations are observed between the grouped sites, with the largest concentrations at cropland sites, as expected, as these are intensively managed agricultural areas dominated by NH3 emissions ( Figure 7A). Borgo Cioffi (IT-BCi) in an intensive buffalo farming region of Southern Italy provided the highest 4-year average of 8.1 µg NH3-N m -3 (cf. 25 group mean = 3.8 µg NH3-N m -3 , n = 10) (Table 4, Supp. Table S4). Next highest in this group are the German Gebesee (DE-Geb) and the Belgian Lonzee (BE-Lon) sites with 4-year average concentrations of 4.9 and 4.8 µg NH3-N m -3 , respectively (Supp . Table S4). At Gebesee, a decrease in NH3 concentrations was observed over the 4 year period, falling almost 2-fold from an annual mean of 8.8 µg NH3-N in 2007 to 4.8 µg NH3-N in 2010 (Supp . Table S4). Annual mean concentrations in 2008 (2.9 µg NH3-N m -3 ) and 2009 (3.2 µg NH3-N m -3 ) were similar, but smaller than in 2010. This illustrates the large inter-30 annual variability in concentrations that can occur even over a short time period. Variability between years may reflect changes in meteorological conditions on emissions from potential sources, with for example warmer, drier years increasing emissions and concentrations, contrasting with lower emissions and concentrations from the same source in a colder and wetter year.
Episodic pollution events can also have a large influence on the annual mean concentration, rather than the direct effects of changes in anthropogenic emissions over this short time scale. This suggests that for compliance assessment, an average over 35 several years would provide a more robust basis than individual years. The assessment of trends also needs a longer time series of at least 10 years (Tang et al., 2018a(Tang et al., , 2018bTorseth et al., 2012;van Zanten et al., 2017).
The markedly high concentrations of NH3 across the NEU network indicates that contributions by emission and deposition of 25 NH3 would be a major contributor to the effects of Nr on sensitive habitats. In comparing the annual averaged NH3 concentration with the revised UNECE 'Critical Levels' of NH3 concentrations (Cape et al., 2009), the lower limit of 1 µg NH3 m -3 annual mean for the protection of lichens-bryophytes were exceeded in 63 % of sites (40 sites in 15 countries) (Supp .   Table S5). Even the higher 3 µg NH3 m -3 annual mean for the protection of vegetation was still exceeded at 27 % of sites (17 sites in 10 countries) (Supp . Table S5). Most notably, all 4 sites from the Netherlands were in exceedance of both the 1 and 30 the 3 µg NH3 m -3 thresholds. The large concentrations in the Netherlands highlights the high levels of NH3 that semi-natural and forest areas are exposed to within an intensive agricultural landscape, where 117 out of the 166 Natura2000 areas were reported to be sensitive to nitrogen input (Lolkema et al., 2015). A recent assessment estimated that critical loads for eutrophication were exceeded in virtually all European countries and over about 62 % of the European ecosystem area in 2016 (EMEP, 2018). In particular, the highest exceedances occurred in the Po Valley (Italy), the Dutch-German-Danish border areas 35 and north-western Spain where the highest NH3 concentrations have been measured in this network. Since NH3 is preferentially deposited to semi-natural and forests (high Vd to these ecosystem types, Sutton et al., 1995), then NH3 will dominate dry NH3-N dry deposition and exert the larger ecological impact. In Flechard et al. (2011), dry NH3-N deposition from the first 2 years of NH3 measurement in the NEU DELTA ® network was estimated to contribute between 25 and 50% of total dry N deposition in forests, according to models. The fraction is larger in short semi-natural vegetation, since Vd for NH4 + and NO3is smaller 40 in short vegetation than to forests (Flechard et al., 2011). https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.

Comparison with NH3 data from the Dutch LML network
The 4-year averaged NH3 concentrations from the Dutch LML air quality network (see Sect.2.7.1) for the period 2007 to 2010 are plotted alongside the NH3 measurements made at the 4 Dutch sites in the DELTA ® network ( Figure 9A). The 4-year averaged concentrations from the 8 LML sites were between 1.5 to 15 μg NH3-N m -3 , highlighting the high concentrations and spatial variability in concentrations in the Netherlands. The mean NH3 concentrations measured at the 4 Dutch sites in the 5 DELTA ® network of 2.9 μg NH3-N m -3 (Horstermeer, NL-Hors; semi-natural) to 5.9 μg NH3-N m -3 (Cabauw, NL-Cab; grassland) were within the range of concentrations measured in the Dutch LML network.

Comparison with NH3 data from the UK NAMN network
The 4-year averaged NH3 concentrations calculated from the 72 sites in the NAMN (see Sect. 2.7.2) for the period 2007 to 10 2010 were smaller than the Dutch LML network, ranging from 0.05 to 6.7 μg NH3-N m -3 that are consistent with smaller NH3 emission from the UK ( Figure 9A). In a joint collaboration between the UK and Dutch networks, inter-comparison of NH3 measurements by the DELTA ® method (monthly) with the Dutch network AMOR wet chemistry system (hourly, van Zanten

Particulate NH4 +
Particulate NH4 + concentrations across the 64 sites were more homogeneous than NH3, varying over a narrower range between 0.13 µg NH4 + -N m -3 at Sodankylä (Finland, FI-Sod) and 2.1 µg NH4 + -N m -3 at Borgo Cioffi (Italy, IT-BCi) (Figure 7, Supp. 20 Table S6). By comparison, the difference in NH3 between the smallest (0.07 µg NH3-N m -3 at Lompolojänkkä, Finland, FI-Lom) and largest (8.1 µg NH3-N m -3 at Borgo Cioffi, Italy, IT-BCi) concentrations varied by a factor of 110 (Figure 7, Supp. Table S4). Secondary aerosols have longer atmospheric lifetimes and will therefore vary spatially much less than their precursor gas concentrations. While the concentrations of NH3 vary at a local to regional level owing to large numbers of sources at ground level, and high deposition in the landscape, NH4 + is less influenced by proximity to NH3 emission sources 25 and varies in concentration at regional scales (Sutton et al., 1998;Tang et al., 2018a).
In the UK, HNO3 data are also available on a wider spatial scale from the AGANet (Tang et al., 2018b, Sect. 2.7.2). The 4year average concentrations of HNO3 from 30 sites in the AGANet are plotted alongside the NEU HNO3 data from the 4 UK sites in its network in Figure 9B. The UK HNO3 data on the UK-AIR database (https://uk-air.defra.gov.uk/) have been 30 corrected for HONO interference with a 0.45 correction factor (see Tang et al. 2018b). For consistency in Figure 9B, the UK raw uncorrected HNO3 data are used for the present comparison. The 30-site mean (0.17 µg HNO3-N m -3 ) was higher than from just 4 UK sites in the NEU network (0.10 µg HNO3-N m -3 ). The range of concentrations were also wider, from 0.03 µg HNO3-N m -3 at a remote background site in Northern Ireland to 0.77 µg HNO3-N m -3 at a central London urban site, where interference from HONO and NOx in HNO3 determination is likely to be larger (Tang et al., 2015;2018b). 35 Like particulate NH4 + , NO3concentrations are also correlated with emission densities of NH3 (R 2 = 0.57, p < 0.001, n = 20) and NOx (slope = 0.15, R 2 = 0.44, p <0.01, n = 20), but not with SO2 (Table 3, Supp. Figure S2). Smallest NO3concentrations were again in Sweden, Norway and Finland with low NH3 and NOx emissions and also smallest concentrations of HNO3, SO2 and NH4 + in the network (Figure 9). Largest NO3concentrations was measured in the Netherlands with a mean of 0.92 µg 40 NO3 --N m -3 , compared with a network average of 0.39 µg NO3 --N m -3 ( Figure 9E, Supp. Table S8). The higher NO3concentrations correlated well with the high NH3, HNO3 and NH4 + concentrations in the Netherlands (Figure 9). This suggests that concentrations of NO3are linked to local formation of NH4NO3, which is dependent on concentrations of NH3 and HNO3, https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License. and also to the influence of meteorology on transport of NH4NO3 between countries on mainland Europe and export out of Europe. Countries in Scandinavia such as Sweden, Norway and Finland and in the British Isles are furthest from the influence of long-range transboundary transport from Europe, with concentrations of NH4NO3 that are smaller than on the continent.

Sulfur (SO2 and SO4 2-) 5
Annual averaged SO2 concentrations measured across the network were between 0.9 and 2.3 µg SO2-S m -3 ( Figure 9C, Supp.  Although the Bugac site is located in a grid with low emissions of all the gases, the higher SO2 (1.2 µg S m -3 ), together with 25 elevated NH3 (2.6 µg N m -3 ) and HNO3 (0.3 µg N m -3 ) concentrations measured at this site suggests that it is likely to be affected by proximity to sources. This contrasts with the BKFores site in the Czech Republic (CZ-BK1) which had smaller NH3 concentrations due to its location away from sources.
Following emission, SO2 disperses and undergoes chemical oxidation in the atmosphere to form SO4 2both in the gas phase 30 and in cloud and rain droplets (Baek et al., 2004;Jones and Harrison, 2011). Particulate SO4 2produced is generally associated with NH4 + and NO3 -(see Sect. 3.4). The regional pattern of SO4 2was similar to, and correlated well with, particulate NH4 + and NO3 - (Figure 9G), suggesting well-mixed air on the continent, since (NH4)2SO4 is stable and long-lived. Countries in the British Isles (UK and Ireland) and in Scandinavia (Sweden, Norway, Finland) have smaller concentrations of SO4 2-(Supp. Table S10). They are located far enough away from sources and activities on continental Europe such that they are less 35 influenced by the emissions from central Europe.
As discussed earlier, particulate NH4 + and NO3concentrations were smallest in the Scandinavian countries, which corresponded with low emission densities of the precursor gases NH3 and NOx. By analogy, since these countries also have the lowest emission densities of SO2 ( Figure 9C), then particulate SO4 2concentrations would be expected to be similarly low. 40 Particulate SO4 2in Finland and Norway (mean = 0.34 µg SO4 2--S m -3 ) and Sweden (mean = 0.37 µg SO4 2--S m -3 ) were however comparable with concentrations on mainland Europe (range = 0.33 to 1.0 µg SO4 2--S m -3 ) and larger than the UK (0.18 µg SO4 2--S m -3 ) and Ireland (0.24 µg SO4 2--S m -3 ) ( Figure 9F). An ion balance check on the ratio of equivalent concentrations of https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License. NH4 + to the sum of NO3and SO4 2-(see next section 3.4) was less than 0.5. Since NH4 + is a counter-ion to NO3and SO4 2formation, the imbalance suggests that SO4 2concentrations may be over-estimated at the sites in Sweden, Norway and Finland.

HCl, Cland Na +
The average concentrations of HCl across the network were of low magnitude, with limited variability, ranging from 0.07 in 5 Russia to 0.36 µg HCl-Clm -3 in Portugal (Figure 9D). At a site level, HCl concentrations varied between 0.06 at Renon (Italy, IT-Reninland location) to 0.48 µg HCl-Clm -3 at Espirra (Portugal, PT-Espcoastal location) (Supp. Table S11). In the UK AGANet network, the highest concentrations of HCl were found in the source areas in SE and SW of England, and also in central England, north of a large coal-fired power station (Tang et al., 2018b). HCl emissions and concentrations in the atmosphere are mostly derived from combustion of fossil fuels (coal and oil), biomass burning and from the burning of 10 municipal and domestic waste in municipal incinerators (Roth and Okada 1998;McCulloch et al., 2011;Ianniello et al., 2011).
Several manufacturing processes, including cement production also emits HCl (McCulloch et al., 2011). At coastal sites, HCl released from the reaction of sea salt with HNO3 and H2SO4 can be a significant source (Roth and Okada 1998;Keene et al., 1999;McCulloch et al., 2011;Ianniello et al., 2011). UK is the only country with available HCl emission estimates (https://naei.beis.gov.uk/data/). Emissions of HCl in the UK (mainly from coal burning in power stations) have declined to 15 very low levels, from 74 kt in 1999 to 5.7 kt in 2015. The 4-year averaged emission density for HCl for the period 2007 to 2010 was just 0.05 tonnes HCl-Clkm -2 yr -1 , although HCl emissions could still pose a threat to sensitive habitats close to sources (Evans et al., 2011). The low HCl concentrations measured in the network would suggest that the shift in Europe's energy system from coal to other sources has contributed to low HCl emissions (UK) and concentrations (observed across the network). 20 Particulate Clon the other hand is predominantly marine in origin, with sea salt (NaCl) as the most significant source (Keene et al. 1999). Molar concentrations of Cland Na + are seen to be similar in most countries, demonstrating close coupling between the two components ( Figure 9H). Largest concentrations of Na + and Cloccurred at coastal countries such as the UK, Ireland, Netherlands and Portugal, with the highest of country-averaged annual concentrations of 1.6 µg Clm -3 and 0.9 µg Na + m -3 25 from Ireland (Supp. Tables S12 and S13). Data from the 30 sites in the UK AGANet network showed a wider range of Cland Na + concentrations ( Figure 9H), with the highest 4-year annual averaged concentrations of 3.8 µg Cl m -3 and 2.0 µg Na + m -3 from the coastal Lerwick monitoring site on the east coast of the Shetland Islands, exposed to the North Atlantic.
Further away from the coastal influence of marine aerosol, the smallest concentrations of Cland Na + were measured in land-30 locked countries such as Germany (mean of all sites = 0.27 µg Clm -3 and 0.15 µg Na + m -3 ). Concentrations in Hungary, Poland, the Czech Republic and Russia were also low, but inferences about these countries are necessarily limited by measurements at a single site in each of these countries. At coastal sites in Norway (NO-Bir) and Sweden (SE-Nor and SE-Sk2), the very low particulate Clconcentrations (< 0.1 -0.3 µg m -3 ), and high Na:Cl molar ratios (3 -5) are anomalous. It is possible for sea salt to be depleted in Cl-(through the loss of HCl gas) by the reaction of NaCl particles with atmospheric acids 35 (Finalyson-Pitts and Pitts, 1999;Keene et al., 1999), leading to high Na:Cl ratios for sea salts transported over long distances.
The coastal locations of these sites (Figure 2) suggests that they are more likely to be influenced by freshly generated marine aerosols (cf. coastal sites in UK and Ireland), and larger concentrations of sea salt (Na + and Cl -) and a 1:1 relationship between Na + and Clare expected. The Clconcentrations are likely to be under-estimated at these sites (see Sect. 3.2.3) and further discussed in the next section (Sect. 3.4). 40 https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.

Correlations between gas and aerosol components
Regression analyses was carried out between the mean molar equivalent concentrations of all inorganic gas and aerosol components measured at each site (n = 66; Fr-FgsP and UK-AmoP excluded) in the NEU network, with summary statistics provided in Table 5. With the exception of SO2 vs HCl (R 2 = 0.05, p > 0.05), the gases were positively correlated with each other, possibly due to similarities in the regional distribution of their emissions and concentrations. Comparing the mean molar 5 concentrations of NH3 with SO2 and HNO3 showed that NH3 was on average 6-fold and 7-fold higher, respectively, whereas molar concentrations of SO2 and HNO3 were similar (Table 6, Figure 11). The molar ratio of NH3 to the sum of all acid gases (SO2, HNO3 and HCl) was on average 3 (Table 6, Figure 11), confirming that there is a surplus of the alkaline NH3 gas to neutralise the atmospheric acids in the atmosphere, similar to that observed in the UK (Tang et al., 2018b). With the more substantial decline in emissions of SO2, compared with a more modest reduction in NOx, the concentrations of SO2 are at a 10 level where it is no longer the dominant acid gas, such that HNO3 and HCl are together contributing a larger fraction of the total acidity in the atmosphere in the present assessment.
Correlation between NH4 + and the sum of anions (NO3 -+ SO4 2-) is an important point of discussion (Table 7), as the ion balance serves as a quality check for the aerosol measurement. Due to some outliers in the comparison, the correlation between NH4 + 5 and SO4 2-(R 2 = 0.28, Figure 12B) is weaker than between NH4 + and NO3 -(R 2 = 0.75, Figure 12C, Table 7). The outliers were measurements made by NILU and CEAM, although these vary according to monitoring locations. The NILU laboratory made DELTA ® measurements for 16 sites in 6 different countries (Belgium, Denmark, Finland, Norway, Sweden and Switzerland).
At 3 sites (Kaamanen FI-Kaa, Laegern CH-Lae, Oensingen CH-Oe1), the ion balance of equivalent concentrations of NH4 + :sum (NO3 -+ SO4 2-) was 1.0, whereas the ratios at the other 13 sites were between 0.4 and 0.7. The CEAM laboratory 10 made measurements for all 3 sites in Spain. For CEAM, the ion balance ratio at Vall de Aliñá (ES-VDA) was 1, whereas the other 2 sites had ratios of 0.5 and 0.6.
Removal of the outlier NILU (7 out of 16) and CEAM (1 out of 3) data points with ion balance ratio < 0.5 improved both the slope (new slope = 0.90) and correlation (new R 2 = 0.78) ( Figure 12C). This indicates either an over-read of the anions (NO3 -15 , SO4 2-) or under-read of NH4 + concentrations by the two laboratories at some sites. Results reported by NILU in the DELTA ® field inter-comparisons (Sect. 3.2) showed that, with the exception of a few high NH4 + and NO3readings, there was on average no overall bias in the NH4 + , NO3or SO4 2measurements by the NILU laboratory that could account for the high SO4 2outliers in the regression (Figure 12). The ion balance checks suggest possible over-read and increased uncertainty in the SO4 2- The regression of Na + and Clalso showed the majority of data points close to the 1:1 line, but with a small group of outliers below the 1:1 line from the CEAM and NILU laboratories ( Figure 12F). Both laboratories performed well in laboratory PT 25 schemes (Sect. 3.1), with more than 80% of reported data agreeing within ± 10% of reference values in both Na + and Cl -, with no bias in the analytical method. The outliers in the ion balance therefore suggests some problems with Na + and Cldetermination on the DELTA ® aerosol filters. Na + and Cldata for some of the field DELTA ® inter-comparisons were omitted from submissions by CEAM and NILU, and submitted data were in poor agreement with other laboratories (Sect. 3.2). Further regression analyses were carried out on individual monthly data, with sites grouped according to measurements made by each 30 of the seven laboratories (Supp. Figure S5). Regressions for CEAM and NILU show the vast majority of data points below the 1:1 line, indicating a systematic under-estimation of particulate Clconcentrations. The other 5 laboratories (INRAE, MHSC, SHMU, UKCEH and VTI) all have data points close to the 1:1 line, with larger scatter both above and below the 1:1 line at lower concentrations. In Figure 12F, a new regression line has therefore also been fitted where outlier data with Na:Cl ratios > 2 from NILU (13 out of 16 sites) and CEAM (all 3 sites) have been removed. Exclusion of the outlier data points provided 35 a regression line that is not significant different from unity (slope = 1.02), with a R 2 value of 0.95 (p < 0.001). The near 1:1 relationship between particulate Na + and Clis consistent with their origin from sea salt (NaCl).
The ion balance checks, together with the regular PT exercises and field inter-comparisons therefore provided the platform against which to assess data quality and comparability of measurements between laboratories. This shows that overall, with 40 the exception of a few identified outlier measurements, the laboratories are performing well and providing good agreement. https://doi.org /10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.

Seasonal variability in gases and aerosol
The time series of monthly averaged concentrations for the period 2006 to 2010 have been plotted to examine seasonality in the different gas and aerosol components according to ecosystem types (crops, grassland, semi-natural and forests) ( Figure 14) and geographical regions (Figure 15). Distinct seasonality were observed in the data, influenced by seasonal changes in emissions, chemical interactions and the influence of meteorology on partitioning between the main inorganic gases and 5 aerosol species.

NH3
Distinctive and contrasting features in the seasonal cycle are observed, with largest concentrations at cropland sites and smallest at semi-natural and forest sites ( Figure 14A). Similar to that observed in the annual mean concentrations (Figure 9, 11), the monthly concentrations are also smallest in Northern Europe and largest in Western Europe ( Figure 15A).

15
Semi-natural sites: There are two distinct peaks in the seasonal cycle of grouped semi-natural sites, in April (mean = 2.2 µg NH3 m -3 , n = 12) and in July (mean = 1.9 µg NH3 m -3 , n = 12) ( Figure 14A). Since these sites are located away from agricultural sources, the seasonality in NH3 concentrations is mostly governed by changes in environmental conditions and regional changes in NH3 emissions. The differences in concentrations between the summer and winter at these sites was by a factor of 3, with smallest 20 concentrations in wintertime (Dec and Jan) when low temperatures and wetter conditions decrease NH3 emissions from regional agricultural sources, while favouring a thermodynamic shift from gaseous NH3 to the aerosol NH4 + phase. Conversely, warm, dry conditions in summer increases surface volatilization of NH3 from low density grazing livestock and wild animals, and favour a thermodynamic shift to the gaseous (NH3) phase, producing the summer peak. Vegetation is another potential source at these background sites under the right conditions (Flechard et al., 2013;Massad et al., 2010). A complex interaction 25 between atmospheric NH3 concentrations and vegetation can lead to both emission and deposition fluxes known as "bidirectional exchange", dependent on relative differences in concentrations. This process is controlled by the so-called "compensation point", defined as the concentration below which growing plants start to emit NH3 into the atmosphere (Flechard et al., 1999;Massad et al., 2010;Sutton et al., 1995). At sites distant from intensive farming and emissions, the bidirectional exchange with vegetation will partly control NH3 concentrations. Inclusion of bi-directional exchange in dispersion 30 modelling of NH3, by incorporating a 'canopy compensation point' is shown to improve model results for NH3 concentrations in remote areas (e.g. Smith et al., 2000;Flechard et al., 1999Flechard et al., , 2011Massad et al., 2010). The larger peak in April at these sites on the other hand suggests the influence of emissions from agricultural sources, e.g. from land spreading of manures.

Forest sites: 35
The average seasonal cycle from the forest sites is similar to that of the semi-natural sites, but diverged over the summer months ( Figure 14A). Here, the seasonal profile is characterised by the absence of any peaks in summer, with concentrations plateauing between May and August. Studies have shown that atmospherically deposited N is taken up by forest canopies, since growth in forest ecosystems is commonly limited by the availability of N (Sievering et al., 2007) and tree canopies are a potential sink for atmospheric NH3 (Fowler et al., 1989;Theobald et al., 2001). The capture and uptake of NH3 during the 40 growing seasons over the summer period could therefore account for the absence of a summer peak in NH3 concentrations at forest monitoring sites, although a similar effect would also be expected for semi-natural sites.

Cropland sites:
Fertilizers and arable crops are significant sources of NH3 emissions and concentrations in an intensive agricultural landscape.
Sites in this group showed considerably higher monthly mean monitored NH3 concentrations than the other groups ( Figure   14A). A more complex seasonal pattern can be seen, with three peaks in NH3 concentrations. Concentrations here are also 5 lowest in the winter, although the wintertime concentrations are 3 times larger than semi-natural and forest sites, reflecting the elevated regional background in NH3 concentrations located within agricultural landscapes. This rises rapidly with improving weather conditions and peaks in the spring to coincide with the main period for manure spreading and fertiliser application before the sowing of arable crops (Hellsten et al., 2007). The distinct springtime maxima in NH3 also reflects implementation of the Nitrates Directive (91/676/EEC), which prohibits manure spreading in winter. In summer, the second peak in NH3 10 concentrations may be associated with increased land surface emissions promoted by warm, dry conditions, and possibly from the application of fertilisers. The smaller autumn peak is also expected to be related to seasonal farming activities/manure spreading. The key drivers for seasonal variability in NH3 concentrations at crops sites are therefore a combination of seasonal changes in agricultural practices (e.g. timing of fertiliser/manure applications) and climate that will affect emissions, concentrations, transport and deposition of NH3. 15 Grassland sites: An additional major source of NH3 in this group of sites is expected to come from grazing emissions and housed livestock (e.g. cattle). Concentrations in this group of sites were generally 2 -3 times larger than semi-natural sites ( Figure 13A), attributed to the increased emissions and concentrations from livestock (Hellsten et al., 2007). The spring peak is related to the 20 practice of fertiliser and manure being spread on grazing fields to aid spring grass growth, which will be cut for hay and silage later in the year. NH3 concentrations in June and July are smaller than in spring or late summer, possibly because grass will be actively growing with possible uptake and removal of NH3 from the atmosphere. The concentrations are also larger in summer than winter, with warmer conditions promoting NH3 volatilization and thermodynamic shift to the gas phase.

25
European regions: The seasonal profiles of NH3 for Central and Western European regions were similar, characterised by a large peak in spring that is likely to be agriculture-related ( Figure 15A), as observed at cropland sites ( Figure 14A). While the peak concentrations in both regions are of comparable magnitude (Central = 2.6 µg NH3 m -3 , Western = 2.8 µg NH3 m -3 ), winter concentrations in the Centre Europe (0.6 µg NH3 m -3 ) were three times smaller than the West (1.5 µg NH3 m -3 ). This may be related to either 30 lower regional background in NH3 concentrations and/or suppressed emissions in colder temperature of Central Europe.in winter. By contrast, Eastern and Southern European regions have a broad peak in summer, although the Eastern region also has a second peak in October (likely agriculture related). Smallest concentrations were found in Northern Europe with the lowest NH3 emissions (Figure 9). The three peaks in the profile shows elevated concentrations in summer driven by warming temperatures, with the spring and autumn peaks attributed to influence from NH3 emissions from agricultural sources. 35

HNO3
The seasonal distribution in HNO3 is similar between the different ecosystem groups, varying only in magnitude of concentrations ( Figure 14C) and reflects the secondary nature of this component that is formed from oxidation of NOx (Fahey et al., 1986;ROTAP, 2012). HNO3 concentrations in the crops group are up to 2 times larger than the grassland group, while 40 the smallest concentrations are in the semi-natural group. This is likely related to proximity of sites in the different groups to combustion sources. A weak seasonal cycle is seen in the secondary HNO3 air pollutant in all cases, with slightly higher https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License. concentrations in late winter, spring and summer and smallest in March and November. The reaction of NO2 with the OH radical is an important source of HNO3 during daytime, whereas N2O5 hydrolysis is considered an important source of HNO3 at night time (Chang et al., 2011). Larger HNO3 concentrations in summer are therefore from increased OH radicals for reaction with NO2 to form HNO3. Similarly, higher concentrations of ozone in spring in Europe (EMEP, 2016) can potentially increase HNO3 concentrations in springtime. Conversely, HNO3 concentrations are lower in winter when oxidative capacity is less. 5 Seasonal variability in HNO3 will also be influenced by gas-aerosol phase equilibrium. In the atmosphere, HNO3 reacts reversibly with NH3 forming the semi-volatile NH4NO3 aerosol if the necessary concentration product [HNO3].
[NH3] is exceeded (Baek et al., 2004;Jones and Harrison et al., 2011). Because of this process, the prime influences upon HNO3 concentrations at sites where NH4NO3 is formed are expected to be ambient temperature, relative humidity and NH3 10 concentrations that affect the partitioning between the gas and aerosol phase (Allen et al., 1989;Stelson and Seinfeld, 1982).
The availability of surplus NH3 in spring (Sect. 3.5.1) would tend to reduce HNO3 and increase NH4NO3 formation, which is reflected in the reduced HNO3 concentrations observed in March when NH3 is at a maximum. In summer, warmer, drier conditions promotes volatilisation of the NH4NO3 aerosol, increasing the gas phase concentrations of HNO3 and NH3 relative to the aerosol phase. Seasonality in HNO3 is therefore complex, related to traffic and industrial emissions, photochemistry and 15 HNO3:NH4NO3 partitioning.
An analysis of the same data grouped according to geographical regions revealed distinctive cycles in HNO3 in Eastern and Southern Europe ( Figure 15C). These two regions showed highest concentrations in summer and smallest in winter, consistent with enhanced photochemistry in warmer, sunnier climates and thermodynamic equilibrium favouring gas phase-HNO3 20 ( Figure 15C). Summertime peak concentrations in NH3 were also observed in these 2 regions ( Figure 15A). In comparison, the seasonal profiles of HNO3 in other regions were similar to that described for different ecosystem types ( Figure 14C).

SO2
Seasonality in SO2 show concentrations peaking in winter at most sites ( Figure 14E), except in Southern Europe where the 25 peak appeared in summer ( Figure 15E). Increased SO2 emissions from combustion processes (heating) in the winter months, coupled to stable atmospheric conditions can result in build-up of concentrations at ground level, thereby contributing to the peak wintertime concentrations. The largest winter concentrations in Central and Eastern regions exceeded summer values on average by a factor of 4, compared with smaller differences in other regions ( Figure 15E). Enhanced oxidation processes in summer also tend to further reduce concentrations of SO2 through the oxidation of SO2 to H2SO4 (Saxena and Seigneur, 1987;30 Sickles and Shadwick, 2007;Paulot et al., 2017). In Southern Europe, the seasonal cycle have winter minima and summer maxima instead, likely from increased combustion sources to meet energy demands for air-conditioning over the hot summer months. It was shown earlier in Section 3.4 that SO2 was spatially correlated to HNO3; differences in relative concentrations between the different ecosystem groups ( Figure 14E) is thus also likely related to relative distance from emission sources.

NH4 + , NO3and SO4 2-
The seasonal profiles of particulate NH4 + (Figures 14B and 15B) were mirrored by particulate NO3 - (Figures 14D and 15D) in all groups, demonstrating temporal, as well as regional (see Sect.3.3.5) correlation between these two components. Since NH4NO3 is more abundant than (NH4)2SO4, the seasonality of NH4 + is likely to be influenced more by the temperature and humidity dependence of the semi-volatile NH4NO3, than by the stable (NH4)2SO4. In summer, warmer and drier conditions 40 promotes the dissociation of NH4NO3, decreasing particulate phase NH4NO3 relative to gas phase NH3 and HNO3. This process https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License. accounts for the summertime minima in NH4 + (Figures 15B and 15B) and NO3 - (Figures 14D and 15D). Conversely, cooler temperatures and higher humidity conditions in winter, spring and autumn shift the equilibrium to the aerosol phase, with observed peaks in concentrations of NH4 + and NO3 -. Since NH3 concentrations are also generally higher in spring than in autumn ( Figure 14A, 15A), the increased availability of NH3 in this period contributes towards the higher concentrations of NH4NO3 in spring than in autumn. In winter, the combination of NH4NO3 remaining in the aerosol phase, combined with the 5 stable conditions that can often develop, maintains high concentrations of NH4 + and NO3in the atmosphere. The peak in NO3in Southern Europe was in February only, compared with broader peaks (Feb-April) in other regions ( Figure 15D) which may reflect differences in climatic conditions. In Figures 14H and 15H, the ratio of the molar equivalent concentrations of NO3to sum (NO3 -+ SO4 2-) are plotted. The ratios were highest in spring and autumn, and smallest in summer, lending support to the importance of NH4NO3 in controlling the seasonality of NH4 + . 10 In the seasonal profiles for particulate SO4 2-, clear summer maxima and winter minima were provided by sites in Southern and Eastern Europe ( Figure 15F). The peaks occurred at different times, in July (Southern Europe) and in August (Eastern Europe) ( Figure 15F) and coincided with the timing of corresponding peaks in NH3 concentrations ( Figure 15A), illustrating the importance of NH3 in driving the formation of the stable (NH4)2SO4. Since (NH4)2SO4 is formed through the preferential and 15 irreversible reaction between the precursor gases (Bower et al., 1997), particulate SO4 2concentrations will be governed by the availability of NH3 and H2SO4 (from oxidation of SO2). As discussed earlier, SO2 concentrations in Southern Europe have a different seasonal cycle from other regions, with higher concentrations in summer than in the winter months ( Figure 15E).
Although the seasonal cycle for Eastern Europe showed smallest SO2 concentrations in the summer, the summer minima here (mean = 1.3 µg SO2 m -3 ) are in fact larger than the summer peak in Southern Europe (mean = 1.1 µg SO2 m -3 ) and 20 concentrations in other regions (0.4 -1.0 µg SO2 m -3 ). Enhanced summertime concentrations in HNO3 were observed in these two regions ( Figure 15B) which also suggests potentially increased oxidative capacity for more of the SO2 to be converted H2SO4 (Sect. 3.5.3). The ready availability of both SO2 (and conversion to H2SO4) and NH3 ( Figure 15A) in Southern and Eastern regions in this period thus coincide to produce the summer peak in particulate SO4 2-.

25
In other regions (Central, Northern, Western), formation of (NH4)2SO4 will be limited by the availability of SO2 which is lowest in summer (Figures 15E) formation of (NH4)2SO4 is also limited in winter. This accounts for the higher concentrations of particulate SO4 2concentrations in winter and in early spring in these regions ( Figure 15F). 30

HCl, Cland Na +
The concentrations of HCl measured at all sites, in all groups, were very small, with monthly mean concentrations varying between 0.1 and 0.3 µg HCl m -3 ( Figures 14G and 15G). There is no discernible seasonality in the data, which suggests either sites in the network are not affected by any large sources of HCl, or that small differences between months are not detectable 35 due to measurement uncertainties at the very low concentrations (method limit of detection ~ 0.1 µg HCl m -3 for monthly sampling). By contrast, Cl - (Figures 14I and 15I) has a distinctive seasonal cycle with higher concentrations in the winter months than summer, similar to that of Na + (Figures 14J and 15J). The temporal correlation in the data therefore lends further support that Na + and Clin the measurements are mainly sea salt (see also spatial correlation in Sect. 3.4). The highest concentrations of Na + and Clduring winter months would be consistent with increased generation and transport of sea salt 40 generated by more stormy weather from marine sources during those periods (O'Dowd and de Leeuw, 2007). https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.

Bulk wet deposition measurements
Annual mean wet deposition of chemical species measured at the NEU bulk sampling sites was estimated by combining measured concentrations with annual precipitation. Site changes also occurred during the operation of the bulk wet deposition network, with some sites closed and new sites added. At Mitra (PT-Mi3), contamination of the rain samples from bird strikes resulted in the rejection of a large proportion of the monthly data and this site was excluded from the data analysis. In total, 12 5 sites provided 2 years of monthly data, with a further 5 sites providing 1 year of monthly data over the period 2008 to 2010. Due to differences in start and end dates for bulk measurements between the sites, the annual mean data derived are for 12 month periods or 2 x 12 month periods, and not from calendar years.

<INSERT FIGURE 16> 10
Annual mean wet deposition data for the 17 sites from 6 countries (Belgium, France, Germany, Italy, Poland, Spain and Switzerland) are summarised in Figure 16. Using Na + as a tracer for sea-salt (Keene et al,. 1986), nss-SO4 2concentrations were also estimated from the total SO4 2-(see Sect. 2.2.2) and are included for comparison. Since the measurements were made at a limited number of sites across Europe, there is insufficient information to make inferences about spatial differences in 15 concentrations. Detailed assessments of extensive precipitation chemistry across Europe are made elsewhere, for example from the EMEP wet deposition networks (EMEP, 2016;Torseth et al., 2012). What the NEU bulk network data clearly shows is that Nr components in rain also exceed that of S (Figure 16), as was observed in the atmospheric data. The mean proportional contribution of total N (NH4 + and NO3 -) to the sum total of all wet deposited species measured (by mass) was 19% (range = 3 -39%), compared with a smaller 9 % (range = 1 -19%) contribution from nss-SO4 2-(Supp. Table S14). Wet deposited N 20 (NH4 + and NO3 -) was on average 2 times higher than nss-SO4 2-, similar to that seen in the relative proportion of total Nr (sum of NH3, NH4 + , HNO3, NO3 -) to total S (sum of SO2, SO4 2-) in the atmospheric data (Sect. 3.3.5). Similar to the atmospheric data (Sect. 3.3.5), a considerable fraction of the wet deposited components was made up of sea salt (Na + and Cl -), with the sum of Na + and Clcontributing on average 50% of the total wet deposited components (range = 20 -84 %, n = 17). Contributions by the other base cations Ca 2+ and Mg 2+ gave a further 20 % (range = 8 -41 %, n = 17) (Supp . Table S14). 25 The intention of the bulk network at the outset was to provide wet deposition data at DELTA ® sites that do not already have such measurements on site. The wet deposition data on NH4 + and NO3 -, combined with a wider precipitation chemistry dataset (e.g. from EMEP and other national precipitation networks) was used to estimate total Nr deposition to a site (Flechard et al., 2011;. Together, the dry (DELTA ® network) and wet Nr estimates (NEU bulk network, combined with data from other 30 national precipitation chemistry networks) are used to compare with EMEP models and to examine the interactions between Nr supply and greenhouse gas exchange at the NEU DELTA ® sites, presented in a separate paper by Flechard et al. (2020).

Implications for a chemical climate dominated by NH3 and NH4NO3 in Europe
International agreements such as the UNECE Convention on Long-Range Transboundary Air Pollution (CLRTAP 1999 Gothenburg Protocol, amended  level where it is no longer the dominant acid gas, such that HNO3 and HCl are together contributing an equal or larger fraction of the total acidity in the atmosphere in the present assessment ( Figure 11). However, SO2 (by mass) has a higher acidification potential (1 kg SO2 = 1.00 kg eq. SO2 than NOx (1 kg NO2 = 0.70 kg eq. SO2 (see Hauschild and Wenzel, 1998), so SO2 will 10 remain important in contributing to exceedances of critical loads for acidification, estimated to be exceeded in 5 % of the European ecosystem area in 2015 (EEA, 2019).
Emissions of NH3 in Europe have increased by about 3% from the agricultural sector between 2013(EEA, 2018 and abatement measures are likely to be needed to meet emission targets set for NH3. (Sutton and Howard, 2018). Thresholds for 15 atmospheric concentrations and deposition of Nr components to semi-natural habitats were exceeded in 63% of the EU-28 ecosystem area in 2016 (EMEP, 2018). In deposition models, oxidised nitrogen species currently included are HNO3, NO2 and aerosol nitrate (NO3 -), with deposition velocities dependent on meteorology and vegetation characteristics (e.g. Flechard et al., 2011). NH3 is the most important individual term in the calculation of total N dry deposition, along with NH4 + and HNO3 dry deposition and wet deposited NH4 + and NO3 -. Although NO2 (not measured in NEU DELTA ® network) will also provide a 20 relevant contribution to dry N deposition, it will (especially for rural semi-natural and forest ecosystems) be smaller than for NH3, based on rather small deposition velocities for NO2 (Smith et al., 2000).
The annually averaged data also show exceedance of the Critical Levels for annual mean NH3 concentrations of 1 and 3 µg NH3 m -3 for the protection of lichens-bryophytes (including ecosystems where they are important for integrity) and other 25 vegetation, respectively, at many of the sites (62 % > 1 µg NH3 m -3 and 27 % > 3 µg NH3 m -3 ) (Supp. Table S5). The widespread exceedance of the Critical Levels for NH3 concentrations across Europe represents an ongoing threat to the integrity of sites designated under the EU Habitats Directive (EU, 1992). In tandem, the growing relative importance of NH3 and NH4 + to total acidic and total nitrogen deposition indicates that strategies to tackle acidification and eutrophication will also need to include measures to abate emissions of NH3. 30 The agricultural sector makes up 92% of the total estimated NH3 emission in Europe (EEA, 2019), with 80 % of that generated by less than 10 % of the farms, so that the largest emission reduction potential could be attained by targeting the small number of industrial-scale farms (Maas and Greenfelt, 2016). A modelling study by Backes et al. (2016) suggested a halving of NH3 emissions could deliver a 24% reduction in total PM2.5 concentrations in northwest Europe, driven mainly by reduced formation 35 of NH4NO3 and that targeting emission reductions during winter had a larger effect than at other times of the year. In recognising the need to tackle NH3, the UNECE has published a guidance document and code of good agricultural practice (COGAP) for reducing NH3 emissions (Bittman et al., 2014), which has also been adopted in the EC NECD and by the UK government in its Clean Air Strategy (Defra, 2019). https://doi.org /10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License. Conclusion The NitroEurope DELTA ® network has provided for the first time a comprehensive quality-assured multi-annual dataset on reactive gases (NH3, HNO3, SO2, HCl) and aerosols (NH4 + , NO3 -, SO4 2-, Cl -) across the major gradients of pollution, ecosystem type and climatic zones of Europe. The harmonised measurement approach of monthly time-integrated monitoring with a simple low-cost DELTA ® method represented an effective use of resources, making it possible to operate a network with a 5 common measurement method across multiple laboratories at a large number of sites. At the same time, the concurrent measurement of the gas and aerosol components permitted an assessment of the atmospheric composition, spatial and seasonal characteristics in the gas and aerosol phase of these components. The dataset has also been used to develop estimates of sitebased Nr dry deposition fluxes across Europe, including supporting the development and validation of long-range transport models. Combined with estimates of wet deposition (NEU bulk wet deposition network and data by other networks) to these 10 sites, an assessment of the interactions between N supply and greenhouse gas exchange was addressed in a separate paper by Flechard et al. (2020), using Nr and CO2 flux data from the co-location of the NEU DELTA ® with CarboEurope Integrated Project sites.
Two key features have emerged in the data. The first is the dominance of NH3 as the largest single component at the majority 15 of sites, with molar concentrations exceeding that of HNO3 and SO2, combined. Changes in the relative concentrations of these gases across Europe suggests that the deposition rates of SO2 and NH3 will increasingly be controlled by the molar ratio of NH3 to combined acidity (sum of SO2, HNO3 and HCl) and deposition models should take these changes into account. As expected, the largest NH3 concentrations were measured at cropland sites, in intensively managed agricultural areas dominated by NH3 emissions. The smallest concentrations were at remote semi-natural and forest sites, although concentrations in the 20 Netherlands, Italy and Germany were up to 45 times larger than similarly classed sites in Finland, Norway and Sweden (< 0.6 µg NH3-N m -3 ), illustrating the high NH3 concentrations that sensitive habitats are exposed to in intensive agricultural landscapes in Europe.
Temporally, peak concentrations in NH3 for crops and grassland sites occurred in spring, reflecting the implementation of the 25 EU Nitrates Directive that prohibits winter manure spreading. The spring agriculture-related peak was seen even at seminatural and forest sites, highlighting the influence of NH3 emissions at sites that are more distant from sources. Summer peaks, promoted by increased volatilisation of NH3, but also by gas-aerosol phase thermodynamics under warmer, drier conditions were seen in all ecosystem groups, except at Forest sites. The seasonality in the NH3 concentrations captured for the different groups is important, both for identifying periods when abatement might be targeted and for model development . 30 Seasonality in the other gas and aerosol components is also driven by changes in emission sources, chemical interactions and by changes in environmental conditions influencing partitioning between the precursor gases (SO2, HNO3, NH3) and secondary aerosols (SO4 2-, NO3 -, NH4 + ). Seasonal cycles in SO2 were mainly driven by emissions (combustion), with concentrations peaking in winter, except in Southern Europe where the peak occurred in summer. HNO3 concentrations were more complex, 35 as affected by photochemistry, meteorology and by gas-aerosol phase equilibrium. Southern and eastern European regions provided the clearest seasonal cycle for HNO3, with highest concentrations in summer and smallest in winter, attributed to increased photochemistry in the summer months in hotter climates. In comparison, a weaker seasonal cycle is seen in other regions, with marginally elevated concentrations in late winter, spring and summer and smallest in March and November.
Increased ozone in spring is likely to enhance oxidation of NOx to HNO3 for forming the semi-volatile NH4NO3 by reaction 40 with a surplus of NH3. Cooler, wetter conditions in spring also favour the formation of NH4NO3 and more of the NH4NO3 remains in the aerosol or condensed phase. This accounts for the higher concentrations of NH4 + and NO3in spring and the https://doi.org /10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License. absence of a HNO3 peak at this time of year. Conversely, increased partitioning to the gas phase in summer decreases NH4NO3 concentrations relative to gas phase NH3 and HNO3.
Particulate SO4 2showed large peaks in concentrations in summer in Southern and also Eastern Europe, contrasting with much smaller peaks occurring in early spring in other regions. The peaks in particulate SO4 2coincided with peaks in NH3 5 concentrations, illustrating the importance of NH3 in driving the formation of (NH4)2SO4. Since NH4NO3 is more abundant than (NH4)2SO4, the seasonality of NH4 + is likely to be influenced more by the temperature and humidity dependence of the semi-volatile NH4NO3, than by the stable (NH4)2SO4. This is supported by similarity in the the seasonal profiles of NH4 + and NO3at all sites, demonstrating temporal, as well as regional correlation between these two components.

10
The second key feature is the dominance of NH4NO3 over (NH4)2SO4, with on average twice as much NO3as SO4 2-(on a molar basis). A change to an atmosphere that is more abundant in NH4NO3 will likely increase the atmospheric lifetimes and extend the footprint of the NH3 and HNO3 gases, due to the potential for the semi-volatile NH4NO3 to act as a reservoir and release NH3 and HNO3 in warm weather. The potential increase in atmospheric lifetime of NH3 suggests that a larger fraction of the reduced and oxidised N will remain in the gas phase as NH3, resulting in a non-linearity in relationship between emissions 15 and concentrations of NH3. Ammonia is an important term in the calculation of total N dry deposition and a significant contributor to the exceedances of thresholds for atmospheric concentrations and deposition of Nr components to sensitive habitats across much of Europe. In the DELTA ® network, the Critical Levels of 1 and 3 µg NH3 m -3 for the protection of lichens-bryophytes and vegetation were exceeded at 62 % and 27 % of the sites, respectively. The importance of NH3 is therefore expected to further increase relative to oxidised N, as NOx emissions continue to decrease.    Equivalent gas concentrations, based on denuder extraction volumes of 3 mL (NH3) and 5 mL (HNO3, SO2, HCl) and air volume of 15 m 3 (typical volume of air sampled by DELTA ® system over a month).            Significance level: * p < 0.05, ** p < 0.01, *** p < 0.001, ns = non-significant (p > 0.05) 20 https://doi.org/10.5194/acp-2020-275 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License. Significance level: * p < 0.05, ** p < 0.01, *** p < 0.001, ns = non-significant (p > 0.05)