Agricultural ammonia dry deposition and total nitrogen deposition to a Swiss mire

Mires are among the ecosystems most affected by eutrophication caused by excessive nitrogen (N) inputs via N deposition of ammonium (NH4 + ), nitrate (NO 3 – ), nitric acid (HNO 3 ), nitrogen dioxide (NO 2 ), and ammonia (NH 3 ). As NH 3 emissions are mostly generated by agriculture, a particular conflict of interest exists in areas where agricultural production systems are located nearby sensitive ecosystems, as often is the case in Switzerland. Therefore, this study aimed at quantifying and evaluating total N inputs to a mire in the Swiss Alpine foothills during 2007 – 2019, focusing on NH 3 dry deposition. Three surface – vegetation – atmosphere transfer models were used to estimate dry deposition of NH 3 based on micrometeorological data in combination with monthly passive sampler NH 3 concentration measurements. NH 3 -N dry deposition summed up to 1.5 – 5.2 kg N ha – 1 a – 1 , while total N loads were 11.5 – 14.2 kg N ha – 1 a – 1 , well beyond the critical N loads for raised bogs (5 – 10 kg N ha – 1 a – 1 ), and also higher than the lower limits for oligotrophic fens and mountain hay meadows (10 kg N ha – 1 a – 1 ). Hence, the mire studied is most likely negatively affected by eutrophication, which could ultimately lead to biodiversity loss and ecosystem instability. This is particularly remarkable as the monthly NH 3 concentrations at the study site (0.4 – 4.7 µ g m – 3 ) are comparatively low


Introduction
Ammonia (NH 3 ) emissions, and their deposition on the Earth's surface, are known to cause environmental problems such as eutrophication or soil acidification, which can ultimately lead to biodiversity loss and ecosystem instability (Erisman et al., 1994;Fangmeier et al., 1994;Spranger et al., 2004;Stevens et al., 2009;Dise et al., 2011;Sutton, 2011Sutton, , 2013;;de Vries et al., 2011;Bobbink and Hicks, 2014).Most NH 3 emissions are caused by agricultural production, in particular livestock farming, which in Switzerland was accountable for around 86% of the national NH 3 emissions in 2017 (Kupper et al., 2018).In the past, multiple strategies were developed to reduce the NH 3 emissions in agriculture, e.g., in 2014, the introduction of resource efficiency payments for farmers who use emission-reducing slurry and manure application techniques, such as trailing hoses (Mann and Lanz, 2013).Nevertheless, Swiss NH 3 emissions per hectare of agricultural land remained at a high level in comparison with other European countries (see e.g.dataset Eurostat, 2021).
Ecosystems such as oligotrophic and ombrotrophic mires, and especially raised bogs, are particularly at risk of such excessive nitrogen (N) inputs (Fangmeier et al., 1994;Sutton, 2011).These sensitive ecosystems are home to highly specialized plant and animal species as the typically low nutrient availability and high soil acidity represent extreme conditions (Lugon et al., 2002).In addition to this great importance for biodiversity, mires provide several other crucial ecosystem services, like exerting a buffer function in times of water surpluses and shortages (Kimmel and Mander, 2010) or acting as a sink for atmospheric carbon dioxide (CO 2 ) (Sutton, 2011;IPCC AR6 WGI, 2021).Therefore, the quantification of N inputs into such sensitive ecosystems is of high importance to evaluate the potential threats by nearby agricultural production and thus the need for action.
The Swiss environmental protection law does not distinguish between wooded and open peatland areas (Moorlandschaftsverordnung, 2017) and thus, we use the Rotherham (2020) definition here for mires: A mire is considered a wetland with at least some peat accumulation and dominated by peat-forming vegetation.The general expectation is that N Abbreviations: CBED, Concentration Based Estimated Deposition; CLNH3, Critical level for ammonia; CLN, Critical load for nitrogen; CLRTAP, Convention on Long-Range Transboundary Air Pollution; EMEP, European Monitoring and Evaluation Programme; IDEM, Integrated Deposition Model; SVAT, Surface-Vegetation-Atmosphere Transfer; UNECE, United Nations Economic Commission for Europe.deposition goes along with a significant loss of biodiversity as plants specialized to grow in N poor environments disappear and make place to generalists coping well with higher N inputs.For example, Blaus et al. (2021) emphasized how important mires or peatlands are, because they host unique biodiversity and provide many valuable ecosystem services (see also Lovegrove et al., 2020).
In Switzerland, mires only occupy a small fraction of the land surface, and hence all types of mires are protected since 1996 (Moorlandschaftsverordnung, 2017; the location of interest here is object #6), but also entire managed landscapes containing partially isolated, degraded mires are protected (Stuber and Bürgi, 2019).In other countries, e.g. in Sweden, mires still form a large part of the boreal landscape (Norstedt et al., 2021), probably the best comparison with past conditions in Switzerland around 1900 (Stuber and Bürgi, 2019; deduced from p. 165) -although locally degraded by historic land use (Tanneberger et al., 2021).Hay harvesting was effectuated on 22% of the Swedish mire area, peaking in the late 1800s.Later, 3% was reclaimed for intensive agriculture, and 40% were forested (after drainage) (Norstedt et al., 2021).Such low-productivity forests are often the only remaining pristine forests in managed forest landscapes in Sweden (Jönsson and Snäll, 2020), however, representing important soil carbon stocks.Thus, mires in Switzerland, although more degraded than in Sweden, assume a similar role.
Nitrogen deposition occurs as wet or dry deposition of N-containing compounds, with dissolved ammonium (NH4+) and nitrate (NO 3 -) as well as dissolved organic N being the main components in wet N deposition.Conversely, the gases NH 3 , nitrogen dioxide (NO 2 ), and nitric acid (HNO 3 ), as well as NH 4 + and NO 3 -in aerosols are part of dry deposition.While wet deposition can be determined by bulk or wet-only samplers of precipitation (Cape et al., 2009), the quantification of dry deposition is more challenging.Previous studies have developed so-called surface-vegetation-atmosphere transfer (SVAT) models to estimate dry deposition of NH 3 using micrometeorological data in combination with NH 3 concentration measurements.The most detailed global overview of satellite-derived surface NH 3 concentrations was provided by Liu et al. (2020) using a 1/8 • × 1/8 • grid resolution from 2008 to 2016, which corresponds to a 14 km and 9 km resolution at the site of interest in W-E and N-S extent, respectively.Until 2016, most studies showed that there was a decline in atmospheric N deposition in developed countries (Du et al., 2019).Balestrini et al. (2019) found that a number of studies have reported decreasing trends of acidifying and N deposition inputs to forest areas throughout Europe and the USA in recent decades, whereas Stevens et al. (2009) found no change or even a slight increase in N deposition in Europe and North America.For Europe, Waldner et al. (2014) quantified a decreasing trend of 2% for inorganic nitrogen in the decade to 2010, while in Switzerland, N deposition has decreased by 28% between 1990 and 2015 (Rihm and Künzle, 2019).
(2) Furthermore, total N input was estimated using openly accessible data on wet deposition of NH4+ and NO 3 -, as well as dry deposition of NH4+ and NO 3 -in aerosols and gaseous HNO 3 and NO 2 .The eutrophication threat was evaluated applying the concept of vegetation-specific critical levels for NH 3 (CLNH3) concentration, and of ecosystem-specific critical loads for (total) nitrogen (CLN) inputs.In addition,(3) this study aimed at identifying potential NH 3 emission reduction measures for farm level fertilization practices, with a focus on grassland forage production, in order to contribute to the long-term goal of protecting vulnerable ecosystems such as mires from excessive NH 3 inputs.

Site description
The Walchwil mire is situated in the Swiss Northern Alpine foothills at 985 m asl and extends over an area of 521 ha of mires (Fig. 1).It consists of raised bogs and fens which are under nature protection, adjacent to permanent grasslands which are mainly used for forage production, with little grazing.Part of this cultivated grassland belongs to the Früebüel agricultural research farm of AgroVet-Strickhof (formerly an ETH research station; see also Gilgen and Buchmann, 2009;Zeeman et al., 2010).The cultivation intensity of the farm's grassland (37.2 ha of fodder supply area until 2019; 30.8 ha since 2020) is classified as medium-intensive to intensive, i.e., the sward is cut three to four times in the months of May/June, July, August, and September/October.Here we use a non-manipulative observational approach (survey-type approach) as opposed to a manipulative experimental approach (according to Eugster and Merbold, 2015).A flux tower (CH-Fru; Fig. 1) was used to systematically observe the business-as-usual management events, which were recorded in detail for the three parcels immediately adjacent to the tower (Rogger et al., 2022).
Average annual precipitation at Früebüel (CH-Fru) in the years 2007-2019 was 1299 mm, with an annual minimum of 693 mm in 2011, and an annual maximum of 1859 mm in 2016.Typically, precipitation was lowest at the beginning of the year (January to March; around 65 mm per month), and highest in summer (June to August; around 168 mm per month).Average air temperature during 2007-2019 was 7.9 • C, with monthly mean temperatures from − 0.1 • C in February to 16.6 • C in July (data taken at CH-Fru site).
Since 2007, atmospheric NH 3 concentrations have been measured at Früebüel with two to three passive samplers of the Radiello type (Martin  (Schmid and Oke, 1990).The source-area covers the three grassland parcels close to the tower (5.2 ha in total; pink color) as well as most of the naturally protected mire (green for forest mire, yellow for open mire).Map sources: map.geo.admin.ch(base map), Kanton Zug (2020).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)E. Tanner et al. et al., 2019;Pan et al., 2020) and the Ferm type (since 2018;Ferm, 1979).Both types of passive samplers use a tube in which the inside is covered with a sticky absorbent which captures a fraction of the NH 3 gas molecules that are moving by via diffusion.Every month, the passive samplers are collected and brought to the laboratory for chemical analyses of the absorbent.The denuders were eluted with ultra-pure water and measured as NH 4 + with flow injection analysis.The calculations for the concentration in the air were determined following the instructions provided by Radiello (https://radiello.com/files-downloads/?num=4) (Thimonier et al., 2019).In summary, the calculation of the monthly concentrations is the mass in ammonium ions (in µg NH 3 ) normalized by the exposure time (in minutes) and an empirical sampling rate (Q 298 , in ml min -1 ).The transition from Radiello-type to Ferm-type denuders at many sites starting in 2018 (at the end of our analysis period) was reported to be in very good agreement between the samplers (Seitler et al., 2021).Since samplers were exchanged monthly, they provided integrated values during the month prior to the exchange.No data were available for January 2015 due to vandalism; thus this gap was linearly interpolated, resulting in a NH 3 concentration of 0.6 µg m -3 .For regional comparisons of monthly NH 3 concentrations, the two closest sites at similar elevation were used: ZB 01 (Zugerberg, 990 m asl, 1.7 km to the NNW of Früebüel) and ZIGE (Zigerhüttli, 989 m asl, 8.5 km to the east in a slightly more intensive production area).

NH 3 dry deposition models
The NH 3 flux (F NH3 ) is defined as the product of deposition velocity (V Dep ) and gradient between NH 3 concentration (χ NH3 ) at reference height (z ref ) above ground level (here z ref = 2.4 m) and the concentration at the plant or soil surface (Eq.1; Flechard et al., 2011): The micrometeorological sign convention is used here, with negative fluxes representing an input from the atmosphere into the ecosystem, i. e., a deposition (e.g., Flechard et al., 2011).A positive sign, on the other hand, stands for emissions from soil or vegetation into the atmosphere.
The NH 3 flux F NH3 in the European context is typically estimated using three widely used models: (i) the Concentration Based Estimated Deposition (CBED) model following Smith et al. (2000), (ii) the European Monitoring and Evaluation Programme (EMEP) model (version rv3.1) following Simpson et al. (2003), and (iii) the Integrated Deposition Model (IDEM) following Erisman et al. (1994).All three models calculate V Dep following a resistance analogy, since V Dep can be considered as the conductivity of a transport, or of gas uptake by the soil or vegetation.Hence, in analogy with Ohm's law, V Dep can be estimated as the reciprocal sum of resistances connected in series along the pathway of NH 3 uptake (Eq.2; Flechard et al., 2011).Thereby, R a is standing for the aerodynamic resistance, R b for the quasi-laminar sublayer resistance, and R c for the surface or canopy resistance.
Each of these resistances represents a section along the transport path through the atmosphere and the uptake at the Earth's surface by the soil or vegetation.The aerodynamic resistance (R a ) represents the efficiency of the gas transfer from the level z ref to d + z 0 , in which the transfer occurs in a turbulent manner, with d representing the aerodynamic displacement height (roughly 2/3 of vegetation height), and z 0 the aerodynamic roughness length (Flechard et al., 2011).R a was calculated by assuming a diabatic wind profile (Eq.3; Erisman et al., 1994): where k represents the von Kármán constant, u * the friction velocity, Ψ H the similarity function for heat which accounts for atmospheric stability effects in turbulent transport near the Earth's surface, and L the Monin-Obukhov length (Monin and Obukhov, 1954;Simpson et al., 2003).The roughness length z 0 was estimated as 0.13 × vegetation height following Jones (2013).
The surface or canopy resistance (R c ) quantifies the affinity of the soil and vegetation for the uptake of NH 3 and is vegetation type-specific (Flechard et al., 2011).The different pathways for uptake are reflected in the various parallel-connected sub-resistances r x (Eq.5; Erisman et al., 1994): where r sto represents the stomatal resistance to the uptake of NH 3 by plants.r sto is influenced by photosynthetically active radiation, minimum stomatal resistance to water vapor, temperature, vapor pressure deficit, water stress, and the nature of the gas of interest (here: NH 3 ).r m is the mesophyll resistance, which represents the transport of the gas through the mesophyll.For NH 3 , it has been shown by experiments that r m tends to be negligible (Erisman et al., 1994).r inc stands for the in-canopy resistance, which represents the transport of the gas through the vegetation canopy towards the ground.r soil represents the soil resistance for gas absorption in the soil.r ns stands for the non-stomatal resistance and represents the NH 3 uptake via the (water-covered) leaf surface, i.e., the non-stomatal uptake by plants.r ns is mostly influenced by temperature, relative humidity, and the presence of other trace substances, such as sulphur dioxide (SO 2 ).In contrast to R a and R b which are equal for the most commonly used models, the parametrization of R c differs among the three models chosen here (see Erisman et al., 1994;Smith et al., 2000;Simpson et al., 2003).
The above-described method for deriving F NH3 is valid when following a deposition-only approach.In reality, however, there exists a bi-directional NH 3 exchange between the vegetation and the atmosphere, as NH 3 deposited in stomata or on plant surfaces can be reemitted (Sutton et al., 1995).Especially (fertilized) croplands and grasslands have been shown to exert bi-directional fluxes, whereas for semi-natural ecosystems, the NH 3 deposition seems to dominate due to lower N status (Sutton et al., 1995).Yet the division is not strict, as bi-directional fluxes have also been measured over semi-natural ecosystems in the past (Sutton et al., 1998;Flechard et al., 1999).Therefore in this study, F NH3 was calculated using a compensation point approach (Eq.6; Flechard et al., 2011): where χ c is the canopy compensation point and represents the net potential for NH 3 emissions from the canopy and determines the direction of F NH3 , as emissions occur when χ c is higher than χ NH3 (z ref ) (Sutton et al., 1998).χ c is dependent on R a (z ref ), R b , r sto , r ns and a stomatal compensation point χ sto (Eq.7; Sutton et al., 1998): χ sto is given by the fact that NH 3 is easily soluble in water, which results in an equilibrium concentration between gaseous NH 3 and dissolved NH4+ in the stomatal cavity (Sutton et al., 1998) leading to E. Tanner et al.
As the compensation point approach enables to determine the actual loads of NH 3 -N, i.e., the net NH 3 -N deposition flux that is actually available for plant growth (gross NH 3 deposition minus concurrent reemissions of NH 3 ), the EMEP and IDEM modelswhich originally did not consider the bi-directional fluxwere modified to include the compensation point approach.Of each model, two different parameter sets were used to represent the two dominant vegetation types at the study site: (i) parameters for (managed) grassland vegetation, hereafter MODEL-GL; and (ii) mire landscape parameters, hereafter MODEL-ML.A detailed description of assumptions made for the models can be found in the Appendices A to D. All these models are well-established for Europe.In the global context, at a much coarser resolution, similar models are being used in combination with annual satellite-derived surface NH 3 concentration estimates (Liu, 2020;Liu et al., 2020).

Evaluation of NH 3 concentrations and N loads
For estimating the impacts of ambient NH 3 concentrations on vegetation, the comparison with the critical levels for NH 3 (CLNH3) defined by the Convention on Long-Range Transboundary Air Pollution (CLRTAP) of the United Nations Economic Commission for Europe (UNECE) is usually made.Critical levels for vegetation have been defined by the CLRTAP as the "concentration, cumulative exposure or cumulative stomatal flux of atmospheric pollutants above which direct adverse effects on sensitive vegetation may occur according to present knowledge" (CLRTAP, 2017a; p. 4).CLNH3 are vegetation type-specific, as not all plants react equally sensitive to NH 3 levels (Table 1).
Similar to the concept of CLNH3, the critical loads for (total) N (CLN), also defined by the CLRTAP, can be used to estimate the impact of the total N loads on vegetation.CLN are a "quantitative estimate of an exposure to one or more pollutants below which significant harmful effects on specified sensitive elements of the environment do not occur according to present knowledge" (CLRTAP, 2017b; p. V-7).
Unlike CLNH3, CLN are not based on the effects of NH 3 alone, but also consider other N-containing compounds.Therefore, our NH 3 dry deposition estimates were complemented by data on wet deposition of NH 4 + and NO 3 -, as well as data on dry deposition of gaseous HNO 3 and NO 2 , as well as NH 4 + and NO 3 -in aerosols (for two years, 2014 and 2019) from a nearby research site ZB 01, roughly 1.7 km NNW of the Früebüel research site (see Seitler et al., 2021).It was assumed that the pollutant concentrations at ZB 01, respectively the N loads, were very similar to Früebüel due to the geographical proximity and similar meteorological conditions.Wet deposition loads at ZB 01 were quantified using bulk collectors, which often lead to an overestimation of the wet deposition load, as not only N in precipitation, but also N-containing gases and particles can be deposited on the bulk collector surface (Cape et al., 2009).To avoid this double accounting of dry deposition to the bulk precipitation deposition, correction factors of Draaijers et al. (1998) were used.These were derived from simultaneous measurements with bulk and wet-only samplers at various locations and can therefore be used to assess the true wet deposition based on bulk deposition data (Draaijers et al., 1998)

Statistical analyses
All statistical analyses were carried out in R version 4.1.2(R Core Team, 2021).For multiple testing, we used the Holm-correction (Legendre and Legendre, 2012) in the p.adjust function of R. To smooth data time series, we used local polynomial regression fitting (loess function in R) with a span of 30% or 90% given in the respective figure caption.The air temperature trend was determined using the nearest-by  Rigi-Seebodenalp research site of the Swiss national air pollution network (NABEL; 7.7 km in SW direction, 47 • 04 ′ 02.6 ′′ N, 8 • 27 ′ 48.0 ′′ , 1030.4 m asl), and the Mann-Kendall test (Libiseller and Grimvall, 2002) in combination with Sen's median slope (Sen, 1968) (mk.test and sens.slopefunctions in R) were used.Orthogonal regressions were computed using the lmodel2 package in R (see Legendre and Legendre, 2012) from which we selected the major axis variant.When comparing NH 3 concentrations, both axes have the same units, and empirical error is not only associated with the y-axis, but equally with the x-axis.To indicate the uncertainty of the linear regression slopes, we used the cplot package in R, which we adopted for the case of major axis regressions (Legendre and Legendre, 2012).

Monthly mean NH 3 concentrations and compensation points
The NH 3 concentrations at Früebüel were compared with those at the ZB 01 and the ZIGE sites to test for regional heterogeneity within the source-area of Früebüel (Fig. 2).As expected, the NH 3 concentrations at Früebüel were similar to those at ZB 01, but slightly lower than those at ZIGE (the major axis regression slopes in Fig. 2  Typical seasonal patterns of NH 3 concentrations were observed over the whole time series, as NH 3 concentrations reached their annual maxima in spring or (early) summer and then decreased towards the end of the year (Fig. 3).The maximum concentration was measured in July 2019, with a value of 4.7 µg m -3 , while the minimum value was recorded in December 2013 with 0.4 µg m -3 .-N for the years 2014 and 2019.The respective ranges of critical loads for nitrogen (CLN) for raised bogs (5-10 kg N ha -1 a -1 , red shaded), oligotrophic fens (10-15 kg N ha -1 a -1 , orange dashed), and mountain hay meadows (10-20 kg N ha -1 a -1 , yellow shaded) are also given.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

E. Tanner et al.
The stomatal compensation points χ sto modelled by the three approaches closely followed the seasonal cycle of air temperature, with values approaching or slightly exceeding 1 µg NH 3 m -3 during peaks in summer, and close to zero in winter (data not shown).The approaches by Nemitz et al. (2000) and Farquhar et al. (1980) led to χ sto that were, on average, 18% higher and lower than the approach by Flechard et al. (1999), respectively.When considering the canopy compensation points χ c , which depend on χ sto , the difference within the model variants for χ c due to the different χ sto approaches was at the most 1.2% Subsequently, the χ sto following the approach of Flechard et al. (1999), and the related χ c , were used to calculate the NH 3 flux F NH3 , since these represented the medium values.Monthly means of the model-dependent χ c ranged between 0.03 and 2.90 µg m -3 with EMEP, resulting in values on average 19% and 58% larger than CBED and IDEM, respectively (Fig. 3).The parametrization of the vegetation type only had a small effect, as the χ c varied typically less than 1% between the GL and ML versions of the same model (data not shown).Being dependent on the atmospheric NH 3 concentration (see Eq. 7), the χ c largely followed the same seasonal pattern as the NH 3 concentration, but on a lower level.

NH 3 -N dry deposition load
Due to the different models, the resistances r ns and r sto differed among the three models, with r ns taking on the highest median value of 86 s m -1 for EMEP, followed by CBED with 31 s m -1 , and finally IDEM with 8 s m -1 .Values of r sto were considerably higher, with a median value of 843 s m -1 for the CBED model, 2524 s m -1 for IDEM, and 9719 s m -1 for EMEP during the daytime.Based on the different values of χ c , r ns and r sto , also the NH 3 -N loads differed among the three models (Fig. 4), thereby reflecting important differences in assumptions made by the three models.Average NH 3 -N deposition at Früebüel stayed between 2 and 3 kg NH 3 -N ha -1 a -1 when modelled with CBED and EMEP, but increased up to 5 kg NH 3 -N ha -1 a -1 when modelled with IDEM.Inter annual variability due to varying NH 3 concentrations was reflected by all three models.While the minimum NH 3 -N depositions were calculated for the year 2016, deposition values were highest in the year 2019.Differences in NH 3 -N deposition between vegetation types were very small for all three models and insignificant (p = 0.97).

CLNH3 and CLN
The mean annual NH 3 concentrations at Früebüel ranged from 1.3 µg m -3 in 2013 and 2014 to 2.5 µg m -3 in 2019.These levels constantly exceeded the CLNH3 for lichens and mosses (1 µg m -3 ; Table 1).In 2018 and 2019, also the range of CLNH3 for higher plants was reached (2-4 µg m -3 ; Table 1), while in the other years, the actual levels at Früebüel remained below the CLNH3 for higher plants.

Total N input measurements
Annual wet deposition obtained at the nearby ZB 01 site was around 2.8 kg NO 3 --N ha -1 and 4.1 kg NH 4 + -N ha -1 in 2014, but slightly higher with 2.6 kg NO 3 --N ha -1 and 5.5 kg NH 4 + -N ha -1 in 2019 (Fig. 5).Annual dry deposition at ZB 01 was similar in both years, with around 0.8 kg HNO 3 -N ha -1 , 0.9 (2014) and 0.8 ( 2019) kg NO 2 -N ha -1 , as well as 0.5 kg NH 4 + -N ha -1 and 0.3 kg NO 3 --N ha -1 in both years.Adding the 2.1 kg N ha -1 and 3.7 kg N ha -1 of NH 3 dry deposition determined in this study (mean of all model runs for 2014 and 2019, respectively) resulted in a total N load of 11.5 kg N ha -1 a -1 in 2014 and of 14.2 kg N ha -1 a -1 in 2019 (Fig. 5).In both years, the lower limit of the CLN for raised bogs (5 kg N ha -1 a -1 ; Table 1) was clearly exceeded already by the wet deposition of NH 4 + and NO 3 -(6.9kg N ha -1 a -1 in 2014, 8.1 kg N ha -1 a -1 in 2019).In both years, the upper limit of the CLN for raised bogs (10 kg N ha -1 a -1 ; Table 1), as well as the lower limits of the CLN for oligotrophic fens and mountain hay meadows (10 kg N ha -1 a -1 ; Table 1) were exceeded when considering total N loads.In contrast, the CLNs for Molinia caerulea meadows (15-25 kg N ha -1 a -1 ; Table 1) and eutrophic fens (15-30 kg N ha -1 a -1 ; Table 1), as well as the upper limit of the CLN for mountain hay meadows (20 kg N ha -1 a -1 ; Table 1) were not exceeded.

Current situation and future perspectives
Against our expectations, we could not find a statistically significant correlation between fertilizer amounts applied in any given month to the three parcels of intensive observation at the Früebüel farm (pink in Fig. 1) and the level of the NH 3 concentration (p = 0.6761; data not shown).This was true, irrespective of whether slurry (with trailing hose) or manure was applied.Since no significant correlation between NH 3 concentration and the application of fertilizers could be found, we compared the background 3 concentration during months without fertilization events and the offset over this background during months with N applications at the Früebüel grassland parcels.The background concentration was 1.34 µg NH 3 m -3 and increased by 0.81-1.08µg NH 3 m -3 (Fig. 6).These NH 3 concentrations were all significantly above the background, irrespective of the type of fertilizer, but no significant difference was found whether fertilizer (both manure and slurry), manure only, slurry only, more manure than slurry (S.) events, or more slurry than manure (M.) events were considered in any given month (Fig. 6).
The detailed time series for fertilization practices available for the intensively observed grassland parcels at the Früebüel farm shows a couple of management changes between 2007 and 2020 (Fig. 7).Fertilization events increased from originally 5 per year to 10 per year (Fig. 7a), which is also a result of increasing the original manure-only fertilization scheme towards a 50% slurry-N fertilization scheme (Fig. 7b).In absolute numbers, distributed over the three parcels, slurry fertilization became a constant annual input of ca. 15 kg N ha -1 a -1 (Fig. 7d), while annual manure fertilization decreased from roughly Fig. 6.Monthly average passive sampler NH 3 concentrations at the Früebüel study site.The background concentration contains all months without fertilizer application on the three closely monitored parcels surrounding the measurement tower.There was no correlation found between months without and with fertilizer applications.NH 3 concentrations during months with fertilizer were significantly increased by the respective offsets (given below each bar; p < 0.05, Holm correction for multiple testing; different letters indicate statistically significant differences).60 kg N ha -1 a -1 to around 30 kg N ha -1 a -1 (Fig. 7c).
This change in management coincided with a pronounced warming trend of 1.1 • C (Sen's median slope over a decade, p = 0.0115, 95% CI 0.4-2.3• C per decade, Sen, 1968) over the observed period (Fig. 7f).Although the increase in total N applied has been corrected in recent years (Fig. 7e), we still see higher NH 3 deposition in 2019 than in 2014 (Fig. 5).

NH 3 -N dry deposition
The NH 3 -N dry deposition was estimated with three different SVAT models driven by local NH 3 concentration measurements at monthly resolution in combination with local micrometeorological observations.Consistently, IDEM yielded higher deposition rates than EMEP and CBED (see Fig. 4), representing the different parametrizations.Because the calculated resistances R a and R b were identical for all models, the model-specific resistances r ns and r sto were the cause of these differences in modelled NH 3 -N deposition loads.The differences in r ns and r sto can, inter alia, be traced back to the various inclusions or omissions of parameters and variables in the models (for details see Erisman et al., 1994;Smith et al., 2000;Simpson et al., 2003).For example, the r ns of the EMEP model takes into account the environmental pollution level by atmospheric sulphur dioxide (SO 2 ), which is hypothesized to influence the non-stomatal co-deposition of NH 3 and SO 2 (Erisman and Wyers, 1993;Simpson et al., 2003;Flechard et al., 2011), whereas this is not done by the other models.However, determining which model yielded the most accurate NH 3 -N dry deposition estimates is not possible, as there exist no direct NH 3 flux measurements for this site.
For all three models, however, the values of r ns were typically lower than those of r sto , so that r ns more strongly influenced χ c and thus F NH3 .
This underlined the importance of non-stomatal uptake of NH 3 via the cuticular leaf surface.Due to the strong dependence of non-stomatal NH 3 deposition on the presence of water films on plant surfaces ton et al., 1995), r ns is also significantly affected by relative humidity.Furthermore, as r ns did not differ among vegetation types, much in contrast to r sto , this translated to rather small differences of NH 3 deposition between GL and ML (Fig. 4).Moreover, all three models employed in this study use rather elaborate parametrizations to simulate the different resistances, whereas simpler approachesas commonly used in Switzerland (e.g., Rihm and Achermann, 2016) tend to use vegetation-specific constant V Dep found in the literature.To compare our results with such literature-based approaches, V Dep values were calculated according to Eq. 2. Generally, V Dep was highest for Rihm and Achermann (2016) with a mean of 8.9 mm s -1 , followed by CBED with 7.1 mm s -1 , and finally EMEP with 5.4 mm s -1 (see Appendices).These V Dep values were in the same order of magnitude as those given by Flechard et al. (2011), who reported for grassland V Dep of about 10 mm s -1 for IDEM, and values of about 6 mm s -1 for EMEP.However, Swiss standard values for simple F NH3 calculations using constant, but vegetation-specific V Dep , are typically higher.For example, Rihm and Achermann (2016) assumed a value of 10 mm s -1 and 12 mm s -1 for intensely cultivated grassland (within crop rotations) as well as for pastures and meadows, respectively.Even bigger differences exist for the mire vegetation type.Rihm and Achermann (2016) assumed a V Dep of 20 mm s -1 , whereas in this study, the resulting V Dep values were 50-75% smaller.Consequently, the estimates of NH 3 -N deposition loads with Swiss standard values for V Dep would have yielded much higher deposition loads than estimated by this study, which is based on local measurements (monthly passive sampler NH 3 concentrations combined by aggregated 30-minute averaged meteorological measurements).
Furthermore, this study implemented a compensation point approach to reflect the bi-directional nature of the NH 3 flux, which reduced NH 3 -N deposition loads compared to Swiss standard estimates even further.In addition to the low NH 3 concentrations measured at Früebüel, this resulted in the rather low NH 3 -N deposition rates of 1.5-5.2kg NH 3 -N ha -1 a -1 .However, these deposition rates might change in the future, as with climate change, drier summers and more hot days are to be expected in Switzerland (Tschurr et al., 2020).This in turn will lead to increasing NH 3 emissions, and thus potentially higher NH 3 deposition, even without changes in agricultural production intensity.Fig. 7. Fertilization practice at the Früebüel study site (CH-Fru site) between 2007 and 2020.Number of fertilization events (a), fraction of slurry application (b), amount of N applied with manure (c) and slurry (d), total amount of N applied (e) as well as mean annual temperature (f).The respective loess fit (bold lines) uses a 90% span in all panels, which means that a moving window of 90% of all years was used for the loess smoothed curves.

E. Tanner et al.
Until 2016, the trend in N deposition in well-developed countries was negative (Waldner et al., 2014;Balestrini et al., 2019;Du et al., 2019), as also reported for Switzerland until 2015 (Rihm and Künzle, 2019).However, this study indicated both increased gaseous NH 3 deposition and wet deposition of NH4+ in 2019 as compared to 2014 (Fig. 4); a factor that has not been implemented in current models.

CLNH3 and CLN
The CLNH3 for lichens and mosses, which are abundant in the mires, were exceeded, even in years with the lowest annual mean NH 3 concentration, indicating that harmful effects on the vegetation must be expected (CLRTAP, 2017a).This is remarkable as the NH 3 concentrations at Früebüel are generally low compared to other Swiss sites (Seitler et al., 2021) and satellite-derived concentrations (Fig. 3; Liu, 2020).Higher plants, on the other hand, are less sensitive to NH 3 emissions, and their CLNH3 were not permanently exceeded, except for the last two years, 2018 and 2019, potentially triggered by high NH 3 concentrations and high air temperatures (Seitler et al., 2021;Rogger et al., 2022) (Fig. 7f), and due to more fertilization events (Fig. 7a).Thus, it remains to be seen how this trend in NH4+-N and NH 3 -N deposition continues.
Considering the evaluation of total N inputs at the landscape scale, it seems reasonable to use the exceedance of the CLN for the most sensitive member present (Klaus, 2007).At Früebüel, this is the raised bog vegetation, whose CLN were exceeded in both years analyzed 5).Also the composition of the total N load is noteworthy.The dry deposition of HNO 3 and NO 2 contributed < 15% to the total N load, most likely due to the fact that NO 2 and HNO 3 are mainly emitted by motorized transport and industry (Kosonen et al., 2019), which are rare at the study site and its surroundings.The dry deposition of NH4+ and NO 3 -in aerosols contributed even less to the total N load, with 7% in 2014, and 6% in 2019, attributed to the rather low V Dep of aerosol deposition (see Table 13 in Seitler et al., 2021).However, the dry deposition loads of NH 3 -N contributed about 36% in 2014, and 39% in 2019.The main contributor to the total N load was the wet deposition of NH4+ and NO 3 -, with 60% and 57% in 2014 and 2019, respectively.This is in contrast to average conditions in Switzerland, where the largest contribution to total N deposition stems from dry deposition of ammonia (Aksoyoglu and Prévôt, 2018;Rihm and Künzle, 2019).The dominance of wet deposition of NH 4 + and NO 3 -at the Früebüel site emphasizes the relevance of long-range transport with precipitation, since this can be transported up to 2500 km (Irwin and Williams, 1988) to avoid exceedance of critical N levels at the landscape scale.

Consequences of excessive N inputs
Because NH 3 concentrations at Früebüel were rather low, direct leaf injuries are unlikely to occur (Bobbink and Hicks, 2014).However, the atmospheric N inputs will have a fertilizing effect on the vegetation since mires are typically deficient in N (Lugon et al., 2002;Sutton, 2011).Potential consequences range from increased growth and shoot-to-root ratios (Fangmeier et al., 1994), higher susceptibility to biotic and abiotic stress, such as herbivory, fungal pathogens or frost, to nutrient imbalances, and increased competition among plant species in these ombrotrophic ecosystems (Fangmeier et al., 1994;Bobbink and Hicks, 2014).These effects can lead to fundamental changes in species composition and even loss of plant diversity (Hettelingh et al., 2014), particularly in raised bogs and oligotrophic fens (Bobbink et al., 2010).Native plant species adapted to oligotrophic conditions lose their competitiveness and are suppressed by nitrophilous plants when nutrient levels increase (CLRTAP, 2017b;Aerts, 1999;Hettelingh et al., 2014).This is especially detrimental as mires are home to around a quarter of all endangered ferns and flowering plants in Switzerlandeven if they only account for about 0.5% of the area (Klaus, 2007).Overall, N deposition globally affects climate and biodiversity (Sutton, 2013).

Reduction measures
About 93% of total Swiss NH 3 emissions can be traced back to agricultural production (Kupper et al., 2018), as in the rest of Europe (Sutton et al., 2014).Thus, further measures need to be implemented to reduce N deposition into sensitive ecosystems.NH 3 mitigation options in agriculture have been the center of attention for decades (see e.g., UNECE, 2014).However, not all methods to reduce NH 3 emissions are appropriate, particularly when agricultural policies but also consumer requests create trade-offs.For example, a promising measure to reduce NH 3 emissions is the promotion of emission-optimized stable constructions, with urine collection channels or dung scrapers.These keep the surface free of manure, so less NH 3 is emitted (UNECE, 2014).However, animal welfare legislation requests that tied-up barns are replaced with animal-friendly playpens, which typically come with a greater manure-contaminated surface and thus higher NH 3 emissions (Schrade and Keck, 2012).There has yet to be found a solution for this trade-off between NH 3 emission reduction and animal welfare.
Another substantial potential for NH 3 emission reductions lies in the choice of slurry application techniques, as for example, NH 3 emissions occurring during the application process account for about 43% of Swiss livestock emissions (Kupper et al., 2018).In grasslands, suitable low-emission applicators are trailing hoses or trailing shoes, which have the potential of reducing NH 3 emissions by around 51-53% (Häni et al., 2016), whereas the direct injection and mechanic incorporation of slurry into the soil are less suited.At the Früebüel site, slurry application was already made by trailing hoses, also facilitated by direct payments (in Switzerland since 2014), and thus no statistically significant difference between manure and slurry applications could be found (Fig. 6).Another measure with big impact is the installation of a fixed cover for slurry, which can reduce storage emissions compared to an uncovered slurry tank storage by 51-90% (Kupper et al., 2020).In Switzerland, this measure will become mandatory in 2022, as part of the revised Swiss clean air act (OAPC, 2020).However, no effects are to be expected at the Früebüel site since this measure is already in place as well.

Conclusion
This study confirmed the eutrophication threat for the Walchwil mire, which is representative for other locations in Switzerland and beyond, as 100% of raised bogs and 84% of fens in Switzerland are considered to be threatened by excessive N inputs.To counteract eutrophication further emission reduction efforts are necessary, which requires further regional, national and coordinated international reduction measures.At local scales, reduction measures for NH 3 emissions are needed in the immediate vicinity to endangered ecosystems.In addition, at larger scales, efforts to reduce long-range transport of wet deposition of NH4+ and NO 3 -are required, because they dominated the total N load at the Walchwil mire.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
sampler measurements used in this study.We also thank the Institute of Applied Plant Biology (IAP Witterswil, Switzerland; Sabine Braun and collaborators) and FUB (Forschungsstelle für Umweltbeobachtung AG, Rapperswil SG, Switzerland) for deposition data from 2014 and 2019, which were collected with funding from the Swiss Federal Office for the Environment (FOEN).All technical staff members at the Grassland Sciences group at ETH Zurich involved in the Früebüel long-term flux measurements and site management are acknowledged for their support over the years.

A. Model Details
The variables leaf area index (LAI) and vegetation height (h) were parametrized for all three models following a scheme proposed in Simpson et al. (2003;p. 25), with linear changes during the growing season.The length of the growing season at the Früebüel site was based on Gilgen (2009), beginning in early May (Day of Year (DOY) = 121) and ending in mid-October (DOY = 288) (Eq.A1): where Y, Y min and Y max are the predicted, minimum, and maximum values of LAI and h, respectively.For the vegetation type grassland (GL), LAI ranged between 1 and 6 based on Gilgen (2009), and between 2 and 3 according to Simpson et al. (2003) for mire landscape (ML).For h, a minimum of 0.1 m and a maximum of 0.6 m was assumed for both GL and ML.For simplicity, the reduction of h in the agriculturally used grassland was not considered.

B. CBED
The CBED version was modelled as described in Smith et al. (2000), with the following assumptions and deviations: The minimal stomatal resistance r sto.min was calculated using the range of 80-300 s m -1 for GL and 140-300 s m -1 for ML given in Smith et al. (2000) applied to the temporal scheme for calculating LAI and h (see Eq. A1).
The ratio of incoming shortwave radiation (SW_IN) to potential incoming shortwave radiation (SW_IN_POT), constrained to values between zero and one, was chosen as the indicator of sky conditions, whereby high values indicated clear sky condition and low values indicated overcast sky conditions.During the night, the indicator was set to zero.

C. EMEP
The EMEP version was modelled as described in Simpson et al. (2003), with the following assumptions and deviations: The molar acidity ratio (a SN ) of SO 2 and NH 3 (Simpson et al., 2003;their eq. 15) was calculated assuming a SO 2 concentration of 0.5 µg m -3 , whereby this represents the measured concentration at the Rigi-Seebodenalp site (see Heldstab et al., 2013), since the SO 2 concentration at Früebüel was not measured.
The stomatal conductance scaling factors (from 0 to 1) accounting for minimum observed stomatal conductance (f min ), light (f light ), leaftemperature (f T ), and leaf-to-air vapor-pressure deficit (f VPD ) were calculated according to Emberson et al. (2000).However, the vegetation type-specific optimal (T opt ), minimum (T min ) and maximum (T max ) temperature for stomatal conductance were taken from Table 8.1 in Simpson et al. (2003) because the values given there better represent the vegetation types at Früebüel.The stomatal conductance factor for leaf phenology (f phen ) was calculated according to the temporal scheme for calculating LAI and h (see Eq. A1), with a range between 0.1 and 1 (range according to Emberson et al., 2000).The stomatal conductance factor for soil-water potential (f SWP ) was set to 1 as proposed by Simpson et al. (2003).

D. IDEM
The IDEM version was modelled as described in Erisman et al. (1994), with the following assumptions and deviations: It was assumed that the surface was snow-covered when temperatures were below 0 • C and precipitation occurred, whereby R c took on a value of 500 s m -1 by default (as soil pH was taken to be <8) (Erisman et al., 1994;their Table 5).
The so-called external leaf resistance (r ext ), corresponding to r ns , was calculated according to the IDEM model description in Flechard et al. (2011), Eq. 7.
The stomatal resistance (r sto ) was calculated according to Appendix A in Erisman et al. (1994), with the minimal stomatal resistance r sto.min parametrized like in the CBED model (see description above).Furthermore, the optimal (T opt ), minimum (T min ) and maximum (T max ) temperature for stomatal uptake were taken from Table 8.1 in Simpson et al. (2003).The influence of water potential (termed g Ψ in Erisman et al., 1994) was set to 1, and the influence of vapor pressure deficit (termed g VPD in Erisman et al., 1994) was calculated according to Simpson et al. (2003) as the vegetation type-specific values given there were more suitable to the vegetation types present at Früebüel.
The soil resistance (rsoil) was set to 0 s m -1 for dry conditions and 50 s m -1 for wet conditions according to Table 5 in Erisman et al. (1994), assuming a pH < 8. Dry conditions were identified as the relative humidity (RH) being below a threshold of 71% as proposed by Wichink Kruit et al. (2008).

E. Deposition velocities
The monthly average V Dep varied between 3.2 mm s -1 (EMEP-GL in April 2007) and 13.2 mm s -1 (IDEM-ML in July 2007) (Figure A1).Depending on model and vegetation type, the V Dep values differed, with IDEM resulting in the highest values, with a mean of 8.75 mm s -1 for grassland, and 9.04 mm s -1 for vegetation type ML (mire landscape).The EMEP model, on the other hand, resulted in the lowest mean values, with 5.34 mm s -1 for grassland, and 5.52 mm s -1 for mire landscape.The V Dep of the GL (grassland) model versions were around 14% smaller than the V Dep of the ML model versions, with a mean of 6.7 mm s -1 compared to 7.6 mm s -1 .Appendix A1.

Fig. 1 .
Fig. 1.Map of the Walchwil mire landscape.The faint reddish area (covering most of the map) is centered at the measurement tower Früebüel (CH-Fru; yellow symbol) and indicates the typical radius (≈ 3.4 km) of the source-area contribution to the NH 3 concentrations measured at the study site(Schmid and Oke, 1990).The source-area covers the three grassland parcels close to the tower (5.2 ha in total; pink color) as well as most of the naturally protected mire (green for forest mire, yellow for open mire).Map sources: map.geo.admin.ch(base map), Kanton Zug (2020).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3 .
Fig. 3. Monthly mean NH 3 concentrations (black solid line) and monthly mean estimates of the canopy compensation points (χ c ) for the three models CBED, EMEP and IDEM (dashed lines) at the study site Früebüel from Jan. 2007 to Dec. 2019.The annual satellite-derived surface NH 3 concentrations from Liu (2020) are shown for reference (years 2008-2016).

Fig. 5 .
Fig. 5.Total nitrogen load at the ZB 01 site, composed of dry deposition of gaseous NH 3 -N, HNO 3 -N and NO 2 -N, NO 3 --N and NH 4 + -N in aerosols, as well as wet deposition of NO 3 --N and NH 4 +
c including heathland, grassland, and forest ground flora.