Atmospheric nitrogen deposition to terrestrial ecosystems across 1 Germany 2

Biodiversity is strongly affected by the deposition of nitrogen and sulfur on terrestrial ecosystems. In this paper 13 we present new quantitative estimates of the deposition of atmospheric nitrogen to ecosystems across Germany. The 14 methodology combines prognostic and empirical modelling to establish wet deposition fluxes and land use dependent dry 15 and occult deposition fluxes. On average, the nitrogen deposition in Germany was estimated to be 1057 eq ha yr. The 16 deposition maps show considerable variability across the German territory with highest deposition on forest ecosystems in 17 or near the main agricultural and industrial areas. The accumulated deposition over Germany of this study is systematically 18 lower (27 %) than provided in earlier studies. The main reasons are an improved wet deposition estimation and the 19 consolidation of improved process descriptions in the LOTOS-EUROS chemistry transport model. The presented 20 deposition estimates show a better agreement with results obtained by integrated monitoring and deposition modelling by 21 EMEP than the earlier results. Through comparison of the new deposition distributions with critical load maps it is estimated 22 that 70% of the ecosystems in Germany receive too much nitrogen. 23 24


Introduction
Anthropogenic activities generate a tenfold more reactive nitrogen (Nr) than in the late 19th century due to increased agricultural production and energy consumption (Galloway et al., 2003).Globally half of the annually fixed nitrogen is due to anthropogenic activities (Fowler et al., 2013).A large part of the reactive nitrogen enters the atmosphere in the form of ammonia (NH3) through animal husbandry and fertilizer use as well as in the form of nitrogen oxides (NOx) through combustion of fossil fuels (Erisman et al., 2011).The remainder is released as nitrous oxide (N2O) or as nitrate (NO3) to the soil-water compartment.In Germany about 26 % of total Nr is emitted as NOx and about 30 % as NH3 (Geupel and Frommer, 2015).Deposition of reactive nitrogen has negative impacts on biodiversity and ecosystem functioning (Sutton et al., 2011).Especially in ecosystems adapted to nutrient poor conditions, a long term and sizeable input of reactive nitrogen may negatively affect plant communities (Bobbink et al., 1998).Field studies have shown an inverse relationship between reactive nitrogen deposition and species diversity (Damgaard et al., 2011).To assess the extent to which an ecosystem is at risk the critical load concept has been developed (Hettelingh et al., 1995).Currently, it is estimated that about half of the European ecosystems receive nitrogen in excess to the critical load (Hettelingh et al., 2013).
flux is derived by estimating the deposition flux of cloud and fog water which is combined with the pollutant concentration in the cloud water.The cloud water concentrations are deduced from the rain water concentrations.The challenge to estimate the occult deposition is to capture the variability in the cloud deposition flux which is strongly dependent on altitude, slopes and local meteorology.Therefore we use high resolution meteorological data available for Germany as a whole, i.e. 7x7 km.Note that this resolution is not able to capture high resolution variability, which means that the occult deposition reflects background values for larger regions and do not reflect the deposition at very exposed sites.
To arrive at the final result the distributions of dry, wet and occult deposition fluxes are simply added.This addition takes place on the high resolution grid of the precipitation (1x1 Km 2 ).Note that although the fluxes are provided on this high resolution the underlying fluxes are not all resolved on this resolution.

Chemistry transport modelling
To assess the land use specific dry deposition distributions across Germany we used the 3-D regional chemistry transport model LOTOS-EUROS (Schaap et al., 2008), which is aimed at the simulation of air pollution in the lower troposphere.
The model is of intermediate complexity in the sense that the relevant processes are parameterized in such a way that the computational demands are modest enabling hour-by-hour calculations over extended periods of several years within acceptable computational time.The model is a so-called eulerian grid model, which means that the calculations for advection, vertical mixing, chemical transformations and removal by wet and dry deposition are performed on a three Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-491Manuscript under review for journal Biogeosciences Discussion started: 30 November 2017 c Author(s) 2017.CC BY 4.0 License.
dimensional grid.The LOTOS-EUROS model has a long history studying the atmospheric nitrogen and sulphur cycles.
Here we outline the main features of the LOTOS-EUROS version 1.10 used in this study.The partitioning between the gas and aerosol phase for ammonia/ammonium and nitric acid/nitrate is treated by ISORROPIA2 (Fountoukis and Nenes, 2007).Reaction of nitric acid with sea salt to form coarse sodium nitrate is included in a dynamical way.This model version also includes a pH dependent cloud chemistry scheme (Banzhaf et al., 2012).The scheme for in-and below-cloud scavenging of gases and particles accounts for droplet saturation (Banzhaf et al., 2012).The LOTOS-EUROS model is one of the few chemistry transport models that uses a description of the bi-directional surface-atmosphere exchange of NH3 (Wichink Kruit et al., 2012).The surface-atmosphere exchange module DEPAC is used for modelling the dry deposition of gases (Van Zanten et al., 2010).The module in LOTOS-EUROS was expanded to include the co-deposition effect of sulphur dioxide and ammonia.The deposition of particles is represented adapting the methodology of Zhang et al. (2001).
For a detailed analysis of the impact of including these process descriptions into LOTOS-EUROS we refer to a dedicated sensitivity study (Banzhaf et al., 2016).
The LOTOS-EUROS model was ran for the year 2009 using ECMWF meteorological data to drive the model.Through a one-way nesting procedure a simulation over Germany was performed on a resolution of 0.125° longitude by 0.0625° latitude, approximately 7 by 7 km 2 .The high resolution domain is nested in a European domain with a resolution of 0.5° longitude by 0.25° latitude, approximately 28 by 28 km 2 .The emissions that were fed into the LOTOS-EUROS model were different for the two modelling domains.For the European background simulation the TNO MACC-II European emission inventory for the year 2009 (Kuenen et al., 2014) was used.For the nest the emission data for Germany were replaced by national data.The available national data contain sector specific emissions for the year 2005 on a regular grid with a resolution of 1/60° longitude by 1/60° latitude (about 1.2 x 1.9 km 2 ).This emission inventory has been produced by the Institut für Zukunftsstudien und Technologiebewertung (IZT) and the University of Stuttgart within the PAREST project (Jörß et al., 2010).This is the most up-to-date spatially distributed inventory for Germany as a whole.Note that the emission data were produced on county basis and that land use information was used to disaggregate the emission information to a higher resolution.This means that the detail in the emission grids is limited, explaining why the modelling was not performed at higher resolutions than 7x7 Km.To account for the emission situation in 2009 the PAREST emissions for Germany were scaled on a sector basis to the officially reported emission totals for 2009 as reported in 2014 by UNECE/CLRTAP (www.uba.de).The temporal variation of the emissions is represented by monthly, day-of-the-week and hourly time factors that break down the annual totals for each source category (Schaap et al., 2004).
For evaluation purposes we use data from the national database maintained by UBA.This database includes data for sulphur dioxide (N=31) and nitrogen dioxide (N=45) at rural background locations.concentrations far above 7 µg/m 3 were removed from the analysis as they were considered hot spot locations.The modelled wet deposition fluxes were compared to observed values as presented below.

Wet deposition estimation
Traditionally, the assessment of wet deposition fluxes to ecosystems in Germany is performed with an empirical approach making use of observed wet deposition fluxes at a large number of stations (Builtjes et al., 2011;Gauger et al., 2008).In this study we derive rain water concentrations at the measurement locations and interpolate these data across Germany to arrive at a nationwide distribution.The distribution of the concentration in rain water is then multiplied with a high resolution precipitation map to arrive at the wet deposition estimates: =   *   (Equation 1) Datasets om precipitation chemistry from various national and regional monitoring programs in Germany were compiled providing information for 260 sites.The national UBA network (n=11) samples on a weekly rhythm, whereas the regional networks (n=249) may operate at a weekly, two-weekly, four-weekly or monthly basis.Unfortunately, the sampling strategies of the regional networks are not synchronised, only allowing an assessment on annual average basis.The majority of the wet deposition data is obtained with bulk samplers as only 40 stations are equipped with wet-only samplers.Hence, the data from the bulk samplers that pass our quality control procedures were corrected for the dry deposition into the funnels using species dependent correction factors (Gauger et al., 2008).As the wet deposition data are obtained from many different sources a common quality assessment and quality control (QAQC) protocol and data selection procedure was applied to the whole database.Following EMEP protocols (EMEP, 1996) the ion balance is calculated for all samples.In case the net ion-charge exceeds ±20%, the measurement is rejected.To remove further outliers a statistical outlier test is performed for the time series of each station using the Grubbs test (Grubbs, 1969).The procedure is iterative in the sense that the procedure is repeated after identifying and removing an outlier until no outliers are found anymore, or too many entries from the series are removed.As we log-transform the data in the interpolation scheme, the procedure is applied to the time series of log-concentrations.All in all, most data flagged invalid are largely due to the ion balance check.
A minimum valid data coverage of 40% for a given year was required to be included in further analyses.This criterion is a compromise between including as many stations as possible and maintaining high data quality.The 40% criterion was established based on a pragmatic approach in which we averaged the concentration in precipitation measured at UBA stations for 1000 random subsets of the available 52 weekly measurements for different data availabilities, i.e. 100%, 80%, 60%, 40% and 20%.As expected, the variability around the annual mean increases when data availability becomes smaller.
At 40% availability the standard deviation is around 15% of the mean concentration values for sulfate, nitrate and ammonium, which we feel is in line with uncertainties in precipitation amounts and other concentration data.
Within this study we used a residual kriging methodology to generate the rain water concentration distribution across Germany for 2009 (Wichink Kruit et al., 2014).Within this procedure the difference between the residual between the observations and an a priori distribution is interpolated.The a priori distribution is the modelled average rain water concentration from the LOTOS-EUROS model.The advantage of using LOTOS-EUROS distributions as a priori is that we use process knowledge in the interpolation, which results in better validation statistics (Wichink Kruit et al., 2014).As shows that for ammonium the differences can be as large as 25%, whereas the differences for nitrate and sulfate are much smaller (~10%).This can be explained by the much smaller gradients across Germany observed in the rain water concentrations for nitrate and sulfate compared to those for ammonium.
Finally, the rain water concentration is multiplied by a high resolution precipitation map for Germany (see Figure 2).This map is derived from precipitation measurements by the German Weather Service using geostatistical approach with a linear regression between precipitation and elevation (Herzog and Muller-Westermeier, 1998).A mean error of 8% was estimated for the annual precipitation amounts by (Herzog and Muller-Westermeier, 1998).We validated this distribution against the independent information on precipitation amounts from the stations with precipitation chemistry.Overall, the comparison is very good with most annual totals within 15% of each other.The higher inaccuracy reported here could well be associated to the host of different samplers and the sometimes long sampling periods (up to one month) used within the wet deposition networks.Field inter-comparison of different bulk and wet-only samplers has found it difficult to estimate precipitation volumes accurately.For instance, an accuracy better than 10% was only reached for 10-20% of the individual samples during a comparison held in the Netherlands with samplers from 20 different countries (Erisman et al., 2003).

Occult deposition estimation
The occult deposition computed within this work refers to nitrogen input by orographic clouds, which is the result of condensation processes in moist air lifted by mountains.Generally, the occult flux Foccult is derived by the multiplication of the deposition flux of fog water FFog and the pollutant concentration in the fog water CFog: The calculation of fog water deposition (FFog) follows the approach by (Katata et al., 2008(Katata et al., , 2011)).In Katata et al. (2008) a simple linear equation for the fog deposition velocity vd based only on horizontal wind speed has been derived from numerical experiments using a detailed multilayer land surface model that includes fog deposition onto vegetation (SOLVEG): where A is the slope of vd that depends on vegetation characteristics (nondimensional), and U the horizontal wind speed [m s −1 ] above the canopy.A is calculated by: 4) where LAI is the Leaf Area Index and h the canopy height [m].The calculations of A using Equation 4 agreed with observations in various cloud forests with LAI/h > 0.2 (Katata et al.,2008) and it was stated that Equation 4 can be widely used to predict cloud water deposition on forests with LAI/h > 0.  6) where ρ is the air density [kg m −3 ], qc is the liquid water content [kg water kg air -1 ] at the lowest atmospheric model layer and u the horizontal wind speed at 10 m [m s −1 ].The elevation of u may be different from that of U in Equation 3 in some cases, but this does not cause a significant error in representative wind speed according to the logarithmic wind profile in the surface boundary layer (Katata et al., 2011).
The approach following Katata (2008;2011) as described above is based on experimental data in forests and hence, provides an estimation of fog water deposition on forests only.Furthermore, the input on vegetation by fog is much more relevant for forests than for other land use categories as e.g. for grassland as the area of incidence is largest for forests when they filter the air mass passing through including fog or clouds.Hence, available studies on the occult input on vegetation are limited on forests and therefore fog water deposition on land use categories other than forest categories are neglected here.
The mean pollutant concentration in fog water (CFog) was estimated from the annual mean concentration in rainwater using so called enrichment factors (=EF): =   *  (Equation 7) Hereby the annual mean concentrations in rainwater per species stem from the interpolated concentration fields derived for the calculation of the wet deposition flux.The enrichment factors for the different species were derived from a compilation of field data from studies that provide simultaneous observations of fog and rain water chemistry (Table 1).The underpinning studies are provided in the supplementary material.Enrichment factors are greater than unity for all species as within all available studies and for all species the concentration in fog water was higher than in rain water.This can be explained by a lower dilution in fog/cloud droplets as these are smaller than rain droplets and contain less water.The variability between the individual studies is large indicating the enrichment factors may be a large source of uncertainty.

Deposition fluxes
The estimated average deposition fluxes for Germany in 2009 are summarized in Error!Reference source not found..
The estimated total deposition of reactive nitrogen amounts to 1057 eq ha -1 a -1 on average across the country.Almost two thirds (64%) of the nitrogen deposition is explained by reduced nitrogen, whereas oxidised nitrogen contributes the rest (34%).The deposition of oxidized and reduced nitrogen show distinct patterns across the country (See Figure 3).Deposition animal density in Germany.For reduced nitrogen, the estimated fluxes indicate that the contributions of dry (337 eq ha -1 a - 1 ) and wet (327 eq ha -1 a -1 ) deposition are almost equal on average.However, the relative contributions show considerable variability as in source areas for ammonia the dry deposition dominates.In more natural regions the wet deposition is about two times more important than the dry flux.For oxidized nitrogen the deposition is highest in the Ruhr area.In addition, a number of other large agglomerations can be recognized, such as Frankfurt, Stuttgart, and Berlin.As opposed to reduced nitrogen, the wet deposition (248 eq ha -1 a -1 ) is more important than the dry deposition (131 eq ha -1 a -1 ) as the dry deposition velocities of NOx are relatively small compared to those of ammonia.The dry deposition flux is strongly dependent on land use category through surface roughness and substance properties such as solubility or reactivity.In Table 2 the land use dependent dry deposition is listed.The comparison between land use classes clearly illustrates that the higher roughness of the forest classes cause increased dry deposition compared to low vegetation classes such as grasslands.The average fluxes for inland surface waters and forests are about a factor 2.5 apart.
Due to the combination of empirical results for the wet and occult deposition and modelling results for the dry deposition it is important to assess the quality of the dry deposition estimates.This can only be done indirectly as observations for dry deposition are hardly available.Of special interest is the consistency between the modelled and observed wet deposition fluxes.Below, we discuss the evaluation of the LOTOS-EUROS model results in more detail.

Evaluation of modelled concentrations
In Figure 5  Moreover, on a short time scale many of the episodes with high concentrations are captured.Similar findings have been reported focussing on total atmospheric NO2 columns in the Netherlands (Vlemmix et al., 2015).The major episode of nitrogen dioxide in January is captured less well which may be due to very stable conditions in parts of Germany.As the Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-491Manuscript under review for journal Biogeosciences Discussion started: 30 November 2017 c Author(s) 2017.CC BY 4.0 License.
exact timing of the plumes is often off by a few hours we calculated the temporal correlation coefficient on the basis of daily averages.The correlation coefficients (r 2 ) are very reasonable with values of 0.71 for NO2 and 0.59 for SO2 (see Table 3).In short, we feel that the distributions of nitrogen dioxide and sulphur dioxide on rural background stations is simulated satisfactorily.
Figure 5c shows the evaluation results for annual mean ammonia concentrations.On average, the model tends to underestimate the observed concentrations slightly and yields an explained spatial variability of 65%.Hence, the model is able to reproduce a large part of the variability and large scale gradients across Germany.Within a given region, e.g.Lower Saxony, still considerable spread around the 1:1 line is observed, which we attribute to the low level of spatial detail in the emission inventory within counties.Overall, the model performance for a regional assessment is promising.In a next step is seems logical to also investigate the seasonal cycles and search for high resolution data sets.As ammonia levels are highly variable more detailed emission information is anticipated to improve the comparison further.

Dry deposition velocity
In Table 4 the average and effective dry deposition velocities to land use classes are tabulated.The effective dry deposition velocities defined as the annual average flux divided by the annual mean concentration are usually lower than those of the average velocity.This is due to the anti-correlation between the dry deposition velocity and the atmospheric concentration of most pollutants.For example, NO2 concentrations show a day time and summer minimum, whereas the dry deposition velocity maximizes at these times.Hence, the annual effective dry deposition velocity is lower than the mean of the hourly velocities.The only exception is nitric acid because its concentration (day time and summer maximum) correlates strongly with the dry deposition velocity leading to a higher effective than average dry deposition velocity.The distribution of the annual mean and effective deposition velocities (at 2.5 m) show little variation across Germany although the seasonal variability in the more continental south is larger than in the north.The deposition velocity of ammonia behaves differently as it includes the impact of the compensation point.Figure 7a clearly illustrates the inverse relationship between the concentration level and the effective dry deposition velocity for coniferous forest for ammonia (left panel).In the large forest areas in Germany velocities up to 2 cm/s are modelled, whereas in ammonia rich areas in Lower Saxony and Bavaria values below 1 cm/s are modelled.The lower dry deposition velocity in the ammonia source areas is a direct consequence of the compensation point approach included in the dry deposition module.In Figure 7b we compare the range of modelled annual mean dry deposition velocities across Germany to a compilation of values reported in literature (Schrader and Brümmer, 2014).Note that this comparison should be considered as indicative as the literature data have been obtained by a host of different methodologies spanning different climatic conditions.Moreover, the modelled deposition velocities refer to 2.5 m height, whereas the literature data often do not specify the representative height.Still, we conclude that the range of the modelled dry deposition velocities for ammonia is plausible and that there are no indications that the modelled values are unrealistic.

Wet deposition
For the evaluation of the wet deposition fluxes of LOTOS-EUROS we compare to the data of 150 stations used for the empirical assessment of the wet deposition flux.The model underestimates the wet fluxes for all components.The underestimation is lowest for reduced nitrogen (21%), see Table 5. Oxidized nitrogen shows an underestimations of 38%.
In absolute terms the underestimation is about 140 eq ha -1 yr -1 for reactive nitrogen.In comparison to the observations the variability of the modelled wet deposition fluxes is rather small.Although models always tend to underestimate observed variability, we feel that one of the main reasons for lower variability is high spatial and temporal variability in precipitation amounts and the general challenge for meteorological models to realistically represent these variabilities.This hypothesis was tested by combining the empirically derived high resolution precipitation map and the modelled rain water concentrations.This exercise showed a considerable improvement for the spatial correlation between the calculated wet deposition fluxes.and station observations, confirming the hypothesis.It should be noted that, as expected, the exercise did not affect the bias.

The impact of empirical calculations
In case the underlying emissions and process knowledge is accurate the total modelled deposition using LOTOS-EUROS should be unbiased and thus highly consistent with the assessment results.Hence, deviations between the two provides hints at areas and components that need improvement in the modelling.The latter is important as a CTM is used to explore the effectivity of mitigation strategies.In Figure 8 we present the relative difference between the final assessed total deposition estimates and the modelled total deposition using LOTOS-EUROS.These ratio maps contain the signature of the highly resolved precipitation map as well as the occult deposition on top of a more general distribution.To remove the first structures it is advised to use higher resolved non-hydrostatic meteorological input data as well as to develop a parameterization for occult deposition in the chemistry transport model.The maps also clearly illustrate our finding that the model system underestimates the deposition of oxidized nitrogen.This underestimation is consistent with the air concentrations of nitrogen oxides.Moreover, this finding is consistent with a recent trend study showing that the oxidized nitrogen components are increasingly underestimated over time since 1995 (Banzhaf et al., 2015).In contrast, our model results for reduced nitrogen do not show indications for a large systematic difference as evidenced for large parts of western and central Germany.Only in the east towards the Polish border there are indications that the wet deposition is underestimated.In the southern half of Bavaria the model overestimates the wet deposition of ammonium and the assessment shows a lower total flux by about 20 %.This exceptional behaviour should be explained and we advise to investigate the emission variability as well as the precipitation statistics in more detail.

Comparison to previous studies
At first we compare our results previously derived nationwide deposition maps obtained for 2007 in the MAPESI project (Builtjes et al., 2011).In principle, in MAPESI the same overall approach was taken as in this study.In comparison to MAPESI the current assessment of total deposition across Germany is lower by 27% (see Table 6).This difference is largely Hence, the reduction in comparison to MAPESI is not a homogeneous reduction across the German territory.
In Table 6 also the results of this study are compared to those of EMEP for 2009 as calculated with the emission reporting of 2014 (www.emep.int).Our total N deposition is very close to EMEP results, with a difference of abut 6%.Altogether, the comparison between the best estimated reduced N deposition in PINETI-2 and the reported total N deposition by EMEP is good.The spatial distributions of the NOy and NHx deposition in the EMEP model are rather similar to ours, although it is obvious that the distributions obtained here show much more structure than the EMEP results due to the higher resolution modelling and high resolution precipitation distribution used here.With respect to oxidized nitrogen the final results for this study are slightly lower than the EMEP model results.However, the LOTOS-EUROS results are significantly lower than the results by EMEP, which is exclusively due to a difference in the wet deposition numbers of both models as the average dry deposition fluxes are almost the same.The systematic underestimation of oxidized nitrogen in precipitation from LOTOS-EUROS is currently under investigation.
To evaluate the total nitrogen deposition one relies on scientific studies that measure wet and dry deposition at a single site.
In Table 7 the N deposition results are compared with the estimates at few research sites in Germany.Forellenbach is an integrated monitoring site and is located in the Southeast of Germany in the Bavarian forest.Neuglobsow is also an integrated monitoring site and is located in the Northeast of Germany.Bourtanger Moor is a Nature2000 area that is located in the Northwest of Germany, close to the border with the Netherlands.Note that the total N deposition at these stations was determined using different methodologies.For Forellenbach and Neuglobsow our estimates are 20 % higher than estimated based on the local observations.At Bourtanger Moor, a variety of methods to determine total N deposition was explored at different locations in the nature area and a large range of total N deposition estimates was found, i.e., values were in a range from roughly 16 till 35 kg ha -1 yr -1 (Mohr, 2013).Our results for Bourtanger Moor using semi-natural vegetation is 20 Kg N ha -1 yr -1 , which is within the observed range although slightly lower than the average of all observations of 25 Kg N ha -1 yr -1 .Overall, these comparisons show differences within the anticipated uncertainty as discussed above.Unfortunately, the number of intensive monitoring stations is rather low, which highlights the need for additional locations where dry deposition fluxes are determined.

Critical loads exceedance
The Critical Load concept delivers effect-based thresholds for the maximum acceptable nitrogen deposition.We compared the established deposition flux for the year 2009 to the Critical Load dataset of Germany for eutrophication (Posch et al., 2012).Regions with rather dry conditions and/or poor sandy soils appear as rather sensitive to nitrogen deposition.About 70% of the receptor area is still at risk in the year 2009 for eutrophication due to nutrient nitrogen deposition (see Figure 9).About half of the receptor area has values up to 10 kg ha -1 a -1 nutrient nitrogen, whereas 20 % shows even larger exceedances.Highest exceedances are found for Lower Saxony, Schleswig Holstein, North-Rhein-Westphalia, Saxony and northern Bavaria.It has to be pointed out, that the critical loads and their exceedances shown here are grid average values for a grid size of 1 km² and thus valuable for a national assessment of eutrophication or acidification only, but do not serve for local assessments.One has to bear in mind that for a certain location the recommended critical loads for such small scale or vegetation type specific assessments can differ substantially from the critical loads shown here.

Conclusions
In this study we have presented the methodology to assess the deposition of reactive nitrogen to ecosystems across Germany.The methodology combines prognostic and empirical modelling to establish land use dependent dry and occult and wet deposition fluxes.On average, the nitrogen in Germany is estimated to be 1057 eq ha -1 yr -1 .Almost two thirds (64%) of the nitrogen deposition is explained by reduced nitrogen.Separate maps are available for the major land use there is considerable variability between observed concentrations at stations at distances close to each other there remains a residual between the observed and optimized distribution.Evaluations of the interpolated fields with the measured data Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-491Manuscript under review for journal Biogeosciences Discussion started: 30 November 2017 c Author(s) 2017.CC BY 4.0 License.
2. Using vd the flux of fog water deposition FFog [kg m -2 s - 1 ] is calculated using:   =   *  *   =  *  *  *   (Equation 5) where ρ is the air density [kg m −3 ], u and qc are the horizontal wind speed [m s −1 ] and the liquid water content [kg water kg air -1 ] near the surface, respectively.The accuracy of Equation 5 in the amount of fog deposition has been validated with Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-491Manuscript under review for journal Biogeosciences Discussion started: 30 November 2017 c Author(s) 2017.CC BY 4.0 License.data on turbulent fog flux over a coniferous forest in Germany (Klemm and Wrzesinsky, 2007) with a prediction error of 13% (Katata et al., 2011).The meteorological input to calculate the occult deposition flux was taken from the COSMO-EU model which is the operational NWP model of the German Weather Service (DWD).COSMO-EU was chosen as it provides the meteorological fields over Germany on a rather high grid resolution of ca.7x7 km 2 .Hourly data of the meteorological fields were used to calculate the annual fog water deposition flux based on Equation 5 with  () = ∑   () *   () *   () =  * ∑ () *   () *   () (Equation of reduced nitrogen maximises in the north west and in the south east of the country, basically mirroring the distribution of Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-491Manuscript under review for journal Biogeosciences Discussion started: 30 November 2017 c Author(s) 2017.CC BY 4.0 License.
the comparison of the modelled and observed annual average concentrations are shown.The model tends to underestimate the observed NO2 concentrations by on average 22%.For NO2 there are many stations that show a close correspondence to observed values near the one-to-one line.However, there are also a number of stations for which the modelled values are about 2-4 µg m -3 lower than those observed.Overall the gradient of NO2 over the country is addressed well.For SO2 the same conclusion can be drawn, albeit that on average a small overestimation is observed.As the modelling of all the processes including deposition occurs on an hourly time resolution it is interesting to see if the model reproduced the seasonality and variability on observation stations.Therefore, in Figure6examples for the time series comparison are shown for two stations in Germany.It can be observed that the model captures the seasonal variability in both components.
determined by two methodological development steps.Firstly, wet deposition QAQC criteria are more strict in this study Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-491Manuscript under review for journal Biogeosciences Discussion started: 30 November 2017 c Author(s) 2017.CC BY 4.0 License.and the geostatistical interpolation was improved from ordinary kriging to residual kriging resulting in a 13 % lower total deposition flux than in MAPESI(Wichink Kruit et al., 2014).Secondly, a series of model developments were consolidated in the LOTOS-EUROS version(Banzhaf et al., 2016) The most relevant improvements were the introduction of the compensation point for ammonia following Wichink Kruit et al. (2012), the update of the parameterization for the dry deposition aerosols followingZhang (2001) and the introduction of a new wet deposition parameterization for below and in-cloud scavenging followingBanzhaf et al. (2012) which accounts for droplet saturation.Whereas the inclusion of these changes hardly affects the modelled total deposition, the new process descriptions reduced the dry deposition efficiency and led to increased wet deposition fluxes for Germany on average.The shift from dry to wet deposition reduced the bias between modelled and observed wet deposition fluxes considerably, especially for reduced nitrogen.As the empirical derived wet deposition maps replace the model results, this shift impacts the resulting assessment of the total deposition across Germany.The newly modelled wet deposition fluxes by LOTOS-EUROS are closer to observations compared to MAPESI which yields a smaller correction for the wet deposition and thus a lower total deposition estimate.Note that within Germany the update of the model parameterizations also causes redistribution from source areas towards natural areas leading to a smaller decline in the assessed total deposition compared to MAPESI in the large forest areas in Germany.

Figure 1 .
Figure 1.Overview of the assessment methodology used in this study.The scheme introduces important input data 577

Figure 2 .Figure 3 .
Figure 2. High resolution precipitation map (left) and its validation against the independent data from the stations with precipitation chemistry

Figure 9 .
Figure 9. Critical load exceedance for reactive nitrogen deposition across Germany 606 607 The oxidation of nitrogen oxides to nitric acid and subsequent formation of particulate ammonium nitrate, especially during winter and spring, favours the long range transport and removal through precipitation.Wet deposition fluxes for both components show (secondary) maxima in areas with high precipitation amounts, i.e. mountainous areas like the alpine region, the Black Forest, the Erz Mountains and the Harz Mountains.The calculated contribution of occult deposition is generally negligible at low altitudes but becomes only important in the mentioned mountainous regions.Surprisingly, the occult deposition in the black forest is estimated to be quite low, which is associated with relatively low values of liquid water content near the surface within the COSMO-EU model during 2009.

Table 4 . Land use dependent annual effective and average dry deposition velocity at 2.5 meter height (above zero-displacement 619 height and roughness length) across land use types in Germany for six components in cm/s.
Biogeosciences Discuss., https://doi.org/10.5194/bg-2017-491Manuscript under review for journal Biogeosciences Discussion started: 30 November 2017 c Author(s) 2017.CC BY 4.0 License.