Quantifying the contribution of shipping NOx emissions to the marine nitrogen inventory – a case study for the western Baltic Sea

The western Baltic Sea is impacted by various anthropogenic activities and stressed by high riverine and atmospheric nutrient loads. Atmospheric deposition accounts for up to a third of the nitrogen input into the Baltic Sea and contributes to eutrophication. Amongst other emission sources, the shipping sector is a relevant contributor to the atmospheric concentrations of nitrogen oxides (NOX) in marine regions. Thus, it also contributes to atmospheric deposition of bioavailable oxidized nitrogen into the Baltic Sea. In this study, the contribution of shipping emissions to the nitrogen budget in the western Baltic Sea is evaluated with the coupled three-dimensional physical biogeochemical model MOM–ERGOM (Modular Ocean Model–Ecological ReGional Ocean Model) in order to assess the relevance of shipping emissions for eutrophication. The atmospheric input of bioavailable nitrogen impacts eutrophication differently depending on the time and place of input. The shipping sector contributes up to 5 % to the total nitrogen concentrations in the water. The impact of shipping-related nitrogen is highest in the offshore regions distant from the coast in early summer, but its contribution is considerably reduced during blooms of cyanobacteria in late summer because the cyanobacteria fix molecular nitrogen. Although absolute shipping-related total nitrogen concentrations are high in some coastal regions, the relative contribution of the shipping sector is low in the vicinity of the coast because of high riverine nutrient loads.


Introduction
The ecosystem of the Baltic Sea is exposed to anthropogenic pressures (Andersen et al., 2015;Korpinen et al., 2012;Svendsen et al., 2015). One major pressure is the high input of nutrients, i.e. bioavailable nitrogen and phosphorus compounds, leading to eutrophication (Svendsen et al., 2015). The eutrophication status has improved over the past decades (Andersen et al., 2017; were only used for the model validation and considered as tagging spin-up. The year 2012 is used for the evaluation of the contribution of shipping-related nitrogen deposition.
The atmospheric biogeochemical forcing data for the MOM-ERGOM simulations was calculated by CMAQ. The CMAQ 95 model is maintained and provided by the U.S. Environmental Protection Agency. For this study, we used CMAQ version 5.0.1 (Nolte et al., 2015;Foley et al., 2010;Appel et al., 2017) with the cb05tucl gas phase chemistry mechanism (Sarwar et al., 2007;Whitten et al., 2010;Yarwood et al., 2005) and aero5 aerosol chemistry, which is based on ISORROPIA v1.7 (Fountoukis and Nenes, 2007;Sarwar et al., 2011). Atmospheric particles are represented by a three-moment scheme containing three size modes (Binkowski and Roselle, 2003). The dry deposition parameterization for particulate matter is an updated version of 100 Binkowski and Shankar (1995), which is based on Slinn and Slinn (1980) and Pleim et al. (1984). The parameterization considers gravitational settling, aerodynamic resistance above the canopy, and surface resistance. The three modes and the three moments are deposited individually. Land based emissions were aggregated with SMOKE for Europe (Sparse Matrix Operator Kernel Emissions; Bieser et al., 2011). Marine shipping emissions were calculated with the STEAM model (Ship Traffic Emission Assessment Model; Jalkanen et al., 2012) based on data of the automatic identification system (AIS). Via AIS 105 modern ships broadcast their location, direction of travel, speed, IMO number (IMO: International Maritime Organization), and further information. Ships are considered to emit NO X but no reduced nitrogen. Sea salt emissions were calculated online (Gong, 2003;Kelly et al., 2010) without surf zone emissions (Neumann et al., 2016).
The CMAQ simulations were performed on two one-way nested model domains with increasing horizontal grid resolution Africa. The lateral boundary conditions were taken from FMI APTA global reanalysis (Sofiev et al., 2018). The first nested model domain (16 × 16 km 2 grid resolution) covered the North Sea and Baltic Sea regions. The latter data were used as atmospheric input data for the biogeochemical modeling experiments. The following CMAQ system variables were summed to obtain oxidized and reduced nitrogen deposition:  Zhang et al. (2012). Detailed DON deposition measurements were not available for the region of interest. Therefore, atmospheric deposition of DON was not considered in this study.
120 Figure 2 shows the resulting annual mean nitrogen deposition in the western Baltic Sea region and the contribution from the shipping sector.
125 Karl et al. (2019a, b) describe the model setup in more detail and present a validation for the simulation results with respect to the atmospheric deposition. The wet deposition of oxidized and reduced nitrogen was systematically underestimated at Baltic Sea stations. The reported underestimation is consistent with results of Vivanco et al. (2017). Nitrogen deposition of CMAQ simulations with very similar forcing data in the same region but in different years was evaluated. The reason for the underestimation could not be fully resolved in Karl et al. (2019a). It is assumed either that NO X to HNO 3 conversion is too 130 slow -possibly because of too low ammonia background concentrations -or that the wet removal of NH + 4 and NO − 3 is too low. Modeled atmospheric concentrations of NO X did properly reproduce measurements at EMEP stations.
The nitrogen deposition data set was bilinearly interpolated onto the MOM-ERGOM model grid resolution and supplied as daily mean values.

135
The ocean physics were simulated with the Modular Ocean Model (MOM) version 5.1 (Griffies, 2004). The whole Baltic Sea was modeled with a horizontal resolution of 3 n.m. × 3 n.m. and 134 vertical layers. Open boundary conditions were provided as climatological data in the Skagerrak, the connection to the North Sea. A dynamic ice model simulates ice cover (fraction of grid cell area), thickness and extent. MOM has been used for several studies of the Baltic Sea and has been extensively validated (e.g., Neumann et al., 2015;Radtke et al., 2012;Schernewski et al., 2015).

140
The marine biogeochemical processes are simulated with the Ecological ReGional Ocean Model (ERGOM), which has been developed at the Leibniz Institute for Baltic Sea Research Warnemünde and is still under active development (Neumann, 2000;Neumann et al., 2002;Kuznetsov and Neumann, 2013;Radtke et al., 2013;Neumann et al., 2015). It was coupled to MOM and shared the same model domain. The nitrogen deposition data were supplied in daily resolution. Riverine nutrient loads were taken from the Updated Fifth HELCOM Baltic Sea Pollution Load Compilation (HELCOM, 2015).

145
In the used ERGOM version, the biogeochemical system is represented by 31 state variables ("tracers"), of which 26 are in the water column and 5 in the surface sediment. Inorganic nutrients -i.e. nitrate (NO − 3 , ammonium (NH + 4 ), and phosphate (PO 3− 4 ) -enter the system via river input, atmospheric deposition, and remineralization of organic matter. They are consumed by phytoplankton that is represented by large phytoplankton, small phytoplankton, and cyanobacteria. Large phytoplankton starts growing at lower temperatures than small phytoplankton but processes nutrients less efficiently meaning that it runs into 150 nutrient limitation more quickly than small phytoplankton. The growth of cyanobacteria only depend on PO 3− 4 and molecular nitrogen (N 2 ), which they fix to cover their nitrogen demand. Phytoplankton, including cyanobacteria, is grazed by zooplankton. Plankton respirates and dies. Dead plankton becomes detritus that sinks to the sediment. The sediment is represented by a one-layer sediment including several relevant sediment processes such as phosphate release under anoxic conditions or denitri-fication. Nutrients may be retained in the sediment, deeply buried, or resuspended. All state variables, processes, and constants 155 are listed in a detailed model documentation in the Supplement.
Shipping related atmospheric nitrogen deposition was tagged by the method described by Ménesguen et al. (2006). It has been implemented in ERGOM and used in previous studies (e.g., Neumann, 2007;Radtke et al., 2012). All state variables containing nitrogen are duplicated: one variable containing all nitrogen in the particular compound and another variable containing only the shipping-related nitrogen. The first type of state variable is denoted as "all NAME" or "NAME all ", whereas the latter 160 type is denoted as "shipping NAME" or "NAME ship ". Process rates are calculated for the original state variables and, then, are linearly scaled according the NAME ship -to-NAME all ratio of the educts (also written as NAME ship /NAME).
Monthly mean concentrations of all state variables were written out in full spatial resolution. Basin mean concentrations were calculated from these data and, hence, are only available as monthly means. Daily mean concentrations were written out in full vertical resolution at the locations of measurement stations (see Sect. 2.5).

Study region
The western Baltic Sea was chosen as study region. It is bordered by land in the south, west, and northeast. Danish islands like Zealand and Funen are located in the center of this region (Fig. 3).
The land use south and west of the study region is dominated by agricultural activities, which lead to nutrient inputs into the Baltic Sea via rivers and the atmosphere. The population density is lower than along the southern North Sea but still high 170 inducing the input of various types of pollutants -i.e. organic pollutants, heavy metals, and plastic litter. The shipping traffic volume is high because a major European shipping route leads through this region connecting harbors in the Baltic Sea to the North Sea and more distant locations. Hence, the deposition of atmospheric shipping emissions and direct discharges of ships import pollutants and nutrients into the Baltic Sea.
The seawater of the Baltic Sea is brackish with a strong gradient in the salinity starting with 20 to 25 g/kg in the Kattegat 175 to salinities below 2 g/kg in the Bothnian Bay and in the eastern parts of the Gulf of Finland. The region of interest is characterized by strong north-south -≈ 17 g/kg in the north and ≈ 10 g/kg in the Bay of Mecklenburg -and west-east gradients -≈ 15 g/kg in the Bay of Kiel and ≈ 8 g/kg in the Arkona Basin. These salinity gradients affect the phytoplankton species composition: cyanobacteria grow only in regions with salinities below ≈ 11.5 g/kg (Wasmund, 1997).
The Baltic Sea surface water is well mixed in the upper 40 m by convection and wind induced turbulence in winter (Feistel 180 et al., 2008). No algal bloom develops as long as the water column is well mixed because algae are mixed too deep where they do not get sufficient sunlight. When the wind speeds decrease in spring, the water column becomes stratified by the development of a thermocline in 25 to 30 m depth and the temperature in the surface water rises. Hence, the beginning of the algal bloom in spring strongly correlates with calmer weather and the emergence of stratification. First, diatoms begin to bloom in the nutrient-enriched surface waters in February to May (Neumann et al., 2002). Nutrient concentrations decrease and flagellates, 185 which are more efficient in their nutrient uptake than diatoms, start blooming in April or May and reach their peak in July.
The bloom declines when one of the required nutrients is depleted in summer. The biogeochemical system is nitrogen limited in most parts of the Baltic Sea indicated by nitrogen-to-phosphorus (N:P) ratios below 16, which is the Redfield ratio (Feistel et al., 2008, Sect. 12.3, Table 12.3). Hence, excess phosphorus remains in the surface water after the diatom and flagellates blooms. The N:P ratio in riverine nutrient loads mostly is larger than 16:1 indicating phosphorus limitation (Svendsen et al., 190 2015). However, the areas affected by river plumes and phosphorus limitation are rather small. Cyanobacteria bloom in late summer. They fix dissolved N 2 and, hence, are not affected by depleted nitrate and ammonium. The algal bloom period ends in autumn when the stratification is broken up by autumn storms.

The year 2012
Only one year (2012) was considered to be analyzed with this study.

195
In 2012 were no exceptionally strong Baltic Sea inflows from the North Sea, which might have affected salinity, temperatures and physical parameters (Mohrholz, 2018a). The precipitation amount in Northern Europe in 2012 was above the long term average EMEP (2014, p. 49). Hence, the freshwater inputs were higher than in the previous years. The riverine nutrient loads were not exceptionally high compared to the long term average. Further, Savchuk (2018)  The nitrogen wet deposition in Northern Europe in 2012 was above the average of the previous ten years due to the increased precipition (EMEP, 2014). The nitrogen dry deposition in Northern Europe in 2012 was lower than in the previous ten years (higher wet leads to lower dry deposition) but the total nitrogen deposion (dry + wet) was still higher. The NO X emissions in Europe in 2012 were lower than in the previos ten years. While the deposition of oxidized nitrogen compounds in Southern

205
Europe was lower compared to previous years due to lower emissions, it slightly increased in Northern Europe due to the higher wet deposition. The ammonia emissions are treated differently in SMOKE for Europe than in the EMEP emission model. Therefore, the information on reduced nitrogen deposition in EMEP (2014) is not applicable here. Unfortunately, the Emissions by SMOKE for Europe were specifically created for the year 2012 and are not fully comparable to previously by SMOKE for Europe create emissions of other years.

210
Summarizing, the year 2012 was no exceptional year with respect to central nutrient dynamics. In the evaluation of EMEP model results, the total nitrogen deposition in Northern Europe was slightly increased due to increased precipitation. However, this might not be the case in this study because nitrogen deposition in the CMAQ data was lower than in the EMEP data.

Validation and evaluation
The results of MOM-ERGOM simulations were validated against observational data at specific stations (see Table 1  A statistical validation of the model results with measurement data is difficult because the number of observations is limited -far below one measurement per month at most stations. Therefore, seven years of data were summarized on monthly basis to a one-year 'climatology'. Climatological median and spread (10 %to 90 %-percentiles) of the measurement and model data 225 time series were then visually compared.
Three stations were chosen to be presented in the results section. They represent different regimes in the considered region: two offshore stations in different basins (OMBMPM2 and BY2) and one station close to the shore (DMU547). Validation plots at three additional stations are presented in the supplement and show a similar outcome.
The atmospheric shipping contribution to the nitrogen budget was assessed on the basis of (a) the listed stations and (b) 230 horizontal mean values per basin. Basin definitions by Omstedt et al. (2000) were used for this study (Fig. 4)  Sea surface temperature is well reproduced by MOM at all stations but sea surface salinity is overestimated at OMBMPM2 and BY2. This is a known issue and has been documented previously (e.g., Neumann and Schernewski, 2008). No measurements at the sea floor were available at DMU547. At the sea floor at OMBMPM2, the modeled salinity exceeds the measure- ments and the amplitude of the seasonal cycle of the modeled temperature seems to be too low. This might point to issues in the vertical transport in the Bay of Mecklenburg.
Modeled sea surface nitrate and phosphate concentrations agree well with the measurements, although phosphate concentrations are slightly underestimated. The seasonal pattern of modeled concentrations is realistic at all stations at the sea surface. At the sea floor, the annual cycle of nitrate does not seem to be captured by the model at OMBMPM2. Modeled nitrate concentra-245 tions increase in spring but measurements shows a decrease. Simulated salinity suggests that stratification is overestimated by the model leading to a lower impact of surface processes on deeper water layers. This also causes the damped amplitude of the annual temperature cycle. At BY2, the annual cycle of nitrate and phosphate concentrations is reproduced by MOM-ERGOM but the nitrate depletion in spring is underestimated.
3.2 Spatial pattern of shipping-related nitrogen  In the Belt Sea, the seasonal variability of the shipping contribution and its spatial variability are very low in all nitrogen fractions. The shipping contribution is between 1 % and 2 %. In the Öresund, it decreases from 1.5 % to 2 % in January to about 1 % in July and then increases again towards the end of the year. Finally in the Arkona Basin, the shipping contribution increases from the beginning of the year until summer and then decreases. The values are in a range between 1 % and 4.5 %.

275
However, there are some places in the Arkona Basin where the shipping contribution remains below 2 %.
Summarizing, the three considered basins represent three different regimes of shipping-related nitrogen deposition and of its contribution to the biogeochemical cycle. However, the relevance of shipping-related nitrogen differs spatially within each basin: the shipping contribution to the nitrogen fractions is much higher in the open ocean than along the coastline. In the surface layer, the relative shipping contribution rises in all fractions and at all stations in spring, peaks in summer, and 285 decreases again. At BY2, the PON ship /PON all ratio decreases already after June and has a minimum in August, after which it increases again. The minimum is caused by the cyanobacteria bloom because the cyanobacteria fixate non-tagged N 2 . The overall shipping contribution at DMU547 is similarly low as in the total basin. At OMBMPM2 and BY2, the TN ship /TN all ratio exceeds 5 %. At BY2, the DIN ship /DIN all ratio even exceeds 10 %. Thus, the shipping-related nitrogen contribution in summer is much higher at individual stations in the center of the basins than on basin average -in the surface layer.

290
The shipping contribution to the nitrogen fractions is much lower in the bottom layer of the three stations. It remains below 2 % in all nitrogen fractions at DMU547 and OMBMPM2. At BY2, the contribution is higher than 2 % but still considerably lower than at the surface due to vertical stratification. The PON ship /PON all ratio peaks with ≈ 6 % at the bottom of BY2 in summer.
A vertically resolved meridional cross section through the Arkona Basin is evaluated in the next section (Sect. 3.4) in order 295 to assess these differences between surface and bottom layer concentrations at station BY2 in more detail.

Vertical distribution of shipping-related nitrogen in the Arkona Basin
In the previous sections, the spatial and temporal distribution of shipping-related nitrogen has been assessed. In this section, a cross section through the Arkona Basin is presented in order to evaluate the vertical distribution of shipping-related nitrogen.  In winter, the Arkona Basin is vertically well mixed. A horizontal gradient clearly exists with low values in the south and high values in the north. In spring, the DIN ship /DIN all ratio increases in the central Arkona Basin at the sea surface and a vertical gradient develops. One to two month later, the TN ship /TN all ratio also develops a vertical gradient. This time lag is reasonable because, first, a signal appears in the DIN due to external DIN input and, then, spreads to PON and DON.
The surface layer DIN ship /DIN all ratio increases until July exceeding 10 % and, then, strongly decreases. The maximum of 305 the TN ship /TN all is at the sea surface until June 2012 when it reaches 6 %. In the subsequent months, the maximum migrates downward and decreases. In July, the maximum is at ≈ 15 m depth and amounts ≈ 5.5 %. In August, it is at ≈ 20 m and amounts ≈ 4.7 %. The downward migration is reasonable: Detritus with high shipping contribution sinks towards the seafloor as a result of the phytoplankton bloom in spring. In the open sea in early summer, production of fresh PON decreases due to nutrient limitation. PON with a high content of non-310 shipping nitrogen is produced in coastal regions (nutrients supplied by rivers), is horizontally mixed from the coast towards the open sea, and sinks. As a result, the maximum of the PON ship /PON all ratio seems to migrating downward (Fig. 11, top row). If the PON concentration is much higher than the DIN concentration, which is commonly the case in summer, the TN ship /TN all ratio will behave similarly to the PON ship /PON all ratio as indicated by the bottom row of Fig. 11.

Summary of resuls 315
The validation of simulations results showed a good agreement of physical and biogeochemical model data with in-situ measurements (Fig. 5).
The wet deposition of oxidized and reduced nitrogen was systematically underestimated at Baltic Sea stations. The reported underestimation is consistent with results of Vivanco et al. (2017). Nitrogen deposition of CMAQ simulations with very similar forcing data in the same region but in different years was evaluated. The reason for the underestimation could not be fully 320 resolved in Karl et al. (2019a).
The deposition of untagged and shipping-related nitrogen was very high along the coastline. Particularly in bights and river estuaries the nitrogen deposition was considerably high. Reasons for this are discussed in the Discussion section below.
The concentration of shipping-related total nitrogen (TN ship ) was relatively homogeneously distributed horizontally (Fig. 7).
A few coastal regions showed increased TN ship concentrations. Relatively, the contribution of shipping-related nitrogen to TN 325 (TN ship /TN all ) was highest distant to the coast due to the lack of riverine nitrogen sources in these areas.
In the surface waters of the Arkona Basin and of the Bay of Mecklenburg, the shipping contribution to all nitrogen fractions was highest in summer and lowest in winter (Fig. 8). The contribution of shipping-related nitrogen to particulate organic nitrogen (PON ship /PON all ) strongly decreased in the Arkona Basin in August caused by a cyanobacteria bloom. In the bottom water, the shipping contribution was quite constant all over the year due to stable vertical stratification during the bloom period.

330
An exception was PON ship /PON all in summer in the bottom water of the station BY2, which is located in the center of the Arkona Basin (Fig. 10). It was caused by large amounts of sinking detritus.
In the Öresund, the annual cycle of the shipping contribution to all nitrogen fractions was inverted to the cycle in the Arkona Basin (Fig. 8) This might affect the vertical transport -too weak vertical mixing -and lead to higher nutrient concentrations at the sea floor.
Another reason might be that nutrients are released from the sediment into the water column in this region. The latter hypothesis 355 cannot be tested because no sediment measurement data with high temporal resolution were available at this station.
The sea surface concentrations and their annual cycle seem to be well reproduced at all three stations -OMBMPM2, OMBMPN1, and OMBMPM. Hence, we assume that the observed deviations at the sea floor of the Bay of Mecklenburg do not negatively affect the general results of this study.
Very high PON and TN concentrations occurred at the station DMU547 in September and December. These were caused by 360 resuspension of detritus through high current velocities at the sea floor (see plots in the Supplement for details).

Discussion of atmospheric nitrogen inputs
The wet deposition of oxidized and reduced nitrogen was systematically underestimated at Baltic Sea stations. The reported underestimation is consistent with results of Vivanco et al. (2017). Nitrogen deposition of CMAQ simulations with very similar forcing data in the same region but in different years was evaluated. The reason for the underestimation could not be fully 365 resolved in Karl et al. (2019a).
The deposition of untagged and shipping-related nitrogen was very high along the coastline. Particularly in bights and river estuaries the nitrogen deposition was considerably high. This is partly of artificial origin and parly a result of specific atmospheric processes as we will describe below.
Atmospheric nitrogen deposition is higher above the land than above the ocean (Seinfeld and Pandis, 2016). Hence, there 370 is a steep gradient in the nitrogen deposition away from the coastline. Coarser horizontal grid resolution of the CMAQ setup compared to the MOM-ERGOM setup and subsequent interpolation of nitrogen deposition data over the land-sea interface cause a smoothing of nitrogen deposition in this region leading to artificillay enhanced deposition into the coastal waters.
The second reason, which is non-artificial, is probably the interaction in the atmosphere between nitrogen oxides (NO X ) from shipping, ammonia (NH 3 ) from agricultural activities and animal livestock, and sea salt particles emitted from the sea 375 surface. Although this topic is not in the focus of this study, we described some details in the subsequent paragraphs.
The NO X reacts to HNO 3 . HNO 3 condenses on wet particles and reduces the pH of the particle water (Reaction R1).
NH 3 condenses on wet particles and increases the pH of the particles' water (Reaction R2). Both processes are equilibrium processes. When both processes take place at the same time, then the pH is kept on a roughly constant level shifting the equilibirum towards the right side Reaction R3.
Additionally, natrium chloride (NaCl; Na + Cl − ) favors the condensation and deprotonization of atmospheric acids, such as HNO 3 (Reaction R4). The condensation of HNO 3 reduces the pH of the particles' water. Hydrochloric acid HCl is a weaker acid than HNO 3 . Hence, Cl − has a higher probability to accept a proton (and to evaporate subsequently) than NO − 3 .

HN O
Sea salt emissions considerably contribute to the atmospheric particle load in the vicinity to the shoreline and favor the for-390 mation of particulate nitrogen compounds. Sea salt particles are relatively large and, hence, have a short atmospheric residence time meaning they are quickly deposited. Therefore, shipping-related nitrogen deposition is expected to be enhanced in some coastal regions through the interaction of shipping-related NO X and sea salt particles.

Discussion of the shipping contribution
The concentration of shipping-related total nitrogen (TN ship ) was relatively homogeneously distributed horizontally (Fig. 7).

395
A few coastal regions showed increased TN ship concentrations. Relatively, the contribution of shipping-related nitrogen to TN (TN ship /TN all ) was highest distant to the coast. The pattern agrees with regions of high shipping activity. This is a coincidence no interdependence. The actual reason for the pattern is the lack of riverine nitrogen sources in offshore regions.
In the surface waters of the Arkona Basin and of the Bay of Mecklenburg, the shipping contribution to all nitrogen fractions was highest in summer and lowest in winter (Fig. 8). The contribution of shipping-related nitrogen to particulate organic 400 nitrogen (PON ship /PON all ) strongly decreased in the Arkona Basin in August caused by a cyanobacteria bloom. In the bottom water, the shipping contribution was quite constant all over the year due to stable vertical stratification during the bloom period.
An exception was PON ship /PON all in summer in the bottom water of the station BY2, which is located in the center of the Arkona Basin (Fig. 10). This is discussed further below.
In the Öresund, the annual cycle of the shipping contribution to all nitrogen fractions was inverted to the cycle in the Arkona 405 Basin (Fig. 8). In the summer, the cycle shows a minimum in the Öresund and a maximum in the Arkona Basin. Particularly in summer, atmospheric deposition is an important nutrient source for large basins such as the Arkona Basin. The fraction between sea surface area and coastline length is lower in the Öresund than in the Arkona Basin. Moreover, the Öresund is considerably impacted by nutrient loads from the Swedish mainland and from Zealand (Mølleåen River). This leads to a lower relevance of atmospheric nitrogen deposition compared riverine nutrient loads and causing the inverted annual cycle.  followed by a steep decline in late summer, which is caused by a cyanobacteria bloom. This result is comparable to this study's result at BY2 were the cyanobacteria bloom had a similar effect. At the other two stations in this study the physical conditions do not allow cyanobacteria blooms and, hence, do not show this result. The spatial pattern of the contribution of shippingnitrogen to NO − 3 and DIN in Raudsepp et al. (2019) and in this study, respectively, is very similar with respect to increased shipping-nitrogen in some coastal regions. This can be expected due to similar nitrogen deposition data sets.

435
In this study, shipping-related nitrogen was tagged in one simulation. The biogeochemical system behaved in the same way as when shipping-related nitrogen had not been tagged. In contrast, two simulations with and without shipping-related nitrogen inputs were performed in Raudsepp et al. (2019). The results were subtracated from each other in a second step and the difference was evaluated ("difference-approach"). The two simulations might reveal different system dynamics because the biogeochemical system is a complex non-linear system. Hence, the results of this study are not one-to-one comparable to those 440 of Raudsepp et al. (2019).
Both valid approaches to assess the contribution of shipping activities to the nitrogen cycle are valid but should be used for different research questions. The difference-approach by Raudsepp et al. (2019) is very useful when we want to assess what would happen if shipping emissions are reduced or totally avoided. The system behaviour might change in this situation and, hence, two distinct simulations should be performed. The tagging-approach would no capture the non-linear changes in 445 the system behaviour. In contrast, the tagging-approach is reasonable when we want to assess the contribution of one or more nutrient sources to state variables in the current situation. We are mainly interested in the latter aspect and, hence, have choosen the tagging approach in this study.

Conclusions
Following Raudsepp et al. (2013), Neumann et al. (2018), and Raudsepp et al. (2019) this is the fourth study dealing with 450 tracing shipping-related nitrogen inputs in the Baltic Sea biogeochemical system. This study focused on the western Baltic Sea using a state-of-the-art biogeochemical model.
The absoluted contribution of the shipping sector to TN was highest along the shoreline, which was caused by the interaction of shipping-related NO X with sea salt particles and ammonia in the atmosphere and subsequent dry deposition. However, the relative contribution of the shipping sector to TN showed an inverted pattern: lowest contribution along the shoreline and 455 increasing towards the open sea. Riverine nutrient inputs led to a relatively low relevance of atmospheric shipping-related inputs along the shoreline . Hence, offshored regions rather than coastal regions might benefit from reduced inputs of shippingrelated nitrogen.
The contribution of shipping-related nitrogen to TN was below 5 % on large scale on annual average. Hence, measures like nitrogen emission control areas, which limit the NO X emissions of ships, are expected to have a low impact on eutrophication 460 on large scale. However, the shipping contribution to TN exceeded 5 % in the centers of the Basins in summer. The shippingrelated DIN was even in a range between 10 % and 15 % in the center of the Arkona Basin. Hence, the shipping sector -and atmospheric deposition in general -is an important nutriert source in offshore regions in summer.
The vertical distribution of nitrogen indicated that sinking of detritus leads to the transport of shipping-related nitrogen into sediment. An assessment of the sedimentary nitrogen composition is not reasonable in this study due to the simple sediment 465 parameterization used. Hence, it is not clear what part of shipping-related nitrogen is buried in the sediment and what part is released back into the water column -either as bioavailable nitrogen or as N 2 . Future studies should focus on the sedimenti.e. with a more sophisticated sediment model.
The contribution of shipping-related nitrogen to TN seems to be low taking values below 5 % on average. However, we do not have comparable numbers of the contribution of other atmospheric nitrogen emission source sectors, i.e. road traffic (NO X ), 470 power production (NO X ), and livestock farming (ammonia/ammonium, NH 3 /NH + 4 ). In this context, this study is rather one case study. Future studies should target several source sectors in order to be able to put the relative contributions of individual source sectos into context. -The CMAQ nitrogen deposition data are available upon request from the co-authors of the HZG. Some results of the CMAQ simulations are available via the SHEBA THREDDS server http://sheba.hzg.de/thredds/catalog/publicAll/WP2-Air/catalog.html.