Nitrate source identification in the Baltic Sea using its isotopic ratios in combination with a Bayesian isotope mixing model

Nitrate (NO−3 ) is the major nutrient responsible for coastal eutrophication worldwide and its production is related to intensive food production and fossil-fuel combustion. In the Baltic Sea NO−3 inputs have increased 4-fold over recent decades and now remain constantly high. NO −3 source identification is therefore an important consideration in environmental management strategies. In this study focusing on the Baltic Sea, we used a method to estimate the proportional contributions of NO−3 from atmospheric deposition, N 2 fixation, and runoff from pristine soils as well as from agricultural land. Our approach combines data on the dual isotopes of NO−3 (δ N-NO−3 andδ O-NO−3 ) in winter surface waters with a Bayesian isotope mixing model (Stable Isotope Analysis in R, SIAR). Based on data gathered from 47 sampling locations over the entire Baltic Sea, the majority of the NO−3 in the southern Baltic was shown to derive from runoff from agricultural land (33–100 %), whereas in the northern Baltic, i.e. the Gulf of Bothnia, NO−3 originates from nitrification in pristine soils (34–100 %). Atmospheric deposition accounts for only a small percentage of NO −3 levels in the Baltic Sea, except for contributions from northern rivers, where the levels of atmospheric NO −3 are higher. An additional important source in the central Baltic Sea is N 2 fixation by diazotrophs, which contributes 49–65 % of the overall NO−3 pool at this site. The results obtained with this method are in good agreement with source estimates based upon δ15N values in sediments and a three-dimensional ecosystem model, ERGOM. We suggest that this approach can be easily modified to determine NO −3 sources in other marginal seas or larger near-coastal areas where NO −3 is abundant in winter surface waters when fractionation processes are minor.


Introduction
Throughout the world, anthropogenic reactive N currently exceeds natural production (Galloway et al., 2003;Gruber and Galloway, 2008).Consequently, riverine nitrogen (N) fluxes have doubled in recent years, which has strongly impacted the marine N cycle and ecosystem health, both at regional and global scales.In coastal ecosystems, the adverse effects of these excess N loads include eutrophication, hypoxia, loss of biodiversity, and habitat destruction (Galloway et al., 2003;Villnäs et al., 2013).For the shallow, brackish, semi-enclosed Baltic Sea, where intense anthropogenic nutrient loadings have been documented since the 1950s (Elmgren, 2001), riverine and atmospheric nutrient inputs are now at least 4-fold higher than a century ago, when anthropogenic influence was low (Schernewski and Neumann, 2005;Stålnacke et al., 1999).Furthermore, cyanobacterial blooms, which can fix N 2 , and thus add nutrients to the surface waters are regular large scale phenomenon each summer (Finni et al., 2001;Vahtera et al., 2007) and the overall increase in nutrient input has supported the expansion of hypoxic zones (Conley et al., 2009(Conley et al., , 2011)).
A main component of the N pool and the one most readily available is nitrate (NO − 3 ) (Nestler et al., 2011;Vitousek et al., 1997), which derives from a wide variety of sources.These can be identified by analysis of the N and oxygen (O) isotopes (δ 15 N-NO − 3 and δ 18 O-NO − 3 ) since the isotopic ratios of NO − 3 from different sources fall within distinct ranges (Kendall, 1998;Kendall et al., 2007).For example, NO − 3 inputs from forested catchments can be discriminated from those coming from agricultural runoff, and the NO − 3 from atmospheric deposition (blue), pristine soils (red), agricultural runoff (green), and N 2 fixation (black), for the Western Baltic Sea, Baltic Proper, Gulf of Finland, Gulf of Bothnia, southern rivers, and northern rivers.Stations are indicated as black dots.For more details see Supplement Table 1.signature of microbial nitrification differs from that of atmospheric deposition (Kendall, 1998;Kendall et al., 2007;Mayer et al., 2002).Source attribution is, however, complicated by N-transformation processes such as denitrification, nitrification, and assimilation, each of which gives rise to significant isotope fractionation.Since heavier isotopes are sequestered more slowly than lighter ones, the reaction product will be isotopically depleted compared to the original NO − 3 source (Kendall, 1998).Alterations of isotope values because of microbial fractionation processes can be minimized by collecting the samples in winter, when low water temperatures reduce microbial activity (Pfenning and McMahon, 1997).
Nonetheless, source attribution is still complicated when there are more than three sources but only two isotopes that describe them (Fry, 2013).SIAR (Stable Isotope Analysis in R), a Bayesian isotope mixing model originally developed to infer diet composition from the stable isotope analysis of samples taken from consumers and their food sources (Moore and Semmens, 2008), was already successfully applied for NO − 3 source identification.Xue et al. (2012Xue et al. ( , 2013) ) were able to estimate the proportional contributions of five potential NO − 3 sources in a small watershed in Flanders (Belgium).Based on their determinations of the isotopes of nitrogen and oxygen they could show that manure and sewage were the major sources of NO − 3 .In the Baltic Sea the NO − 3 pool present in the surface waters in spring originates from the previous growth season and is consumed during the onset of the phytoplankton spring bloom, in February/March.Stratification in summer hinders circulation down to the halocline, thus atmospheric deposition and N 2 fixation are the major N sources, whereas in coastal areas riverine discharge dominates (Radtke et al., 2012;Voss et al., 2011).Yet, to what extent the various NO − 3 sources add to the overall pool of NO − 3 in the Baltic as a whole is still a matter of debate.In this study, a source attribution for four major sources is presented.Taking the Baltic Sea as an example we will show, that the use of the isotopic composition of NO − 3 (δ 15 N-NO − 3 and δ 18 O-NO − 3 ) in combination with SIAR can be used elsewhere for source identification on an ecosystem scale level.

Field sampling
Surface water samples from the Baltic Sea were collected in February 2008 (n = 22) and 2009 (n = 17) before the onset of the phytoplankton spring bloom aboard the RV Alkor and in November 2011 (n = 1) aboard the RV Meteor using a Seabird CTD system with attached water bottles.Samples from the Nemunas River (55 • 18 5.5 N, 21 • 22 53.9E; 55 • 41 25.6 N, 21 • 7 58.4E; n = 4) and the Kalix River (65 • 56 4.2 N, 22 • 53 9.2 E; n = 1) (Fig. 1) were taken between November 2009 and February 2010.Values for NO − 3 in which atmospheric deposition was the source were obtained from wet deposition samples collected at three stations around the Baltic Sea: Warnemünde, Germany (54 • 10 N, 12 • 5 E,); Majstre, Sweden (57 • 30 N, 18 • 31 E); and Sännen, Sweden (56 • 13 N, 15 • 17 E) from December 2009 until February 2010 (Table 1).In Warnemünde, precipitation was collected on an event basis, and retrieved daily to limit microbial degradation, using a sampler consisting of a plastic funnel (diameter: 24 cm) connected to a 1 L polyethylene Here, the sampler consisted of a plastic funnel (diameter 20.3 cm) connected to an 8-L polyethylene bag.All samples were filtered through pre-combusted Whatman GF/F filters (4 h at 400 • C) and stored frozen until further analysis.

3
Samples were analyzed following a standard protocol for the determination of NO − 3 and nitrite (NO − 2 ) (Grasshoff et al., 1983); the precision of the method is ± 0.02 µmol L −1 .Dual isotope analysis of NO − 3 (δ 15 N-NO − 3 and δ 18 O-NO − 3 ) was carried out using the denitrifier method (Casciotti et al., 2002;Sigman et al., 2001), in which NO − 3 and NO − 2 are quantitatively converted to nitrous oxide (N 2 O) by Pseudomonas aureofaciens (ATTC 13985), a bacterial strain that lacks N 2 O reductase activity.In brief, N 2 O is removed from the sample vials by purging with helium and then concentrated and purified in a GasBench II prior to analysis with a Delta Plus mass spectrometer (ThermoFinnigan).NO − 2 was not removed since its concentrations were always less than 2 % (referring to the procedure described in Casciotti et al., 2007).N and O isotope measurements of roughly 30 % of the samples were replicated in separate batch analyses.Two international standards, IAEA-N3 (δ 15 N = 4.7 ‰ vs. N 2 ; δ 18 O 25.6 ‰ vs. VSMOW) and USGS 34 (δ 15 N −1.8 ‰ vs. N 2 ; δ 18 O −27.9 ‰ vs. VSMOW) (Böhlke et al., 2003), were measured with each batch of samples.Samples with NO − 3 / NO − 2 concentrations as low as 1 µmol L −1 were analyzed.The sample size for the actual stable isotope measurements was 20 nmol for samples with concentrations > 3.5 µmol L −1 and 10 nmol for those with concen-trations < 3.5 µmol L −1 .Isotope values were corrected after Sigman et al. (2009) for δ 18 O-NO − 3 ; single point correction was referred to IAEA-N3 for δ 15 N-NO − 3 .The precision was < 0.2 ‰ for δ 15 N and < 0.6 ‰ for δ 18 O.Together with the samples, a culture blank was analyzed to which no sample was added.The isotope ratios are reported using the delta notation in units of per mil (‰).

NO − 3 sources
To estimate the contribution of different NO − 3 sources, two isotopes δ 15 N-NO − 3 and δ 18 O-NO − 3 (j = 2) from the four major NO − 3 sources: (1) atmospheric deposition, (2) runoff from pristine soils, (3) runoff from agricultural land and (4) N 2 fixation were applied (Table 2).In this context, N 2 fixation was defined as NO − 3 originating from the degradation and remineralization of nitrogen fixers and therefore carried their low isotopic signal.Thus, for NO − 3 from N 2 fixation, δ 15 N values of ∼ −2 to 0 ‰ were assumed, since N 2 fixation produces organic material that is only slightly N depleted against air nitrogen (Carpenter et al., 1999(Carpenter et al., , 1997;;Montoya et al., 2002).The δ 18 O values were estimated to be between −3.8 ‰ and 2.0 ‰, based on measurements in the subtropical northeast Atlantic where N 2 fixation was the main source of N (Bourbonnais et al., 2009)  To expand the data set, we included NO − 3 isotope data from river water samples, ground water samples, and samples from tile drain outlets collected in 2003 and published in Deutsch et al. (2006).In that study, the Warnow River (n = 2) was sampled twice, in January and February 2003.These sources were likewise sampled in winter, since marked seasonal shifts in the isotopic composition of NO − 3 can occur due to shifts in the origins of the sources (Knapp et al., 2005).Samples from tile drain outlets were used to represent NO − 3 from agricultural runoff and were obtained from the catchment of the Warnow River, whose waters are strongly influenced by agricultural land use (Pagenkopf, 2001).High δ 15 N-NO − 3 values of 9.9 ± 1.5 ‰ and lower δ 18 O-NO − 3 values of 4.6 ± 1.0 ‰ are typical for areas that are influenced by agricultural activities and are similar to studies of Wankel et al. (2006) and Johannsen et al. (2008).Johannsen et al. (2008) found in the rivers Rhine, Elbe, Weser and Ems, with comparable high agricultural activities, δ 15 N-NO − 3 values between 8.2 and 11.2 ‰ and δ 18 O-NO − 3 values from 0.4 to 0.9 ‰ in winter.However, a differentiation between NO − 3 from mineral fertilizers and sewage/manure was not done; rather a mixed signal from rivers that are mainly influenced by agricultural activities was taken.Groundwater samples were used as the source of NO − 3 from pristine land (Deutsch et al., 2006).Their δ 15 N-NO − 3 and δ 18 O-NO − 3 values significantly differed from those of agricultural runoff (p < 0.05) but were similar to the values of other areas, such as Biscuit Brook (Burns et al., 2009) and the San River (Koszelnik and Gruca-Rokosz, 2013), where pristine soils were sampled and reflect nitrification activity in soils unaffected by human activity.
The dual isotopes of NO − 3 values presented in Deutsch et al. (2006) were analyzed according to Silva et al. (2000).In this method, NO − 3 is chemically converted via anion exchange resins to AgNO − 3 and the δ 15 N-NO − 3 and δ 18 O-NO − 3 values are measured via pyrolysis and isotopic ratio mass spectrometry (for a detailed description, see Deutsch et al., 2006).A normal distribution of the isotopic data from the four sources was confirmed by applying the Shapiro-Wilk normality test.δ 15 N-NO − 3 and δ 18 O-NO − 3 values from NO − 3 from atmospheric deposition of 0.3 ± 1.4 ‰ and 76.7 ± 6.8 ‰, respectively, are also in line with literature values.The δ 15 N values of atmospheric NO − 3 are usually between -15 to +15 ‰ and the δ 18 O between 63 and 94 ‰ (Kendall et al., 2007).
Six regions within the catchment of the Baltic Sea were investigated for their potential NO − 3 sources (Fig. 1).According to the topography of the Baltic Sea, the samples were assigned to four major areas: Western Baltic Sea, Baltic Proper, Gulf of Finland, and Gulf of Bothnia.Additionally, three rivers differing in their degree of anthropogenic impact were included in this study and divided into two groups: northern and southern rivers.Rivers with high nutrient loads drain mainly into the southern Baltic Proper and were represented here by the Nemunas and Warnow Rivers, whose NO − 3 concentrations in winter can be as high as 260 µmol L −1 (Deutsch et al., 2006;Pilkaityte and Razinkovas, 2006).The Gulf of Bothnia receives large amounts of fresh water from rivers represented by the Kalix River.These rivers drain mainly pristine, forested land and have maximum NO − 3 concentrations of around 20 µmol L −1 (Sferratore et al., 2008).

SIAR mixing model
The applied mixing model is described by the following equations: where X ij is the observed isotope value j of the mixture i; i = 1, 2, 3, . .., I are individual observations; and j = 1, 2, 3, . .., J are isotopes.s j k is the source value k of isotope j (k = 1,2,3, . .., K) and is normally distributed, with a mean of µ j k and a standard deviation of ω j k .p k is the proportion of source k that needs to be estimated by the model.c j k is the fractionation factor for isotope j on source k and is normally distributed, with a mean of λ j k and a standard deviation of τ j k .ε ij is the residual error representing additional unquantified variations between mixtures and is normally distributed, with a mean of 0 and a standard deviation of σ j .Detailed descriptions of the model can be found in Jackson et al. ( 2009), Moore andSemmens (2008), andParnell et al. (2010).As noted above, by collecting samples between November and February we minimized the influence of fractionation processes such as assimilation and denitrification that can alter the isotopic signal of NO − 3 .Therefore in Eq. ( 1) we assumed that c j k = 0.
Two different runs of the SIAR model were performed.In the first, for the Western Baltic Sea, Baltic Proper, and Gulf of Finland, all four sources were included in the calculation.In the second, for the Gulf of Bothnia, the southern rivers, and the northern rivers, N 2 fixation as a potential NO − 3 source was excluded since in these areas there is no N 2 fixation by diazotrophs because the Gulf of Bothnia is phosphorus limited, in contrast to the Baltic Proper (Graneli et al., 1990).
Highest nitrate concentrations in the Nemunas River also corresponded to the highest δ 15 N-NO − 3 with 10.0 ‰ and vice versa, with lowest concentrations and nitrogen isotope values in the Baltic Proper (1.5 ‰).The δ 18 O-NO − 3 values ranged from −2.8 ‰ in the Gulf of Bothnia to 10.6 ‰ in the Northern River, Kalix (Fig. 2, Supplement 1).

3
SIAR calculated that in the southern Baltic Sea, agricultural runoff was the main NO − 3 source with the highest contribution in the western Baltic Sea with up to 67 % (mean 53.5 ± 3.2 %) and in the southern rivers with up to 100 % (mean 93.5 ± 4.1 %) (Table 3, Fig. 1).NO − 3 from atmospheric deposition was negligible with 3.5 % (mean 1.1 ± 0.5 %) and NO − 3 from pristine soils lower with up to 42 % (mean 7.5 ± 5.9 %) in the western Baltic Sea (Table 3, Fig. 1).In the Baltic Proper, NO − 3 from N 2 fix-ation was the dominant NO − 3 source with up to 65.3 % (mean 58.8 ± 2.0 %) (Table 3, Fig. 1).In the northern Baltic Sea NO − 3 from atmospheric deposition is only important in the northern rivers with a contribution of up to 23.4 % (mean 11.8 ± 1.5 %) (Table 3, Fig. 1).NO − 3 from pristine soils is mainly transported by the northern rivers (75.3 ± 7.9 %) to the Gulf of Bothnia, where SIAR calculated that 99.0 ± 0.9 % stems from the runoff from pristine soils (Table 3, Fig. 1).

Comparison of isotope patterns in the water column and sediments
The δ 15 N values from surface water correlated significantly with those from surface sediments, as reported in Voss et al. (2005) (p < 0.001) (Fig. 3).Stations for sediment sampling were in close vicinity to stations from water column sampling (Fig. 4).In the Baltic Proper, the δ 15 N of the surface water NO − 3 was indistinguishable from the δ 15 N of the sediment surface (3.6 ± 1.0 and 3.5 ± 0.6 ‰, respectively; Table 4).In the near-coastal areas of the Baltic Proper and the Gulf of Finland, the δ 15 N of surface water NO − 3 was 7.9 ± 1.8 ‰, slightly higher than the surface sediment value for the same area of 7.3 ± 2.1 ‰ (data in Voss et al., 2005) but still not significant different (p < 0.01) (Table 4).

NO −
3 in the Baltic Sea The measured winter surface water concentrations of up to 259 µmol L −1 are typical for eutrophic systems and similar values have been reported from the Chesapeake Bay and the coastal areas of the North Sea (Dähnke et al., 2010;Francis et al., 2013).The concentrations of nutrients in the sub-basins of the Baltic Sea reflect the densities of the human populations in the vicinity of the adjacent sub-catchments.Thus, in the near-coastal area of the southern Baltic Proper, NO − 3 concentrations were higher than in the northern parts, since the catchment areas of Germany, Poland, and the Baltic States are much more densely populated (> 500 inhabitants km −2 ) and the land is intensively used for agricultural purposes.The northern regions are dominated by boreal forests and less populated (< 10 inhabitants km −2 ) (Lääne et al., 2005;Stepanauskas et al., 2002;Voss et al., 2011).Consequently, for the southern Baltic Proper a relationship between fluvial NO − 3 loads and NO − 3 concentrations in coastal waters could be established that indicates a direct impact of riverine nutrients on coastal waters (Voss et al., 2011;HELCOM, 2009).However, there was no similar correlation between riverine N loads and nutrient concentrations either for the coastal areas of the Gulf of Bothnia or for the open waters of the Baltic Proper (Voss et al. 2011).The Gulf of Bothnia is the only sub-basin in which the effects of eutrophication are so far minor, although Lundberg et al. (2009) and Conley et al. (2011) reported a degradation in the water quality from north to south and from the outer to the inner coastal area of the Gulf, with seasonal hypoxia at many sites.Trends of increasing nutrient levels should be interpreted as a warning signal for the future and highlight the need for management approaches based on sound knowledge of the many potential sources of NO − 3 .In the Gulf of Finland, which is regarded as the most heavily eutrophic sub-basin of the Baltic Sea, a consequence of high receiving nutrient loads from the Neva River and the city of St. Petersburg (Lundberg et al., 2005), NO − 3 concentrations were about 2-fold higher (7.6 ± 0.9 µmol L −1 ) compared to the rest of the Baltic Sea sub-basins, where concentrations in winter were almost identical.This shows that NO − 3 concentrations alone cannot be used to identify NO − 3 sources for the sub-basins; rather, stable NO − 3 isotopes values allow for accurate source determination, as we will show in the following sections.

3
The use of NO − 3 stable isotopes for source identification is complicated when the mixing of multiple N sources with overlapping isotopic ranges occurs together with microbial processes such as nitrification, assimilation, and denitrification (Kendall, 1998;Wankel et al., 2006).In this study, we assumed that the effects of fractionation by microbial processes were negligible because all our samples were collected in winter, at a mean temperature of 3.1 ± 1.3 • C (data not shown), when microbial activity is low (Pfenning and McMahon, 1997), as confirmed in a study of nitrification in the Baltic Sea by Jäntti et al. (2011).They showed that in the Gulf of Finland although nitrification potentials may be high during cold months, in situ nitrification is undetectable, whereas the rate increases progressively towards the summer.
We are aware that the variability of the source signals must be taken into account in source attributions.Both Xue et al. (2012Xue et al. ( , 2013) ) and Yang et al. (2013) showed that SIAR can be applied in NO − 3 source identification, although the resolution of this model is largely determined by the uncertainty of the isotopic composition of the sources.In the studies of both groups, the means and variances of the sources were calculated mostly from literature values, which were not obtained in the investigation areas, nevertheless they received consistent results.In contrast, in our study, the isotopic composition of the sources, except NO − 3 from N 2 fixation, was determined from samples obtained within the study area.In our calculations we considered the impact of the variability of the sources and report not only mean values and error estimates, but also minimum and maximum contributions, as suggested by Fry (2013) (Table 3).

NO − 3 from agricultural runoff
The isotopic values of riverine NO − 3 were previously shown to be enriched when agricultural land is the source of inputs 3 from surface water samples.δ 15 N values from sediments were taken from Voss et al. (2005).The positive slope suggests a tight coupling between δ 15 N-NO − 3 in surface waters and δ 15 N in sediment samples.(Johannsen et al., 2008;Mayer et al., 2002;Voss et al., 2006).Catchments with high percentages of agricultural and/or urban land use export NO − 3 with δ 15 N-NO − 3 values of around 7 ‰.In the same study, the oxygen isotope ratios of NO − 3 were almost uniformly 13 ± 1 ‰ (Mayer et al., 2002).Johannsen et al. (2008) measured δ 15 N-NO − 3 values of 11.3 ‰ in highly eutrophic rivers draining into the North Sea, whereas the highest δ 18 O-NO − 3 value was 2.2 ‰.In the Oder River outflow, a main NO − 3 contributor to the Baltic Sea, δ 15 N-NO − 3 of 7.6 ‰ and δ 18 O-NO − 3 of 2.9 ‰ were determined (Korth et al., 2013).Our measurements for the Warnow and Nemunas Rivers fall in the expected range, with a mean δ 15 N-NO − 3 of 9.2 ‰ and a mean δ 18 O-NO − 3 of 3.1 ‰, and are consistent with the high percentages of agricultural land in the river catchment areas: 50 % for the Warnow River (Pagenkopf, 2001) and 50 % for the Nemunas River (C.Humborg, personal communication, 2011).For both, SIAR calculations indicated that 75.2-100 % (mean 93.5 ± 4.2 %) of the NO − 3 pool is from agricultural runoff.NO − 3 with this signature seems to be transported to the central Baltic Sea, since SIAR-based estimates showed significant percentages of agriculturally derived NO − 3 in the Western Baltic Sea (41.0-66.5 %; mean: 53.5 ± 3.2 %), the Baltic Proper (32.8-45.5 %; mean: 39.0 ± 1.6 %), and the Gulf of Finland (40.9-63.4%; mean: 51.9 ± 3.0 %).However, high percentages were only expected for the Gulf of Finland and the Western Baltic Sea, where large N loads from agricultural land have been documented (Hong et al., 2012).Indeed, for the Baltic Proper, the sizeable contribution of agricultural NO − 3 (39.0± 1.6 %) was surprising and contrasted with previous findings that nearly excluded riverine NO − 3 as a major nutrient source for the central Baltic Sea (Voss et al.,  2005, 2011).However, Neuman (2000) estimated that 13 % of the N input of the Oder River is transported to the central Baltic Sea, while Radtke et al. (2012) could show, using a source attribution technique in the three-dimensional ecosystem model ERGOM (Ecological ReGional Ocean Model), that at least a part of the dissolved inorganic nitrogen (DIN) load from the Vistula River, the main NO − 3 contributor to the Baltic Sea (Wulff et al., 2009), enters the Baltic Proper.This 3-D model comprises a circulation model, a thermodynamic ice model, and a biogeochemical model and utilizes the Modular Ocean Model, MOM3.1 (Radtke et al., 2012).
Another explanation for the high estimated agricultural influence in our study could be the intrusion of water containing NO − 3 with similar NO − 3 isotope values as our agricultural NO − 3 source during mixing/advection from below the halocline.Deep-water NO − 3 in the Baltic Sea has a δ 15 N of about 7 ‰ (Frey et al., unpublished data), which is higher than the average deep-water ocean NO − 3 signature of 5 ‰ (Sigman et al., 2000).This elevated δ 15 N in NO − 3 mainly comes from water column denitrification in the oxic-anoxic interface in water at a depth of about 100 m (Dalsgaard et al., 2013).However, the year-to-year variations in DIN due to vertical mixing and advection from below the halocline are sensitive to hydrographic conditions.When the halocline is weak and well ventilated, oxygen conditions improve, resulting in

Figure 1 .
Figure 1.Station Map of the Baltic Sea and percent contribution of the four nitrate sources, NO −3 from atmospheric deposition (blue), pristine soils (red), agricultural runoff (green), and N 2 fixation (black), for the Western Baltic Sea, Baltic Proper, Gulf of Finland, Gulf of Bothnia, southern rivers, and northern rivers.Stations are indicated as black dots.For more details see Supplement Table1.

Figure 3 .
Figure 3. δ 15 N from sediment samples vs. δ 15 N-NO − 3 from surface water samples.δ 15 N values from sediments were taken from Voss et al. (2005).The positive slope suggests a tight coupling between δ 15 N-NO − 3 in surface waters and δ 15 N in sediment samples.

Figure 4 .
Figure 4. Station map for the comparison of isotope patterns in the water column and sediments.Gray circles are the stations referred to in Voss et al. (2005) and black crosses are those from this study.Isotope values were compared at stations with the same number.

Table 2 .
Means and standard deviations of the δ 15 N-NO − 3 and δ 18 O-NO − 3 values of the NO − 3 sources used in the SIAR mixing model.For further details, see Material and Methods, SIAR mixing model.

Table 3 .
Source attribution results: Mean, standard deviation, and minimum and maximum values for the potential contributions of four potential NO − 3 sources for the areas Western Baltic Sea, Baltic Proper, Gulf of Finland, Gulf of Bothnia, southern rivers, and northern rivers.

Table 4 .
Comparison of δ 15 N-NO − 3 values from surface water samples and δ 15 N values from sediments samples in sub-regions of the Baltic Sea.