High spatial‐resolution monitoring to investigate nitrate export and its drivers in a mesoscale river catchment along an anthropogenic land‐cover gradient

Nitrate monitoring is commonly conducted with low‐spatial resolution, only at the outlet or at a small number of selected locations. As a result, the information about spatial variations in nitrate export and its drivers is scarce. In this study, we present results of high‐spatial resolution monitoring conducted between 2012 and 2017 in 65 sub‐catchments in an Alpine mesoscale river catchment characterized by a land‐use gradient. We combined stable isotope techniques with Bayesian mixing models and geostatistical methods to investigate nitrate export and its main drivers, namely, microbial N turnover processes, land use and hydrological conditions. In the investigated sub‐catchments, mean values of NO3− concentrations and its isotope signatures (δ15NNO3 and δ18ONO3) varied from 2.6 to 26.7 mg L−1, from −1.3‰ to 13.1‰, and from −0.4‰ to 10.1‰, respectively. In this study, land use was an important driver for nitrate export. Very strong and strong positive correlations were found between percentages of agricultural land cover and δ15NNO3, and NO3− concentration, respectively. Mean proportional contributions of NO3− sources varied spatially and seasonally, and followed land‐use patterns. The mean contribution of manure and sewage was much higher in the catchments characterized by a high percentage of agricultural and urban land cover comparing to forested sub‐catchments. Specific NO3− loads were strongly correlated with specific discharge and moderately correlated with NO3− concentrations. The nitrate isotope and concentration analysis results suggest that nitrate from external sources is stored and accumulated in soil storage pools. Nitrification of reduced nitrogen species in those pools plays the most important role for the N‐dynamics in the Erlauf river catchment. Consequently, nitrification of reduced N sources was the main nitrate source except for a number of sub‐catchments dominated by agricultural land use. In the Erlauf catchment, denitrification plays only a minor role in controlling NO3− export on a regional scale.


| INTRODUCTION
The European Community has been taking measures concerned with nitrate (NO 3 À ) water pollution for over 40 years. In 1991, the European Union (EU) introduced the Nitrates Directive, which aimed at reducing water quality deterioration caused by nitrate pollution related to agriculture. Almost 10 years later, in 2000, the EU adopted a Water Framework Directive (EU, 2000) that called for water quality management at the river basins scale. However, despite introducing respective directives and setting up the obligation to develop action plans to prevent nitrate concentrations above 50 mg L À1 , many EU countries still exceed this threshold (Schumacher, 2016). Not only EU member states encounter problems related to NO 3 À pollution. Nitrate contamination is a worldwide environmental issue. High NO 3 À concentrations in water threaten not only human health but also the environment as they lead to, among other things, eutrophication and toxic algal blooms (Lee et al., 2008;WHO, 2011). Upper mentioned facts emphasize the importance of a reliable assessment of the interaction between anthropogenic activities and the nitrogen dynamics at the river basins scale. Nevertheless, the basin-scale management approaches in large catchments are hindered due to the poor understanding of the spatial and seasonal variability of nitrate sources, transport and turnover pathways (Lintern et al., 2020). Major uncertainties and knowledge gaps exist with respect to (1) the transport of nitrate from different sources through different compartments of the hydrological system and (2) the degree of transport-related alterations of nitrate source signatures (Lintern et al., 2020;Schlesinger et al., 2006).
A prerequisite for designing effective water pollution mitigation programs is in-depth knowledge of the drivers controlling the spatial and temporal variability in nitrate export at relevant management scales (Wall et al., 2011). Therefore, many researchers addressed the problem of characterization and quantification of nitrate sources and in-stream nitrate processing along with a characterization of its controls (Nestler et al., 2011;Rose et al., 2015;Schwientek et al., 2013).
Coupled nitrogen and oxygen stable isotopes of nitrate have proven to be useful to identify nitrate sources and transformations within catchments qualitatively (Xue et al., 2009) not only on a hillslope scale (McAleer et al., 2017) but also in large river basins, such as the Mississippi (Panno et al., 2006), Illinois (Panno et al., 2008), Seine (Sebilo et al., 2006) and Sawa (Vrzel et al., 2016). Stable isotope monitoring in large and mesoscale river catchments is usually conducted at the outlet only or with low-spatial resolution. However, nitrate sources and transformations may display substantial spatial and seasonal variations that can be missed while using traditional, low-spatial resolution monitoring. Moreover, conducting monitoring with a low spatial resolution in large river catchments does not provide insights into how hydrochemical variations at the catchment outlet are linked to the variations in headwater catchments, that are known to be zones of nutrient-and carbon processing (von Schiller et al., 2017) and are recognized for their high ecological value (Bishop et al., 2008;Van Meerveld et al., 2020). Recently, a high spatially resolved investigation of nitrate dynamics has been performed in mesoscale river catchment using coupled dual nitrate isotopes and river discharge . The results suggest that such an approach can provide more insights into nitrate dynamics in response to land use and flow regimes.
Stable nitrate isotope signatures coupled with mass balance mixing models were used to quantitatively assess nitrate sources Voss, Deutsch, et al., 2006). Nevertheless, the usage of mass balance mixing models to trace nitrate sources is constrained to well-defined systems and does not take into account several substantial sources of uncertainties related to (1) multiple nitrate sources, (2) overlapping isotopic compositions of nitrate sources, and (3) the fact that nitrate sources are rarely defined by narrow isotopic ranges. The introduction of the combined usage of Bayesian updating and a Monte Carlo search routine opened new opportunities for tracing sources and the fate of nitrate in river basins. This approach is implemented in programs like SIAR-stable isotope analysis in R (Parnell et al., 2010), and MixSIR (Moore & Semmens, 2008) or the concept of Soulsby et al. (2003). Since first applications of Bayesian stable isotope mixing models (BSIMMs) for estimating the probability distributions of proportional contribution of nitrate sources to the nitrate mixture in surface water (Xue et al., 2012), they have been successfully applied to study nitrate sources under various land uses, including urban (Divers et al., 2014), agricultural (Ding et al., 2014), irrigated (Zhang et al., 2018) and multiple land-use areas (Jin et al., 2018;Li et al., 2019;Xing & Liu, 2016). Advances in the usage of BSIMMs include incorporation of fractionation factors related to denitrification (Li et al., 2019;Xia et al., 2017;Yue et al., 2015) and combination of BSIMM with nitrate export and flux calculations at the catchment outlet in large scale (>350 000 km 2 ) (Li et al., 2019) or with low-spatial resolution (four locations) in small scale (<16 km 2 ) (Divers et al., 2014). Moore and Semmens (2008) proposed to improve BSIMMs estimates by the usage of informative prior distributions. To the best of our knowledge, this approach has not yet been used to investigate the contribution of nitrate sources in surface water.
Except for that, BSIMMs were not exploited to investigate nitrate export by combining them with nitrate export calculations in high spatial resolution in a mesoscale river basin.
The objectives of this study are to (1) quantitatively determine NO 3 À sources in a mesoscale river catchment with a high spatial reso-  Figure S1).
F I G U R E 1 Map of the Erlauf catchment and its tributaries with land cover (left) and all surface water sampling locations monitored from 2012 to 2018, each representing sub-catchment (right). WWTP states for wastewater treatment plants

| Sampling
We conducted our monitoring program in the Erlauf catchment between 2012 and 2017. Our sampling methodology involved a combination of regional and event-based surface water sampling campaigns. We performed regional sampling campaigns seasonally with a high spatial resolution, including the main river and all its major tributaries. Regional sampling campaigns were complemented by four event-based (event-triggered) sampling campaigns conducted at selected sampling locations. We carried out event-based campaigns during or immediately after precipitation events no more than 3 days after scheduled regional sampling campaigns. Altogether, we performed 19 surface water sampling campaigns and monitored 65 sampling locations. The details regarding the dates and locations of the surface water monitoring are provided in the supplementary information (Table S1). Regional sampling campaigns were performed generally during base flow conditions with daily average precipitation during campaigns not exceeding a few mm (Table S2, and Figure S2).
However, conditions antecedent to regional sampling campaigns were variable (Table S2), resulting in varying average discharges calculated at the outlet for the regional sampling campaigns (Table S2).
At each site, using the portable meter, we measured the basic physicochemical properties of water, including electrical conductivity (EC), temperature (T), and pH. Furthermore, we collected water samples for major ion and stable isotope analysis of nitrate and water. All samples were kept in high-density polyethene (HDPE) bottles without headspace. Besides the surface water monitoring, we collected precipitation samples of rainfall and snow in the study catchment. Precipitation water was collected as monthly composite samples in evaporationfree precipitation collectors (PALMEX, Croatia; Gröning et al., 2012) with a volume ranging from 30 ml to 2.5 L. Additionally, we collected 1 L snow grab samples during winter. In total, 61 precipitation samples and 691 stream water samples were collected and analysed for nitrate and water stable isotopes, as well as nitrate concentrations.

| Laboratory analysis
Water samples for stable isotope and chemical analysis were filtered through cellulose acetate filters with a pore size of 0.22 and 0.45 μm, respectively. Samples were stored refrigerated prior to analysis within 1 month of collection. The denitrifier method with bacteria strains of Pseudomonas chlororaphis (ATCC #13985) was used to measure the isotopic composition of dissolved nitrate (Casciotti et al., 2002;Sigman et al., 2001). The isotopic composition of N 2 O (δ 15 N NO3 and δ 18 O NO3 ) produced from sample NO 3 À was measured by gas isotope ratio mass spectrometry. A DELTA V Plus mass spectrometer, in combination with a GasBench II from Thermo Scientific, was used for the nitrate isotope determination. Stable isotope ratios are expressed in the delta (δ) notation as follows: (Kohl et al., 1971): where R sample and R standard are the sample's and the standard's ratio of  (Bateman & Kelly, 2007;Curt et al., 2004;Fogg et al., 1998;Heaton, 1986;Kreitler & Browning, 1983;Li et al., 2007;Mariotti et al., 1988;Mayer et al., 2001;Panno et al., 2008;Rapisarda et al., 2010;Rennie et al., 1976;Rogers, 2008;Spoelstra et al., 2007;Vitòria et al., 2004;Wassenaar, 1995;Widory et al., 2005;Williard et al., 2001;Zhang et al., 2008). We have described in detail the procedure of calculating the mean isotopic signatures of nitrate sources in the supplementary information (Text S1).
We used NO 3 À /Cl À molar ratios as an indicator of nitrate sources since different NO 3 À sources have different levels of NO 3 À /Cl À ratios. The mineral fertilizers are characterized by high NO 3 À /Cl À molar ratios and low concentrations of Cl À , while untreated effluents from MS have relatively low NO 3 À /Cl À ratios and high Cl À concentrations (Liu et al., 2006). The NO 3 À /Cl À ratios of sewage will increase after being treated in wastewater treatment plants (Xia et al., 2017). In the Erlauf watershed, neither halite nor sylvite deposits occur. Therefore, we expect no influence of evaporites on NO 3 À /Cl À ratios.

| Stable isotope mixing model
We used MixSIAR, a Bayesian tracer (e.g. stable isotope) mixing model framework (Moore & Semmens, 2008;Parnell et al., 2010;Stock et al., 2018), to calculate the proportional contribution of NO 3 À sources in surface water in all sub-catchments. Supplementary Information (Text S2) includes additional information about the model.

| Discharge and precipitation data
The daily discharge data were provided by the state government office of Lower Austria ('Amt der Niederösterreich Landesregierung').
In the Erlauf catchment, there are 12 gauging stations. Five of them are located along the main river, and the other seven are on the main tributaries ( Figure 1).
Daily precipitation data were obtained for eight stations located within the catchment and in the close neighbourhood (Table S3)

| Regionalization of river discharge and nitrate loads
To calculate the nitrate export from all investigated sub-catchments and to understand the spatial heterogeneity of the nitrate export and its controls, we interpolated the runoff in our sampling locations from discharge gauging stations on a daily time step using the top-kriging technique. Top-kriging, developed by Skøien et al. (2006), is a geostatistical method that allows interpolating runoff characteristics along with the stream network. In comparison to traditional deterministic or geostatistical interpolation approaches, it accounts for the river network hierarchy. More details can be found in Skøien and Blöschl (2007).
As the regional sampling campaigns were conducted within 4 days during steady flow conditions with only small intra-campaign changes in discharge, and the event-based sampling campaigns were conducted within a one-time step, discharge variability due to the travel time of the event wave between up-and downward river crosssections is assumed to be small. Therefore, we used spatial kriging (time-independent kriging). We leveraged the rtop package (Skøien et al., 2014) in the statistical environment R (R Core Team, 2018) to apply the top-kriging approach. We tested the predictive accuracy of the interpolation using ordinary cross-validation. In the ordinary cross-validation, each gauged station is, in turn, treated as an ungagged station, and the runoff is interpolated from the other gauged stations. The interpolated runoff is then compared with the observed daily runoff. We quantified the predictive accuracy with the Nash-Sutcliffe Efficiency (NSE) (Nash & Sutcliffe, 1970 (Merz et al., 2008;Skøien et al., 2006).
Previous studies (Lark, 2000;Skøien et al., 2014) show that the quality of the predictions is relatively insensitive to the choice of the variogram, at least as long as there are several observations within its range. In the Erlauf catchment, a high density of stations (12 stations per 631.5 km 2 ) is available.
We calculated the specific discharge (Q spec ) by dividing the dis-

| Discharge separation
To shed more light on the dynamics of hydrological processes, we divided streamflow into the base flow and quick flow components.
The contribution of these components is likely to affect nitrate concentrations and isotopic compositions because various pathways for the mobilization of nitrogen from natural and anthropogenic sources may be associated with certain hydrological events. Therefore, we conducted a separation of total discharge with a simple smoothing method proposed by the Institute of Hydrology (1980) (2) a decrease in NO 3 À concentrations larger than 0.2 mg L À1 . Likewise, to obtain spatial information on the potential impact of denitrification, we analysed the changes in δ 18 O NO3 , δ 15 N NO3 and NO 3 À along the main river.
The field-based, apparent enrichment factors for nitrogen ( 15 ε) and oxygen ( 18 ε) were calculated using the simplified Rayleigh equation: where ε stands for the isotopic enrichment factors for nitrogen and oxygen, d stands for the δ 15 N and δ 18 O values, respectively, and C stands for the nitrate concentration.

| RESULTS
3.1 | Qualitative determination of nitrate sources The ratios of NO 3 À /Cl À varied widely from 0.12 to 14.06 in the catchment, suggesting a mixture of multiple sources of nitrate. In general, high Cl À concentrations and low NO 3 À /Cl À ratios were found in catchments with a high percentage of agricultural land use. This could be caused by the high contribution of manure and effluents from sewage. Contrary, low Cl À concentrations and a wide range of NO 3 À /Cl À ratios were found in catchments with a low percentage of agricultural land use, reflecting the impact from atmospheric deposition and RNS ( Figure S4).

| Temporal variability
We investigated the temporal variations of NO 3 À concentrations and stable isotope signatures on a seasonal basis. We observed a significant difference in δ 15 N NO3 between April and all other months ( Figure 4a). Similarly, we observed a significant difference between samples collected in November and the other months for δ 18 O NO3 ( Figure 4b). We detected a significant difference in NO 3 À concentrations between samples collected in April and samples collected in December or January (Figure 4c). We found the highest coefficient of variation in June (102%) and the lowest in April (88%).

| Correlation analysis
Results of the pairwise Spearman's correlation analysis of all collected surface water samples ( Figure 5)  All sub-catchments show a negative dependency of BFI and δ 18 O NO3 , of which one-fifth has a very strong or strong correlation (Table S6). A similar general tendency was observed neither between BFI and δ 15 N NO3 nor between BFI and NO 3 À concentrations.
The analysis of the relationship between nitrogen and oxygen isotope signatures of nitrate on a seasonal basis ( Figure S3) shows no positive linear correlation with a slope from 1:1.3 to 1:2.1. During the location-wise analysis of data from all sampling campaigns (see Section 2.10), 60 out of 62 analysed locations did not show significant linear regressions specific for denitrification (Table S4). Nine sets of sampling campaigns (Table S5) matched the prerequisites of intercampaign analysis (see Section 2.10). For those nine sets, calculated 15 ε ranged from À6 to À46 while 18 ε ranged from À8 to À69 (Table S5).
Spatial analysis along the main river shows that increased isotopic signatures were generally accompanied by increased nitrate concentrations and nitrate loads ( Figure S7).

| Spatiotemporal variations of nitrate sources and loads
The model output ( Figure 6) shows that calculated probability distri- with 99.7% forest, and 0.3% agriculture), the primary source of NO 3 À calculated for the whole monitoring period was RNS with the mean contribution of 95%, followed by AD (5%), and the contribution of two other sources close to zero. At the outlet of the river (location E19), the primary source of NO 3 À calculated for the whole monitoring period was also RNS with the mean contribution of 46%, followed by MS (44%), NF (8%), and AD (2%). In the sub-catchment characterized by the highest percentage of agriculture (location B7; 3 km 2 , with 100% of agricultural areas), the primary source of NO 3 À calculated for the whole monitoring period was MS with the mean contribution of 83%, followed by NF (9%), RNS (6%), and AD (3%).
We found the highest SL NO3 in April and June (Figure 7, columns 1 and 2). During all analysed months, SL NO3 were generally increasing from south to north. We found the highest SL NO3 in small agricultural sub-catchments with a high percentage of arable land located next to the Erlauf river outlet. The analysis of the SL NO3 derived from each of the four sources (Figure 7, rows 2-5) reveals that in those catchments higher total SL NO3 were caused mainly by increased SL NO3 derived from MS. Generally, the SL NO3 were higher in catchments characterized by higher LC AGR , especially those characterized by a high percentage of arable land. The SL NO3 derived from AD were very low.
The SL NO3 derived from RNS did not follow a spatial pattern and were highest in April and June. The SL NO3 derived from NF were highest in April and June, especially in catchments with a high percentage of arable land.

| Event-based monitoring
We did not observe any significant difference in mean NO 3 À concentrations nor mean δ 15 N NO3 between pre-event and event-based sampling campaigns (  (Aravena & Robertson, 1998;Fukada et al., 2003;Mengis et al., 1999). The analysis of the relationship between nitrogen and oxygen isotope signatures of nitrate on a seasonal basis does not provide any evidence for a strong impact of denitrification ( Figure S3). During the location-wise analysis of data from all sampling campaigns, most locations did not show significant linear regressions specific for denitrification (Table S4). However, two sampling locations displayed slopes in the dual-isotope plot that might be indicative of the occurrence of denitrification with weak correlations. Their further analysis showed that the correlation between δ 18 O NO3 and δ 15 N NO3 in one of the locations is driven rather by the mixing of different sources than denitrification ( Figure S5). However, the possibility of a weak denitrification signal could not be rejected in the second location ( Figure S6). Nevertheless, on average, less than 0.6% of the nitrate load at the outlet of the Erlauf River is delivered from this location. Therefore, even if denitrification occurred in this location, it had a negligible impact on the isotopic composition of the F I G U R E 5 The spearman pairwise correlation matrix between nitrate isotopes signatures (δ 15 N NO3 , and δ 18 O NO3 ), nitrate concentrations (NO 3 À ), specific nitrate loads (SL NO3 ), specific discharges (Q spec ), and percentages of the agricultural (LC AGR ) and the forested land cover (LC FOR ), based on all samples collected from all sampling locations. Correlations with the p value >0.001 are specified on the plot main river. The location-wise analysis of data from all sampling campaigns did not provide any evidence for a significant role of denitrification in nitrate turnover on the entire river system. Previous research has shown that denitrification may occur in so-called hot spots and hot moments when conditions are favourable (e.g. Harms & Grimm, 2008;Palta et al., 2014;Peter et al., 2011). Therefore, we performed further inter-campaign analyses (see Section 2.10), during which nine sets of sampling campaigns (Table S5) matched the prerequisites. However, except for one location, subsequent analysis of 15 ε and 18 ε (Table S5) did not confirm straightforward denitrification as the sole process controlling the isotopic composition of nitrate. The 15 ε of denitrification reported in the literature vary between À40‰ and À5‰ (Kendall & McDonnell, 1998) with typical values reported for groundwater denitrification between À8‰ and À5‰ (Mariotti et al., 1988). The oxygen isotope enrichment factors for denitrification ( 18 ε) fall between À18‰ and À8‰ (Xue et al., 2009). The apparent enrichment factors calcu-  (Kendall et al., 2007;Mayer et al., 2001), (2) a higher exchange of oxygen atoms between intermediate N-compounds (nitrite) and water (Casciotti & Buchwald, 2012;Kool et al., 2011), or higher evaporation causing a positive water oxygen isotope shift.
Overall, our analyses showed that nitrate concentrations are likely to be controlled by nitrification processes. On a regional scale, bacterial denitrification plays only a minor role in the Erlauf catchment. Previous studies used stable isotopes to evaluate large-scale patterns in N turnover processes showing that they can be variable and site-specific. Denitrification was found to have a negligible impact on a regional scale nitrate turnover in Bode catchment characterized with similar land use and topographic gradients Mueller, Zink, et al., 2016). Dual isotope evaluation of nitrate proved F I G U R E 6 Proportional contributions of four potential NO 3 À sources estimated by the MixSIAR model are presented for all sub-catchments seasonally and for all collected data (last row). White colour refers to a lack of data (Table S1) that denitrification was a main regional nitrogen sink in forested subtropical catchments (Yu et al., 2019).

| Dynamics of NO 3 À export and land cover
Contrary to δ 18 O NO3 , we found obvious correlations between both percentages of land cover (LC) and NO 3 À concentrations and d 15 N NO3 , as well as between the two last themselves ( Figure 5). This indicates that the changing land use resulted in increased NO 3 À concentrations what was likely attributed to the increasing contribution of MS, as this source is characterized by relatively high δ 15 N NO3 and nutrient content. We excluded the possibility of increased δ 15 N NO3 caused by the higher impact of denitrification because the correlation between δ 18 O NO3 and δ 15 N NO3 was insignificant. This is consistent with the concluded low impact of denitrification on a regional scale. Our findings align with the results of other studies involving stable isotopes in catchments characterized by high LC AGR and LC URB . For example, in F I G U R E 7 Row 1 (top) presents specific NO 3 À loads (SL NO3 ) calculated for each sub-catchment, and rows 2 to 5 present SL NO3 derived from each of four NO 3 À sources. In the figure presented are SL NO3 calculated for five selected seasonal sampling campaigns from April 2013 to January 2014, based on NO 3 À concentrations measured during respective sampling campaign, discharge data calculated by the usage of the topkriging (Section 2.7), and mean NO 3 À sources contributions calculated by the usage of MixSIAR ( Figure 6, columns 1 to 5, respectively). White colour refers to a lack of data (Table S1) their meta-analysis of 33 river catchments, Johannsen et al. (2008) found a strong positive correlation (R 2 = 0.71) between δ 15 N NO3 values and the proportion of arable and urban land caused by increased inputs of nitrate from sewage and manure.
The presence of a strong correlation between Q spec and SL NO3 , not observed between both percentages of LC and SL NO3, suggests that Q spec plays a more important role than LC in controlling SL NO3 .
Both percentages of LC were also weakly correlated with Q spec , with lower Q spec found in sub-catchments with higher LC AGR . This is likely to be related to the fact that most of the catchments with the highest LC AGR are located on lower altitudes and receiving less precipitation than the mountainous sub-catchments. show that most of the nitrate originating from AD is not contributing directly to the stream water, but it is processed through the biota and then nitrified before entering the stream. Many other studies that used stable isotopes in forested catchments report low proportional contribution (mean~10%) of nitrate from atmospheric deposition contributing to streams during baseflow (Rose et al., 2015).

| Spatiotemporal variations in NO 3 À loads and their sources
The highest SL NO3 were found for the two campaigns with relatively high discharges (Figure 7, columns 1 and 2) emphasizing the importance of Q spec role in controlling SL NO3 . The analysis of the SL NO3 derived from each of the four sources (Figure 7, rows 2-5) suggests that in catchments characterized by higher LC AGR , especially those characterized by a high percentage of arable land, higher total SL NO3 were mainly related with increased SL NO3 derived from MS but also from NF. Very low SL NO3 derived from AD were related to T A B L E 1 Mean stable isotope and NO  The lack of significant differences in mean NO 3 À concentrations and δ 15 N NO3 between pre-event and event-based sampling campaigns (Table 1) is consistent with our previous finding that denitrification has a low impact in the investigated catchment on a regional scale.
Moreover, no significant difference in mean NO 3 À concentrations suggests a low direct contribution of AD into the stream water and a low impact of dilution during event-based campaigns.  The observed correlations between LC AGR and event-related Nisotope shifts further confirm that the mobilization of RNS during precipitation events is driving the changes in δ 15 N NO3 , especially in the agricultural catchments where the MS contribution is high (see Figure 6). Stable isotope signatures imply that agricultural and forested catchments react differently to the rainfall-runoff events with respect to changes in pathways of nitrate pool mobilization. Lack of correlation between the LC AGR and the difference between δ 15 N NO3 measured during autumn pre-event and event sampling campaigns ( Figure 8d) is likely related to the fact that the pre-event sampling campaign was not performed during pure base flow conditions (BFI = 0.89) and that the interflow and surface runoff flow paths were already contributing to the streamflow during the preevent sampling campaign.
4.3.3 | Relationship between discharge, NO 3 À isotopic signatures, and NO 3 À export through the monitoring period The results of the pairwise Spearman's correlation analysis between Q spec , δ 18 O NO3 , and NO 3 À concentrations ( Figure 5) indicate a low impact of dilution during periods of higher specific discharge in the investigated catchments. Considering the correlation between Q spec and δ 15 N NO3 , one can conclude that periods of higher specific discharge in the investigated catchment are likely to be associated with a higher contribution of RNS or NF sources rather than a direct contribution of AD. Stronger correlations between SL NO3 and Q spec compared to the correlation between SL NO3 and NO 3 À concentrations emphasize the impact of discharge changes on NO 3 À export from catchments.
Results of the correlation analysis between BFI and nitrate concentrations and isotopes (Table S6) support the assumption that nitrate pools with elevated δ 18 O NO3 are mobilized from the unsaturated zone. We found that during the performed sampling campaigns, most of the atmospheric nitrate was not contributing directly to the stream. Instead, it was first cycled through the biota and fixed in the soil storage before it was nitrified and released to the stream. Our findings are consistent with the observations of Burns and Kendall (2002), who report a major contribution of atmospheric nitrate only during high flow events that exceeded the annual flow. None of the rainfall-runoff events investigated in our study exceed the annual flow. Moreover, our results suggest that different nitrate export pathways occur during different hydrological conditions.

| CONCLUSIONS
Our integrated approach provides valuable insights into the nitrate export from a mesoscale river catchment and its main controls: landuse, hydrological, and nitrate transformation processes. The studied Erlauf catchment is a typical representative of a mesoscale Alpine foothill catchment characterized by a land-use gradient from forested headwaters to agricultural lowlands. Therefore, our site-specific findings may be utilized to improve the interpretation of the data from low spatial resolution water quality monitoring in catchments with similar characteristics. In the Erlauf river system, nitrate from external sources is stored and accumulated in soil storage pools instead of being directly mobilized and dislocated. Nitrification of reduced nitrogen species in those pools plays the most important role for the Ndynamics in the study catchment. Consequently, nitrification of reduced N sources was the main nitrate source except for agricultural sub-catchments. In this study, land use is the important driver of nitrate export. The agricultural land cover was tied to elevated nitrate concentrations and changes in the proportional contribution of nitrate sources, especially a significantly higher contribution of manure and sewage. In the Erlauf catchment, nitrate degradation potential is not high enough to solely control nitrate export, and on a regional scale, denitrification plays only a minor role. Therefore, high specific nitrate loads in small agricultural catchments may only be managed by reducing the nitrogen surplus. One option would be the utilization of precision farming techniques allowing a spatially distributed application of organic and chemical fertilizers according to the actual fertilization demand.