Fate of nitrate during groundwater recharge in a fractured karst aquifer in Southwest Germany

Over the past decades, fractured and karst groundwater systems have been studied intensively due to their high vulnerability to nitrate (NO3−) contamination, yet nitrogen (N) turnover processes within the recharge area are still poorly understood. This study investigated the role of the karstified recharge area in NO3− transfer and turnover by combining isotopic analysis of NO3− and nitrite (NO2−) with time series data of hydraulic heads and specific electrical conductivity from groundwater monitoring wells and a karstic spring in Germany. A large spatial variability of groundwater NO3− concentrations (0.1–0.8 mM) was observed, which cannot be explained solely by agricultural land use. Natural-abundance N and O isotope measurements of NO3− (δ15N and δ18O) confirm that NO3− derives mainly from manure or fertilizer applications. Fractional N elimination by denitrification is indicated by relatively high δ15N- and δ18O-NO3− values, elevated NO2− concentrations (0.05–0.14 mM), and δ15N-NO2− values that were systematically lower than the corresponding values of δ15N-NO3−. Hydraulic and chemical response patterns of groundwater wells suggest that rain events result in the displacement of water from transient storage compartments such as the epikarst or the fissure network of the phreatic zone. Although O2 levels of the investigated groundwaters were close to saturation, local denitrification might be promoted in microoxic or anoxic niches formed in the ferrous iron-bearing carbonate rock formations. The results revealed that (temporarily) saturated fissure networks in the phreatic zone and the epikarst may play an important role in N turnover during the recharge of fractured aquifers.


Introduction
Contamination of karstified aquifers is of major concern since these systems supply up to 25% of the world's population with drinking water (Ford and Williams 2007). Karst aquifers are known to be particularly vulnerable to anthropogenic contaminants such as nitrate (NO 3 − ) (Einsiedl and Mayer 2006;Katz 2012;Husic et al. 2019). Karstified groundwater systems mainly develop via the natural dissolution of carbonate rocks (i.e., limestones, dolomites) resulting in the formation of a heterogeneous flow system characterized by sink holes, sinking streams, caves, fractures and fissures. Water flow within the system is modulated by highly permeable karst structures (enlarged fractures and conduits), as well as the less permeable rock matrix (small pores and fissures) (Ford and Williams 2007;Hartmann et al. 2014). The conduit network allows rapid response to rain events, resulting in quick intrusion of surface water (and potentially also pollutants) into the aquifer, as well as rapid transport within it (Einsiedl 2005;Pronk et al. 2006). Although the sources of nitrogen (N) inputs for non-karstified and karstified aquifers are similar (e.g., agriculture, waste water; Wakida and Lerner 2005;Matiatos 2016), N pathways in karst systems were shown to be much more diverse due to heterogeneous flow pathways (Husic et al. 2019).
While fissure and fracture networks, particularly the epikarst zone, have been shown to play a major role in storing and locally rerouting vertical infiltrating waters (Jones 2003), they are also known to promote the formation of zones of enhanced biological activities (Pipan and Culver 2007;Lian et al. 2011). Hence, besides processes occurring within the overlying soil (e.g., Munch and Velthof Gerard 2007), N turnover is also mediated by terrestrially derived and sedimentbound microbial communities present within the karstified system (Simon and Benfield 2002;Heffernan et al. 2012;Henson et al. 2017). Previous studies showed that bacterial communities located either in caves or the major conduit network can drive various reactions such as mineralization, immobilization, nitrification and denitrification (Farnleitner et al. 2005;Katz 2012;Henson et al. 2017). For example, Herrmann et al. (2017) provided evidence for the presence of denitrifying bacterial communities suspended in groundwater monitoring wells and/or attached to the parent rock matrix of a limestone aquifer. Which process is favoured by the microbial community (and in turn the fate of e.g., fixed N) strongly depends on the availability of suitable electron donors (e.g., organic carbon or ferrous iron bearing minerals such as pyrite), as well as the ambient redox conditions (Henson et al. 2017;Herrmann et al. 2017).
Natural abundance measurements of NO 3 − and NO 2 − isotope ratios ( 15 N and 18 O) represent a powerful tool to gain deeper insights into the sources and fate of fixed (i.e., bioavailable) N in groundwater. Common pollutant sources of NO 3 − in groundwater (e.g., soil-N, manure, synthetic fertilizer and sewage effluents) have distinct isotopic signatures, which can be used to semiquantify the relative importance of the single N sources (Panno et al. 2001;Katz et al. 2004;Musgrove et al. 2016;Vystavna et al. 2017). At the same time, N compounds do not necessarily behave conservatively in groundwater, and processes such as denitrification have been demonstrated to alter the primary isotopic signatures in, e.g., karstic systems (Panno et al. 2001;Husic et al. 2020). In turn, the N and O isotopic signatures of NO 3 − (and other dissolved N species) can be used to assess various N turnover processes in groundwater (Kendall and Aravena 2000;Einsiedl and Mayer 2006;Grimmeisen et al. 2017). So far, most groundwater studies focused on the investigation of N and O isotopic signatures of NO 3 − , since concentrations of intermediates such as NO 2 − are commonly low (e.g., Herrmann et al. 2017). The few field studies that also included isotope analyses of NO 2 − (Smith et al. 2004;Wells et al. 2016) demonstrated that multi-species measurements provided additional process-level insight in N transformation pathways. The dual N and O isotope approach for both, NO 3 − and NO 2 − , is based on the fact that during most biotic N-elimination processes, the lighter N and O isotopologues are preferred, resulting in distinct parallel enrichment of the heavy isotopes (i.e., 15 N, 18 O) of the remaining pool of NO x -species (Kendall and Aravena 2000;Granger et al. 2008). In contrast, NO 3 − regeneration may lead to a differential behaviour of the NO 3 − and NO 2 − isotope ratios, and thus allows to disentangle NO 3 − -producing and consuming processes when they occur simultaneously (e.g., Lehmann et al. 2004; Wankel 2016), in, for example, the recharge zones of karst aquifers.
How N turnover in the recharge areas of karst aquifers contributes to NO 3 − elimination along the flow path, remains unclear (Jones and Smart 2005;Musgrove et al. 2016;Husic et al. 2019). Therefore, the main goal of this study was to investigate possible causes promoting NO 3 − variability within the recharge area of karst groundwater systems, and thus to unravel the particular role of the recharge area and epikarst zone with regard to NO 3 − transfer, turnover, and storage. To achieve this goal, natural-abundance isotope ratio measurements of NO 3 − , as well as of NO 2 − (δ 15 N and δ 18 O), were combined with hydraulic, hydrochemical and biogeochemical approaches. Many studies investigating N turnover processes in karstified aquifers focussed on the larger conduit network (i.e., spring-based studies), and thus on compartments that are known to quickly respond to rain events (e.g., Birk et al. 2004;Kaufmann et al. 2014;Asante et al. 2018). In order to capture the occurring N transport processes, the presented study focuses on several groundwater monitoring wells, which are connected to different flow paths within a fractured and karstified aquifer, but are not dominated by karst conduits; for comparison, one of the main springs of the karst aquifer was also examined. The sampling locations were situated in the recharge area of a karstified and fractured limestone aquifer in SW Germany, which is characterised by a high variability in NO 3 − concentrations.

Study catchment
The studied fractured and karstified Upper Muschelkalk aquifer is situated in the "Oberes Gäu", a landscape within the Southwest German Triassic Scarp lands, located approximately 30 km southwest of Stuttgart. Part of this regional aquifer underlies the catchment of the River Ammer, which originates from karstic springs close to the town of Herrenberg. The focus of this study is on the~30 km 2 large western section of the aquifer, situated west of Herrenberg (Fig. 1a). Due to outcropping of the carbonate rocks of the Upper Muschelkalk, this region constitutes the main recharge area to the groundwater system (Fig. 1b). Topographically, the area forms a slightly undulating plateau with small but steeply incised valleys dipping towards the Ammer Valley in the southwest (Fig. 1c). The elevation ranges from 400 to 590 m asl between the Ammer springs and the plateau. The overlying soils are usually shallow (0.3-1.8 m) and consist partly of loess, rendzinas developed from carbonate rocks, and more clayey soils in places where mudstones of the Lower Keuper formation cover the Muschelkalk. Karstification of the Upper Muschelkalk formed numerous dolines within the study area (see Fig. 1b), providing pathways for rapid infiltration of water to greater depths. Other karstic structures present in the study area include dry valleys, stream sinkholes, and two major karstic springs (i.e., Ammer springs 1 and 2 (AMQs); see Fig. 1b,c).
Mean annual precipitation for the studied period 2015 to 2018 at stations nearby (Herrenberg and Bondorf) range between 550 and 690 mm/year, and the average annual temperature was 9.9°C, which is about 1.9°C higher than the longterm mean value since the 1980s. Estimates of mean annual Fig. 1 a Location of the study area close to the town of Herrenberg in Germany (DE) [country codes from https://www.iso.org/obp/ui]; b lithostratigraphic map of the study area with the four major field sites (Sulz am Eck (Sul), Haslach (Has), Ammer springs (AMQs), and the artesian well Altingen (ArtAlt)), based on the Gauß-Krüger coordinate system; the dotted black line indicates the cross section below; c geological cross section of the main geological formations in the study area, including the major field sites groundwater recharge rates (20-year average) range between 200 and 300 mm/year (LUBW 2016;D'Affonseca et al. 2020).
Access to the groundwater is provided via the karstic springs (Ammer springs, AMQs), six groundwater monitoring wells near Haslach (Has1-Has6), and seven groundwater monitoring wells near Sulz am Eck (Sul1-Sul8), which were used for physical and chemical monitoring. An additional artesian monitoring well (ArtAlt) in the confined part of the aquifer east of the town of Herrenberg was investigated for comparison. The AMQs consist of at least two large pools (elevation 403 m asl), where groundwater is directly upwelling from the uncovered Muschelkalk rocks with an average discharge of about 40-50 L/s. Due to the location of the monitoring wells in Haslach and Sulz on the plateau (elevation 436-588 m asl), groundwater levels at these wells ranged between 35 and 100 m below the surface (490-515 m asl). All wells have diameters of 10.2 or 12.7 cm, and screened sections ranging from 8 to 10 m length, except for the deepest wells in Sulz featuring a longer screen length of 20 m.
Land use at the study site is dominated by agriculture (about 67%), followed by urbanized areas (about 15%), and forested land (about 18%) (Grathwohl et al. 2013;Liu et al. 2018). Intense agriculture leads to the release of NO 3 − and other compounds into the karstic groundwater system. Other possible sources of N-contaminants include leakages from sewer systems from farms or urban areas; however, the majority of which are located downgradient of the monitoring wells. Two operating quarries of 0.16 and 0.24 km 2 at the plateau (location see Fig. 1b) provide access to the carbonate rocks of the Muschelkalk aquifer, allowing rock samples to be taken from freshly exposed rock faces. The study area is part of a large water protection zone, supplying more than 150,000 people with drinking water from the Upper Muschelkalk aquifer.

Geological and hydrogeological setting
The study area (see Fig. 1b (Harreß 1973). In addition to karstic features, regional groundwater flow is influenced by numerous tectonic structures comprising conjugated fault patterns formed by two large synthetic WSW-ENE fault zones, with antithetic SE-NW faults in between (D'Affonseca et al. 2020).
The Upper Muschelkalk outcrops in the study area and gently dips to the SE, where it is increasingly covered by claystones, dolomites, sandstones, and marlstones of the Lower Keuper (Fig. 1b,c), which leads to unconfined conditions on the plateau whereas towards the SE the Muschelkalk aquifer becomes confined or even artesian (D'Affonseca et al. 2020). Generally, the groundwater flow direction is from NW to SE following the dipping of the Triassic formations. In the lower part of the Upper Muschelkalk, a 6-8 m thick, marlstone-claystone sequence (Haßmersheimer layers) hydraulically separates the aquifer into a basal and an upper part (Villinger 1982;Ufrecht and Hölzl 2006).

Groundwater monitoring and sampling
In December 2015, six automatic probes and data-loggers (CTD Diver, Schlumberger) were installed in selected monitoring wells (Has2, Has4, Has5, Sul1, Sul3, Sul4) to continuously record pressure (hydraulic head, H; range: 0-10 m; error: ±0.1%; resolution: 0.2 cm), water temperature (Tw; range: -20 to +60°C; error: ±0.1°C; resolution: 0.01°C), and specific electrical conductivity (SEC; range: 0-120 mS/cm; error: ±1%; resolution: 1 μS/cm) at intervals of 30 min. Additionally, oxygen probes (miniDO 2 T, PME) were installed in the same monitoring wells to measure dissolved oxygen (DO; range: 0-150% saturation; error: ± 5%; resolution: 0.1 mg/L) in groundwater at intervals of 30 min. The oxygen probes were deployed from Dec 2015 to Jun 2016 for Has4, Sul3, and Sul4, and from Dec 2015 to Mar 2017 for Has5 and Sul1, respectively. Unfortunately, the DO probe in well Has4 had to be installed close to the bottom of the well to ensure permanent water coverage, and due to sporadic sediment contact of the probe near the bottom DO data were compromised. Larger than expected water-table fluctuations in the range of 30 m led to pressure maxima in Feb/Mar 2016 and particularly in Jan 2018, which exceeded the maximum permissible value of the probe. Therefore, the CTD probes in all wells were replaced in Feb 2018 by new probes (4 TD and 2 CTD Diver, Schlumberger), with an extended measurement range of up to 50 m water column.
Groundwater samples for chemical analysis (major ions, dissolved N compounds) were taken at the monitoring wells in the groundwater recharge area in three different sampling campaigns in Aug 2014, Apr 2016, and Jan 2018. The Apr 2016 campaign also included sampling for isotopic analysis (δ 15 N, δ 18 O of NO 3 − and NO 2 − ). Additional data on major ions including NO 3 − were available from a previous sampling campaign in Sep 2010. Data on groundwater chemistry from the karstic Ammer spring (AMQ) and the artesian well (ArtAlt) downgradient of the recharge area were used for comparison. Different submersible pumps were used for sampling the monitoring wells depending on the diameter of the wells (10.2 or 12.7 cm) and the available water column. The pumps were usually operated with pumping rates between 0.01 and 0.20 L/ s. Groundwater samples were taken after exchanging the water volume inside the well at least 1.5 times. Dissolved oxygen, pH, Tw and SEC were measured in the field using handheld probes (WTW GmbH, DO resolution 0.1%, accuracy ± 0.5% of value; pH accuracy ± 0.1% of value; SEC range 0.0 to 500 μS/cm, error: ± 0.5% of value; Tw range −5.0 to +105.0°C, error: ±0.1 K) inserted to a flow-through cell, which was directly connected to the pump via PVC or PTFE tubes of 50-100 m length. Due to the relatively low pumping rates in some wells, the measured water temperatures were up to 5°C higher compared to the original water temperature and were therefore not used for interpretation.
Water samples from the Ammer spring were collected as grab samples directly at the outlet. Groundwater samples from the artesian well ArtAlt were either taken from the artesian outflow or by using a submersible pump when the pressure head dropped below the well top during the summer season.

Capture zone delineation
Capture zones of the groundwater monitoring wells were determined by means of backward particle tracking using a recently established numerical groundwater flow model for the Muschelkalk aquifer (D'Affonseca et al. 2020). The code MODFLOW NWT (Niswonger et al. 2011), an autonomous version of the three-dimensional (3D) code of finite differences MODFLOW-2005 (Harbaugh 2005), was employed to simulate the regional flow regime within the aquifer, assuming steady-state conditions for simplicity. Using the groundwater flow data simulated by this model, conservative advective mass transfer calculations were conducted to estimate the capture zones of the monitoring wells by backward particle tracking with MODPATH, a MODFLOW postprocessing module developed to calculate 3D flow paths (Pollock David 1994). In order to avoid setting particles into dry cells, particles started near the filter screen bottom of each considered monitoring well. The starting point consisted of a horizontal circle of 0.3 m in radius containing 1,000 particles.

Solute concentration measurements
NO 3 − and NO 2 − concentrations were determined by standard spectrophotometric methods using a continuous flow analyser (CFA, Seal Analytics AA3), equipped with an additional set of pre-mounted membranes to filter out precipitates. Briefly, NO 2 − is reacted with an acidic sulphanilamide (SAN) solution forming a diazonium ion which binds to the N-1napthylethylenediamine dihydrochloride solution (NEDD), resulting in a pinkish colorimetric complex (azo dye), detectable at 540 nm. NO 3 − is reduced first by a hydrazine sulphate solution and total NO x (NO 3 (Sawicki and Scaringelli 1971); additionally, most ions were quantified by ion chromatography (Thermo Scientific Dionex DX-120, ANIONS: Dionex IonPac A523 analytical column with a Dionex IonPac AG23 guard column; CATIONS: Dionex IonPac CS12A-5 μm analytical column with a Dionex IonPac CG12A-5 μm guard column; LOQ = 0.1 mg/L). In order to determine the concentrations of the single iron species, a modified ferrozine assay, using 40 mM amidosulfonic acid (SFA) instead of 1 M HCl (Stookey 1970;Klueglein and Kappler 2013) was applied. The samples were diluted 1:10 in 1 M HCl at the field site, and then directly pipetted into a 96 well plate or in 1 ml cuvettes, respectively. Following the ferrozine assays protocol, the samples were photometrically at 562 nm (Thermo Scientific Multiskan GO).

Nitrate and nitrite isotope analysis
Natural abundance measurements of N and O isotope ratios in NO 3 − were performed using the denitrifier method (Coplen et al. 2004;Granger et al. 2006;Granger and Sigman 2009 ) were added to 12 ml vials, which were then closed with grey butyl stoppers and crimp sealed. Under a fume hood, the acetic acid/azide mix (100 μl per 3 ml sample) was added to each vial with a syringe. The next morning, 100 μl of a 10 M NaOH was added to stop the reaction. Until measurement (CF-IRMS, see the preceding), the samples were stored in the dark and upside down. For calibration, NO 2 − isotope standards -N-7373 and N-10219 (Casciotti and McIlvin 2007)-were prepared on the day of isotope analysis and processed the same way as samples. All N and O isotope data are expressed using the common δ notation and given in permille deviation (

Hydro(geo)chemistry
A hydrochemical facies analysis of the karst aquifer (Fig. S1 of the electronic supplementary material (ESM)) reveals that most of the investigated groundwater can be classified as Ca 2 + /Mg 2 + -HCO 3 − type waters. As expected, the hydrochemical composition of the groundwater reflects the dissolution of the calcite and dolomite minerals dominating the aquifer matrix as has been reported for other carbonate aquifers (e.g., Einsiedl and Mayer 2005). Concentrations of Cl − and SO 4 2− in the majority of groundwater samples were in the range of 0.1 to 0.65 mM and 0.1 to 0.94 mM, respectively. Both, the observed Cl − concentrations and low molar Na + / Cl − ratios (Fig. S7 of the ESM) are common indicators supporting the anthropogenic origin of chloride in groundwater (e.g., Böhlke 2002). In contrast, sulfate can be expected to be predominantly of geogenic origin due to the presence of gypsum bearing formations overlying the Muschelkalk sediments. Two wells of Sulz am Eck (Sul3, Sul5) show elevated concentrations of Cl − and particularly SO 4 2− , which is probably due to the admixture of water exposed to the underlying evaporite layers of the Middle Muschelkalk (e.g., Blanchette et al. 2010;Warren 2016), since the filter section of these wells starts just above these layers. A general summary of the average concentrations of the major and minor ions detected is shown in Table S1 of the ESM. CTD-based temperature measurements ranged from 10.0 to 11.0°C (except Sul3, 12.0°C). Overall, the temperatures are close to the mean annual temperature of the study area. Average values of specific electrical conductivity (SEC), pH and DO for the four different major sites are shown in Table 1. A circumneutral to slightly alkaline pH has been determined for all groundwater samples. Both, SEC and pH values fall within the range observed and commonly associated with karstified areas (Pronk et al. 2006;Ford and Williams 2007;Ravbar et al. 2011). Average DO concentrations at the field sites were usually close to saturation (except for ArtAlt), with occasionally concentrations significantly lower than 200 μM (Fig. S10 of the ESM). This implies that the groundwater is persistently influenced by oxic conditions in the recharge area and becomes anoxic under confined conditions at the location of the artesian well (ArtAlt).

Hydraulic and chemical groundwater response
Medium to large fluctuations in the hydraulic head (H) have been observed in most of the investigated wells, which is in accordance with findings from other karst aquifers (e.g., Ravbar 2013). Groundwater dynamics are presented in Fig. 2, including time series of daily precipitation, H and SEC during the period of Jan 2016 to Mar 2019. A direct comparison between H and daily precipitation indicates pronounced differences in hydraulic responsivity of the individual wells towards precipitation events. The highest responsivity could be observed at Sul4 (Fig. 2a), where hydraulic heads increased within a few hours after rain events of about 10 mm or more, although amplitudes remained usually small (for details see Fig. S2 of the ESM). The lowest responsivity (but at highest amplitudes) was observed at Has5 (Fig. 2b), which seemed responsive to very strong precipitation events only, primarily during the recharge period in the winter season when evapotranspiration is low. In the majority of the wells, water-table reactions after rain events of at least 20 mm occurred within 1 or 2 days; however, extended but lower intensity rain events sufficed to trigger a hydraulic response during the winter season. All groundwater wells, except Has5, show a clear reaction of SEC to rain events (Fig. 2) within 1-4 days after the events, which suggests that the observed increase in H was at least partly governed by inflow of a different water component, and not solely a pressure response. At Has2 and Sul1, the SEC response (i.e., dilution peaks) happened nearly synchronous with the rise in hydraulic heads, followed by slightly delayed (2-10 days) but fast increases in SEC. SEC dilution dominated at Sul4, where almost all hydraulic responses were accompanied by a small decrease of SEC. The 10-60% increase in SEC after rain events at Sul1, Has2, and Has4 indicates a contribution of a water component with higher mineralization than under baseflow conditions. A similar increase in SEC was observed at Has2 after a strong rain event in Mar 2020 (25 mm within 1 day), which was accompanied by increases of about 10-20% in the concentrations of Na + , Ca 2+ , Mg 2+ , and Cl − , respectively (see Table S2 of the ESM). The much higher variability of SEC at Sul3 is most likely related to the more saline groundwater component from the Middle Muschelkalk, which is strongly diluted after rain events. Overall, the combined hydraulic and chemical responses observed suggest that at least two water components are mobilized during stronger rain events, which reach the groundwater wells within a few days.   (Fig. 3). The median NO 3 − concentrations of the four major field sites show an apparent increase from Sulz am Eck towards Haslach and the AMQ (Fig. 3). By contrast, NO 3 − concentrations were relatively low or undetectable (<0.002 mM) at ArtAlt. In contrast to the pronounced spatial variability at Sulz am Eck and Haslach, seasonal NO 3 − concentration fluctuations appear to be minor (see Fig. 4; Fig. S3 (Fig. 5). The NO 3 − δ 15 N values ranged from~+8 to +25‰ (Fig. 5a), whereas δ 18 O-NO 3 − values ranged between 0 and +20‰ (Fig. 5b) (Fig. 6). δ 15 N-NO 2 − values vary widely between the wells and range from roughly −9 to +20‰ (Fig. 6a). The NO 2 − δ 15 N values  Table S3 of the ESM). Standard deviation for technical replicates represented by error bars production/accumulation and fractional denitrification within the groundwater, or due to method-inherent O-isotope exchange during storage and/or analysis with the azide method, cannot be excluded.

Land use in the well capture zones
Groundwater capture zone delineation was used to evaluate the potential agricultural impact on NO 3 − concentrations at the single wells. Capture zones of all wells located at the two major field sites Sulz am Eck and Haslach were delineated by backward particle tracking, using a recently established groundwater flow model of the Muschelkalk aquifer (D'Affonseca et al. 2020). All capture zones followed the general groundwater flow direction across forested and arable land (Fig. 7). While groundwater flow at Sulz is mainly controlled by the aquifer dipping direction (ESE), the flow regime at Haslach is increasingly influenced by the Ammer springs, causing the trajectories of the Haslach wells to be more SSE oriented; therefore, the length of the capture zones ranged from <0.5 km (Sul4, Sul7) to~6 km (Has3, Has4). Based on these delineations, the capture zone area (ha) was estimated for the major land use categories using aerial images of the study period (Fig. S5 of the ESM). Overall, wells at both field sites appear to be dominated by water deriving from agricultural land. Proportions (%) of the main land use categories associated to each well are shown in Fig. S6 of the ESM. Wells such as Sul1, Sul2a and Has5 are instead proportionally more affected by forested areas.

Discussion
The groundwater in the recharge area of the karst aquifer displayed a relatively large spatial variability of NO 3 − concentrations, exceeding the temporal variations at individual sites in most cases (see Fig. 3). In the sections that follow, deeper insight into the hydrogeological and biogeochemical processes that govern the observed NO 3 − variability, both from a hydrochemical as well as from an isotopic-tracer perspective, are provided. The complementary information provided by monitoring wells connected to different portions of the heterogeneous flow system, as compared to karstic springs, allow in turn the refinement of the conceptual models of N transport in the karst aquifer.

Nitrogen sources and nitrate variability
Since arable land comprises a large proportion of land use in the recharge area, most groundwater NO 3 − can be expected to originate from agricultural practices (Einsiedl and Mayer 2006;Gurdak and Qi 2012). Other potential anthropogenic N sources in the study area include septic systems of isolated houses and workshops. However, a screening of typically sewage-derived organic micro-contaminants (i.e., carbamazepine, clofibric acid, metoprolol, sulfamethoxazole; e.g., Heberer 2002;Whelehan et al. 2010) in the study area did not show concentrations above 1 ng/L for any of these substances, suggesting that N-compound contamination associated with wastewater is negligible. Inorganic constituents derived from agricultural additives such as Cl − or SEC, are commonly considered useful indicators for agriculturally derived NO 3 − inputs in groundwater (Böhlke 2002;Boy-Roura et al. 2013). The Na + /Cl − molar ratios of <1 (Fig. S7 of the ESM) indicate that Cl − mainly derived from sources other than NaCl such as KCl, which is commonly applied as artificial fertilizer (Böhlke 2002). Analysis of dissolved Cl − and NO 3 − concentrations performed for all four field sites revealed a positive correlation of the two solutes at Sulz and Haslach (Fig. 8), supporting the influence of agriculture on NO 3 − concentrations. A different pattern was found for AMQ, where no clear relationship between NO 3 − and Cl − was discernible, and a regression analysis even yielded a negative slope (Fig. 8c). However, NO 3 − concentrations at AMQ showed a positive trend with elevated discharge, whereas Cl − was diluted at higher discharge (Fig. S8 of the ESM). Such opposing relations are usually observed when discharge is controlled by rapid infiltration and transport of event water through Fig. 6 a δ 15 N- (Musgrove et al. 2016;Asante et al. 2018;Husic et al. 2019 (Yue et al. 2017;Nikolenko et al. 2018), which bypassed the organic soil-N pool during infiltration of rain water through larger karstic structures. However, SEC data show that water mobilized during rain events towards the monitoring wells comprises only a small fraction of a quick flow component. Therefore, it seems unlikely that NO 3 − from atmospheric deposition reaches the monitoring wells undiluted and without further (bio-) geochemical processing.

Impact of land use on nitrate concentrations
Agricultural land use has been generally recognized as the major control of NO 3 − concentrations in groundwater (Nolan and Stoner 2000; Gurdak and Qi 2012). At the study sites, a corresponding trend is indicated by increasing NO 3 − contents with a larger proportion [fraction, %] of agricultural land use (arable land and intensive grassland) in the capture zone of the wells (Fig. 9). However, for wells with NO 3 the different wells suggest that agricultural N inputs massively contribute to the overall NO 3 − loading of the studied groundwater, other more local processes appear to affect spatial variability ( Fig. 7) additionally, and in some instances to a greater extent.

Nitrate transport to the water table
Various conceptual models describing contaminant (NO 3 − ) transport in karstified aquifers have been proposed in previous studies Bakalowicz 2005;Hartmann et al. 2014). The conceptual model presented here (Fig. 10) differentiates between flow and transport towards groundwater monitoring wells at two distinct locations with respect to the fracture and conduit network, and karstic springs directly connected to karst conduits. Flow within the fractured aquifer is usually characterized by a broad distribution of flow paths ranging from quick or shaft flow (in conduits, joints, or larger  (Perrin et al. 2003;Einsiedl 2005;Page et al. 2017). The epikarst comprises the near-surface weathering zone of the carbonate rocks immediately below the soil and consists of fractures and joints widened by dissolution (Ford and Williams 2007;Medici et al. 2019). How exactly N compounds are transported by infiltrating water towards, and below, the water table (e.g., the transport paths indicated in Fig. 10) can be expected to also influence NO 3 − concentrations in groundwater (Husic et al. 2019).
The combined hydraulic and chemical response observed at the study site suggested that at least two water components are mobilized during stronger rain events, reaching the groundwater monitoring wells within a few days. Evidence for the contribution of quick flow infiltrating the karst system along shafts and larger fractures is provided by the observed SEC dilution peaks occurring nearly simultaneously (<1 day) with major rain events in Sul1, Sul4 and Has2. This is further supported by the hydraulic head response at these wells, showing high responsivity yet with only moderate amplitudes, which might best be explained by the close proximity to large-volume karst structures (Ford and Williams 2007;Medici et al. 2019) (see well A, Fig. 10). However, the response of hydraulic heads is not necessarily caused by water fluxes at the well locations and hence, in contrast to spring hydrographs, cannot be used to separate flow components directly.
In some of the wells, the hydraulic-head response was accompanied with distinct subsequent increase in SEC, indicating the mobilization of a water component with higher mineralization within only a few days. Similar increases in SEC observed at karstic springs have frequently been attributed to pulses of water from the subcutaneous or epikarst zone (Williams 1983;Lakey and Krothe 1996;Ravbar et al. 2011). The displacement of water stored temporarily in the epikarst or other compartments by large rain events is therefore a plausible cause explaining the SEC variations. This is supported by the increase in Ca 2+ und Mg 2+ concentrations observed after a strong rain event in Has2 (Table S2 of the ESM). In Fig. 10, such a scenario would correspond to the transport path to well B. Due to a decreasing permeability with depth towards the unweathered and less karstified bedrock, the epikarst is known to act as transient storage of infiltrating Fig. 10 Conceptual model of potential nitrate sources, transport, and turnover in the recharge area of karstic aquifers. Wells close to major conduits (A) can be expected to respond to nearly all major rain events (though at low amplitude). In contrast, wells influenced by waters deriving from the fracture and fissure system (B) will show lower responsivities yet with higher amplitudes due to the low porosity of the carbonic rocks. Zones promoting denitrification are possibly present in oxygen-depleted niches located within the epikarst or the fissure/matrix network of the phreatic zone. Isotope values are based on Nikolenko et al. 2018 water (Jones 2003;Fidelibus et al. 2017). The resulting longer residence times, as well as high CO 2 and organic carbon contents in the nearby soil, not only promote carbonate dissolution but possibly may also cause transient O 2 deficiency and in fractional N turnover (Fidelibus et al. 2017). Besides the epikarst, transient storage could also be induced by diffusive exchange between the rock matrix and small fractures and fissures in the phreatic zone. These at least transiently watersaturated zones might also develop the potential to promote organotrophic denitrification, e.g., in anoxic niches (Husic et al. 2019), although the overall contribution to the spatial variability of NO 3 − may be limited. Given the relatively large increase of SEC after rain events, as compared to dilution of SEC, the groundwater wells in the recharge area of the karst aquifer appear to be predominantly influenced by water from transient storage rather than by quick flow.
In contrast to the monitoring wells of Sulz am Eck and Haslach, former tracer tests gave clear evidence that the AMQ is directly connected to the karstic conduit system (Harreß 1973). Hence, the steady NO 3 − levels at AMQ represent an average value integrating over the catchment's different water flow paths and storage compartments drained by the conduit system. NO 3 − originating from the near soil surface is considered to dominate recharging water since it is mobilized by intense rainfalls resulting in high NO 3 − but low Cl − concentrations during peak discharge of the spring (Huebsch et al. 2014).

Evidence for N turnover
In general, karstified aquifers and particularly karstic springs are known to contain oxygen-rich waters (Kendall and Doctor 2005;Ford and Williams 2007;Benk et al. 2019). DO concentrations above 200 nM have been shown to impact denitrification-related gene expression (Dalsgaard et al. 2014) and concentrations above 5 μM has traditionally been assumed to inhibit denitrification (Codispoti et al. 2001); therefore, the vadose zone and the epikarst are generally considered as storage compartments for NO 3 − , and a zone of nitrification rather than denitrification Fidelibus et al. 2017;Ascott et al. 2017). However, previous findings on the O 2 sensitivity of denitrifying enzymes have partially been revised (Takaya et al. 2003;Schreiber et al. 2012;Kuypers et al. 2018) and various microorganisms have been shown to successfully perform dissimilatory NO 3 − reduction to NO 2 − also under hypoxic and even oxic conditions (Roco et al. 2016;Zhou et al. 2019;Cojean et al. 2019). This is in accordance with studies providing evidence for ongoing denitrification within the vadose and/or epikarst zone (Brahana et al. 2005;Kuniansky and Spangler 2014;Panno et al. 2001), as well as at oxic and anoxic groundwater monitoring wells of a limestone aquifer (Henson et al. 2017;Wegner et al. 2018). The data from this study support these previous findings, suggesting the presence of microoxic to anoxic niches located within the epikarst (close to the organic rich soil) and/or the phreatic zone (omnipresent Fe(II)-bearing carbonates). Some of the investigated wells show large temporal changes of dissolved oxygen concentration, with minimum DO values <~51 μM at Sul1 (Fig. S10 of the ESM). The low DO concentrations in Sul1 and other wells (see Table 1) indicate that parts of the water in these wells, at least temporarily, undergo enhanced oxygen consumption. While O 2 respiration obviously does not lead to complete O 2 consumption and anoxia in the major conduits, it is reasonable to hypothesize that suboxic and even anoxic conditions can occur locally, providing important microenvironments for denitrification in waters already relatively low in O 2 .
In order to elucidate possible N turnover processes, δ 15 Nand δ 18 O-NO 3 − values versus molar NO 3 − concentrations are plotted on an antiproportional (Keeling plot) and a logarithmic scale (Rayleigh plot) (Fig. 11a,b, after Kendall and Doctor 2005). A fractionating process like denitrification would lead to an exponential relationship, and plotting δ 15 N-and δ 18 O-NO 3 − values versus the natural logarithm of the NO 3 − concentrations will produce a straight line. If an exponential relation is not observed and a straight line is produced on a Keeling plot, the NO 3 − concentration and isotope composition trends are governed by simple mixing of two endmembers. The available data trends are explained equally well by the two regression models, suggesting that the two processes (mixing with a high-δ N source and fractionation during denitrification) cannot be clearly distinguished at the investigated sites.
Mixing of water and NO 3 − from different sources is the natural result of sample collection by extracting groundwater from extended well screens, thus including contributions from different groundwater flow paths. However, the presence of N turnover processes in the groundwater system is clearly indicated by NO 2 − concentrations of up to 0.14 mM (see Fig. S4 of the ESM), which were observed in wells at Sulz am Eck and Haslach during the sampling campaign in Apr 2016. NO 2 − , a highly reactive intermediate, is commonly produced enzymatically either during nitrification or denitrification (Knowles 1982;Kendall and Aravena 2000;Zhu et al. 2013 Although the results indicate that denitrification is ongoing in the aquifer, there is no clear evidence on which electron donor is involved in NO 3 − reduction. Organic matter, due to proximity to soils, may be important in the epikarst; however, DOC concentrations were generally relatively low (<0.3 mM; Fig. S12 of the ESM), and no correlation was observed between NO 3 − and DOC concentrations (Fig. S13 of the ESM), suggesting that metabolic pathways other than organotrophic denitrification might play a role, e.g., involving Fe(II)-bearing minerals (Wegner et al. 2018;Schwab et al. 2019) such as pyrite (up to several weight% in the Muschelkalk). In light of the overall high sulfate concentrations at the studied sites (0.8 mM; due to the dissolution of gypsum), it is impossible to detect the production of sulfate via sulphide-dependent denitrification (via its anti-correlation with NO 3 − ), even if rates were quite high (Fig. S14 of the ESM). Dissolved Fe 2+ concentrations were usually below 0.02 mM, but reached significantly higher levels at least at Has5 and Sul3 (0.03 and 0.4 mM, respectively; Fig. S15 of the ESM ). The presence of free Fe 2+ may be taken as evidence for microbiallymediated pyrite oxidation (Hayakawa et al. 2013) and may serve as indirect evidence for chemolithotrophic and/or mixotrophic denitrification. Although the pyrite-bearing carbonate formations provide plenty of these niches in the rock matrix, turnover rates are likely to be very slow, since substrate diffusion is limited and thus the accessibility to microorganisms in the micritic limestones is also limited.

Conclusions
A strong spatial NO 3 − variability in the groundwater of a fractured and karstified aquifer was observed. These variations can most likely be attributed to (1) local variances in agricultural practices within the capture zones of the Fig. 11 Analysing δ 15 N-(circles) and δ 18 O-(triangles) NO 3 − values for a source mixing using a Keeling plot and b denitrification using a Rayleigh plot. Standard deviation from technical replicates for CF-IRMS measurement is included as error bars (n = 2). ArtAlt, which is commonly anoxic, is added for comparison investigated groundwater monitoring wells, (2) a variable proportion of (often insufficiently known) karstic entryways below arable land and forests, and (3) fractional NO 3 − loss by denitrification in niches of the aquifer with near anoxic conditions within the otherwise oxygen-rich karstic formations of the recharge area. The multi-species (NO 3 − , NO 2 − ) isotope approach allowed to better constrain the N turnover processes which potentially occur during transient storage of water and solutes in the temporarily saturated network of the epikarst or the fissure network of the phreatic zone. Karstic springs, as natural outlets of the entire conduit and fracture network of at least parts of the aquifer, are known to provide integrated information on water and solute inputs and transport along different flow paths. The information gained by the additional investigation of several monitoring wells in the recharge area turned out to be an important contribution for the development of the presented conceptual model on NO 3 − transport and turnover. This was due to the observed spatially varying chemical and hydraulic responses reflecting the connection of the monitoring wells to different parts of the fracture and fissure network which might favour transitory anoxic or suboxic conditions. Hence, the findings here suggest that future studies might need to consider possible impacts of near anoxic niches in the oxygen-rich karst recharge areas and to include not only karstic springs but also local groundwater observation points to better understand N transport and turnover dynamics within the karst groundwater systems.
Funding information Open Access funding enabled and organized by Projekt DEAL. This study was supported by the German Research Foundation (DFG)-funded RTG 1708 "Molecular Principles of Bacterial Survival Strategies" and the Collaborative Research Centre 1253 CAMPOS (Project 5: Fractured Aquifer) (DFG Grant Agreement SFB 1253/1).
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.