Corresponding author: S. Krause, School of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham, Edgbaston, B15 2TT, UK. (firstname.lastname@example.org)
 Biogeochemical turnover in hyporheic zones is known to have the potential to affect the chemical signature of surface water cycling through shallow streambed sediments. This study investigates the impact of streambed physical properties on the fate of nitrate and dissolved oxygen in groundwater upwelling through the streambed of a lowland river. For analyzing depth-dependent patterns and zonation of nitrogen concentrations, diffuse gel probes in shallow (top 15 cm) streambed sediments have been deployed in a nested setup together with multilevel minipiezometers for streambed sediments of 15–150 cm. Spatial heterogeneity of groundwater upwelling was controlled by patterns of low-conductivity peat and clay strata that caused locally confined conditions, suggesting increased streambed residence times. Nitrate concentrations in the upwelling groundwater changed by up to 68.06 mg L−1 within the top 15 cm of streambed sediments and by up to 107.47 mg L−1 at depths of 15–150 cm, indicating that significant nitrogen turnover was not restricted to shallow streambed sediments. Intensive reduction of nitrate concentrations was found, in particular, in vicinity of low-conductivity streambed strata. The coincidence of confined groundwater upwelling and reduced oxygen concentrations at these locations suggests that increased residence times and associated depletion of dissolved oxygen create conditions favorable for nitrate reduction. Our results highlight that increased nitrogen turnover at aquifer-river interfaces is not necessarily limited to shallow streambed zones, where surface water is mixing with groundwater, but can affect upwelling groundwater in reactive hot spots that extend to greater streambed depths and beyond hyporheic mixing zones.
 Anthropogenic alterations of the nitrogen cycle have been found to often cause adverse impacts on groundwater and surface water quality and habitat conditions [McMahon and Bohlke, 1996; Brunke and Gonser, 1997; Duff and Triska, 2000; Robertson and Wood, 2010; Krause et al., 2011a, 2011b, 2011c], and are considered to represent a growing environmental problem of global concern [Seitzinger, 1988; Vitousek et al., 1997; Seitzinger et al., 2006; Schlesinger et al., 2006; Diaz and Rosenberg, 2008; Galloway et al., 2008].
 River corridors have been recognized for their increased potential for nitrate uptake where riparian floodplains [Pinay et al., 1994; Vidon and Hill, 2004a, 2004b] and hyporheic zones (HZs) [Holmes et al., 1996; Wondzell and Swanson, 1996; Hill and Lymburner, 1998; McClain et al., 2003] represent hot spots of nitrate turnover, leading to a reduction of stream nitrogen loads. However, HZs have been found to not always function as a sink, where nitrate fluxes are attenuated due to denitrification [Pinay et al., 1994, 2009; Hill et al., 1998], but also to be a potential source of nitrate, where processes of ammonification and nitrification can increase nitrate concentrations in the HZ and the stream, respectively [Jones et al., 1995; Duff and Triska, 2000; Storey et al., 2004].
 Nitrate turnover rates in hyporheic environments have been shown to be controlled by the presence of labile bio-available oxidizable organic carbon [Seitzinger, 1994; Chafiq et al., 1999; Hill and Cardaci, 2004; Zarnetske et al., 2011b], as well as streambed residence times and associated oxygen depletion along hyporheic flow paths [Zarnetske et al., 2011a; Pinay et al., 2009]. Surface water residence times in HZs and associated nitrate uptake at small scales are strongly affected by streambed topography and channel bed form [Hill et al., 1998; Kasahara and Wondzell, 2003; Kasahara and Hill, 2006; Cardenas et al., 2004, 2008; Boano et al., 2010], whereas at larger (stream reach) scales, hyporheic exchange is predominantly controlled by valley and channel geometry, ambient groundwater flow, and heterogeneities of aquifer properties [Malcolm et al., 2005; Fleckenstein et al., 2006; Cardenas and Wilson, 2006; Boano et al., 2006, 2007; Frei et al., 2009]. Zarnetske et al. [2011a] found hyporheic residence times within a streambed gravel bar to have significant impact on nitrogen processing rates. Although nitrification coincided with shorter residence times until dissolved oxygen (DO) was sufficiently reduced, longer residence times in the then oxygen-depleted hyporheic environment caused an increase of denitrification rates. Further studies have also confirmed a carbon limitation of microbial denitrification in riparian and hyporheic environments [Baker et al., 1999; Pinay et al., 2002; Zarnetske et al., 2011b].
 The efficiency of nitrogen processing in streambed environments has previously been described to be enhanced in small headwater streams and midstream sections [Peterson et al., 2001; Alexander et al., 2000; Mulholland et al., 2008; Zarnetske et al., 2011a, 2011b] where the relative ratio of hyporheic exchange fluxes between groundwater and surface water compared to stream discharge is highest. The majority of previous research has focused on the biogeochemical impact of hyporheic cycling with a primary perspective on the transport and transformation of surface water that is downwelling into the streambed and mixes with groundwater before it upwells further downstream.
 The importance of the streambed passage for the fate of nitrogen in the upwelling groundwater is still insufficiently understood, although it is widely recognized that groundwater represents a major source of nitrogen for many lowland streams in agriculturally dominated catchments [Foster, 2000; Gooddy and Besien, 2007; Krause et al., 2009a, 2009b]. The often rather high DO content and only limited availability of electron donors strongly delimits the denitrification capacity of many aquifers, resulting in a minor natural attenuation potential. Coupled with the relatively long groundwater residence times, the discharge of groundwater nitrogen therefore represents a long-term risk for surface water chemical status, environmental health, and ecosystem services [Duff and Triska, 2000; Rivett et al., 2008a; Wang et al., 2012].
Kennedy et al. [2009a, 2009b] provided important insight into the flow controls on streambed nitrogen cycling within upwelling groundwater of an agricultural stream. They identified nitrogen concentration changes in relation to hydraulic conductivity and head gradient–controlled streambed residence times along several hundred meter long stream reaches. Investigations of Krause et al. [2009b] indicated that the rather low-moderate groundwater nitrate concentrations in the river Leith (UK) changed substantially in the streambed. Nitrate concentration changes were correlated with the thickness of shallow (max 100 cm) drift deposits overlying the sandstone bedrock and spatially variable upwelling rates, suggesting a direct impact of groundwater residence times in the organic enriched streambed sediments. Initial investigations at a UK lowland river (river Tern) revealed that high spatial variability of aquifer-river exchange flow patterns was caused by spatially disconnected lenses of peat and clay in the streambed [Krause et al., 2012a] that controlled spatial patterns of groundwater upwelling, enhancing shallow hyporheic exchange and potential streambed residence times [Angermann et al., 2012; Krause et al., 2012a].
 This study aims to investigate how nitrogen-rich groundwater is affected by the dynamic upwelling passage across heterogeneous streambed sediments. It therefore uses the river Tern as an example to represent typical agriculturally dominated lowland rivers to test the following hypothesis: Significant changes in nitrate concentrations and chemical turnover are not restricted to hyporheic streambed sediments where surface water is mixing with groundwater, but extend to depths beyond the hyporheic exchange zone where the transformation of nitrogen in upwelling groundwater is not affected by mixing with the surface water but controlled by the high spatial variability in physical streambed properties.
 Therefore, this study will (1) quantify the patterns and extent of nitrate concentration changes at small-scale with high spatial resolution (2 cm) for the top streambed sediments of 0–15 cm, (2) identify the fate of nitrate and DO in upwelling groundwater at a 250 m long stream reach along nested piezometer profiles at streambed depths ranging from 15 to 150 cm, (3) evaluate the nitrate variability and concentration changes as a function of streambed depths and in correlation to DO concentration changes with depth, and (4) investigate the correlation of nitrogen and oxygen concentration changes with depth, and analyze the impact of streambed physical properties on the fate of nitrate and oxygen in upwelling groundwater with a particular focus on flow-confining peat and clay strata within the streambed.
2 Materials and Methods
2.1 Field Site
 The field site is an approximately 250 m long meandering stream section of the river Tern (2°53′W, 52°86′N) in the UK (Figure 1a). The field-site geology is dominated by drift deposits overlying Permo-Triassic sandstone, which represents one of the UK's major groundwater aquifers [Wheater and Peach, 2004]. The river channel is 5–8 m wide and surrounded by an extensive, mainly agricultural floodplain (Figure 1b). Channel morphology is dominated by steep river banks, pool-riffle-pool sequences (Figure 1c), and partly vegetated channel bars. The river Tern has been selected by the UK Natural Environment Research Council as a study area representative for agricultural lowland rivers under the Lowland Catchment Research Programme [Wheater and Peach, 2004], and it is monitored by the Environment Agency as part of the Shropshire Groundwater Scheme [Streetly and Shepley, 2002], providing monitoring infrastructure and baseline data (groundwater observation boreholes, borehole logs, flow gauge) used in this study.
 Streambed sediments in the research area vary between midsized gravels, different sizes of sands, and fine silty material with hydraulic conductivities ranging from 10−3 to 10−5 m s−1 [Krause et al., 2012a; Angermann et al., 2012]. Clay and peat layers in the streambed are characterized by significantly lower hydraulic conductivities of 10−8 to 10−9 m s−1 [Krause et al., 2012a]. Abundance, thickness, and depth of peat or clay strata in the research area varies substantially, as shown in Figure 2a for two example streambed profiles. The low hydraulic conductivities of these structures, which are common streambed features in lowland rivers, have been shown to have significant impact on the regional groundwater upwelling, causing flow confinement and increased streambed residence times [Krause et al., 2012a]. This study focused on summer base flow conditions between May and October 2009. Meteorological data were recorded at the nearby Keele weather station (2°16′12.90″W, 52°59′55.86″N), and stream discharge data were provided by the Environment Agency from the Ternhill gauging station (2°55′12″W, 52°87′92″N). Precipitation during the observation period was dominated by an extended wet period at the end of July and early August 2009 (Figure 3a), with the sum of monthly rainfall exceeding 100 mm.
 Within the observation period, mean daily river discharge for Ternhill (Q) averaged 0.91 m3 s−1 (Figure 3a), matching long-term mean discharge records of 0.85 m3 s−1 with a 95% exceedance (Q95) of 0.4 m3 s−1 and a 10% exceedance (Q10) of 1.38 m3 s−1 (data period 1972–2011, UK National River Flow Archive, http://www.ceh.ac.uk/data/nrfa/data/station.html?54044). Discharge dynamics were characterized by intensive temporal variability with minimum base flow discharges of 0.51 m3 s−1 to maximum storm discharges of 1.77 m3 s−1 (Figure 3a). The general summer base flow period was interrupted by a major discharge event in the end of July to early August, with more than 20 days with discharges higher than 1.0 m3 s−1 (Figure 3a), resulting from prolonged rainfalls in July 2009 (Figure 3a).
2.2 Experimental Infrastructure
 Differential GPS-based topographic surveys covering the channel, streambed, and riparian floodplain of the field site were conducted in June 2009, providing a high-resolution digital elevation model with a vertical precision of 1 cm and a horizontal precision of 25 cm, and exact surveyed heights of installed groundwater boreholes and piezometers.
 A network comprising ten 3 m deep groundwater boreholes covering the drift deposits of the riparian floodplain and two stream gauges (Figure 1b) was installed in May to June 2008. Water levels were recorded at five selected groundwater boreholes and two surface water observation points (Figure 1) by automatically recording pressure transducers at 5 to 15 minute intervals (Table 1). Groundwater and surface water heads monitored at the observation boreholes were corrected for barometric pressure fluctuations using an atmospheric pressure sensor located above the groundwater level (GW) in groundwater borehole GW7 (Figure 1b). Manual head measurements at all boreholes were conducted during all sampling dates (Table 1) for validation of automatically recorded hydraulic heads.
Table 1. Spatial and Temporal Resolution and Accuracy of Observed Parameters for the 21 May 2009 to 30 September 2009 Experimental Monitoring Period
Hydraulic head–SW/Hydraulic head–GW/Barometric head
5 min/15 min/5 min 12/06/2009–30/09/2009
Solinst LT M5/F15 diver, combined water level and temperature logger/Solinst BaroLogger
Hydraulic head streambed surveys 2009
21/05, 02/06, 19/06, 30/06, 31/07, 21/08, 30/09
Streambed piezometer and graduated dipmeter
1 h 12/06/2009–30/09/2009
Keele, meteorological station (18 km distance)
03/08/2009 (200 × 1 min measurements)
Sensornet Halo (2 m sampling resolution)
1 h 12/06/2009–30/09/2009
EA gauging station Ternhill
Multilevel minipiezometer and diffuse equilibrium in thin film (DET) probe sampling and analysis for nitrate, dissolved oxygen
21/05, 02/06, 19/06, 30/06, 31/07, 21/08, 30/09 (analysis within 48 h), DET probes deployed 07/08/2009–21/08/2009
Nitrate: Dionex ICS2500 ion chromatograph (Dionex UK Ltd., Camberley, Surrey, UK)
Limit of detection (LOD) of 0.3 mg/L N, precision of 2.5%
Dissolved oxygen: Hannah Instruments HI-9828/20 Multi Parameter Probe
 A network of 27 multilevel mini piezometers (P1–P27) was installed for observing interstitial pore water pressure head distributions at sediments depths of 15–150 cm below the streambed surface (Figure 2c). The piezometer design followed those published by Rivett et al. [2008b] with a central High-density Polyethylene (HDPE) monitoring tube of 10 mm inner diameter used for water level observations and up to seven outer Polytetrafluoroethylene (PTFE) sampling tubes of 1 mm inner diameter (Figure 2c). The PTFE sampling tubes at depth intervals of 15–25 cm allowed for accurate sampling of small pore water volumes from discrete depth profiles with higher monitoring resolution and lower risk for sample overlaps than state of the art piezometers of larger diameter. Samples were extracted into 40 ml syringes after purging the sampling tubes from trapped water at seven sampling dates (Table 1), reducing the risk for gas exchange. Collected samples were filtered with 0.45 µm cellulose acetate filters before the analysis.
 Piezometers were driven into streambed sediments covering a network comprising a longitudinal transect along the stream reach and several cross sections extending the observations toward the river banks (Figure 1b). Hydraulic heads in streambed piezometers were monitored manually during all 2009 sample dates (Table 1) using a graduated electric contact meter. Observations of hydraulic head differences between interstitial pore water in streambed sediments and surface water were used to calculate vertical hydraulic gradients (VHGs), indicating the intensity and direction of aquifer–river exchange fluxes. VHGs were calculated as Δh/Δl, with Δh given by the difference of water tables observed inside and outside of the piezometer, and Δl given by the distance between the midscreen depth and the surface water-sediment interface. Based on previous experience [Krause et al., 2009b; Kaeser et al., 2009], the accuracy of dipmeter-based hydraulic head observations was assumed to be ±3 mm head, which also accounts for measurement uncertainties related to turbulent flow conditions at the outside of the piezometers [Kaeser et al., 2009].
 A high density of streambed cores for the characterization of locations and extent of low-conductivity strata would have critically altered the streambed sediments and, in particular, created preferential flow pathways where confining peat and clay lenses would have been disturbed. Spatial information on peat or clay layers was therefore obtained during the installation of piezometers and associated with depths of persistent failure of sample extraction [Krause et al., 2012a]. Fiber-optic distributed temperature sensing (FO-DTS) of streambed thermal anomalies provided spatially distributed information on thermal anomalies that were interpreted as the results of hot spots or inhibition of groundwater upwelling as caused by streambed peat or clay structures. Patterns of detected peat or clay structures and resulting groundwater upwelling fluxes, as well as the design, deployment, and setup of the FO-DTS monitoring, are explained in detail by Krause et al. [2012a]. The comparative analysis of VHG- and FO-DTS–derived streambed temperature anomalies provided evidence that particularly high VHGs in the research area did not necessarily reflect high groundwater flow rates through the streambed, but rather indicated confined conditions and upwelling inhibition beneath low-conductivity peat and clay strata [Krause et al., 2012a].
 For sampling interstitial pore water nitrate concentrations in superficial streambed sediments, diffusive equilibrium in thin film (DET) probes [Davison et al., 1991; Harper et al., 1997; Ullah et al., 2012; DGT Research Ltd. (www.dgtresearch.com)] were used in this study. DET probes allow pore water sampling at high spatial resolution (millimeters to centimeters scale), by far exceeding the capabilities of multilevel minipiezometer observations. DET pore water sampling deploys plastic probes loaded with a hydrogel based on 5% polyacrylamide and 95% water [Davison et al., 1994]. A filter membrane (polysulfone) covers and protects the gel within the probe, whereas at the same time enabling fast diffusion of pore water solutes into the gel until equilibrium between interstitial pore water and gel concentrations is reached.
 The dimensions of the applied DET sediment probes were 240 × 40 × 5 mm (Figure 2), with a 0.8-mm-thick diffusive gel layer and 0.135 mm filter membrane. DET probes had a 20-mm-wide-and-150-mm-long window, which enabled the gel to be in contact with the sediment during deployment in the streambed. To allow the streambed sediments to resettle after probe deployment and to ensure optimal contact time for the diffusion of solutes, the probes were left within the streambed sediments for 14 days before sampling. In total, eight DET probes were installed at the center section of the investigated stream reach (Figure 1). Four probes were installed along a pronounced pool-riffle-pool sequence (Figure 1) in the thalweg of the investigated river section. The other four probes were installed within a shallow stream bar without substantial topographic variation at the far Western side of the same stream section. After extraction of the gel probes from the streambed sediments, the gels were cut into slices of 2 cm within 15 min and stored in centrifuge vials to avoid lateral diffusion. For analysis, the gel slices were back-equilibrated with known quantities of deionized water. After the analysis, the gel slices were weighted to determine their exact storage volume and back-calculate the in situ pore water concentrations. DET sampling was applied at 2 cm vertical resolution in this study, limited by the lower level of detection and the minimal sample volume of the used DIONEX ICS2500 ion chromatograph (Dionex UK Ltd., Camberley, Surrey, UK).
 Pore water, groundwater, and surface water samples were analyzed by a HANNAH HI-9828/20 multiprobe for DO concentrations before being analyzed by ion chromatography. All samples were cooled during transport and storage, and analyzed within 48 hours after sampling. All borehole and piezometer samples, as well as back-equilibrated DET sample solutions, were analyzed within 48 hours.
3.1 Aquifer-River Exchange Flow Patterns
 Groundwater levels at shallow riparian observation boreholes (Figure 1) exceeded surface water levels for most of the observation period (Figure 3b), indicating upwelling conditions with riparian groundwater discharging into the stream. Temporal variability in groundwater and surface water levels coincided and was dominated by base flow conditions until mid-July and again from mid-August. A more than 3 week episode of precipitation events in July (Figure 3a) was characterized by increased groundwater (up to 59.45 m asl (above sea level)) and surface water (59.38 m asl) levels. Surface water recharging the groundwater was indicated by inverse head gradients at GW3 during short (<2 days) storm events (Figure 3b). Groundwater heads at observation boreholes in the central floodplain (GW7) remained above surface water heads all the time, indicating that the inversion of aquifer-river exchange fluxes did not affect central parts of the riparian floodplain. General groundwater upwelling was furthermore confirmed by VHG observations, which were positive throughout the streambed piezometer network (Figure 4a). Patterns of time-averaged VHGs at respective piezometer locations varied substantially in space with a VHG range of 0.54 (min VHG = 0.03, max VHG = 0.57, and mean VHG = 0.28). Time-averaged VHGs at piezometer locations where peat or clay strata within the streambed have been identified by Krause et al. [2012a] were characterized by generally high VHGs (Figure 4a), in particular between P18 and P23 (Figure 1).
 At these locations, the existence and the lateral extent of low-conductivity peat and clay lenses, which strongly limited the upwelling of groundwater, was furthermore confirmed by FO-DTS monitoring of streambed temperature anomalies. FO-DTS surveys in summer conditions identified discrete streambed hot spots where the groundwater upwelling was inhibited. In contrast, at locations where the colder groundwater was upwelling through highly conductive sandy sediments, streambed temperatures were colder (Figure 4b) and significantly lower than the spatial mean (Figure 4c). Areas of highest streambed temperatures coincided spatially with high VHG locations, evidencing the flow confining impact of streambed peat or clay layers (Figure 4a).
3.2 Streambed Cycling of Nitrate: Concentration Ranges in Surface Water, Groundwater, and HZ
 The ranges of nitrate and DO concentrations in groundwater (floodplain boreholes), surface water, and interstitial pore water (streambed piezometers) varied significantly during the observation period. The range of nitrate concentration in interstitial pore water (127.4 mg L−1) was significantly higher than in surface water (20.65 mg L−1) and groundwater (63.40 mg L−1). Maximum concentrations in interstitial pore water were higher than any of the two end-members mixing in the streambed (Figure 5). Groundwater nitrate concentrations (average = 25.95 mg L−1) ranged from 3.03 to 66.43 mg L−1, whereas concentrations in surface water ranged from 21.49 to 42.14 mg L−1 (average = 31.41 mg L−1). Average interstitial pore water concentrations of 51.75 mg L−1 (min 3.01 = mg L−1; max = 130.41 mg L−1) were higher than average groundwater or surface water concentrations (Figure 5). Variation in nitrate concentrations was highest in interstitial pore water with standard deviations (SDs) of 25.31 mg L−1, followed by groundwater with SDs of 17.81 mg L−1 and surface water with SDs of 7.06 mg L−1.
 DO concentrations ranged from 1.07 to 7.07 mg L−1 (average = 3.79 mg L−1) in groundwater, 3.23 to 7.62 (average = 5.25 mg L−1) in surface water, and 0.57 to 9.73 (average = 4.55 mg L−1) in interstitial pore water. Although average DO concentrations were highest in surface water, the range of observed concentrations was higher in groundwater (6 mg L−1, with maximum concentrations below surface water) and in interstitial pore water (9.36 mg L−1, with maximum concentrations above surface water) (Figure 5). In comparison to nitrate concentrations, SDs of DO varied less between different water sources with SDs in surface water of 1.36 mg L−1, SDs in groundwater of 1.11 mg L−1, and interstitial pore water SDs of 0.98 mg L−1. For testing the statistical significance of differences in nitrate and DO concentration in groundwater boreholes, streambed piezometers, and surface water samples, a one-way analysis of variance was carried out, confirming a statistical difference at 0.05 significance level.
3.3 Small-Scale Variability in Interstitial Pore Water Nitrate Concentrations (DET Profiles)
 Profiles of nitrate concentrations in superficial streambed sediments exhibited considerable spatial variability horizontally as well as vertically. In the Eastern transect along the pool-riffle-pool sequence, nitrate concentrations increased within flow direction (Figure 6). The largest concentration difference was found between front pool (with a vertical profile average of 40.73 mg L−1) and the three following profiles with 88.70 mg L−1 at the riffle head, 93.20 mg L−1 at the riffle tail, and 90.13 mg L−1 at the downstream end pool. Vertical concentration ranges along the profiles varied from 33.29 mg L−1 (min = 32.96 mg L−1; max = 66.25 mg L−1) at the front pool to 29.46 mg L−1 (min = 73.22 mg L−1; max = 102.68 mg L−1) at the riffle head, 68.06 mg L−1 at the riffle tail (min = 60.43 mg L−1; max = 128.49 mg L−1) and 35.06 mg L−1 (min = 79.33 mg L−1; min = 114.39 mg L−1) at the downstream end pool (Figure 6).
 In the Western transect along the shallow stream bar, nitrate concentration ranges within the vertical profile were slightly higher than along the Eastern transect with 55.53 mg L−1 (min = 44.70 mg L−1; max = 100.23 mg L−1) at probe I to 53.93 mg L−1 (min = 92.08 mg L−1; max = 146.01 mg L−1) at probe II, 35.84 mg L−1 (min = 44.09 mg L−1; max = 79.93 mg L−1) at probe III, and 51.21 mg L−1 (min = 64.10 mg L−1; max = 115.31 mg L−1) at probe IV at the downstream end (Figure 6). Horizontal variability did not follow similarly clear patterns as along the neighboring Eastern pool-riffle-pool section with highest profile averages of 128.3 mg L−1 at probe II and 65.05 mg L−1 at probe I, 61.12 mg L−1 at probe III, and 82.26 mg L−1 at probe IV at the downstream end (Figure 6).
 Variability of concentrations along single vertical profiles was higher in the Western transect with SDs of 21.94 mg L−1 at the front probe I to 16.43 mg L−1 at probe II, 15.51 mg L−1 at probe III, and 19.49 mg L−1 at probe IV at the downstream end than at the Eastern pool-riffle-pool sequence with SDs of 11.22 mg L−1 at the front pool to 9.08 mg L−1 at the riffle head, 20.31 mg L−1 at the riffle tail, and 11.83 mg L−1 at the downstream end pool.
3.4 Intermediate-Scale Variability in Piezometer Nitrate Concentrations
 Figure 7 highlights the changes of nitrate (Figures 7a and 7c) and DO (Figures 7b and 7d) along the streambed passage by indicating average concentrations and concentration ranges within the deepest (Figures 7a and 7b) and the shallowest piezometer sampling points (Figures 7c and 7d). Overall, average nitrate concentrations in the shallowest piezometer sampling points (43.22 mg L−1) exceeded concentrations at the deepest locations (37.45 mg L−1). Similarly, the average concentration range observed at the deepest piezometers was lower with 20.15 mg L−1 in contrast to an average range of 25.30 mg L−1 at the shallowest piezometer sample points. The horizontal variation between piezometer locations at the same depth was higher at the deepest sampling points with average SDs over all sampling dates of 28.71 mg L−1 in contrast to SDs of 24.40 mg L−1 at shallow observation points. Average DO concentrations did vary only insignificantly between deepest (4.64 mg L−1) and shallowest sampling locations (4.34 mg L−1), and also in the horizontal dimension, spatial variability was comparable with SDs of 0.83 mg L−1 in deep sampling points and SDs of 0.80 mg L−1 at shallow locations. The lowest nitrate and DO concentrations at the deepest piezometer sampling points were found at locations with evidence of peat or clay layers along the profile (e.g., P10, P18, P19, P22, P23). Concentrations at these locations were generally lower, with average NO3 of 23.03 mg L−1 and average DO of 4.00 mg L−1, than at piezometers without flow confining peat or clay strata, with average NO3 of 43.53 mg L−1 and average DO of 4.91 mg L−1 (Figures 7c and 7d). Although concentrations at some of the piezometers with peat or clay strata changed significantly from the deepest to the shallowest sampling points (e.g., P23, P19, P10), others varied only marginally (e.g., P18, P22).
 Overall, nitrate concentrations at piezometers without peat or clay layers along the sampling profile ranged from 1.64 to 130.43 mg L−1 (average = 42.15 mg L−1), whereas concentrations at piezometers with peat or clay layers ranged from 3.11 to 79.83 mg L−1 (average = 17.60 mg L−1). DO at piezometers with no evidence for peat or clay layers varied between 1.64 and 9.39 mg L−1, with an average of 4.59 mg L−1. At the same time, DO concentrations varied from 1.64 to 7.73 mg L−1 (average = 4.24 mg L−1) at locations with known peat or clay layers along the piezometer profile. Generally, average concentrations of nitrate and DO at piezometers without peat or clay layers exceeded the concentrations observed at locations with peat or clay strata. Also, the variability of nitrate concentrations was higher at locations without peat and clay (SD = 28.38 mg L−1) than at piezometers with peat or clay (SD = 19.59 mg L−1). In contrast, SDs of DO concentrations were higher at locations with peat or clay layers (1.33 mg L−1) than at piezometers without (0.99 mg L−1).
 Vertical changes of nitrate and DO concentrations along the profiles of piezometer observation points varied significantly for different piezometer locations (Figure 8). While at some piezometers, concentrations (and ranges) changed only marginally along the vertical profile (e.g., P4, P6, P8) followed a clear trend along the upwelling gradient (e.g., P9, P10, P16, P18) (Figure 8). Piezometers with peat or clay layers were frequently characterized by significantly reduced nitrate and DO concentrations beneath and above the flow-inhibiting strata that exceeded the range of observed concentration changes in piezometer without peat or clay (Figure 8b). Infrequent concentration changes along the vertical profile at locations without peat or clay layers (e.g., P16 and, to a lesser degree, P6) represented mainly upward reductions of nitrate concentrations and increases in DO concentrations (Figure 8a).
 Figure 9 identifies the spatial variability of local minimum nitrate concentration in the streambed in comparison to surface water concentrations to compare them (a) to the receptor that is impacted by the groundwater nitrate contributions, and (b) in relation to the same absolute base, which allows consideration of their spatial alignment with respect to streambed morphology (Figure 1) and groundwater-surface water exchange fluxes (Figure 4). Although minimum streambed concentrations, in particular in the most upstream and downstream sections, as well as at piezometers P17, P23, and P24, were higher than river concentrations, nitrate concentrations at all other piezometers were up to 60–70 mg L−1 lower than surface water concentrations, in particular, in the vicinity of piezometers P14 to P22. In its majority, minimum nitrate concentrations were observed at streambed depths between 50 and 125 cm.
 Nitrate concentrations within streambed piezometers increased with higher DO concentrations, with both concentrations being related by a R2 of 0.64 for a second-order regression function (Figure 10a). The rather linear relation between nitrate and DO changed noticeably when DO concentrations dropped below 3 to 4 mg L−1, coinciding with a significant reduction in nitrate. That correlation, however, was weaker if only the uppermost piezometer sampling points were considered where R2 was only 0.45 (Figure 10b), but stronger (R2 = 0.73) for the deepest piezometer sampling points (Figure 10c).
4.1 Exchange Fluxes Between Groundwater and Surface Water
 Head gradients between surface water and groundwater observation boreholes (Figure 3), as well as throughout positive VHGs at the piezometer network (Figure 4a), indicated general groundwater upwelling during the observation period. Low-conductivity peat and clay lenses in the streambed had significant impact on the exchange fluxes between aquifer and river. VHGs underneath peat or clay layers were increased (Figure 4a), whereas FO-DTS observations at these locations did indicate no groundwater upwelling-associated temperature anomalies (Figures 4b and 4c), suggesting substantial flow confinement. Such conditions have been shown to increase streambed residence times of upwelling groundwater [Conant, 2004; Kennedy et al., 2009a, 2009b].
 These findings coincide with a range of previous heat tracer studies at the field site that confirmed the complex impact of streambed structural heterogeneity on aquifer–river exchange fluxes. As shown by Angermann et al.  for an area coinciding with the high-resolution DET sampling of this study, the reduction of groundwater upwelling caused by peat or clay lenses resulted in enhanced surface water downwelling and hyporheic mixing at these locations.
4.2 Depth-Dependent Nitrate Concentration Changes in the Streambed
 The passage of upwelling groundwater through the streambed evidently had significant impact on the fate of nitrate and DO. Both concentration ranges in shallow pore water exceeded the values of groundwater and surface water end-members (Figure 5), implying that observed streambed concentrations cannot be purely interpreted as the result of mixing. High nitrate concentrations above and low oxygen concentrations below potential mixing end-member concentrations (Figure 5) indicated oxygen-consuming, nitrate-producing processes within the streambed.
 Nitrate concentrations changed significantly at small scales in the top 15 cm sediments (Figure 6). In the Eastern transect, nitrate concentrations increased in stream flow direction (Figure 6), whereas concentrations in the Western section first increased and then decreased in downstream direction. Angermann et al.  identified the Eastern section as an area of increased horizontal hyporheic flow with surface water downwelling at the crest of a riffle and re-exfiltration at the riffle tail, suggesting increased hyporheic residence times. Zarnetske et al. [2011a] found similar increases in nitrate concentrations to result from residence time-controlled increases in turnover rates that caused increased nitrification until oxygen was depleted to a critical level after which denitrification rates increased. The observed concentration patterns suggest that, in comparison to the results of Zarnetske et al. [2011a], denitrification after oxygen depletion seems to be less relevant for at least the Eastern transect.
 Although multilevel minipiezometers did not capture spatial concentration changes at similarly detailed (cm) scale as DET gel probes, they detected substantial spatial variability in nitrate and DO concentrations (Figure 7) at larger scales of several tens of centimeters at locations by far exceeding depths of surface water downwelling and hyporheic mixing. As evidenced by VHG observations, nitrate concentration changes in upwelling groundwater at streambed depths between 15 and 150 cm cannot be explained by the mixing of groundwater and surface water. Furthermore, average vertical variability of nitrate concentrations within single piezometers profiles of 15 to 150 cm depth (SD = 23.99 mg L−1) exceeded the average vertical variability in the top 15 cm (SD = 15.73 mg L−1). Substantial alteration of nitrate concentrations within the top 15 cm of maximum 68.06 mg L−1 (average = 45.30 mg L−1) was exceeded by even higher concentration changes of maximum 107.47 mg L−1 (average = 48.99 mg L−1) along single piezometer profiles. This suggests that the spatial extent of increased biogeochemical turnover in upwelling groundwater in the research area is not restricted to the HZ where upwelling groundwater mixes with surface water, which coincides with the findings of Kennedy et al. [2009a, 2009b]. Our results complement previous studies [e.g., Storey et al., 2004; Pinay et al., 2009; Zarnetske et al., 2011a, 2011b] that found increased nitrate concentration changes and biogeochemical processing predominantly in a shallow hyporheic exchange zone. The detected high intensities of nitrogen cycling at depths exceeding hyporheic exchange support the hypothesis that the impact of hyporheic exchange, which can be of high importance for the self-purification of stream water, can be amplified by deeper transformation processes in the streambed when nutrient inputs are groundwater dominated.
4.3 Impact of Flow-Confining Peat and Clay Layers
 The spatial structure of low-conductive streambed peat and clay lenses did not only affect aquifer–river exchange flow patterns, but also significantly impacted on nitrogen and DO concentrations in the streambed pore water. Persistently lower nitrate and DO concentrations at streambed piezometers with peat or clay strata within the vertical profile indicated that these structures impact on the fate of nitrate and DO within groundwater upwelling through the streambed. Nitrate concentrations in the vicinity of peat or clay lenses were frequently reduced below groundwater or surface water values. In contrast, in piezometer profiles without peat or clay, nitrate concentrations in upwelling groundwater broke through without significant changes, or in rare cases showed some evidence of dilution with surface water in the uppermost sample points levels as, for instance, at P16 (Figure 8a). Nitrate concentration differences of up to 60 to 70 mg L−1 between streambed and surface water (Figure 9) spatially coincided with locations of identified upwelling inhibition by confining streambed peat or clay lenses (Figure 4), suggesting that streambed physical conditions controlled not only groundwater–surface water exchange fluxes, but also promoted the development of biogeochemical streambed hot spots.
 Increased residence times underneath flow-confining layers, as suggested by the observation of increased VHG and streambed thermal anomalies (Figure 4), are likely to directly affect the efficiency of chemical turnover rates as previously shown by Hill and Lymburner , Pinay et al. , or Zarnetske et al. [2011a]. However, although previous studies have focused on the residence times of surface water along the hyporheic or riparian passage, the results of this study provide comparable evidence for residence time impacts on nitrogen turnover in upwelling groundwater, which is independent from the spatial extent of surface water downwelling and hyporheic mixing.
 Similar to HZs, the intensity of nitrogen cycling at greater depths is affected by flow path–dependent residence times. However, in contrast to HZs, where flow pathways and residence time distributions have mainly been described as a function of streambed morphology [Kasahara and Wondzell, 2003; Cardenas et al., 2004; Kasahara and Hill, 2006; Boano et al., 2007], fluxes and residence times at these depths are predominantly controlled by spatial patterns in streambed permeability [Krause et al., 2012a]. Kennedy et al. [2009a, 2009b] have shown correlations between streambed hydraulic conductivity and water fluxes on residence time distributions and subsequent nitrate transformation rates for several stream reaches. The results of this study expand previous mechanistic process understanding by highlighting the particular impact of streambed peat or clay lenses on groundwater upwelling patterns (coinciding with the conceptual model derived by Conant ) and nitrate concentration changes.
 Substantially reduced DO concentrations around flow-confining streambed strata (Figures 7 and 8b) furthermore suggest that due to increased residence times, oxygen concentrations in the streambed have been depleted. The resulting rather anoxic conditions are more favorable for reductive denitrification. In addition, peat and clay strata provide more bio-available organic carbon and, hence, increase the availability of the electron donor required for denitrification. Such interpretation coincides with the findings of previous studies indicating that denitrification rates and nitrate concentrations were interlinked with the availability of oxidizable labile organic carbon [Hedin et al., 1998; Sobczak et al., 1998; Zarnetske et al., 2011b; Barnes et al., 2012].
 Direct comparison of DO and nitrate concentrations at streambed piezometers suggest a moderate (R2 = 0.64) relationship between both concentrations (Figure 10). This correlation was stronger at 150 cm depth (R2 = 0.73), where there was evidence for no surface water downwelling, than at 15 cm depth (R2 = 0. 45), which was at several locations evidently affected by hyporheic exchange. The rather linear relationship between DO and nitrate concentrations changed drastically when oxygen concentrations dropped below 3 to 4 mg L−1, coinciding with a significant reduction in nitrate (Figure 10). These patterns were similar to the threshold behavior described by Zarnetske et al. [2011a], who observed nitrate production along horizontal hyporheic exchange flow paths to be correlated with increased hyporheic residence times until the residence time–dependent depletion of DO caused denitrification. Our results indicate a similar relationship between oxygen depletion and nitrate reduction, but at greater depths and along a groundwater upwelling instead of a hyporheic flow path as low nitrate and oxygen concentrations in this study coincided with the location of peat and clay lenses (Figure 9) and associated increased residence times. The fact that hot spots of nitrate and DO concentration changes were associated with flow-confining peat or clay lenses highlights that intensive streambed nitrogen cycling exists at depths beyond the hyporheic mixing zone, and emphasizes the potential for substantial alteration of nitrate concentrations in upwelling groundwater.
 The results of this study provide evidence that high intensities of nitrogen transformation at the aquifer–river interface occur at multiple spatial scales including streambed depths beyond the HZ. Concentrations of DO and nitrate decreased by up to an order of magnitude in regions of low-conductivity streambed peat and clay layers. The observed nitrate reduction within confined streambed zones is interpreted as the result of increased residence times around low-conductivity strata, which evidently caused a depletion of DO concentrations, leading to the anoxic conditions required for denitrification. The observed relationship between oxygen and nitrate suggests that similar nonlinear behavior as observed for shallow hyporheic exchange fluxes also occurs during groundwater upwelling, which bears substantial consequences for integrated river basin and aquifer management.
 We acknowledge support of this work by the UK Natural Environment Research Council (NE/I016120) and the Nuffield Foundation. We also thank Kevin Voyce (Environment Agency of England and Wales) and Ian Wilshaw (University of Keele) for the provision of hydrometeorological data, as well as Tim Millington (University of Keele) for the development of the FO-DTS visualization script. We thank Aaron Packman and an anonymous reviewer for their very helpful comments that greatly improved this article.