Diffusive equilibrium in thin films provides evidence of suppression of hyporheic exchange and large‐scale nitrate transformation in a groundwater‐fed river

The hyporheic zone of riverbed sediments has the potential to attenuate nitrate from upwelling, polluted groundwater. However, the coarse‐scale (5–10 cm) measurement of nitrogen biogeochemistry in the hyporheic zone can often mask fine‐scale (<1 cm) biogeochemical patterns, especially in near‐surface sediments, leading to incomplete or inaccurate representation of the capacity of the hyporheic zone to transform upwelling NO3−. In this study, we utilised diffusive equilibrium in thin‐films samplers to capture high resolution (cm‐scale) vertical concentration profiles of NO3−, SO42−, Fe and Mn in the upper 15 cm of armoured and permeable riverbed sediments. The goal was to test whether nitrate attenuation was occurring in a sub‐reach characterised by strong vertical (upwelling) water fluxes. The vertical concentration profiles obtained from diffusive equilibrium in thin‐films samplers indicate considerable cm‐scale variability in NO3− (4.4 ± 2.9 mg N/L), SO42− (9.9 ± 3.1 mg/l) and dissolved Fe (1.6 ± 2.1 mg/l) and Mn (0.2 ± 0.2 mg/l). However, the overall trend suggests the absence of substantial net chemical transformations and surface‐subsurface water mixing in the shallow sediments of our sub‐reach under baseflow conditions. The significance of this is that upwelling NO3−‐rich groundwater does not appear to be attenuated in the riverbed sediments at <15 cm depth as might occur where hyporheic exchange flows deliver organic matter to the sediments for metabolic processes. It would appear that the chemical patterns observed in the shallow sediments of our sub‐reach are not controlled exclusively by redox processes and/or hyporheic exchange flows. Deeper‐seated groundwater fluxes and hydro‐stratigraphy may be additional important drivers of chemical patterns in the shallow sediments of our study sub‐reach. © 2015 The Authors. Hydrological Processes Published by John Wiley & Sons Ltd.

Diffusive equilibrium in thin films provides evidence of suppression of hyporheic exchange and large-scale nitrate transformation in a groundwater-fed river

INTRODUCTION
Of the major chemical elements necessary to sustain life (nitrogen, carbon, phosphorus, oxygen and sulphur), nitrogen (N) has the greatest total mass (4 × 10 21 g) in Earth's hydrosphere, atmosphere and biosphere (Galloway et al., 2003). Anthropogenic acceleration of global N cycling has almost doubled 'reactive' (all N species except N 2 gas) N levels (Kulkarni et al., 2008) by the spread and fertiliser-based intensification of agriculture, by atmospheric pollution, and now by climate change. In the UK, the widespread application of slurry and fertilisers in the 1960s/1970s has led to extensive contamination of groundwater resources by nitrate (Butcher et al., 2006;Stuart et al., 2007). Recent research suggests that peak nitrate loading is now being recorded in many catchments and that it will take decades for the leached nitrate to discharge into freshwaters due to storage and slow transport times in groundwater aquifers (Wang et al., 2012). This may be particularly problematic in groundwater-fed rivers as predicted warmer and drier summers as a result of climate change (Wilby et al., 2006) will mean groundwater contributions to surface water will become more important, potentially leading to nitrate contamination of surface waters from groundwater. Predicted higher winter rainfall may offset groundwater-sourced nitrate contamination by dilution during short-term storm events. However, extended periods of winter rainfall in groundwater-dominated catchments may enhance groundwater flux and nitrate transport to surface water thereby counteracting any potential improvement in water quality due to dilution. Too much reactive N in surface freshwaters has environmental, economic and human health implications. High levels of nitrite (NO 2 À ) and nitrate (NO 3 À ) can cause algal blooms and oxygen depletion (Burt et al., 2011), and stream acidification in catchments with poorly buffered soils (Curtis et al., 2005), all of which reduce biodiversity (Houghton et al., 2011). The removal of NO 3 À from drinking water costs the UK water industry approximately £58 million per year (DEFRA, 2006). Excess NO 3 À in drinking water can reduce the ability of the blood to carry oxygen (Bryan, 2006) and has been linked to increased risk of cancer (Yang et al., 2007).
In recent years, the potential of riverbed hyporheic sediments to attenuate NO 3 À from polluted groundwater has received much attention Lansdown et al., 2014). The hyporheic zone is typically understood to be the region of groundwater-surface water mixing in riverbed sediments (Valett et al., 1993), although other definitions exist related to the ecological (Hancock et al., 2005;Boulton, 2007) and hydrogeological (White, 1993) significance of this zone. As the interface and zone of mixing of biogeochemically distinct water bodies, the hyporheic zone is often characterised by steep physical and chemical gradients where transport and transformation of chemical species occurs (Hill et al., 1998). Hyporheic exchange flows (HEF) generated by geomorphological features in the riverbed facilitate the dynamic flux of oxygen and organic matter in hyporheic sediments, which can create biogeochemical 'hotspots' in the streambed where NO 3 À attenuation (and release) can occur (Rivett et al., 2008). However, measuring pore water chemistry in the hyporheic zone at scales relevant to N cycling presents a problem for most sampling methodologies. For example, standard multi-level pore water sampling techniques (Rivett et al., 2008) are only effective generally to within 10 cm of the riverbed surface providing very coarse measurements, which relate to a discrete sediment stratigraphy rather than a biogeochemical zone. Pore water sippers (Huettel, 1990) have greater spatial resolution but can modify the subsurface flow field by pumping leading to mixing of pore waters and surface waters (D'Andrea et al., 2002). In comparison, in situ passive sampling technologies such as diffusive equilibrium in thin films (DET) (Davison et al., 1991;Krom et al., 1994;Mortimer et al., 1998) and dialysis peepers (Doig and Liber, 2000) allow for high spatial resolution (down to mm-scale) analysis of important redox-sensitive parameters without modifying subsurface flow fields.
Dialysis peepers can be used to measure at a similar resolution to DET; however, DET gels are preferred over peepers as they equilibrate with solutes at a faster rate permitting shorter deployment times (Fones et al., 2011). Wang et al. (2013) have estimated historical peak nitrate loading from the Penrith Sandstone has arrived in most parts of the River Eden catchment, north-west England. Several areas including parts of the River Leith sub-catchment will experience peak nitrate loading in the next 30-40 years highlighting the need to resolve spatial and temporal patterns in environmental risk associated with groundwater nitrate fluxes. In this study, we focus on a stretch of the River Leith that has been the subject of intensive research as a zone of dynamic ground-surface water exchange and N transformation. Reach-scale (200 m) observations of subsurface vertical and lateral water fluxes, together with riverbed geophysical surveys, have identified a sub-reach (50 m) preferential discharge location characterised by direct connectivity to the underlying sandstone aquifer ) (see Supplementary Material 1). The concentrations of pore water anions (from 100 to 10 cm below the riverbed) in this sub-reach (including NO 3 À ) are elevated with respect to the wider river reach, and show limited evidence of attenuation under baseflow conditions either by physical mixing or chemical transformation Heppell et al., 2013;Ullah et al., 2013;Lansdown et al., 2014) (see Supplementary Material 2). In the absence of hydrological or chemical data for the upper 15 cm of sediments, Munz et al. (2011) demonstrated in modelling experiments that surface water infiltration into the sediments is unlikely to occur under baseflow conditions, thereby suppressing the potential for HEF to attenuate NO 3 À migration to surface water from the underlying aquifer. However, this hypothesis could not be verified with direct measurements under the previous experimental setup which utilised multi-level pore water samplers. The purpose of this research was to target this 'monitoring gap' using the technique of diffusive equilibrium in thin films to capture vertical chemical profiles of NO 3 À and redox-sensitive solutes at high spatial resolution. Our specific objectives were the following: (i) establish detailed vertical and longitudinal patterns of NO 3 À and redox-sensitive solutes in the upper 15 cm of armoured riverbed sediments and (ii) identify if chemical transformation and/or subsurface-surface water mixing acts to attenuate upwelling NO 3 À -rich groundwater.

Study site
This investigation took place in a gaining reach of the River Leith, a tributary of the River Eden, Cumbria, England ( Figure 1a). The River Leith catchment lies 1386 P. BYRNE ET AL.
primarily on Permo-Triassic Sandstone overlain by glacio-fluvial sediments to a depth (in the riverbed) of approximately 50 cm (Kaeser et al., 2009). Previous research as part of a parent project investigating hyporheic exchange and N transformations in the hyporheic zone (NERC Reference: NE/F006063/1) focussed on a 200 m reach set in a narrow floodplain comprising agricultural and pastoral landscape. Surface water nitrate concentrations are typically 2 mg N/L under baseflow conditions . A 50 m subreach (without tributaries) transitioning from a riffle to a pool environment was chosen for more intensive subsurface investigations described here (Figure 1b). This sub-reach has been identified as a zone of locally elevated  . Furthermore, pore water chemical profiles (from 100 to 10 cm depth) have identified locally elevated pore water solutes (6.2-7.4 mg N/L) associated with the zone of enhanced upwelling with no evidence of attenuation by physical mixing Heppell et al., 2013).

Experimental design
The technique of DET (Davison et al., 1991) was used to obtain concentration profiles of redox-sensitive parameters (NO 3 À , SO 4 2À , dissolved Fe and Mn) from the upper 15 cm of the riverbed. DET samplers equilibrate with the pore water chemistry and are used here to provide concentration profiles at cm-scale. DET technology has been successfully applied in a range of 'soft' sediment environments including peatlands, lakes, estuaries and coasts (Davison et al., 1991Zhang and Davison, 1995;Mortimer et al., 1998;Bottrell et al., 2007). Readers are referred to Ullah et al. (2012) for more detailed information on DET gel preparation. A previously described protective stainless steel cover (Ullah et al., 2012) was utilised in this study to allow the deployment of DET in the armoured riverbed sediments. DET sample sites were established along the sub-reach to allow measurement of shallow sediment pore water chemistry along the length of the anomalous hydrological and chemical zone. Three probes were deployed in a riffle environment (site C) at the upstream end of the sub-reach ( Figure 1c). Six further probes were deployed in pool environments (three at site D and three at site E).

Field deployment and retrieval of diffusive equilibrium in thin-films probes
Diffusive equilibrium in thin-films probes were deployed on one occasion in September 2011 towards the end of the summer baseflow period. Deployment of DET probes in the armoured riverbed was achieved by first driving a stainless steel drive point into the sediments to a depth of 15 cm. The DET probe in the stainless steel holder was then inserted next to the drive point, which was then carefully removed allowing the sediment to collapse around the probe. The probes were allowed to equilibrate with the sediment pore waters for 72 h before retrieval to allow equilibrium between the gel and the sediment pore water (Mortimer et al., 1998).
Baseflow conditions were stable in the river over the 72 h period during which time a mean discharge of 0.69 m 3 /s was measured at the Environment Agency Cliburn weir (N54:37:03; W2:38:23), approximately 200 m downstream of the study reach. Once removed, the probes were placed in zip-lock bags and stored on ice to minimise potential diffusion in the gel. Probes were then transferred to the laboratory and the gel slicing and extraction processes started within 2 h of retrieval.

Measurement of vertical flux and multi-level sampler (coarse-scale) pore water chemistry
The DET probes were deployed as close as possible to an existing network of nested channel piezometers installed to allow contiguous measurement of subsurface water flux and coarse-scale pore water chemistry ( Figure 1c). Readers are referred to Binley et al. (2013); Heppell et al. (2013) and Byrne et al. (2013) for more detailed information on the design of the piezometer network and the measurement of subsurface water flux and pore water chemistry. In-channel piezometers were screened at 20, 50 and 100 cm below the riverbed. Vertical hydraulic gradients were measured upon retrieval of the DET probes and calculated by dh/dl, with dh being the elevation difference between local stream and piezometer water level, and dl being the distance between the mid-screen depth of the piezometer and the riverbed surface. Vertical flux was calculated using Darcy's Law as q v = K * dh/dl, where K is the hydraulic conductivity calculated from slug tests in the same piezometers used to compute hydraulic gradients .
Each 100 cm piezometer was fitted with multi-level pore water samplers at 10, 20, 30, 50 and 100 cm depths. Pore water samples were extracted at the time of DET probe retrieval using a syringe and flexible plastic tubing. Sampling lines and syringes were flushed with pore water before collection of samples. A sample of surface water near the piezometers was taken at the same time as pore water samples. All samples were filtered (0.45 μm surfactant-free cellulose acetate membrane) in the field. Samples for anion (NO 3 À and SO 4 2À ) analysis were collected in polycarbonate bottles. Sample bottles were rinsed with pore water three times before filling completely with sample to avoid the presence of air pockets. Samples for metal analysis (Fe and Mn) were collected in pre-rinsed (Milli-Q water) polypropylene tubes and acidified (2%) with concentrated HCl. All samples were transferred on ice to the laboratory within 2 h of collection and analysed within 24 h of collection. One travel blank and two filter blanks were used for all analytes.
Laboratory processing of diffusive equilibrium in thin-films gels and multi-level pore water samples Diffusive equilibrium in thin-films gels were sliced into 1 cm sections using a Teflon-coated razor blade. The DET slices were then transferred to 1.5 ml vials and 1 ml Milli-Q water was added to each vial. The samples were then placed on frozen ice packs and shaken overnight on a reciprocating shaker to allow back equilibration of NO 3 À and SO 4 2À in the gel with the Milli-Q water. After 1388 P. BYRNE ET AL.
shaking, the gels were transferred to another set of 1.5 ml vials for Fe and Mn extraction. Samples for NO 3 À and SO 4 2À analysis (DET and standard pore water samples) were analysed by ion chromatography (DIONEX) within 48 h of retrieval. The limit of detection based on 14 blank samples was 0.07 mg N/L and 0.69 mg S/L. The analytical accuracy of repeat control samples was ± 6.2% for NO 3 À and ± 5.1% for SO 4

2À
. Gel slices for Fe and Mn extraction were eluted with 1 ml of 1M HCl and shaken overnight on a reciprocal shaker. The extracts (and standard pore water samples) were then analysed by ICP-MS (Thermo X series). The limits of detection were 0.09 μg/l Fe and 0.03 μg/l Mn. The analytical accuracy of repeat control samples was ± 3.2% for Fe and ± 3.6% for Mn. Samples from the initial Milli-Q extraction were analysed for Fe and Mn to verify there was no loss of these solutes in this extraction phaseboth Fe and Mn tested below detection limits. The fidelity of DET measurements of solute gradients in pore waters has been examined in detail (Harper et al., 1997). Prior to slicing the gel, solutes will diffuse from high to low concentrations and so narrow maxima (1-2 mm) in concentrations will be slightly under estimated when gels are divided. However, with the relatively coarse slicing resolution of 1 cm, where only 1-2 cm wide maxima can be identified, relaxation effects are greatly diminished, especially as probes were transported on ice.

Statistical data analysis
Analysis of variance (ANOVA) was used to test for significant differences in vertical flux and pore water solutes (DET and standard pore water samplers) between sample locations and depths. The environmental data failed the normality assumptions for parametric analysis (Kolmogorov-Smirnov test) even after transformation; therefore, the non-parametric Kruskal-Wallis one-way analysis of variance test and the Mann-Whitney U-test were used to test for differences between sample locations and depths. All statistical calculations were performed in the program PASW Statistics 18.

Longitudinal and vertical trends in fine-scale pore water chemistry
Solute concentration profiles of shallow sediments (0-15 cm) within the experimental sub-reach are illustrated in Figure 2 and summarised in Table I. Probe three from site C (black circles) illustrates data from 1 to 11 cm only as the DET gel was damaged during retrieval from the sediments. The vertical DET NO 3 À concentration profiles exhibited a degree of variability both within and between sample sites. At site C, NO 3 À profiles were generally invariable with depth (Table I); an exception to this linear trend being probe 3 (black circles), which showed high variability from the aggregated mean (5.7 mg N/L) with two maxima of 9.3 and 8.7 mg N/L at 1 and 8 cm depths, respectively (Figure 2a). Site D showed the most within site variability of the study sites for NO 3 À ; however, this was superimposed on a generally consistent linear trend for individual probes (Figure 2b). Mean NO 3 À at site D was 5 mg N/L with two maxima of 7.3 and 7.4 mg N/L at 10 and 15 cm depths, respectively. Site E exhibited the least within site variability in NO 3 À concentrations with mean concentrations of 2.6 mg N/L. Concentrations of SO 4 2À in the sediment pore waters were higher than NO 3 À and generally exhibited more within site variability than NO 3 À . Sites C and D exhibited high variability in SO 4 2À concentrations over the 15 cm probe depth, ranging from 4.7 to 17.9 mg S/L and 3.1 to 20.8 mg S/L, respectively. Site E showed the least within site variability in NO 3 À and SO 4 2À concentrations (Figure 2c) with mean concentrations of 2.6 mg N/L and 9.17 mg S/L. Mean site values indicate a trend of decreasing NO 3 À and SO 4 2À concentrations in the shallow sediments from sites C to E ( Table I). The distinctiveness of NO 3 À concentrations between sites is confirmed by Mann-Whitney U-tests, which show that concentrations are significantly different (p ≤ 0.01) between sample sites. Although mean SO 4 2À concentrations decrease from sites C to E (Table I), the within site variability of the SO 4 2À data results in a significant difference (p ≤ 0.05) only between sites C and E.
Mean dissolved Fe and Mn concentrations at site C were 1.86 and 0.05 mg/l, respectively. The vertical concentration profiles of Fe and Mn at site C were generally invariable with the exception of probe 3 (black circles) (Figure 2a), which exhibited high dissolved Fe (11.3 mg/l) and Mn (0.18 mg/l) concentrations at the sediment-water interface. Two smaller Fe and Mn maxima occurred at 7 and 10 cm depths, respectively. Dissolved Fe and Mn at sites D and E showed a generally similar vertical trend (Figure 2b and c) although mean concentrations at site E (2.68 mg/l Fe and 0.53 mg/l Mn) were higher than at site D (1.09 mg/l Fe and 0.11 mg/l Mn) ( Table I). The vertical profile was almost parabolic in shape with elevated concentrations in the upper 1-5 cm, a steep decrease to relatively consistent concentrations from 6 to 10 cm and an increase in concentrations from 11 to 15 cm. For both dissolved metals and at both sites, the maximum concentration occurred at 15 cm depth. The highest dissolved Fe (12.18 mg/l) and Mn (1.99 mg/l) concentrations across all three sample sites were recorded at 15 cm depth at site E (probe 1, open circles). A spatial trend contrary to that of NO 3 À and SO 4 2À was observed for dissolved Fe and Mn. Generally, mean dissolved Fe and Mn concentrations increased from sites C to E ( Table I); the exception being Fe at site C, which was influenced by the high variability in probe 3 (black circles). Mann-Whitney U-tests of significance show dissolved Mn was significantly different (p ≤ 0.01) between samples sites. Dissolved Fe was generally significantly different (p ≤ 0.01) between sample sites (except between sites C and D).

Longitudinal and vertical trends in vertical flux and coarse-scale (multi-level sampler) pore water chemistry
Vertical flux calculations derived from saturated hydraulic conductivity measurements and vertical hydraulic gradient measurements taken at the same time as DET probe retrieval are illustrated in Figure 3 alongside summary data for baseflow periods (July to September) collected as part of the parent project in 2009 and 2010. These data demonstrate a longitudinal trend in vertical flux measured at 100 cm; flux decreases in the downstream direction from sites C to E. Although no significant difference in vertical flux was observed between samples sites for aggregated depth data, flux at site C was significantly greater ( p ≤ 0.05) than flux at site E. A significant ( p ≤ 0.05) increase in vertical flux occurred in the sediments above 100 cm (20 and 50 cm) at sites C and E whereas the vertical flux profile was more uniform with depth at site D.
Measurements of surface water and pore water solutes obtained from the multi-level samplers and at the same time as DET probe retrieval are also presented in Figure 3 alongside summary data for baseflow periods (July to September) in 2009 and 2010. Table II presents a summary of observed coarse-scale pore water NO 3 À and SO 4 2À concentrations and a comparison with DET-derived values. Pore water NO 3 À concentrations were significantly greater ( p ≤ 0.01) than surface water concentrations at all three sample sites; the reverse relationship ( p ≤ 0.01) was observed for the SO 4 2À data. Pore water NO 3 À exhibited a significant ( p ≤ 0.01) decrease in concentration from sites C to E. At sites C and E, the NO 3 À concentration profile was relatively stable between 10 and 100 cm depths; whereas, at site D, pore water NO 3 À gradually decreased in concentration towards the riverbed surface; although values were similar at 10 and 20 cm. Pore water SO 4 2À concentrations were not significantly different between sample depths at individual sites and also between sites.

Vertical concentration profiles of NO 3 À and redox-sensitive parameters
Nitrogen transformation in the hyporheic zone has been shown to be controlled by infiltrating surface water that delivers oxygen and organic matter for metabolic processes; typically NO 3 À production occurs where oxygenated surface water enters the riverbed and NO 3 À consumption occurs farther along the flow path when oxygen has been respired (Zarnetske et al., 2011). In the present study, most of the DET probes showed variability in solute concentrations with depth; however, this variability did not conform to the archetypal biogeochemical zones that represent the degradation of organic matter using successively less energy-efficient terminal electron accepting processes. Instead, the DET probes exhibited cm-scale changes in solute concentrations which suggests the absence of largescale attenuation of NO 3 À from upwelling groundwater in our sub-reach. Extensive investigations of 'soft' riverine, lacustrine and marine sediments using DET technology have revealed similar, highly localised micro-scale changes in solute concentrations (Fones et al., 1998;Docekalova et al., 2002;Mortimer et al., 2002).
The NO 3 À concentration profiles from the DET probes suggest surface water infiltration into the riverbed sediments was not significant during the study period. If surface water infiltration is significant, we would expect to see strong vertical concentration gradients due to the difference in groundwater and surface water NO 3 À concentrations. In addition, several studies utilising conservative tracers Byrne et al., 2013;Heppell et al., 2013) and numerical groundwater modelling (Munz et al., 2011) have demonstrated groundwater-surface water mixing is unlikely to be occurring below 5 cm sediment depth in our sub-reach, at least under baseflow conditions. This suggests that the observed cm-scale variability in nitrate concentrations occurs at depths greater than surface water infiltration. Krause et al. (2009) observed similar variability in nitrate concentrations in deeper sediments (20-40 cm) of the same sub-reach. This suggests that nitrogen transformation in this sub-reach may be temporally removed from surface water infiltration under baseflow conditions. Instead, nitrogen transformation may be related to the temporary expansion of the hyporheic zone during high flow events and to the delivery of organic matter and oxygen to deeper sediments during these times . Nitrate concentrations increased above 5 cm sediment depth in probe 3 (black circles) at site C. This could be  (Ullah et al., 2012, and subsequent nitrification (Mortimer et al., 2002). Of the three sample sites investigated, site C (riffle) is the most likely to experience HEF as a result of pumping exchange (Boano et al., 2008) and/or changes in the hydrostatic head gradient across the riffle (Wondzell et al., 2009). If the upper 5 cm is indeed an oxygenated zone, it is curious then why we see a concomitant increase, rather than decrease, in the concentration of the dissolved metals. It may be that elevated Fe and Mn at the same location may not simply represent reduced forms, but may be more reflective of oxidised colloidal complexes that are able to pass through the membrane filter and gel (Ahmed et al., 2010). However, we believe that the true cause of this artefact is disturbance to the gel during deployment (the gel strip was squashed meaning division into discrete cm bands was difficult), which may have affected the volume and hence final concentration of the sliced gel strips. This artefact aside, it is clear from the low concentration gradients from the other DET probes that no large-scale changes (losses or gains) of NO 3 À are occurring in the shallow sediments, ensuring that elevated groundwater NO 3 À is transported to the surface water. An interesting and consistent trend in vertical chemical profiles occurred for Fe and Mn at sites D and E. The vertical profile at these sites was parabolic in shape with elevated concentrations from 1-5 and 10-15 cm. In this pool section of the sub-reach where HEF is less likely to occur , elevated Fe and Mn in the upper 5 cm may be related to the accumulation of leaf litter and other organic matter. At the time of the DET experiments in mid-September, the river was still in baseflow conditions; however, organic matter had begun to accumulate in the pool (sites D and E) because of the dieback of extensive channel margin Bur-reeds (Sparganium spp.) and leaf fall from bank-side trees. The surface peak in Fe and Mn could have resulted from the metals being major electron acceptors for the decomposition of this organic matter (Davison et al., 1991;Zhang et al., 1999). Leaf litter accumulation on the streambed can also modify channel hydraulic properties (Argerich et al., 2011) leading to increases in water residence times, which extend contact times between solutes and stream microbial communities (Haggard and Storm, 2003;Argerich et al., 2008). Calculations of water residence times at 20 cm depth based on Darcy-derived vertical flux do indeed suggest higher water residence times at sites D and E. The combination of the increase in water residence times and input of organic matter could create conditions favourable for anaerobic microbial metabolism resulting in Fe and Mn reduction (Fones et al., 1998). It is interesting then why we do not see evidence for denitrification in the upper sediments, given bacterial NO 3 À reduction is energetically more favourable than Fe or Mn reduction. We suggest that denitrification is probably occurring but that the resultant changes in nitrate concentration are of the order of ng/l or μg/l; these are difficult to observe when the background NO 3 À concentrations are mg/l. In contrast, the Fe and Mn background concentrations are generally ng/l or μg/l so it is possible to detect smaller concentration changes resulting from bacterial activity. Of course, reduction of all three electron acceptors (NO 3 À , Fe and Mn) could appear to occur together, especially in heterogeneous systems such as riverbed sediments where micro-niche activity is likely to be important (Stockdale et al., 2009).
The largest observed increases in Fe and Mn concentrations occurred at 10-15 cm at sites D and E. Others have noted the occurrence of high dissolved Mn concentrations at depths of 10-20 cm in the River Leith (Ullah et al., 2012;Byrne et al., 2013). Iron and Mn peaks at similar depths have been observed also in lake and estuarine sediments (Fones et al., 1998;Shuttleworth et al., 1999;Docekalova et al., 2002). Byrne et al. (2013) attributed these high Mn bands in the River Leith sediments to Mn reduction enhanced by the supply of DOC by down welling surface water at high river stage. Lateral flows from the riparian zone were also postulated to be an important source of DOC in deeper (20-50 cm) sediments Heppell et al., 2013). Lateral flux at 20 cm depth within our sub-reach is greatest around site E (see Supplementary Material 1). Therefore, the lateral movement of carbon-rich water originating from the riparian zone may be an important driver of biological reduction in the shallow sediments at site E.

The interaction of hydro-stratigraphy and redox-sensitive chemistry
The hyporheic zone of riverbed sediments has been proposed as a possible medium for the removal of NO 3 À À from upwelling groundwater. However, the present research provides evidence that NO 3 À attenuation in the upper sediments may be limited or non-existent where strong groundwater upwelling retards the development of a hyporheic zone (defined here as a region of groundwater-surface water mixing). This raises the question of the possible role of hydrogeology and hydro-stratigraphy in controlling the chemical dynamics and concentration profiles of pore water solutes including NO 3 À . In this study, we have identified from DET samplers (1 cm resolution) and multi-level pore water samplers ( 10-50 cm resolution) a longitudinal trend of generally decreasing pore water anion (NO 3 À and SO 4 2À ) concentrations in the downstream direction. Previous waterborne geophysical surveys identified a local high of pore water electrical conductivity persisting to 5 m depth centred on the upstream section (site C) of our sub-reach (Clifford and Binley, 2010;Binley et al., 2013). This high electrical conductivity has been associated with elevated concentration of pore water solutes (including NO 3 À ) to a minimum depth of 100 cm. Moreover, pore water dissolved organic carbon was lower and dissolved oxygen was higher at site C compared with a vegetated sediment stretch (~8 m upstream of site C), suggesting limitation of nitrate reduction potential by upwelling water at site C . Putting this evidence together, it has been suggested that this sub-reach is a zone of preferential flow of NO 3 À -rich groundwater Heppell et al., 2013). The groundwater could be sourced from a fracture in the underlying bedrock, which effectively acts as a concentrated point source of NO 3 À , where groundwater flow from the Permo-Triassic Sandstone is typically diffuse in nature (Wang et al., 2012).
The movement of NO 3 À -rich groundwater is undoubtedly aided by the local hydro-stratigraphy and its influence on subsurface flux and water residence times. The same longitudinal trend of decreasing (downstream) anion concentrations is evident for vertical flux at 100 cm depth; although this spatial pattern is modified at shallower depths (20 and 50 cm) most likely by the influence of lateral subsurface fluxes from the riparian zone . Elevated flux (at 100 cm) might be aided by local stratigraphy. Particle size analysis of the riverbed sediments reveals site C to have a significant fraction of pebbles, gravel and coarse sand (40-90%) persisting to at least 70 cm below the riverbed . At sites D and E, this coarse zone terminates at 30-40 cm; thereafter, the sediment is composed mainly of medium and fine sand (35-90%). Hydraulic conductivity measurements reflect the particle size data with hydraulic conductivities decreasing from sites C to E . Together, these data perhaps explain the longitudinal trend in vertical flux at 100 cm. Calculations of water residence times based on Darcy-derived vertical flux show water residence time at all depths (100, 50 and 20 cm) to decrease in the downstream direction from sites C to E (assuming a predominantly vertical flow path). Water residence times at site E are significantly higher than at the other sites at all sample depths. Water residence time is known to significantly affect the rate of transformation of NO 3 À with higher water residence times corresponding to increased rates of denitrification (Pinay et al., 2009;Zarnetske et al., 2011).

A methodological note
As this paper has used both DET and multi-level pore water samplers to present and interpret biogeochemical patterns in riverbed sediments, a discussion of the merits of the two techniques is perhaps warranted. Comparing DET measurements (NO 3 À and SO 4 2À ) over 15 cm with multi-level sampler data at 10 cm shows that the two measurements were not always consistent with each other (Table II). The largest discrepancy occurred for SO 4 2À whereas measurements of NO 3 À were more comparable (except site D). Ullah et al. (2012) also observed comparable NO 3 À concentrations from DET and pore water samplers. The DET SO 4 2À measurements were approximately 50% greater than measurements obtained from the multi-level samplers.
The observed discrepancy in solute concentrations is most likely due to the contrasting scale and method of sample collection between the two techniques. Multilevel pore water samplers operate at a minimum resolution of 5 cm and effectively sample a volume of pore water, modifying the flow field and possibly integrating oxic and anoxic micro sites in the sediments (Kalbus et al., 2006;Krause et al., 2009). In the present study, multi-level samplers missed the considerable variation in solute concentrations captured by DET samplers in the upper sediments, the result being the loss of important information on the dynamics of chemical interactions occurring at cm-scale. DET can measure solutes at sub-cm resolution providing accurate information on nutrient dynamics and biogeochemical activities in river bed sediments Mortimer et al., 2002). Simple interpretation of DET measurements at high resolution from shallow sediments as a continuation of data from deeper multi-level pore water samplers may not be feasible because of the different scale and mechanism of sampling. Multi-level pore water samplers are an invaluable method for providing broad-scale assessment of subsurface chemistry. However, DET is especially useful when chemical data are needed from river sediments, which are highly heterogeneous in nature and where dynamic subsurface fluxes can exert a significant influence on pore water chemistry. 1394 P. BYRNE ET AL.

CONCLUSIONS
Although DET technology has been widely applied in 'soft' lake, marine, estuarine, freshwater and coastal sediments, the utilisation of a novel delivery mechanism in this study for armoured riverbed environments permitted investigation of nitrogen dynamics in a coarse sediment lotic environment. By analysing vertical concentration profiles captured by the DET samplers, our study has revealed cm-scale changes in the concentration of redox-sensitive solutes at depths below surface water infiltration and demonstrated the important control of subsurface water flux on nitrogen biogeochemistry. Our deployment of the DET samplers in the upper 15 cm of river sediments filled a pre-existing 'monitoring gap' and provided strong evidence for the absence of surfacesubsurface water mixing in our study reach under baseflow conditions. The significance of this is that strongly upwelling NO 3 À -rich groundwater is not attenuated in the river sediments as might occur where hyporheic exchange flows deliver organic matter to the sediments for metabolic processes. It would appear that the general biogeochemical patterns observed in the shallow sediments of our sub-reach are not controlled exclusively by redox processes and/or hyporheic exchange flows. Deeper-seated groundwater fluxes and hydro-stratigraphy may be additional important drivers of biogeochemical patterns in our study reach. Whilst we have only investigated three sites in a 50 m stretch of river, our results suggest that chemical attenuation of groundwater-sourced NO 3 À in gaining river systems may be affected by the relative magnitude of subsurface water flux. This is important in the context of predicted increases in the concentration of NO 3 À in the Permo-Triassic Sandstones of north-west England and predicted warmer, drier summers as a result of climate change.