Spatial and temporal dynamics of nitrogen exchange in an upwelling reach of a groundwater‐fed river and potential response to perturbations changing rainfall patterns under UK climate change scenarios

We report the complex spatial and temporal dynamics of hyporheic exchange flows (HEFs) and nitrogen exchange in an upwelling reach of a 200 m groundwater‐fed river. We show how research combining hydrological measurement, geophysics and isotopes, together with nutrient speciation techniques provides insight on nitrogen pathways and transformations that could not have been captured otherwise, including a zone of vertical preferential discharge of nitrate from deeper groundwater, and a zone of rapid denitrification linking the floodplain with the riverbed. Nitrate attenuation in the reach is dominated by denitrification but is spatially highly variable. This variability is driven by groundwater flow pathways and landscape setting, which influences hyporheic flow, residence time and nitrate removal. We observed the spatial connectivity of the river to the riparian zone is important because zones of horizontal preferential discharge supply organic matter from the floodplain and create anoxic riverbed conditions with overlapping zones of nitrification potential and denitrification activity that peaked 10–20 cm below the riverbed. Our data also show that temporal variability in water pathways in the reach is driven by changes in stage of the order of tens of centimetres and by strength of water flux, which may influence the depth of delivery of dissolved organic carbon. The temporal variability is sensitive to changes to river flows under UK climate projections that anticipate a 14%–15% increase in regional median winter rainfall and a 14%–19% reduction in summer rainfall. Superimposed on seasonal projections is more intensive storm activity that will likely lead to a more dynamic and inherently complex (hydrologically and biogeochemically) hyporheic zone. We recorded direct evidence of suppression of upwelling groundwater (flow reversal) during rainfall events. Such flow reversal may fuel riverbed sediments whereby delivery of organic carbon to depth, and higher denitrification rates in HEFs might act in concert to make nitrate removal in the riverbed more efficient.

the reach is dominated by denitrification but is spatially highly variable. This variability is driven by groundwater flow pathways and landscape setting, which influences hyporheic flow, residence time and nitrate removal. We observed the spatial connectivity of the river to the riparian zone is important because zones of horizontal preferential discharge supply organic matter from the floodplain and create anoxic riverbed conditions with overlapping zones of nitrification potential and denitrification activity that peaked 10-20 cm below the riverbed. Our data also show that temporal variability in water pathways in the reach is driven by changes in stage of the order of tens of centimetres and by strength of water flux, which may influence the depth of delivery of dissolved organic carbon. The temporal variability is sensitive to changes to river flows under UK climate projections that anticipate a 14%-15% increase in regional median winter rainfall and a 14%-19% reduction in summer rainfall. Superimposed on seasonal projections is more intensive storm activity that will likely lead to a more dynamic and inherently complex (hydrologically and biogeochemically) hyporheic zone. We recorded direct evidence of suppression of upwelling groundwater (flow reversal) during rainfall events. Such flow reversal may fuel riverbed sediments whereby delivery of organic carbon to depth, and higher denitrification rates in HEFs might act in concert to make nitrate removal in the riverbed more efficient.  Galloway et al., 2004) to biologically unavailable N 2 gas and return it to the atmosphere (Bernhardt et al., 2005;Mulholland et al., 2008;Zhao et al., 2015). This N sink capacity includes processes occurring in the hyporheic zone of groundwater catchments (e.g., Burns et al., 2019;Krause et al., 2015;Schlesinger, 2009;Stelzer et al., 2020;Trimmer et al., 2012) where a mosaic of redox conditions is supported . The permeability of groundwaterfed riverbeds enables the advection and supply of reactants (such as organic matter and nitrate) to the microbial communities which drive nitrogen processing (Lansdown et al., , 2016. Evidence suggests the N sink capacity of rivers is compromised by the net accumulation of N in agricultural subsoils (e.g., Van Meter et al., 2016;Worrall et al., 2015) that moves slowly through the vadose zone of groundwater systems (Ascott et al., 2017;Cuthbert et al., 2019;Stuart et al., 2011). Land management policies have sought to improve freshwater quality by reducing nitrate loading in rivers (e.g., Kanter et al., 2020;Sinha et al., 2019) but the timeframes involved may be too slow to offset other stressors on freshwater ecology (Birk et al., 2020) such as climate-driven water temperature increases (e.g., F. L. Jackson et al., 2020;Leach & Moore, 2019;Ouellet et al., 2020), as proposed by Vaughan and Gotelli (2019) in the context of offsetting climate debt.
Nitrate loading must also be understood in the context of projected changes in river flows under a changing climate, which, for the UK is similar or increased average winter river flows and reduced average summer river flows with increased storm activity (UKCP18, www. metoffice.gov.uk). While the effects of climate change impacts on river flows across the range of geologies found in UK aquifers are yet to be observed (Garner et al., 2017;Hannaford, 2015;C. R. Jackson et al., 2015;C. Murphy et al., 2019;Prudhomme et al., 2013), the chalk aquifer of the south east UK already shows evidence of an increased frequency of groundwater drought due to elevated evapotranspiration (Bloomfield et al., 2019); and intense summer storm activity, predicted to increase under UK climate change scenarios, has been shown to impact stream ecology (e.g., Hutchins et al., 2020;Woodward et al., 2015). Oscillatory climate system drivers such as the North Atlantic Oscillation (NAO) (e.g., Kuss & Gurdak, 2014) are also important. Recent work by Rust et al. (2020) shows the NAO can be statistically detected with a 7-year periodicity in UK river flow. These climate drivers have the potential to impact on regional rainfall distributions, water resource and nutrient yields and will influence groundwater-surface water interactions including hyporheic exchange flows (HEFs) (e.g., Azizian et al., 2017;Singh et al., 2019). All these physico-chemical processes have consequence for river nitrogen sink capacity (e.g., Stelzer et al., 2020) with which this paper is concerned.
Early hydrological research on the hyporheic zone (e.g., Bencala et al., 1990;Fuller & Harvey, 2000;Haggerty et al., 2002;Hill et al., 1998;Malcolm et al., 2003) focused mainly on the relationship between river water and the upper few centimetres of the sediments of the riverbed. Work by Stelzer (e.g., Stelzer et al., 2011;Stelzer & Scott, 2018) among others, revealed the importance of nitrate processing in deeper sediments. As well as downward flux from the river into the sediments of the riverbed, upward flows from groundwater through the hyporheic zone and into the river are important in understanding the evolving chemistry of groundwater as it moves through the hyporheic zone (e.g., Brunke & Gonser, 1997;Conant, 2004), and in particular the capacity for N attenuation under baseflow conditions. This paper synthesizes the physical hydrology and biogeochemistry process-based understanding for a lowland groundwater-fed river, representative of systems where the river is continuously recharged by groundwater throughout the year. The synthesis includes new data to extend understanding of the relationship between the river channel and its riparian zone under different flow conditions. Through synthesis and new data, the broader goal is to explore potential changes to the N dynamics of groundwater-fed river systems due to external perturbations arising from changing rainfall patterns and temperatures predicted under UK climate change scenarios. To support this goal, additional research is drawn from Kay et al. (2020)  Alterations to groundwater flows and groundwater quality arising from the climate change are likely to impact on river ecology. For example in rivers associated with chalk aquifers there is evidence that low flows may result in decline of vegetation, such as Ranunculus pseudofluitans, associated with priority habitats (Westwood et al., 2020). We also know that warming may impact biogeochemistry by changing the way rivers couple and transform carbon and nutrients (Hood et al., 2017;Preiner et al., 2020), and work by Kurylyk et al. (2014) and Kurylyk et al. (2015) suggests that although groundwater-fed rivers are likely to remain buffered to temperature changes in the regional groundwater, this buffering capacity may decline over time (Leach & Moore, 2019). Similar "buffering" or chemostatic behaviour such as that observed in the Eden catchment may mask aquifer response to climate drivers where large mass stores (i.e., parent material and legacy nutrient stores) result in fairly constant intra-annual nitrate concentrations despite large variations in river flow (Butcher et al., 2008).

| DRAWING ON DETAILED INSIGHTS FROM A STUDY REACH TO ILLUSTRATE THE SPATIAL COMPLEXITY OF PROCESSES INVOLVED IN SURFACE-SUBSURFACE WATER INTERACTION AND NITRATE EXCHANGE
The study area (Figure 1), a 200 m groundwater-fed river reach, is part of the River Eden catchment, Cumbria UK, and comprises a loose gravel alluvium overlying unconsolidated (Permo-Triassic) sandstone with a sequence of pools and riffles. The reach has limited drift cover thus providing direct contact with the sandstone and regional aquifer flow. Our conceptual framework was akin to that proposed by Stelzer and Bartsch (2012), in that it was assumed that nitrate would be removed by denitrification from nitrate-rich, oxic groundwater as it upwelled into near-surface sediments containing buried particulate organic matter. Further, discrete patches of net nitrification and denitrification were anticipated, driven by dissolved organic carbon (DOC) supplied to the riverbed via HEFs (Zarnetske et al., 2011). To capture the effect of small and larger spatial scale processes, detailed in-channel process-based understanding was gained by combining non-standard measurement techniques including a patch-scale isotope tracer "push-pull" (centimetre scale) technique Lansdown et al., 2016), centimetre scale riverbed nutrient profiles Ullah et al., 2014), and decimetre to metre resolution advanced geophysical applications Clifford & Binley, 2010;McLachlan et al., 2017). New data building on work by Dudley-Southern and  and Käser, Binley et al. (2014) extends the analysis beyond the channel and over time to refine our conceptual understanding of ecosystem control points (Bernhardt et al., 2017) or "hotspots" and "hot moments" within a broader framing of hyporheic zone processes for N dynamics under climate-driven changing river flows.
Steady state modelling of conservative solute transport undertaken by Käser et al. (2013) proposed that HEFs can be created by both macroforms (>1 m length e.g., due to riffles and emergent veg- the River Leith study reach shows that denitrification was the dominant nitrate attenuation process in the riverbed, with dissimilatory nitrate reduction to ammonium (DNRA) of secondary importance, and no evidence of anammox. Denitrification was observed to occur throughout the riverbed, even under predominantly oxic conditions. According to Lansdown et al. (2014) nitrification occurred across a far narrower chemical gradient in the lower river reach and was inhibited by the low oxygen conditions associated with horizontal flows. The spatial pattern of nitrate removal can be described according to the dominant flow pathway, with three zones identified in Figure 3b.
The majority (>80%) of nitrate removal (via denitrification) occurs in the gaining reach (Z1) within sediments not exposed to HEFs under baseflow conditions. A second zone (Z2), comprising c. 20% of the reach area, is characterized by preferential groundwater discharge (PGD) and dominated by vertical upwelling that enabled nitrate from regional groundwater to be rapidly transported to surface water with little opportunity for denitrification. Data reported by Heppell et al. (2014) suggest this zone of PGD contributes 4%-9% of the total surface water nitrate load on an instantaneous basis and c. 2% of total denitrification within the riverbed. A third, constrained zone (Z3) occupies around 2.5% of the reach by area, where anoxic lateral subsurface flow derived from the riparian zone generates a hotspot, contributing 8% of total denitrification within the riverbed .
Damköhler numbers have been proposed as a useful means of examining the capacity of different landscape units to remove nitrate F I G U R E 2 Modelled illustration of representative flow patterns arising from (a) macroforms and (b) microforms for the R. Leith riverbed geomorphology, modified after Hu et al. (2014). The red lines show hyporheic exchange flows (HEFs) and the blue lines show upwelling groundwater flow paths. River flow is left to right (Gu et al., 2007;Pinay et al., 2015). Work reported by Lansdown et al. (2015) applied Damköhler analysis to the riverbed sediment as groundwater moved from 100 cm depth to the surface. Here the same method is used to report dimensionless Damköhler numbers for denitrification as where D a,den is the dimensionless Damköhler number, τ T is the transport timescale (d) and τ R is the reaction timescale (d). Figure 3b applies Damköhler numbers to the three water flux zones described above.
The majority of the experimental reach (Z1 + Z2) was characterized by a Damköhler number for denitrification, D a,den <1, indicating that nitrate removal in upwelling groundwater was limited by slow reaction rates compared to advection. The data suggest that water flux was too fast relative to the denitrification rate for complete nitrate removal from groundwater to occur. Hotspots of denitrification that had a D a,den >1 were associated with horizontal (lateral and longitudinal) water fluxes such as those found in the lower reach (Z3). The importance of near-surface lateral flow pathways in transferring nutrients from land to water in agricultural systems has been reported earlier by Heathwaite and Dils (2000) observed for the majority of the riverbed, including the zone of preferential discharge (Z2) described above. Here, organic matter supplied by surface water in the top 10 cm of the sediment is being mineralised to NH 4 + (see Figure 4a), with highest denitrification and nitrification rates recorded in the uppermost part of the riverbed within the hyporheic zone (see Figure 4c). The second conceptualisation shows a hotspot where bioavailable organic carbon is being delivered to the riverbed from riparian areas (see Figure 4b).
Here the zones of maximum denitrification activity and nitrification potential occur deeper in the riverbed due to the lateral flows that supply dissolved organic matter and ammonium due to mineralisation (see Figure 4d). Thus, ammonium concentrations are elevated at depth, and not just within the top 10 cm of the sediment. Maximum denitrification and nitrification rates are generally thought to occur under contrasting oxygen settings, with denitrification optimal in reduced conditions, and nitrification in oxidized conditions (Seitzinger et al., 2006). Our measurements indicated that zones of high nitrification potential and denitrification activity overlap in net oxic sediments without the discrete patches that we had originally anticipated . Although denitrification rates increased towards the sed-

| ECOSYSTEM CONTROL POINTS IN GROUNDWATER-FED RIVERS
The synthesis of results for the river reach described above show that the nitrate sink capacity depends on the relative balance between transport versus reaction controls (in this case for denitrification).
Where water moves faster than nitrate can be reduced, nitrate is controlled primarily by hydrological processes such as mixing and dilution. Where nitrate is reduced faster than water transport, denitrifica- accounting of nutrient fluxes in a river reach to contextualize the relative importance of "hot spots and hot moments" in contributing to, for example, nitrogen exchange. This spatial complexity is described above for the different zones of nitrogen removal (e.g., Figure 3b). (see e.g., Kaandorp et al., 2019;Orr et al., 2015) and will depend on a combination of landscape structure, groundwater flow path length, type (e.g., fracture or inter-granular), and depth and residence time (Briggs & Hare, 2018;Tetzlaff et al., 2009). Over a timeframe of decades, future summer low flow conditions derived from preferential discharges from deep aquifers may manifest as areas of markedly lower temperature relative to the surrounding riverbed offering thermal refugia for fish (Geist et al., 2002;Kurylyk et al., 2014). Over the same timeframe, the past legacy of high N-fertilizer usage (see e.g., Ascott et al., 2017) is likely to continue to contribute significant flux of nitrate to the riverbed with little opportunity for nitrate removal by microbial processes. However, it is also possible that preferential discharge from shallow aquifer environments may warm faster in response to air temperature changes (Eggleston & McCoy, 2014) and nitrate concentrations may decline more rapidly in response to diffuse nitrate management strategies (e.g., introduction of Nitrate Vulnerable Zones, which currently cover 55% land in England) than their deeper aquifer counterparts characterized by longer lag times.
For Permanent Control Points, data for the study reach previously reported by Heppell et al. (2014) found that lateral connectivity con- interventions. By integrating in-river/below riverbed measurements with "out of channel" observations in riparian piezometric networks, it is possible to extend understanding beyond the river channel to evaluate the role of "permanent control points" and to examine integrated spatial controls on system biogeochemistry. Such integration enables exploratory analysis of changing external drivers such as climate change on groundwater-fed systems to be explored.

| THE IMPORTANCE OF HYPORHEIC EXCHANGE FLOWS FOR NITROGEN CYCLING UNDER A CHANGING CLIMATE
The UK weather and climate is highly variable. Observed annual mean rainfall over England and Wales has not changed significantly since records began in 1766 (Jenkins et al., 2008), although the most recent decade (2010-2019) is wetter (>3% 1981-2010; >7% 1961-1990) for the UK overall (Kendon et al., 2020). The seasonal pattern of observed rainfall has changed. The proportion of winter rainfall falling F I G U R E 6 Evidence of out-of-channel behaviour. (a) Soil electrical conductivity survey results in the adjacent floodplain showing areas of relatively high electrical conductivity associated with fine textured moisture retentive soils, and low electrical conductivity associated with elevated (drier) ground. The circular symbols show the piezometric network. (b) Floodplain topography analysis carried out by Käser, Graf et al. (2014) in heavy rainfall events has increased over all regions of the UK over the past 45 years (Jones et al., 2013;Sanderson, 2010). Observations suggest a long-term trend towards decreasing mean summer rainfall in all regions of the UK but the relative contribution of heavy events to total summer rainfall has increased (Jones et al., 2013). A pattern corroborated by Burt and Ferranti (2012) for northwest England, where the study catchment is located. Changes in rainfall patterns, including higher rainfall volumes and intensities have been projected by climate models for some time (Fischer & Knutti, 2016). There is a growing consensus that extreme daily rainfall rates and rainfall events are becoming more intense (Slingo et al., 2014) and the effect of climate change makes events like those recorded in northwest England in December 2015 -"storm Desmond" that caused an estimated £500 M of damage, about 40% more likely (Otto et al., 2018). For the river Eden catchment, climate projections indicate further pronounced seasonal changes in future rainfall, with a 14%-15% increase in median winter rainfall predicted by UKCP09-WG for the 2050s high emissions scenario, and a 14%-19% reduction in summer rainfall (Ockenden et al., 2017). Prolonged low summer flows are projected to be interspersed with convective high-intensity rainfall events (see Ockenden et al., 2016). Groundwater flux is driven by many factors including surface water discharge and groundwater recharge. Given the climate projections described above, we anticipate a change in the magnitude of these variables that will likely vary between summer and winter, potentially changing the pattern of groundwater flux and consequently nitrogen dynamics within the riverbed. Simulation modelling of the implications of a changing climate for annual surface water nitrogen flux has been undertaken for the river Eden catchment (Ockenden et al., 2017) using a high-resolution (1.5 km grid) regional climate model (RCM-1.5 km) for the UK and from the UK Climate Projections 2009 Weather Generator (UKCP09-WG) combined with high-frequency surface water quality data from the River Eden Demonstration Test Catchment (www.edendtc.org.uk). Modelled predictions estimated a 71% probability of future surface water nitrate flux increasing by 2050 owing to elevated rainfall-driven nitrogen transfers from agricultural land (Ockenden et al., 2016) unless mitigation measures are put in place. The impact on surface water nitrogen may be compounded by legacy nitrogen stores in groundwater (Briggs & Hare, 2018;Butcher et al., 2008) as wetter winters lead to increased groundwater recharge rates andoutside major storm events (see below) -lead to greater groundwater upwelling. The research synthesized here along with that of others, has found that upwelling groundwater reduces hyporheic exchange Gomez-Velez et al., 2014), and a recent modelling study has shown that strongly upwelling groundwater also reduces nitrate processing within HEFs (Azizian et al., 2017). The stronger the vertical flux of groundwater the less nitrate is removed along the upwelling groundwater pathway in deeper riverbed sediments and within the overlying HEFs near the sediment surface (Azizian et al., 2017). Conversely, dryer summers may weaken upwelling groundwater fluxes. Bloomfield et al. (2019) show evidence of an increased frequency and magnitude and duration of groundwater droughts (defined as mean periods of below-normal illustrate that climate warming could lead to large and difficult-topredict changes in river metabolism and its coupling to nutrient cycles. The authors suggest that responses to warming will emerge from interactions between population-, community-, and ecosystem-scale properties that presently cannot be predicted from theory. For all scenarios, the patterns and mode of delivery of nutrients and fine sediments from the catchment to the riverbed, as well as physical disturbance effects from transient high flows, will alter the biofilm microbiome in the hyporheic, which is critical to biogeochemical functioning (Battin et al., 2016). Although effort is now being made to improve models that couple the effects of groundwater and HEFs on nitrogen cycling under steady and dynamic discharge conditions, we have yet to develop predictive models that incorporate these types of hydroecological responses to climate change. Such modelling advances are also dependent on approaches to enable the extrapolation from detailed process-based understanding.
The experimental work synthesized here has focussed on detailed understanding of processes at centimetre to metre scales within in a impractical to extend such a spatially detailed experimental program to characterize an entire river reach and, therefore, we need to consider ways of scaling such investigations. Such a challenge is not new to hydrological science. We believe that an appropriate way forward is to use a combination of large-scale modelling tools alongside reconnaissance type field methods to identify what we term "critical points" within an entire reach. These critical points should exhibit processes that have a major influence on nitrogen exchange and may, for example, be areas of substantial groundwater recharge to the river, or zones with extensive connectivity to the riparian zone. Reconnaissance methods may include in-stream and ground-based geophysical methods, as proposed by Binley et al. (2013), alongside more traditional spatial sampling of water chemistry (including isotopes) and flow accretion. The second challenge relates to developing experimental infrastructure that can permit the monitoring of processes over longer time scales (e.g., decades) in order to reveal informed insight into long term changes to the hydrological and biogeochemical function. We are often constrained to research funding and instrumentation lasting only a few years. Although some attempts have been made to establish longer term studies (in the UK, the longest running example is the Plynlimon observatory), these remain uncommon. For such investigations we also need to consider alternative approaches to measurement. We cannot measure everything, everywhere, all of the time, and so need to develop suitable survey designs that target key dynamic signals, perhaps coupling ground-based approaches with the rapidly advancing capacity of airborne sensor technologies, while remaining sustainable and also adaptable to future change as our conceptual models develop.

| CONCLUSIONS
The research and new analysis synthesized here illustrates the critical importance of incorporating hydrogeological process understanding both beneath the riverbed, and from the wider landscape setting, in F I G U R E 7 Photographs of the study site during a major flood and under normal flow conditions. The 2005 photograph is taken during the recession limb of a rainfall event due to 200 mm (3 months equivalent) rainfall over a 36-hour period beginning 6 January predictive tools if we are to capture appropriately the role of the hyporheic zone for nitrate processing and its consequences for groundwater-fed river metabolics under a changing climate. The review and synthesis of previous reports coupled with new insights shows how a unique combination of physical hydrology and biogeochemical tools applied in detail to a river reach, can capture systematically the process understanding and complexity of the hyporheic zone. Few studies are able to spend the many person-years looking at a 200 m section so the challenge -and opportunity -lies in translating the insight and the understanding gained to frame research questions that target larger scale predictive physically-based and statistical models, which need to account for this process understanding in groundwater-fed rivers. To gain understanding of the potential impacts of a changing climate on future nitrate loads for groundwater-fed rivers, we need to improve the coupling of hydro(geo)logical, geomorphological and ecological process-based understanding through integrated field experimentation, advanced remote sensing technologies, and dynamic modelling.

DATA AVAILABILITY STATEMENT
This article is an analysis and synthesis of research undertaken throughout the NERC-funded project. The data that support the findings of this study are available on request from the authors.