Experimental observations of this study were conducted at the Wood Brook at the Birmingham Institute of Forest Research (www.birmingham.ac.uk/bifor) field site in Staffordshire, UK, between December 2016 and November 2017 (Figure 1). The second-order stream drains a 3.1 km² catchment characterised by a mixture of arable farmland and young (trees planted as “whips” in 2014) and mature (ages >100 year-old) mostly deciduous woodland. The catchment is underlain by Permotriassic sandstone with superficial deposits of glacial till up to 10 m thick and organic rich, sandy clay top soils between 0.15 and 0.6 m thick (Blaen et al., 2017).

A 1000 m study reach was established in an area of mature deciduous woodland upstream of the outflow of the catchment (Figure 1). The woodland is dominated by English oak (Quercus robur) with an understory of coppice hazel (Corylus avellana), hawthorn (Crataegus spp.), and sycamore (Acer pseudoplatanus). Common alder (Alnus glutinosa), goat willow (Salix caprea), and wych elm (Ulmus glabra) are also present adjacent to the stream (Hart et al., 2020). Dense canopy cover resulted in intensive shading throughout the reach during periods of full leaf-out (April to October; Day of year [DOY] ~90–300). The land cover of the contributing catchment at the lower end of the reach consisted of 19.3% broadleaf woodland, 45.4% arable farmland, and 33.0% improved grassland. While the lower end of the reach consisted of 15.4% broadleaf woodland, 47.2% arable farmland, and 35.3% improved grassland.

The study reach is characterised by steep incised banks for the majority of its length. Under baseflow conditions, the mean stream width along the reach is 1.6 m and the median travel time (derived from conservative solute tracer tests; Supplementary Figure 1) was 2 h. The streambed sediments are dominated by fine silt up to 0.8 m deep in the upper half of the study reach, and coarser sands and gravels up to 0.3 m deep further downstream. Consistent with the geology of the area, streambed core samples indicated the presence of an underlying clay layer throughout the study reach (except for one location where an outcrop of red sandstone was observed) and as such stream losses to groundwater were expected to be minimal. Similarly, substantial groundwater inflows were unlikely because the local water table was several metres below the streambed (Blaen et al., 2017). See Supplementary Figure 2 for further details.

Stream monitoring stations were established 1000 m apart, at the upper and lower ends of the study reach (Figure 1). The downstream station was operated for the full duration of the study and was equipped with a pressure transducer (Adcon, Austria) for water depth measurements and a Manta 2 multi-probe (Eureka, TX, USA). The Manta 2 housed sensors for measurement of water temperature, electrical conductivity, pH, and dissolved oxygen (DO). A stage-discharge relationship (R² = 0.88; Supplementary Figure 3) was established from salt dilution gauging measurements (Hudson and Fraser, 2005; Blaen et al., 2017). The upstream station was operated over three periods in 2017 during spring (DOY 74–123), summer (DOY 146–200), and autumn (DOY 259–306). During these periods, OPUS UV spectral sensors (TriOS GmbH, Germany) for measurement of NO3--N and DOC concentrations were deployed at both upstream and downstream stations. Discharge was not measured continuously at the upstream station. However, repeated dilution gauging measurements over a range of flow conditions through the study period showed that discharge at the upstream station could be predicted reliably from the discharge time series at the downstream station (see Supplementary Figure 4 for details).

Due to shallow water depth for much of the study period the sensors could not be reliably submerged in the river channel. To overcome this problem all sensors (except the pressure transducer) were housed in insulated kiosks 1 m from the stream bank. At both the upstream and downstream stations, 1 L of water was pumped every hour from intake points in the thalweg of the stream using ISCO 3710 automatic samplers (Lincoln, NE, USA). Water was passed through silicone tubing to flow cells containing the water quality sensors. The intakes were covered with coarse (1 mm) nylon mesh to reduce damage to the pump tubing from large particulates. Sensors remained wet between sample readings, and the volume of the tubing and flow cells was <15% that of the pumping volume (i.e., flow cells were purged sufficiently on each sampling occasion). Instruments were programmed to acquire sample readings 3 min after the completion of each pumping cycle. Data at the upstream station were stored locally on a CR1000 datalogger (Campbell Scientific, Loughborough, UK). Records from the downstream station were uploaded every 3 h via a telemetry system to an internet server.

Sensors were cleaned weekly using acetone on optical detection windows and a mild detergent on other components. Pump tubing was flushed on each cleaning occasion with 10% HCl to inhibit biofilm development. In addition to the manual cleaning regime, the Manta 2 probe had an automatic wiper that cleaned all sensors every hour. Sensors within the Manta 2 were calibrated every 2 months. The OPUS UV spectral sensors were cross-calibrated in the laboratory at the beginning and middle of the study period, and measurements exhibited strong linear relationships over a range of NO3--N (R² = 0.997) and DOC (R² = 0.993) concentrations. The downstream OPUS UV spectral sensor was also calibrated against laboratory reference samples following Blaen et al. (2017). The Limit of detection (i.e., 3 SD of lowest concentration standard) was 0.31 and 0.03 mg L for DOC and nitrate, respectively. Previous studies, both in this catchment and elsewhere, have demonstrated these instruments show close agreement with reference samples analysed in the laboratory and produce stable measurements during long-term deployments (Rode et al., 2016; Blaen et al., 2017). In addition, soil porewater samples (n = 6) were collected from shallow piezometers adjacent to the stream channel throughout the study period and analysed for NO3--N and DOC in the laboratory following standard methods as detailed in Blaen et al. (2017).

In addition to the stream monitoring stations, 102 towers, each ~25 m high, have been established in the woodland immediately adjacent to Wood Brook, to support a Free-Air Carbon Enrichment (FACE) facility (BIFoR FACE; Hart et al., 2020). On one of these towers, ~100 m from the study reach (Figure 1), LI-190R (LI-COR, NE, USA) photosynthetically active radiation (PAR) sensors were installed at 10 m (below the main oak canopy and above the hazel coppice canopy) and at 25 m (just above oak canopy) height. PAR measurements were acquired at 15 s resolution and 1 min average values were recorded to a CR1000 datalogger to provide a relative comparison of light availability above and below the canopy.

The median stream discharge over the 11 month monitoring period was 8.6 L s⁻¹ (Table 2). Streamflow was highest in early 2017, declined through summer months, and then increased slightly again in autumn (Figure 2). A total of 44 storm events were delineated from the hydrograph. These were distributed relatively evenly over the monitoring period, with the largest flow events observed in winter and early spring. Low amplitude diel cycles in streamflow were observed under baseflow conditions (e.g., July 2017), but were not apparent under event flow conditions (Figure 2). Diel flow variation under baseflow conditions (median 16.9%) was substantially lower than during storm flow periods (34.5%).

Water temperature and DO both exhibited strong seasonal and diel patterns (Figures 3A–C). On a seasonal basis, water temperature was lowest in winter and highest in summer, while DO followed a less pronounced inverse pattern. At the diel scale, water temperature typically peaked in mid-afternoon (1500 h local time), while the highest mean PAR and DO values occurred at 1200 and 1230 h, respectively. The magnitude of the diel variability in DO was consistent for most of the monitoring period, although a marked increase in diel range was observed for ~6 weeks between late March and early May (Figures 3B,C). Rapid changes in discharge (i.e., storm events) had little influence on water temperature, but were associated with short-term (hours to days) reductions in DO concentration and a dampening of the diel DO signal.

Daily PAR values below the canopy were lowest in winter, peaked in early spring with leaf-out of the main oak canopy, declined until late summer (August) and then remained low for the rest of the monitoring period (Figure 3D). In contrast, daily PAR values above the canopy followed a similar trend during winter and spring, but then continued to rise to a peak in June, before declining throughout late summer and autumn.

Across the whole monitoring period, mean NO3--N concentrations at the upstream and downstream stations were 6.2 ± 1.2 and 6.0 ± 1.2 mg L⁻¹, respectively, and followed similar temporal trends at both stations. NO3--N concentrations were highest in spring and declined through the monitoring period (Figure 4). On a seasonal basis, NO3--N concentrations were higher at the upstream station compared to the downstream station in summer (median values 6.3 and 6.2 mg NO3--N L⁻¹, respectively) and autumn (median values 5.1 and 4.4 mg NO3--N L⁻¹, respectively), but not in spring (median values 7.2 and 7.2 mg NO3--N L⁻¹, respectively). Diel fluctuations in NO3--N concentrations were relatively weak and changes in concentrations were not associated with changes in discharge (see Supplementary Figure 7). Storm events induced rapid changes in NO3--N concentrations, with the majority of storms associated with a short-term decrease in NO3--N concentration followed typically by a recovery to pre-event concentrations within 1–3 days.

The mass balance for NO3--N during spring was mostly positive (mean = +8.2%), indicating in stream release of NO3--N along the study reach (Figure 4; Table 3). During summer net release of NO3--N was lower than in spring (+0.6%) and shifted to net uptake across the reach during autumn (−2.6%). Storm events induced increases in release of NO3--N during all seasons relative to baseflow conditions, albeit with considerable variability (Figure 5). Some evidence of diel fluctuations in NO3--N uptake was observed in spring (Figures 4, 6), with increased uptake apparent in the late afternoon, although this was not as pronounced as for DOC.

Mean DOC concentrations at the upstream and downstream stations were 13.3 and 11.9 ± 3.0 mg DOC L⁻¹, respectively, over the monitoring period. As with NO3--N, DOC concentrations showed similar temporal dynamics between stations. DOC concentrations were highest in spring and autumn, with the lowest concentrations observed in summer (Figure 4). DOC concentrations were higher at the upstream station relative to the downstream station during spring (median values 12.49 and 11.00 mg DOC L⁻¹, respectively), summer (median values 10.5 and 10.3 mg DOC L⁻¹, respectively), and autumn (median values 13.2 and 12.0 mg DOC L⁻¹, respectively). In contrast to NO3--N, DOC concentration dynamics showed clear diel patterns under baseflow conditions at both the upstream and downstream sites (Figure 4). DOC concentration dynamics mirrored stream discharge, with the highest concentrations observed during large storm events, although on a diel basis discharge typically peaked 2 h ahead of DOC (see Supplementary Figure 7). Mass balance for DOC was negative for most of the monitoring period, with the exception of one period at the beginning of spring and another at the beginning of summer (Figure 4; Table 3). Storm events increased variability in DOC uptake, but did not lead to a consistent increase or decrease across seasons (Figure 5). Diel variability in mass balance was observed during spring for baseflow conditions, with the highest uptake rates occurring at ~1100 h and the lowest at ~1800 h (Figures 4, 6). In contrast, no diel pattern was evident during summer and much of autumn.

The DOC:nitrate ratio recorded at the down-stream monitoring site was lowest in spring and highest in autumn (Figure 7A). There was generally a positive relationship between discharge and the DOC:nitrate ratio (Figure 7B) that was most pronounced during spring (β = 0.04 ± 0.005, R² = 0.48, p < 0.01), with no significant relationship apparent in autumn. A weak, albeit significant, relationship (β = 0.058 ± 0.009; R² = 0.18, p < 0.001) was found between NO3--N uptake and DOC uptake under baseflow conditions. However, this relationship varied by season (Figure 8), with NO3--N uptake and DOC uptake negatively related in spring (β = −0.14 ± 0.011) and positively related in summer (β = 0.16 ± 0.02). In autumn, net NO3--N uptake and DOC uptake rates became less coupled a weaker negative relationship apparent (β = −0.12 ± 0.019).

Across the whole monitoring period, NO3--N uptake dynamics were predicted by a relatively small subset of the variables (top model; weight = 0.38; adjusted R² = 0.73; Table 4), with mean daily DOC:nitrate ratio (standardised β = 0.64), water temperature (−0.18) and discharge (−0.17) the most important variables. The relationship between DOC:nitrate ratio and N uptake presented a clear break point (3.3 ± 0.15 SE) which was identified using segmented regression (Davies test P < 0.001; Supplementary Figure 8). For DOC uptake the relationship with the predictor variables was weaker (top model; weight = 0.33; adjusted R² = 0.12), but PAR (0.31), mean daily discharge (−0.35), and magnitude of previous storm event (0.31) were the most important variables (Table 4). Clear seasonal variability in the drivers of N uptake were apparent (Figure 9; Table 4) and a strong correlation with DOC:nitrate ratio was apparent in spring and summer, but was reduced in autumn. Higher discharge was associated with higher (lower) uptake rates in summer (autumn) but the inverse was true for water temperature (Figure 9, Table 4). For DOC uptake discharge and storm events appeared to be the main controls across all seasons, with PAR only important in spring (Table 4).

Patterns of hourly NO3--N or DOC net uptake rates immediately following storm events showed few consistent changes relative to pre-event conditions (Figure 10). Storms in spring and autumn did not appear to induce major short-term changes in nutrient rates, with the exception of NO3--N uptake in spring that was noticeably higher than pre-event conditions. In contrast to spring and autumn events, summer post-event net uptake rates of both NO3--N and DOC were markedly higher with strong reductions in nitrate between upstream and downstream compared to pre-event conditions.

The median mass balance for NO3--N was positive (136 mg m⁻² day⁻¹), with the study reach acting as a net source for 60% of the monitoring period. This was most probably due to high N loading from the surrounding landscape (i.e., agricultural legacy) that led to saturation of whole-stream NO3--N uptake capacity and the export of surplus NO3--N downstream (Bernot and Dodds, 2005; Sylvester-Bradley and Kindred, 2009). However, it is worth noting that in some cases DOM mineralization can result in ammonification and subsequent denitrification (Shogren et al., 2019). At times when mass balance was negative and removal of NO3--N occurred (i.e., early summer and autumn), the observed rates of between 0 and 1000 mg m⁻² day⁻¹ were similar to, and often higher than, rates observed in other headwater streams in agricultural catchments in temperate areas of the world. For example, Webster et al. (2003) reported a mean gross NO3--N uptake rate of 111 mg m⁻² day⁻¹ (range: 0–7299 mg m⁻² day⁻¹) across 11 headwater streams in N. America, while in a meta-analysis of instream nutrient retention, Ensign and Doyle (2006) reported a mean NO3--N uptake rate of 170 mg m⁻² day⁻¹ across 29 s-order streams. The general pattern we observed of increasing NO3--N uptake from spring to autumn is consistent with other studies (Jarvie et al., 2018; Reisinger et al., 2019) and was also identified by Comer-Warner et al. (2020) in isotope study of the sediments of Wood Brook. Hence, we suggest heterotrophic denitrification may have become an increasingly dominant pathway for removal during autumn in our shaded study system. Also, a study by Yue et al. (2020) highlights how coupled isotope and high frequency monitoring is a particularly useful technique for unpicking N sources and pathways. Recent studies from river reaches with less riparian shading have recorded greater uptake rates across spring—autumn monitoring campaigns, with peaks of >2000 mg m⁻² day⁻¹ and study reaches acting as net sinks for NO3--N (Reisinger et al., 2019; Preiner et al., 2020). This highlights the importance of assimilatory uptake for regulating NO3--N fluxes particularly during spring and summer (Jarvie et al., 2018), a period when shading was high in our system and net export of NO3--N was apparent.

Our study reach was a net sink for DOC (median uptake = −258 mg m⁻² day⁻¹) with removal of DOC from the water column apparent for 75% of the study days. We suggest this was due to the shaded nature of the site (i.e., strong heterotrophic demand for DOC) and leaf fall inputs from riparian trees acting as a source through much of the year. It is worth noting that less research has been conducted into instream DOC uptake relative to NO3--N uptake (Mineau et al., 2016). Most previous work in headwater streams has highlighted the strong control of DOC composition on uptake rates, yet this has been based on experimental injections of organic carbon (Bernhardt and McDowell, 2008; Fellman et al., 2009). Studies measuring uptake using mass balance approaches have highlighted moderate uptake in headwater systems with rates comparable to that observed in our study (e.g., −380 mg m⁻² day⁻¹; Wollheim et al., 2015) or less (e.g., ~0 mg m⁻² day⁻¹; Lupon et al., 2020) than observed in this study. However, our results provide further empirical data to support previous hypotheses that low order river corridors can represent important sites of DOC turnover at the landscape scale (Bertuzzo et al., 2017), particularly compared to more downstream reaches (Huntington et al., 2019). Our results showed that the association between N and C uptake is subject to high day-to-day variation and could be driven by variability in DOC composition (Lupon et al., 2020). However, at seasonal scales coupling between N and C was observed, particularly during summer, thus underlining the utility of long-term, high frequency datasets for unpicking river ecosystem functioning.

No clear diel patterns in NO3--N mass balance were observed (cf. Kunz et al., 2017). Previous studies have reported strong diel variation in NO3--N uptake rates (Heffernan and Cohen, 2010), although most have been conducted in systems with high assimilatory demand for NO3--N by primary producers, which are governed by diel changes in light intensity. Therefore, if heterotrophic uptake processes (e.g., denitrification) are dominant in this system, it is conceivable that rates of net NO3--N uptake are more time-invariant than in other streams dominated by autotrophic processes. In contrast to NO3--N, distinct diel cycles in net DOC uptake were observed during late spring, although these were much less pronounced at other times of the year. As peak DOC uptake typically occurred around midday, one plausible explanation for these patterns is that diel changes in light intensity increased photodegradation of DOC in spring when the canopy was still relatively open. Photodegradation can account for a significant fraction of DOC removal from surface waters. For example, Worrall et al. (2015) reported removal rates up to 0.1 mg C L⁻¹ h⁻¹ for the River Dee, UK. In upland river systems where riparian shading is limited, and DOC concentration can be elevated due to highly organic soils, photodegradation can account for ~60% of losses in the water column (Moody and Worrall, 2016). This light induced uptake/loss occur through a number of processes including: (1) direct photo-oxidation/aggregation of DOC, (2) an increase in DOC bioavailability for microbial uptake (Lu et al., 2013; Moody and Worrall, 2017), and (3) priming effects associated with stimulation of autochthonous carbon production (Guenet et al., 2010). Conversely, daily cycles in DOC uptake may have been driven by diel variation in discharge with source waters containing different concentrations of DOC. The relatively strong positive relationship between discharge and DOC indicates DOC changes were not driven by a dilution effect. Moreover, during several periods the study reach acted as a net sink of DOC during the day and a net source at night, indicating that single time point uptake experiments could lead to erroneous interpretations of instream uptake processes, particularly if diel variability is not adequately characterised.

For this forested headwater reach, we found stronger relationships with the variables hypothesised to be the key drivers of nutrient uptake for NO₃-N, relative to DOC. Nutrient stoichiometry was the best predictor of N mass balance and reflects that, under high DOC concentrations, demand for inorganic nitrogen increases, thus driving uptake rates (Taylor and Townsend, 2010). While this relationship has been previously highlighted on the basis of discrete uptake experiments based with field additions (Wymore et al., 2016), our study is the first to explore links across the growing season (e.g., spring—autumn). Interestingly, we observed a weakening of the relationship between DOC:nitrate ratio and N uptake in autumn when the ratio with highest. A break point in the relationship was identified at a ratio of ~3.3 (Supplementary Figure 8), which corresponds to the point at which nitrogen assimilation is no longer carbon limited (Taylor and Townsend, 2010). PAR was only related to nutrient uptake in spring suggesting assimilatory processes may have only operated as a mechanism for nutrient removal before seasonal riparian shading reduced instream primary production (Fellows et al., 2006). However, our results do not allow us to explicitly assess the mechanisms of the nitrogen cycle driving NO3- uptake. During summer, discharge also became an important predictor of N uptake reflecting the transport limitation of DOC during this time. It is likely uptake was stimulated by the flushing of fresh DOC into the channel (coupling of C and N uptake; Figure 8) from surrounding riparian soils, which also altered C:N stoichiometry (Heppell et al., 2017). This process of solute transport during higher flows is probably also coupled with an increased potential for biochemical reactions in new or “young” water storage zones within the riverbed (Ward et al., 2019). The molar ratio of DOC:nitrate did not appear to limit N uptake during autumn, rather discharge (negative correlation) and time since previous storm event, our proxy for habitat disturbance (positive correlation), were the main drivers. This suggests a shift from stoichiometric controls, as DOC availability is not limiting, to physical controls in autumn.

DOC uptake in this study was associated with PAR (positive), discharge (negative), and magnitude of previous event (positive). The relationship with PAR is interesting as this could represent DOC uptake due to light availability stimulating metabolism (Demars et al., 2020) or conversely could be related to increased solar radiation reaching the stream and greater potential for photodegradation of DOC (Moody and Worrall, 2017). Further work is required to explicitly test this however the elevated delta DO during early spring (Figures 3A,C) suggests the former mechanism is more likely. The role of hydro-climatology in controlling DOC uptake was also apparent with higher flows associated with reduced rates of uptake. This could be a function of DOC source shifting as ambient DOC composition (i.e., humic vs. protein compounds) can influence uptake (Lupon et al., 2020). Microbial communities respond most strongly to flushing in streams with low ambient humic DOC (Catalán et al., 2018), which was not the case for our study system. Despite the negative relationship with discharge, the magnitude of the previous storm event appeared to stimulate uptake and could reflect the potential for storms to restructure microbial communities through mobilisation and transport of new microbial species with potential implications for community functioning (Kan, 2018). Interestingly, there was no relationship between DOC uptake and the DOC:NO3- ratio suggesting nitrate never limited DOC uptake. This is in contrast to recent studies which have found negative relationship between DOC uptake and increasing DOC: nitrate ratios (Catalán et al., 2018), albeit across a range an order of magnitude greater than observed in this study. Hence, the result we observed is likely due to the agricultural legacy of the catchment and relatively high availability of N.

Seasonal and diel patterns in nutrient mass balance were modified by stochastic storm events throughout the monitoring period. Uptake and release rates of both NO3--N and DOC during storm events exhibited higher variability than during baseflow conditions, although no consistent increase or decrease in uptake during storm events was evident. This higher variability is likely to be explained by rapid changes in both NO3--N and DOC concentrations associated with storm events, as observed in both this study and previous research (Blaen et al., 2017; Vaughan et al., 2017). In addition, increased flow velocities and turbidity levels through the study reach during storm events reduced the time available for instream nutrient processing and diminished light intensity in the water column (Kraus et al., 2017).

High flows associated with storm events can cause substantial disturbance to instream habitat conditions and lead to major changes in aquatic ecosystem structure and function (Poff et al., 1997; Milner et al., 2013). In particular, previous studies have observed storms to suppress rates of instream metabolism (both ecosystem respiration and gross primary production) in many systems across urban, forested and alpine environments (Uehlinger et al., 2003; Beaulieu et al., 2013; Rode et al., 2016; Reisinger et al., 2017). Given that stream metabolism is often coupled with nutrient uptake (e.g., Hall and Tank, 2003), it has been hypothesised that storm-driven reductions in metabolic activity would have corresponding negative impacts on rates of instream nutrient uptake (Walsh et al., 2005; Wenger et al., 2009). In this study, we observed little evidence to support this hypothesis: post-event mass balances for NO3--N and DOC did not differ consistently from pre-event during spring and autumn. This may indicate that aquatic microorganisms in this system are adapted to be highly resistant to disturbances associated with high flows, and supports previous studies that have suggested instream communities dominated by heterotrophs are less susceptible to storm-related disturbances than those dominated by autotrophs (Uehlinger et al., 2002; Roberts et al., 2007). In summer, post-event net uptake rates were higher than those prior to the event, which may be attributable to potential stimulation of heterotrophic activity by the input of nitrogen and labile organic carbon from source areas within the catchment that became activated during storm events (Roberts et al., 2007; Beaulieu et al., 2013; Blaen et al., 2017). The limited number of storms that occurred during the monitoring period precluded a more systematic analysis of short-term responses of nutrient uptake rates to storm events. For example, a larger number of cases would facilitate insights into the effects of flood magnitude on nutrient uptake rates (Reisinger et al., 2017; Bernhardt et al., 2018). Similarly, capturing nutrient dynamics during larger winter storms, or coupling measurements with isotope sampling (cf. Yue et al., 2020) would lead to a better understanding of community responses to disturbance over an annual or inter-annual context. Thus, we suggest the application of high-frequency nutrient sensors will prove a valuable approach to gain detailed understanding of these processes in future.