The impact of rainfall events, catchment characteristics and estuarine processes on the export of dissolved organic matter from two lowland rivers and their shared estuary

• Rapid transport of DOC from soils to river after rainfall events • DOC andDONprocessingwithin estuary


Introduction
Dissolved organic matter (DOM) is an important source of carbon and nitrogen to aquatic ecosystems and a great deal has been learnt about the role of DOM in global biogeochemical cycles over the last few decades (e.g. Yamanaka andTajika, 1997, Hansell andCarlson, 2014 andrefs within). Human activity, in particular the type of land use, has been shown to determine both the source and the composition of DOM, and new processes such as atmospheric deposition have been recognised as important inputs of both carbon and nitrogen to nutrient cycles (e.g. Cornell et al., 2003;Muller et al., 2008). The role of rivers in the transport of DOM to estuaries and coastal seas is also becoming increasingly better understood (e.g. Raymond et al., 2016;Casas-Ruiz et al., 2017;Drake et al., 2018). Once considered simply to be passive transporters of terrestrially-derived DOM from soils to the sea, rivers are now recognised as dynamic systems where both production and loss processes can potentially alter both the concentration and the composition of DOM during transport (Battin et al., 2008;Bertuzzo et al., 2017;Graeber et al., 2018;Harris et al., 2018). The balance of such processes can differ, however, even between reaches of the same river and hence the flux of both dissolved organic carbon (DOC) and dissolved organic nitrogen (DON) is hard to predict (Wymore et al., 2018). The development of models and management tools to predict the composition of DOM entering rivers as well as to quantify the flux of carbon and nitrogen from riverine systems into estuaries has therefore been limited despite recent advances (e.g. Anderson et al., 2019;Yates et al., 2019).
As it has also been demonstrated that terrestrial DOM is more bioavailable than previously believed (Autio et al., 2016;Wiegner et al., 2006), both DOC and DON are implicated as potential contributing factors to problems such as eutrophication and hypoxia within estuaries (Seitzinger and Sanders, 1997;Paerl et al., 1998;Wiegner et al., 2009).
Stochastic rainfall events that lead to rapid and sustained increased river flow rates are expected to have a disproportionate impact on these estuarine nutrient burdens. For example, much of the export of DOM from soils to streams occurs during brief periods of high river flows following intense rainfall events (Inamdar and Mitchell, 2007;Morel et al., 2009;Hitchcock and Mitrovic, 2013). As these events are expected to increase in temperate latitudes over the coming century and beyond (IPCC, 2014), it is important to understand now how they may impact upon concentrations of DOC and DON both within rivers and downstream in estuaries under different flow conditions. Much work has been done on the impact of rainstorms on DOC and DON concentrations in rivers and streams with various different types of watershed characteristics (e.g. Buffam et al., 2001;Inamdar and Mitchell, 2007;Morel et al., 2009). The majority of these studies, however, have focussed on systems that have low inorganic nutrient loads (nitrogen typically b100 μM) with few studies investigating the role DOC and DON will play in rivers and estuaries already burdened by high inorganic nutrient loads and at imminent risk of eutrophication.
The Christchurch Harbour Macronutrients Project was designed to investigate the impact of stochastic rainfall events on the transport and biogeochemical cycling of macronutrients in two temperate UK south coast rivers, the Hampshire Avon and the Stour, as well as in their shared estuary, Christchurch Harbour (UK). The Hampshire Avon has been previously identified as a river of national importance due to its predominantly agricultural watershed and elevated inorganic nutrient loadings (mean nitrate concentration ≈400 μM; Jarvie et al., 2005). We present here a unique data set of frequent observations from the lowest river gauging stations and estuary impacted by high nitrate concentrations. The goals of our study were to ascertain the annual variability of DON and DOC in the context of high inorganic nitrogen loads, to examine the impact of rainfall events leading to rapid but sustained increases in river flow on the potential fate of DOC and DON, and to investigate the role of the shared estuary in determining the organic and inorganic nutrient flux into coastal waters.

Study area
Christchurch Harbour is a shallow microtidal estuary on the south coast of England with a single outflow into the English Channel (Fig. 1). The mean tidal range during spring tides is 1.2 m, and the mean water depth outside of the main channel is approximately 0.5 m (Huggett et al., 2020). Two rivers, the Hampshire Avon (hereafter referred to simply as the Avon) and the Stour, drain into the estuary with a total catchment area of 2779 km 2 . The mean flow of the Avon and Stour at the lowest gauging stations on each river is 19.5 m 3 s −1 and 13.8 m 3 s −1 respectively (Centre for Ecology and Hydrology (CEH), 2008). A third small river, the Mude, drains into the estuary near the outlet ( Fig. 1) but only has a mean flow of 0.1 m 3 s −1 (CEH, 2008) and was not included in this study. The predominant land use types in the catchments of these two rivers are similar with over 75% of each catchment being a mixture of grassland and arable/horticultural land, but also with some woodland and small areas (1-2%) of heathland and urban areas (Table 1, CEH, 2008). A catchment map with land use classification is available in the Supplementary information (Supplementary Fig. 1). The geology between the two catchments does differ however with the Avon draining from predominantly chalk and the Stour draining from a mixture of chalk (50%) and clay (30%). These geological differences are reflected in their Baseflow Indices (BFI)the Avon has a BFI of 0.90 indicating a high groundwater component in river discharge, and the Stour has a BFI of 0.65 suggesting that a lower proportion of river discharge originates from stored catchment sources (CEH, 2008).

Sampling regime
Water samples were collected from Environment Agency gauging stations at Knapp Mill (50.744 N, −1.782 W) on the Avon and Throop (50.764 N, −1.842 W) on the Stour, as well as at Mudeford Quay (50.724 N, −1.7409 W) at the mouth of the Christchurch Harbour estuary at 5-8 day intervals between May 2013 and April 2014. The two river flow gauging stations were the closest to the estuary on each of the respective rivers located 12.7 km (Throop) and 6.2 km (Knapp Mill) upstream from the mouth of the estuary. Surface water was collected using a clean bucket and immediately decanted into acidcleaned HDPE bottles for later inorganic nutrient analysis or into combusted (450°C for a minimum of 4 h) glass bottles for total dissolved nitrogen (TDN) and dissolved organic carbon (DOC) analysis. At each site surface water temperature and conductivity (salinity at the estuarine site) were measured in situ using an EXO2 multiparameter sonde (Xylem, UK). At the estuary site, water temperature and salinity were also measured at depth just above the sediment. Samples from the estuary mouth were collected at low tide.
Boat transects were carried out at high tide in Christchurch Harbour fortnightly between 27th May 2014 and 4th September 2014 at 6 sites along a salinity gradient from the mouth of the estuary to an upstream site within the Stour (Fig. 1). A YSI 6600 sonde (Xylem, UK) was used to measure salinity and temperature profiles and the depth of highest chlorophyll fluorescence was sampled using a 5 L Niskin bottle except on 27th May 2014 when surface water was sampled using a clean bucket.
On return to the lab, water samples for later nitrate (NO 3 − ) plus nitrite (NO 2 − ) analysis were filtered through a 25 mm diameter GF/F filter using an inline syringe unit, preserved with 0.015 M HgCl 2 (100 μL per 20 mL), and stored in the dark at room temperature (Kirkwood, 1996). Samples for DOC and TDN analysis were filtered through a combusted 47 mm diameter Whatman GF/F filter (nominal pore size 0.7 μm) on an acid-washed glass filter rig under low vacuum (b10 mmHg) and 20 mL of filtrate stored in combusted glass vials with acid-washed Teflon septa. Phosphoric acid (60 μL of 50% (v/v)) was added to each sample vial before storing at 4°C (Badr et al., 2003). Samples for ammonium and urea analyses were also collected during this filtration and were pipetted into 25 mL push top glass vials. Reagents for ammonium (NH 4 + ) and urea analyses were added immediately and vials were incubated at room temperature in the dark for up to 24 h (ammonium) and for between 3 and 5 days (urea). Urea samples were not collected during the boat transects.

Analytical methods
Concentrations of nitrate plus nitrite were determined at the University of Portsmouth on a QuAAtro segmented flow nutrient analyser (SEAL Analytical, UK). Ammonium and urea concentrations were measured according to the method of Holmes et al. (1999) and Goeyens et al. (1998) respectively at all sites from late August 2013. Dissolved inorganic nitrogen (DIN) concentrations were calculated by adding nitrate plus nitrite and ammonium concentrations. Concentrations of DOC and TDN were measured with a TOC-V CPN analyser (Shimadzu, Japan) calibrated with a mixed standard of potassium hydrogen phthalate and glycine. Certified Reference Materials (DSR from University of Miami, USA) were used to validate results by comparing against the certified concentrations for DOC (41-44 μM; analytical mean 45.8 μM) and TDN (31-33 μM; analytical mean 30.5 μM). Concentrations of dissolved organic nitrogen (DON) were quantified by subtracting DIN concentration from TDN concentration. DON data from the first 5 sampling dates are not available.

Additional data and calculations
Daily mean river flow data for both river sites was provided by the Environment Agency with data for the study period as well as historic flow data from the period 2000-2010 inclusive from which an 11-year daily mean flow was calculated for each river. Limited data for DOC and DIN concentration during the sampling period were also available from the Environment Agency for the Knapp Mill site only and are presented for comparison purposes. Rainfall data was acquired from the Meteorological Office station at Bisterne situated 6.3 km north of Knapp Mill on the Avon. The flushing time of the estuary was calculated using a simple tidal prism method as described by Huggett et al. (2020). Instantaneous fluxes for the river sites were calculated by multiplying the measured concentration of each nutrient by the daily mean flow for the same day at that site. For the estuary, the daily mean flow for both rivers were summed together to determine the total daily mean flow and then multiplied by the nutrient concentrations measured at the estuary mouth. Annual fluxes were calculated by linearly interpolating between known concentration data points to obtain daily concentrations and then calculating flux as above before summing all daily values for each site. Annual fluxes were divided by catchment size to allow direct comparison of annual yield to other rivers and estuaries. The baseflow contribution to daily river flow was estimated using the smoothed minima technique of Gustard et al. (1992) as detailed in Jordan et al. (1997).
Nitrate and river flow data from the Avon weekly sampling campaign have previously been published in Pirani et al. (2016) where the data were combined with water quality and phosphate data to develop a model to determine past nutrient fluxes based on historical river flows.

River flow and estuarine flushing time
Water temperature in both rivers and the estuary followed a seasonal cycle with minimum temperatures between 5.7°C and 6.2°C observed in winter and maximum summer temperatures reaching between 22.8°C and 23°C in July 2013 (data not shown).
Daily mean flow in both rivers decreased from the start of sampling (April 2013) to a summer low flow state equal to the estimated baseflow for each river by around mid-July 2013 and remained low until a sharp increase towards the end of October 2013 (Fig. 2a, b). This period of elevated flow lasted for approximately 3 weeks before flow decreased again during a dry period from mid-November to mid-December 2013, although it remained above the summer low flow values. A second period of elevated flow started with sharp increases in flow in both rivers on the 16th December 2013 with sustained flows above both the 11-year mean and the highest flow values from the earlier period of elevated flow until the sampling finished on 10th April 2014 (total duration 150 d). Both periods of elevated flow were associated with increased rainfall locally (Fig. 2a). Flow data was unavailable at Knapp Mill for several days in early January 2014 after the gauging station was struck by lightning, but river flow rates both immediately before and after the loss of data were substantially elevated from background and so the flow was assumed to remain elevated across the period of missing data. Mean flow rates were more variable in the Stour than in the Avon, but mean flow during low flow periods was typically higher in the Avon. Daily mean flow rates for both rivers as well as the summed daily river flow and daily rainfall for the period of the estuarine transect sampling in summer 2014 also show an increase in river flow after rainfall events (Fig. 2c).
In each river, the daily mean flow during the study period was significantly different (Mann-Whitney U test, Avon U 60449, Stour U 58501, both p b 0.05) to the historic daily mean river flow from 11 years of Environment Agency data ( Fig. 2a, b). These differences are particularly evident in the second elevated flow period where daily mean flows were up to 3 times greater in the Avon and 7 times higher in the Stour than the 11-year daily mean. Low rainfall in early December 2013 also leads to a period where mean daily flow was as low as 12% (Stour) or 28% (Avon) of the 11-year daily mean.
Flushing times within the estuary ranged from 0.1 days with combined river flows of 100 m 3 s −1 to 1.5 days with minimum summer combined flows of 10 m 3 s −1 . The flushing time was consistently b1 day over the second elevated flow period from the 17th December 2013 until the end of sampling in April 2014.

Inorganic and organic nutrient concentrations
Concentrations of nitrate plus nitrite throughout the sampling period were higher in the Stour (mean 502 μM) than in the Avon (mean 381 μM) and at the estuary mouth (mean 328 μM; Fig. 3a-c). The highest concentrations at all sites were seen during the periods of decreasing flow following each of the elevated flow periods. The impact of low nutrient coastal waters can be seen in the estuary at Mudeford over the periods of low river flow when nitrate plus nitrite concentrations are lower than in either river (Fig. 3c). Spearman correlations revealed a significant relationship between DIN concentrations in both rivers as well as between each river and the DIN concentration in the estuary (ρ 0.533-0.606, p b 0.001 for all). Ammonium concentrations ranged from b1 μM to 9.8 μM and were lower on a weekly basis in the Avon (mean 2.4 μM) than in the Stour (mean 3.8 μM) or the estuary (mean 3.8 μM). Concentrations of ammonium were always b2.5% of DIN in the estuary and b1.8% of DIN in the rivers.
DOC concentrations ranged from 167 to 486 μM (mean 249 μM) in the Avon, 156-1119 μM (mean 353 μM) in the Stour, and 162-676 μM (mean 273 μM) in the estuary (Fig. 3d-f). A general pattern of relatively low DOC concentrations (b300 μM) was observed at all three sites ( Fig. 3d-f) between May and October 2013 before the first period of elevated river flow, and subsequently concentrations at all sites increased during elevated river flow and later decreased as flow declined. Again, Spearman correlations revealed a significant relationship between DOC concentrations in both rivers as well as between each river and the DOC concentration in the estuary (ρ 0.680-0.769, p b 0.001 for all).
DON concentrations ranged from 0 to 83 μM (mean 32 μM) in the Avon, 0-155 μM (mean 40 μM) in the Stour, and 0-54 μM (mean 17 μM) in the estuary (Fig. 3g-i). There was no clear pattern in DON concentration at any one of the sites and the response to elevated river flow events differed between sites and between events. Concentrations of DON increased in both the Avon and the Stour just prior to the start of the first elevated flow period, but this increase was not observed in the estuary. Over the course of this first elevated flow period DON concentrations in both rivers decreased and then increased steadily again, but in the estuary concentrations remained low. The greatest concentrations in the estuary and in the Avon occurred on the same date (7th March 2014) during the second elevated flow period, but the highest concentration of DON in the Stour was observed during the lower flow conditions between the two elevated flow periods in early December 2013 (Fig. 3h). Spearman correlation revealed a significant relationship between DON concentration in both rivers (ρ 0.504, p b 0.005), but there was no relationship between either of the rivers and the DON concentration in the estuary. The proportion of DON in TDN ranged from 0 to 26% across the three sites with a mean proportion of 5% in the estuary, 8% in the Avon, and 7% in the Stour.
Urea concentrations ranged from b1 μM to 3.6 μM but there was no clear pattern between sites. Mean urea concentration was 1.2 μM in the Avon, 1.2 μM in the estuary, and 1.4 μM in the Stour.

Relationships with river flow
Overall there was a pattern of increasing DIN with river flow in both rivers up to a flow of approximately 25 m 3 s −1 in the Stour and 35 m 3 s −1 in the Avon, after which DIN concentration decreased as river flow increased further (Fig. 4a, b). DOC concentrations in both rivers increased to a peak as river flow increased at the lower range of river flows (up to approximately 25 m 3 s −1 in the Avon and up to approximately 50 m 3 s −1 in the Stour; Fig. 4c, d). Above these river flows there appears to be a positive relationship between river flow and DOC concentration. In the Avon the relationship between DON and river flow is complicated with one peak in DON concentration below 20 m 3 s −1 and another peak at approximately 70 m 3 s −1 (Fig. 4e). DON concentration in the Stour was highest at low river flow but there was a smaller second peak at approximately 60 m 3 s −1 (Fig. 4f).
During dry periods the calculated baseflow is equal to or very close to the measured river flow, but rainfall events can decrease the proportion of measured flow contributed by baseflow (see Fig. 2). When DIN, DOC and DON are plotted against the proportion that baseflow contributes to measured flow then some relationships become clearer (Fig. 5). The concentration of DIN in both rivers appears to increase when the proportion of baseflow increases (Fig. 5a, b), whilst the opposite is true for DOC with the highest concentrations observed when baseflow is contributing less to river flow (Fig. 5c, d). Again, the variability in DON concentrations does not appear to have a clear relationship with baseflow contribution (Fig. 5e, f).

Estuarine results
In the estuary it is clear that periods of elevated river flow restricted the inflow of more saline coastal waters (Fig. 6). At a combined river flow of 40 m 3 s −1 or greater the salinity at the mouth of the estuary (both surface and bottom) was typically b5 and frequently b1, resulting in Christchurch Harbour becoming essentially a freshwater lake under very high river flow conditions. These total river flows correspond to all dates sampled from 17th December 2013 until the end of the weekly sampling programme and so reflect the second elevated flow period in its entirety.
Surface salinity within the estuary decreased from the mouth to the upstream sites. There is a clear conservative relationship between nitrate plus nitrite concentration and salinity in both the weekly sampled data at Mudeford (crosses; Fig. 7a) and the estuarine transects (coloured circles). Ammonium concentration is relatively low (b3 μM) at a salinity of 20 or greater, but there is more variability at lower salinities (Fig. 7b). The dominance of nitrate and nitrite in this system is evident in the clear relationship between DIN and salinity (Fig. 7c). The relationships between DOC and salinity (Fig. 8a) and between DON and salinity (Fig. 8b), however, are not as clear. In the weekly DOC samples and some of the estuarine transect samples there is the suggestion of a conservative relationship, but there are other estuarine samples with high concentrations of DOC at each end of the salinity range (e.g. 7th August 2014) or with peaks in the mid-salinity range (e.g. 10th July 2014). Interpretation of the DON data is complicated by the large number of values at concentrations below the limit of detection, but a general relationship between salinity and DON concentration appears  to be present with lower DON at higher salinities (Fig. 8b, coloured circles).

DOC: nitrate ratios
Ratios of DOC: nitrate were b2.5 at all sites throughout the weekly sampling and the highest ratios (1.5 to 2) were observed during high flow periods in late October and in December/ January (Fig. 9a). DOC: nitrate ratios were also low across the estuarine transect sampling study with a mean ratio of 1.87 (Fig. 9b). Only 4 samples had a DOC: nitrate ratio N 3 and 3 of these samples occurred in the mid to low estuary on the same date (7th August 2014) when DOC concentrations were high and DIN concentrations were amongst the lowest observed. The maximum ratio observed on this date was 18.8 at a mid-estuary site.

Fluxes of DIN, DOC, and DON
When the instantaneous flux of DIN, DOC, or DON is calculated the importance of increased flow events becomes evident (Fig. 10 Table 2.

Discussion
Two periods of sustained rainfall with associated elevations in river flow were captured during the year of sampling with an initial wetting up period in late October 2013 being followed by a prolonged period of several months duration between December 2013 and March 2014. These events allow the dynamics of DIN, DOC and DON concentrations within the two rivers and the estuary to be examined under a range of hydrological conditions.

DIN dynamics and sources
Concentration of DIN in both rivers was consistently high (N290 μM) throughout the study period, reinforcing the status of these rivers and the shared estuary as an impacted system. A substantial proportion of each catchment (N35%) is used for arable or horticultural purposes, and thus a major source of DIN is likely to be agricultural in nature. Both Heppell et al. (2017) and Yates et al. (2019) report a positive relationship between the percentage of arable land use and nitrate or total nitrogen concentrations in the upper reaches of the Hampshire Avon catchment. Rainfall events should result in an increase in the flux of this diffuse-source DIN from land into the rivers (Withers and Lord, 2002). The finding that DIN concentrations are high when river flow is dominated by baseflow, however, implies that there are also point sources contributing to the DIN load (e.g. effluent from sewage treatment works), or that the groundwater may also be high in DIN, and finally that in-stream processes such as macrophyte and microalgal uptake at these relatively downstream sites use only a small portion of the total DIN. Jarvie et al. (2005) calculated that the effluent load of nitrate to the Avon at Knapp Mill was around 11% of river load and also identified that groundwater is a major source of nitrate in the Avon. Approximately 75% of nitrate in U.K. groundwater is believed to be from agricultural sources, although other sources such as atmospheric deposition, discharges or leaks from septic tanks and sewers, and the spreading of sewage sludge on land may also contribute (Rivett et al., 2007). In the Hampshire Chalk aquifer, isotopic analysis has shown that denitrification is an insignificant process within the unsaturated zone and, as a result, the nitrate concentrations in groundwater have been increasing since the 1970s (Rivett et al., 2007). Little has been published on the water chemistry of the Stour but, as the Stour has a lower BFI and is less dependent on groundwater contributions, the majority of the DIN is suspected to be from agricultural or sewage treatment sources. The conservative relationship between DIN and salinity again suggests that biological processing within the estuary has minimal impact on total DIN concentration with the result that Christchurch Harbour exports the majority of the DIN to the coastal seas of the English Channel.

DOC dynamics and sources
Whilst daily flow in the two rivers behaved differently with the Stour displaying greater variability and more rapid fluctuations in flow than the Avon, the DOC concentrations in both rivers behaved in a similar manner over the year as demonstrated by a significant Spearman correlation over time. Little variability in concentration was observed during the lowest flow summer months, potentially reflecting that instream production-loss processes at these downstream sites on each  river counterbalance any terrestrial inputs of fresh DOC (Creed et al., 2015). This proposed excess in DOC was also observed in the estuarine concentrations over the same period with constant concentrations throughout the summer baseflow period being transported to the coastal zone.
The DOC dynamics under elevated flow conditions throughout the system, however, show a different pattern. Rapid increases in DOC concentration in both rivers as river flows increased and during subsequent peaks within the longer period of prolonged high flows reflect flushing of terrestrial organic matter, which is transported rapidly downstream by the increased flows, subsequently escaping any significant upstream biogeochemical processing. The maximum DOC concentration measured during these pulses in the Stour was more than double that measured in the Avon, reflecting the different catchment characteristics of the two rivers. The Avon is groundwater-dominated in a predominantly chalk catchment (Jarvie et al., 2005;Yates et al., 2016) whilst the Stour has less permeable clay soils where surface runoff can result in rapid transport of DOC into the river. In contrast, groundwater DOC concentration in chalk aquifers is typically low. Rivett et al. (2007) reported a mean DOC concentration of 60.8 ± 19.2 μM from 1725 groundwater samples across the major Cretaceous Chalk aquifer in England, and Stuart and Lapworth (2016) determined the mean baseline concentration of groundwater DOC in Hampshire chalk at 67.8 μM. Point sources such as sewage treatment works (STWs) are present in both catchments with over 140 STWs and 30 fish farm discharges in the Avon catchment alone (Jarvie et al., 2005), and sources such as these are likely to have contributed to the rapid pulses in DOC observed. Other sources such as groundwater seepage and ditch drainage occur on longer timescales and may have contributed to the continued elevation of DOC concentrations during prolonged high river flow events (Morel et al., 2009). Atmospheric deposition from rainwater could also contribute to the DOC concentration, with studies reporting mean DOC concentrations in rainwater ranging from approximately 8 μM in mid-Wales (Wilkinson et al., 1997) to 50 μM in Greece (Pantelaki et al., 2018) and up to120 μM in the coastal United States (Willey et al., 2000). When DOC concentrations are plotted against the number of days since rainfall last occurred, it can be seen that concentrations are high within the first two days then stabilise at around 300 μM or lower at all sites (Fig. 11). This is further evidence that a major source of DOC to these rivers during rainfall events was the flushing of superficial soils. This is commonly Table 2 Annual yields of dissolve inorganic nitrogen (DIN), dissolved organic nitrogen (DON), and dissolved organic carbon (DOC) from study sites and a selection of streams and rivers across the UK, Europe, and the world. Units are kg C or N km −2 y −1 .  seen in wetland systems rich in organic soils (Inamdar and Mitchell, 2007;Worrall et al., 2012) but has also been observed in more agricultural catchments to the one studied here (Royer and David, 2005;Morel et al., 2009). Within the estuary there is some evidence of DOC production during the estuarine transect sampling, especially on 10th July 2014 where one mid-estuary sample stands out as higher than the surrounding samples. The chlorophyll a concentration on the same day at this site was 93 μg L −1 and a dinoflagellate bloom was later confirmed using inverted microscopy. These production processes appear to remain relatively localised within the estuary, however, as DOC concentrations both upstream and downstream of the site were over 100 μM lower. On the 7th August 2014 DOC concentrations were elevated throughout the estuary (Fig. 8a, green circles) but the lowest concentration was observed at mid-salinity. There was no associated increase in either chlorophyll a or river flow (data not shown). Ammonium concentration was also relatively high on this date, especially at salinities between 10 and 15 which correspond to the upper estuary sites on that date. Discharge from Holdenhurst sewage treatment works on the lower Stour (Fig. 1) could be the cause of this elevated DOC and ammonium before mixing with the higher salinity waters (e.g. Maier et al., 2012). There may be evidence of localised DOC production near the mouth of the estuary on this date also, as there is an increase in DOC at a salinity of 32. There are shallow sand banks between sites 1 and 2 at the mouth of the estuary. Sand is a highly permeable sediment and at depths of 1-2 m benthic production processes could be tightly coupled to the water column (Huettel et al., 2014). The source of this DOC peak could therefore be the sediments rather than water column processes with the incoming tide or waves driving pore-water exchange and flushing the DOC into the water column (Huettel et al., 2014).
The mean estuarine flushing time estimated for low summer flows is around 1.5 days which may be too short a period for significant uptake of DOC relative to the total concentration to occur within the estuary before the DOC is exported to the coastal zone. Annual DOC yield was at the higher end of ranges reported for global rivers and estuaries excluding the Nushagak River (Table 2), and is comparable to the range of export values estimated by Worrall et al. (2012) and Jarvie et al. (2017) for rivers in the United Kingdom based on land use characteristics and predominant soil types.

DOC: nitrate relationship
As the accumulation of nitrate in aquatic ecosystems has been identified as a major environmental concern, the importance of carbon as an essential nutrient coupled to the microbial processing of nitrate has become more evident (Taylor and Townsend, 2010). The molar ratio of DOC: nitrate is being increasingly used as an indicator for the potential fate of nitrate within a system (e.g. Sandford et al., 2013;Wymore et al., 2016;Heppell et al., 2017), with heterotrophic nitrogen assimilation proposed to be carbon-limited at a DOC: nitrate threshold ratio of 3.5 across all systems (Taylor and Townsend, 2010). During the weekly sampling the DOC: nitrate ratio was b2.5 at all times and at all sites, and the ratio only increased above 3.5 in the estuarine transects on 2 sampling dates (25th June 2014 and 7th August 2014). This would imply that in-stream processing of nitrate by heterotrophic organisms is carbon limited throughout the majority of the year, and is further evidence that control of anthropogenic nitrogen inputs to this system is needed. Interestingly the highest DOC: nitrate ratios during the weekly sampling were observed during the high flow events over the winter period, implying that the increased DOC delivery to the river during rainfall events may have the potential to relieve some of the carbon limitation if nitrate concentrations were decreased. Localised areas of DOC production within the estuary in summer months, such as algal blooms, are enough to raise the ratio and may result in areas of nitrate drawdown.

DON dynamics and sources
In contrast to the DOC dynamics, DON concentrations showed greater variability between sites with a significant relationship observed between the two rivers but no relationship between either of the rivers and estuarine DON concentrations. Concentration-discharge plots (Fig. 4) for each river are suggestive of a point source, such as a sewage treatment works, acting as the major source of DON in the rivers with concentration generally decreasing with increased flow. As flow in each river increases past~40 m 3 s −1 , however, concentrations increase, and this was particularly obvious in the Avon. This may reflect increased transport of DON from 'new' sources such as septic tanks connected via localised flooding. The highest concentration of DON in the Stour occurred at low river flow rates in early winter between the two high flow periods and at a time when relatively low water temperatures (7.4°C) would be expected to limit biological production processes within the river. This is further evidence for an external source contributing to DON concentrations within the Stour.
DON concentrations were lower in the estuary over the dry summer months during the weekly sampling than in either river, reflecting the mixing of higher salinity low nutrient waters but also perhaps some DON removal in the lower reaches of the rivers and the estuary. The concentration range for DON in the estuary was higher during the summer boat transect work than during the weekly sampling campaign (Fig. 8b, coloured circles), and whilst the overall pattern appears to be one of decreasing concentration as salinity increases, there are certain dates that show interesting patterns. For example, the samples on 12th June and 25th June 2014 (Fig. 8b, white and red circles) both appear to have higher DON concentrations in the upper estuary at the 3 sites in lower salinity waters than would be expected in a simple conservative mixing relationship. This would imply that there is removal of DON within the estuary between sampling sites 4 and 3 (Fig. 1). There are large areas of sand banks between these two sites and at the time combined river flows were relatively low at around 20 to 25 m 3 s −1 which would result in an increase in the estuarine flushing time. The estuary is shallow and so, in addition to water column processes such as phytoplankton uptake, bacterial respiration, and photooxidation (Seitzinger and Sanders, 1997;Wiegner et al., 2006;Badr et al., 2008), it is possible that these sandy sediments were acting as a sink for DON. Hopkinson et al. (1999) observed sediments, including sands, acting as a sink for DON in an estuary in Massachusetts with a degree of both temporal and spatial heterogeneity across a seasonal cycle, and Agedah et al. (2009) also report uptake of DON by sediments in the anthropogenically-impacted Colne estuary (UK). The rates reported by Agedah et al. (2009), however, were slow in relation to the residence time of the estuary and thus they proposed that the majority of DON in the system was exported to the coastal zone. In contrast Burdige and Zheng (1998) reported estuarine sediments in Chesapeake Bay to act as a source of DON to the water column. As DON concentrations were substantially lower than DOC or DIN concentrations throughout this study the impact of any potential removal or production process is more likely to be seen reflected in the total DON concentration at the estuary mouth. Unfortunately the data presented here is not sufficient to fully resolve the role of the sediments in the processing of DON in Christchurch Harbour, but further work is ongoing on sediment-water exchanges in this system.
A key feature of the two rivers studied was the very high concentrations of nitrate plus nitrite (N290 μM) throughout the year. As the concentration of DON is derived by subtracting the DIN concentration from the corresponding total dissolved nitrogen concentration, it is possible to get values that are negative or below the limit of detection when DON concentrations are low relative to DIN. Vandenbruwane et al. (2007) found the likelihood of this occurring to be of particular importance in samples where the proportion of DON to TDN is ≤15%. The mean DON: TDN ratios in both rivers and the estuary were below 10% which may explain in part the number of DON samples that were found to be below the limits of detection in this study. This in turn may explain the lack of relationship observed between DON concentrations and flow. Yates et al. (2019) determined the proportion of DON to TDN in intensively farmed arable catchments underlain by chalk, such as our Avon catchment in particular, to be b10%. However, the magnitude of the river flows during rainfall events still result in thousands of kilograms of nitrogen being transported to the estuary and beyond on a daily basis as DON. Whilst this flux was an order of magnitude lower than the flux of DIN in this system it is still a considerable yield of nitrogen when compared to many other global estuaries (Table 2).
There was no clear relationship found between DOC and DON at any of the sites or across all of the sites. The potential reasons for this are twofoldeither the factors controlling DOC and DON dynamics in these systems are different, or the components of the DOM pool are utilised by the microbial community at different rates (Wiegner et al., 2006(Wiegner et al., , 2009. Several studies have shown that the dynamics of DON and DOC within the same river can differ with DON being cycled faster within rivers than DOC (Stepanauskas et al., 2000;Solinger et al., 2001;Wiegner et al., 2006;Inamdar and Mitchell, 2007). In addition the ultimate fate of the organic carbon and nitrogen can differ as it can be either incorporated into bacterial biomass or oxidised and excreted (Hopkinson et al., 1999).

The impact of rainfall events
The resolution of sampling in this study was not high enough to fully resolve the behaviour of DOC or DON within the Rivers Avon or Stour or within Christchurch Harbour during storms using hysteresis curves (e.g. Lloyd et al., 2016), but weekly sampling over the course of a year has demonstrated that local increases in rainfall result in sudden increases in both river flow and riverine DOC concentration and thus flux. Elevated river flows also increase the DIN flux to the coastal zone (Supplemental Fig. 2). DON dynamics are more complicated to resolve possibly due to the interplay of different sources as well as currently unidentified removal processes within the estuary. Whilst the river flows observed during the year of study could be considered atypical in comparison to the 11-year mean flows for each river, they are certainly relevant when considering the estimated fluxes of each river under future climate change conditions of drier summers and more frequent stochastic storm events (IPCC, 2014).

Conclusion
There are few studies detailing the yields of dissolved organic nutrients in rivers and estuaries that are already known to be anthropogenically-impacted with elevated inorganic nutrient loads, despite their potential contribution to causing problems such as eutrophication and hypoxia. Annual yields of dissolved inorganic and organic nutrients at the lowest gauging stations of the Hampshire Avon and the Stour, as well as in their shared estuary of Christchurch Harbour, are comparable to and often greater than yields documented from other riverine and estuarine systems both within the UK and globally. Whilst the yield of DON was typically an order of magnitude smaller than the corresponding yield of nitrate plus nitrite at each site, the range of 118-198 kg N km −2 y −1 is still an ecologically important load of nitrogen potentially available to the aquatic microbial community. The processes controlling the dynamics of DOC and DON differed in both the rivers and the estuary, highlighting the importance of considering the component parts of DOM when investigating the role of DOM in aquatic systems.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.