Nitrogen transformations along a shallow subterranean estuary

The transformations of chemical constituents in subterranean estuaries (STEs) control the delivery of nutrient loads from coastal aquifers to the ocean. It is important to determine the processes and sources that alter nutrient concentrations at a local scale in order to estimate accurate regional and global nutrient fluxes via submarine groundwater discharge (SGD), particularly in boreal environments, where data are still very scarce. Here, the biogeochemical transformations of nitrogen (N) species were examined within the STE of a boreal microtidal sandy beach located in the Magdalen Islands (Quebec, Canada). This study revealed the vertical and horizontal distribution of nitrate (NO−3 ), nitrite (NO−2 ), ammonia (NH + 4 ), dissolved organic nitrogen (DON) and total dissolved nitrogen (TDN) measured in beach groundwater during four spring seasons (June 2011, 2012, 2013 and 2015) when aquifer recharge was maximal after snowmelt. Inland groundwater supplied high concentrations of NOx and DON to the STE, whereas inputs from seawater infiltration were very limited. Non-conservative behaviour was observed along the groundwater flow path, leading to low NOx and high NH+4 concentrations in the discharge zone. The long transit time of groundwater within the beach (∼ 166 days), coupled with oxygen-depleted conditions and high carbon concentrations, created a favourable environment for N transformations such as heterotrophic and autotrophic denitrification and ammonium production. Biogeochemical pathways led to a shift in nitrogen species along the flow path from NOx-rich to NOx-poor groundwater. An estimate of SGD fluxes of N was determined to account for biogeochemical transformations within the STE based on a N-species inventory and Darcy’s flow. Fresh inland groundwater delivered 37 mol NOx yr−1 per metre of shoreline and 63 mol DON m−1 yr−1 to the STE, and NH+4 input was negligible. Near the discharge zone, the potential export of N species was estimated around 140, 1.5 and 33 mol yr−1 per metre of shoreline for NH+4 , NOx and DON respectively. In contrast to the fresh inland groundwater, the N load of beach groundwater near the discharge zone was dominated by NH+4 and DON. Our study shows the importance of tidal sands in the biogeochemical transformation of the terrestrial N pool. This local export of bioavailable N probably supports benthic production and higher trophic levels leading to its rapid transformation in surface sediments and coastal waters.

the STE appears to be a sink of terrestrially-derived N. The net transformations of N in the STE led to N removal along the groundwater flow path along the groundwater flow path.

Introduction
Land-ocean interfaces are critical transition zones that may affect the ecology and quality of coastal ecosystems (Schlacher and Connolly, 2009). Chemical constituents in submarine groundwater discharge (SGD) are now widely recognized to have 5 a significant impact on coastal ecosystems (Knee and Jordan, 2013;McCoy and Corbett, 2009;Null et al., 2012;Slomp and Van Cappellen, 2004). SGD is conventionally defined as "any flow of water out across the seafloor without regards to its composition and its origin" (Burnett et al., 2006). Thus, before entering coastal waters, fresh groundwater travels through the shallow subterranean estuary (STE) (Moore, 1999), a region where mixing between fresh and marine groundwater promotes biogeochemical processes that can lead to rapid changes in nutrient concentrations and induce non-conservative input or 10 removal (Gonneea and Charette, 2014). The STE supports extensive chemical reactions near the discharge interface and is often assumed to be a non-steady-state system (Kroeger and Charette, 2008). Continental factors (e.g., local hydrogeology, recharge, precipitation) as well as marine factors (e.g., tidal and wave pumping, hydrography, and density) induce temporal and spatial variability in biogeochemical conditions (see Santos et al. 2012 and references therein). The mixing zone is subject to oscillating conditions, with rapid changes in oxygen saturation, redox potential, and organic matter input 15 controlled by tidal stage and amplitude as well as seasonal water-table fluctuations (Abarca et al., 2013;Heiss et al., 2014;Robinson et al., 2014). These physical processes are likely to impact the distribution and biogeochemical reactivity of many dissolved constituents (Beck et al., 2007;Kroeger and Charette, 2008). In this context, the STE can either be a source of nutrients or act as a barrier and limit nutrient discharge to coastal environments. Assessing the role of the STE in nutrient transformations is crucial to better quantifying global chemical fluxes via SGD (Moore, 2010). 20 Rivers have long been considered the main conveyors of N to the ocean (Seitzinger et al., 2005, and references therein). Beusen et al. (2013) recently provided evidence that SGD also plays an important role in regional and global marine N cycles. N loads from SGD to near-shore ecosystems were estimated to be 4 Tg N y -1 (Voss et al., 2013), and the role of SGD in coastal eutrophication has also been demonstrated (Valiela et al., 1990). These N loads may be an important factor in the Biogeosciences Discuss., doi: 10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License. development of harmful algal blooms in coastal waters (Anderson et al., 2008;Glibert et al., 2014). Fresh groundwater is often rich in nutrients and others materials from anthropogenic inputs due to coastal development (agriculture, urbanization) (Howarth and Marino, 2006;Null et al., 2012;Rocha et al., 2015), and models predict a 20% increase in N loads from SGD within the next few decades due to coastal development (Beusen et al., 2013).
Estimates of SGD nutrient loads to the coastal ocean have often been based on nutrient concentrations in fresh groundwater, 5 with the assumption that nutrient transport through the STE is conservative (Burnett et al., 2006). However, numerous studies have demonstrated that concentrations of dissolved N change throughout the STE because of biological and chemical reactions (Beck et al., 2007;Loveless and Oldham, 2010;Moore, 2010;Robinson et al., 2007b;Santos et al., 2009).
Variations in oxygen and organic matter input along the hydraulic gradient lead to a combination of heterotrophic processes that can enhance or attenuate the export of N to the coastal ocean (Santoro, 2010). For example, in the Gulf of Mexico 10 (Turkey Point,Florida), the STE acts as a source of ammonium because remineralization of marine organic matter through the STE provides nutrients to the SGD exported to the embayment . In Waquoit Bay (Cape Cod, Massachusetts), Kroeger and Charette (2008) demonstrated that ammonium accumulates in the STE because remineralization of organic matter transported by marine and fresh groundwater outpaces nitrification. In contrast, based on the N attenuation observed in a shallow STE due to denitrification processes (Cokburn Sound, Australia), Loveless and 15 Oldham (2010) calculated nitrate loads to coastal waters that were 1-2 times lower than previous estimates based on nutrient concentrations from fresh groundwater. As these studies show, ignoring non-conservative mixing can lead to an over-or under-estimation of nutrient loads to coastal waters (Beck et al., 2007).
STEs are transient systems where steady state, and thus the sequence of redox reactions, are rarely achieved (Sundby, 2006).
In transient systems, diagenetic reactions reflect redox oscillations and environmental conditions far from steady state. 20 Redox oscillations, with alternating oxic and anoxic conditions in sediments, allow coupled nitrification-denitrification to take place in the same location within the sediment (Aller, 1994). Alternative pathways of nitrate reduction, such as dissimilatory nitrate reduction to ammonium (DNRA) and ANAMMOX, have also been reported in the STE (Erler et al., 2014;Kroeger and Charette, 2008;Rocha et al., 2009). Many of these processes transform dissolved inorganic nitrogen Biogeosciences Discuss., doi: 10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License.
(DIN) and dissolved organic nitrogen (DON) along the groundwater flow path. This N can then be exported to the coastal ocean or removed by denitrification.
Martinique Beach, located in the Îles-de-la-Madeleine (Québec, Canada) in the southern limit of the boreal climatic zone, contains a boreal STE that is exposed to little or no external contamination. Because climate and hydrology change rapidly in this environment, the role of boreal STEs remains to be accurately elucidated (Hinzman et al., 2005), and studies in these 5 cold environments are scarce. The objective of this four-year study was to evaluate the sources and pathways of N transformations in a microtidal STE that modify the groundwater dissolved N pool, including inorganic (nitrate, nitrite, ammonium) as well as organic (DON) forms of N. DON is generally assumed to be from natural rather than anthropogenic sources and is often neglected (Hansell and Carlson, 2014). Nevertheless, DON concentrations can be high in SGD and should be considered (Kroeger et al., 2007;Santos et al., 2014). In addition, to evaluate the role of the beach groundwater as 10 a N source or sink, SGD fluxes of N species were estimated to evaluate the potential impact of this boreal STE on the local coastal embayment.

Study area
Martinique Beach is located on the main island of the Îles-de-la-Madeleine archipelago in the Gulf of St. Lawrence (Québec,15 Canada; Fig.1). The Martinique Beach system originates from a recent transgression sequence. Rapid rates of sea-level rise along the Atlantic coast of Canada over the middle to late Holocene buried the unconfined Permian sandstone aquifer that is now covered by tidal sediment (Gehrels, 1994;Scott et al., 1995aScott et al., , 1995b. The site undergoes semi-diurnal tides with a mean range of 0.8 m and a maximum range of 1.7 m during spring tide. The archipelago has no rivers, thus the aquifer recharge is only from rain and snow, with the highest recharge during spring snowmelt. The mean yearly recharge is about 230 mm 20 (Madelin'Eau, 2004). Because groundwater constitutes the only source of drinking water in the archipelago, the hydrogeology is well known and the aquifer constantly monitored (Chaillou et al., 2012;Madelin'Eau, 2007, 2009, 2011. Since anthropogenic pressures like urbanization and agriculture are limited on the archipelago, the main sources of N Biogeosciences Discuss., doi: 10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License. contamination are from residential and recreational areas. Therefore, Martinique Beach is an ideal system in which to study N transformations in a boreal microtidal subterranean estuary.
The Martinique Beach STE acts as a shallow unconfined aquifer at the nearshore limit of the Permian Aquifer; it releases both fresh and recirculated saline groundwater to the coastal embayment . It is a low-energy beach under a micro-tidal regime (Jackson et al., 2002;Masselink and Short, 1993). The upper meters of the beach consist of 5 marine sands with a median particle size of 0.30 mm (silt content <5%), mainly composed of quartz (95%). The hydraulic conductivity of this sedimentary unit is about 11.40 ± 4.40 m d -1 . Lower hydraulic conductivity was measured in the underlying sandstone aquifer (K ~1.80 m d -1 ; Madelin' Eau, 2007), which is composed of fine silicate and aluminosilicate sands with Fe-coated silicate grains (Chaillou et al., 2014). These two layers are organic-poor (total organic carbon [TOC] < 0.20% weight percent (w.t.) and total nitrogen [TN] < 0.10% w.t.; Chaillou et al., 2014). In the landward 10 part of the beach, however, an old-age soil horizon dated to ~900 B.P. ( 14 C dating; Juneau, 2012) occurs a few centimeters below the beach surface. This horizon is carbon-rich (TOC > 20% w.t.) but has a low nitrogen content (TN < 0.50% w.t.; Chaillou et al., 2014). Based on Darcy's law, Chaillou et al. (2016) estimated a fresh inland groundwater flow rate (Qinland) of 1.75x10 -5 m 3 m -1 s -1 at the landward boundary of Martinique Beach. The fresh groundwater flow rate through the beach face (Qbeach), from the landward part to the intertidal zone were also estimated assuming an isotropic system in which vertical and 15 horizontal flows are uniform. This mean Qbeach is 2.41x10 -5 m 3 m -1 s -1 and represents about 25% of the total volumetric SGD flux (fresh + marine) to the coastal waters .

Groundwater sampling
Sampling was carried out in June 2011June , 2012June , 2013, and 2015 along a 50 m cross-shore transect. In 2011 and 2012, groundwater samples were collected in the landward part of the STE. In 2013 and 2015, we focused on the intertidal and 20 discharge zone, where fresh meteoric groundwater comes in contact with recirculated seawater. Groundwater extraction was done using multi-level samplers in 2.5 m long PVC pipes (Fig. 1), similar to those described by Martin et al. (2003).
Groundwater was collected at 10,30,50,80,100,150,190, and 230 cm below the beach surface. Samplers were re-inserted at the same locations each year using DGPS coordinates. To allow sediments around the samplers to reach equilibrium, sampling started two days after their insertion. Groundwater was pumped using a peristaltic pump, and physicochemical 25 Biogeosciences Discuss., doi: 10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License. parameters (pH, temperature, oxygen, salinity) were measured directly using an on-line flow cell with a calibrated multiparameter probe (600QS, YSI Inc.). Oxygen measurements are not available for 2015, due to sensor malfunction. After stabilization of physico-chemical parameters, all groundwater samples were filtered through a 0.2 µm polypropylene capsule filter. Samples for nutrient analyses (NH4 + , NO3 -, and NO2 -) were stored in acid-washed polyethylene tubes that were rapidly frozen for later analysis; samples for total dissolved iron and manganese were stored at 4°C in 50 mL acid-washed 5 polyethylene tubes and acidified with 50 µL of 10% nitric acid; and samples for dissolved organic carbon (DOC) and total dissolved nitrogen (TDN) were stored in baked 7 mL vials and acidified with 25 µL of high purity 10% HCl. TDN measurements were only performed in 2012. Groundwater end-member samples (n=10) were collected in the manner described above from private and municipal wells located 50 to 2000 m landward of the most inland sampler. Seawater endmember samples (n=6) were collected about 50 cm above the seabed using a submersible pump at about 900 m offshore in 10 Martinique Bay.

Chemical analyses
NH4 + samples were measured by flow injection gas exchange-conductivity analysis based on the method described by Hall and Aller (1992). The precision was ± 5% with a detection limit of 0.1 µmol L -1 . NO3and NO2 -, referred to as NOx, were analyzed by the colorimetric method developed by Schnetger and Lehners (2014) and measured with a powerwave XS2 15 microplate spectrophotometer. The precision was 2% and the limit of detection was 0.4 µmol L -1 . DIN was calculated as the addition of NH4 + , NO3 -, and NO2 -. TDN was analyzed in 2012 by high temperature combustion (HTC) using a Total Organic Carbon analyzer (TOC-vpn, Shimadzu) with a TNM-1 module, and a precision of 2%. Dissolved Organic Nitrogen (DON) was calculated as the difference between TDN and DIN (i.e., DON = TDN -[NH4 + + NOx]). DON calculations were only possible in 2012 based on TDN measurements. The DON measurement is still problematic since it combines the analytical 20 errors and uncertainties of the three analyses. Nevertheless, there is currently no single accepted method for the measurement of DON (Hansell and Carlson, 2014). Here we estimate the precision to be around 10%.

Distribution of salinity and oxygen saturation
Previous studies have already discussed the distribution of physico-chemical parameters along the groundwater flow path at Martinique Beach based on 2012 and 2013 data (Chaillou et al., 2014Couturier et al., 2016). Here, we will briefly present an overview of the salinity and redox conditions in the STE (

Nutrient distribution along the STE
In the fresh groundwater end-member, NOx (SNO3 -+NO2 -) concentrations were elevated (65.52 ± 26.70 µmol L -1 ; Table 1). NH4 + concentrations were low, with concentrations below 1 µmol L -1 . The groundwater end-member was rich in TDN, with DON making up 53% of the TDN (i.e., DON = 110.96 ± 3.42 µmol L -1 ). Compared to the fresh groundwater end-member, 20 the seawater end-member was largely depleted in NOx (0.52 ± 0.55 µmol L -1 ), and NH4 + concentrations were also low (0.87 ± 0.50 µmol L -1 ). DON was still the main nitrogen species (7.30 ± 0.84 µmol L -1 , ~80% of TDN). Overall, TDN concentrations in the seawater end-member were 20 times lower than in the groundwater end-member. These results are summarized in Table 1 µmol L -1 ), and were mainly observed in the saline circulation cell. In contrast to NOx, NH4 + concentrations increased in the 5 STE, from ~20 µmol L -1 to > 500 µmol L -1 (Fig. 3B), up to 1056 µmol L -1 . Ammonium (NH4 + ) concentrations measured in the STE were 1 to 1000 times higher than end-member values (Table 1, Fig. 3b). In 2013, an area of high concentrations was observed in the mixing zone, in front of the saline circulation cell, where NH4 + concentrations reached values greater than 400 µmol L -1 (Fig. 3B). NH4 + concentrations were still high in the saline circulation cell (e.g., 84-92 µmol L -1 ), and these were also high compared to the overlying seawater end-member (Table 1). NH4 + concentrations decreased sharply with 10 depth below the saline circulation cell and below the mixing zone. For example, NH4+ concentrations were around 150 µmol L -1 at 30 cm below the beach surface and decreased to 50 µmol L-1 at 230 cm (Fig. 3B). NH4 + was the main TDN species in the STE (NH4 + represents on average 60% of TDN in all samples). Thus, the TDN distribution was quite similar to the NH4 + distribution (Fig. 3B), with high values in the mixing zone. These distributions were consistent from year to year. TDN decreased sharply below the mixing zone and the saline circulation cell; values ranged from 50 to 100 µmol L -1 and 15 dropped below detection deeper in the saline circulation cell. DON represented 31% of the TDN in the beach groundwater, and the highest concentrations were observed in the mixing zone (>200 µmol L -1 , Fig. 3D). DON levels decreased below the saline circulation cell, with concentrations close to 0. N species showed different distributions relative to the groundwater salinity and DO saturation along the STE (Fig. 4). N species were characterized by non-conservative behaviour relative to the theoretical two-end-member mixing between 20 seawater and fresh groundwater. NOx declined from 60 µmol L -1 in fresh inland groundwater to concentrations below detection in saline groundwater (S > 15) (Fig. 4A). The highest concentrations of NOx were encountered when DO saturation was below 60%. While dissolved NOx showed removal in the flow path, NH4 + exhibited excess concentrations relative to conservative mixing between the two end-members (Fig. 4B). NH4 + concentrations clearly showed strong production. The highest concentrations of NH4 + occurred mainly under suboxic conditions (DO < 20%) and decreased significantly with 25 Biogeosciences Discuss., doi: 10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License.
increased DO (p value < 0.05). Both NH4 + and NOx were observed in 81 of 245 samples (~33% of the data set). These samples were mainly from just below the saline circulation cell and the associated mixing zone, where oxygen-depleted conditions prevailed ([DO] < 20%). Nevertheless, the coexistence of NH4 + and NOx was observed in 15% of samples.
In contrast to the behavior of NOx and NH4 + , TDN and DON exhibited two distinct trends along the salinity gradient: (i) they fell below the theoretical mixing line in fresh and brackish waters (salinity 0 -10) and this removal occurred in suboxic-5 anoxic conditions, and (ii) their concentrations increased above the theoretical mixing line in saline waters (> 10).

Steady or unsteady biogeochemical conditions?
Throughout the study period, the saline circulation cell was small both in length (up to 10 m) and depth (up to 60 cm below the beach surface), while fresh groundwater was present in the landward part of the STE and below the saline circulation 10 cell. After the snow melt, the water table is high in the Permian sandstone aquifer (Madelin'Eau, 2004) and in the adjacent beach aquifer . Under these hydroclimatic conditions, the saline recirculation cell and its associated mixing zone are spatially limited, and the inland hydraulic gradient is the main control of total SGD (Heiss and Michael, 2014;Robinson et al., 2007a).
Despite this apparent hydrogeological stability, one cannot consider flows and biogeochemical conditions to be at steady 15 state in the STE. Tidal cycles and waves induce variability in the groundwater inflow while changing discharge fluxes alter the mixing between fresh and saline groundwater (Abarca et al., 2013). The redox oscillation induced by the input of oxygen-rich seawater advected by tides (and waves) in the swash zone impacts the transformations of chemical constituents including nitrogen (Charbonnier et al., 2013;Huettel et al., 2003).

Biogeochemical controls of DIN concentrations along the groundwater flow path 20
The non-conservative behaviour of DIN along the groundwater flow path influences the nutrient concentration in discharging groundwater, while at the same time making it difficult to estimate the flux of groundwater-derived DIN to the coastal ocean (Johannes, 1980;Moore, 2010;Valiela et al., 1990). The calculation of chemical fluxes using samples from inland wells may result in significant errors in estimated chemical fluxes. Processes occurring in the STE must be elucidated Biogeosciences Discuss., doi: 10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License.
to improve our understanding of the STE's role in altering groundwater-derived N. The DIN pool changes from NOx-rich groundwater in the aquifer to NH4+-rich groundwater in the STE. In our study, NOx represented 99% of the DIN pool in the groundwater end-member and dropped to 37% in the seawater end-member. In the next section, the biogeochemical mechanisms controlling the N pool along the flow line are explored.

Nitrate loss along the STE
NOx concentrations were low within the STE despite high inputs from fresh inland groundwater transport to the STE. There was a rapid and strong attenuation in NOx levels, with mean concentrations of 60 µmol L -1 in inland wells dropping to 1.90 µmol L -1 in the STE.
Under oxygen-depleted conditions, denitrification may be the major process responsible for rapid NOx loss. Electrons needed 10 for denitrification originate from the microbial oxidation of organic carbon when the amount of organic carbon present as DOC in groundwater is not too low (Cannavo et al., 2004). DOC levels were high in Martinique Beach groundwater, with concentrations in the range of 0.14-4.68 mmol L -1 in the STE (mean 1.94 mmol L -1 , Couturier et al., 2016). The high DOC levels and the oxygen-depleted conditions may support heterotrophic denitrification. The stoichiometry of nitrate reduction and the oxidation of organic matter by denitrification, given by Jørgensen et al. (2014), is as follows: 15 (1) 5CH2O + 4NO3 -+4H + = 2N2 + 5CO2 + 7H2O According to this stoichiometry, the mean concentration of DOC observed in the STE (i.e., 1.94 mmol C L -1 ; Couturier et al., 2016) could be used to reduce 1.55 mmol L -1 of nitrate to dinitrogen by denitrification. With concentrations around 0.19 mmol L -1 in inland wells, this means that all groundwater-borne NO3may conceivably be reduced by DOC. The use of DOC as an electron donor for denitrification depends on its bioavailability, but little is known about the role of OC origin on 20 denitrification. Organic carbon in aquifers mainly comes from soil leaching, and its bioavailability is often described as limited (Korom, 1992). Couturier et al. (2016) showed that DOM had a strong terrestrial signature along the STE in Martinique Beach. This OC was characterized by a high molecular weight and was enriched in lignin-derived compounds. In an alluvial aquifer, Baker and Vervier (2004) confirmed that the rate of denitrification was best predicted by the concentration of low molecular weight organic acids compared to high molecular weight compounds. In an unconfined 25 Biogeosciences Discuss., doi: 10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License. sandy aquifer, Postma et al. (1991) reported that nitrate reduction was minimal when OC was present as lignin and lignite fragments (i.e., as high molecular weight compounds). Thus, the terrestrial DOC present in the Martinique Beach STE may not promote high rates of denitrification in the study site.
Although denitrification via heterotrophic and autotrophic processes could explain the complete depletion of NOx along the flow path, the occurrence of high NH4 + concentrations also suggests the development of dissimilatory nitrate reduction to ammonium. DNRA occurs under strictly anaerobic conditions. The partitioning of nitrate between denitrification and DNRA 15 is related to the availability of organic carbon and NOx (i.e., Corg : NOx ratio) (Postma et al., 1991). DNRA is favoured when the electron acceptor (NOx) becomes limiting (i.e., high Corg : NOx ratio) and under reducing conditions, while under high NOx availability and electron donor limitation (i.e., low Corg : NOx ratio), denitrification is the thermodynamically favoured pathway (Kelso et al., 1997;Korom, 1992;Strohm et al., 2007). In coastal sediments, DNRA is a key process in the benthic N cycle and can be more important than denitrification (Gardner et al., 2006;Song et al., 2013). In the Martinique Beach 20 STE, the high DOC level and low NOx concentrations (i.e., Corg : NOx ratio > 400) suggest that DNRA could be a key pathway consuming groundwater-borne NOx. Based on the following chemo-organoheterotrophic DNRA pathway (Megonigal et al., 2004): (4) 2H + + NO3 -+ 2CH2O = NH4 + + 2CO2 + H2O The reduction of one mole of NO3produces one mole of NH4 + . This indicates that the NOx supplied by groundwater is not 25 Biogeosciences Discuss., doi: 10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License. sufficient to explain the strong concentrations of NH4 + measured along the transect. The same is true for chemolithoautotrophic DNRA, which links the reduction of NOx to the oxidation of inorganic electron donors like sulfide (An and Gardner, 2002;Brunet and Garcia-Gil, 1996;Sayama, 2001) and Fe 2+ (Hou et al., 2012;Roberts et al., 2014;Weber et al., 2006), and is also thermodynamically favourable under the conditions encountered in the STE. Whatever the electron donor, NH4 + production by DNRA is limited by the NOx supply to the STE. To produce such high concentrations of NH4 + , other 5 sources of N must be invoked. Further studies should be performed in the STE of Martinique Beach to quantify the different pathways that control NOx loss and NH4 + production.

Ammonium production along the STE
Mineralization of organic matter is likely the most important source of NH4 + in the Martinique Beach STE. DON measurements in 2012 were high (0-1481.84 µmol L -1 ), with a mean value of 80.29 µmol L -1 . DON is a complex mixture of 10 primarily uncharacterized compounds, of which 10 to 70% are estimated to be bioavailable (Seitzinger et al., 2002). DON bioavailability is often reported to be dependent on the nature of compounds (Sipler and Bronk, 2014). In the beach groundwater, DON represented 39% of the TDN, so its mineralization by heterotrophic micro-organisms could be responsible for part of the NH4 + production in the STE (Kroeger et al., 2006). Because ammonification is highly dependent on the bioavailability of DON, it is difficult to estimate what fraction of NH4 + could be derived from DON mineralization. 15 Based on the estimation that 10 to 70% of DON is bioavailable, mineralization of DON could lead to the production of 8 to 56 µmol L -1 of NH4 + , which represents between 2 and 10% of the NH4 + concentration observed in beach groundwater.
In coastal sediments, where sulfate is not limiting, sulfate reduction may also produce NH4 + according to the following reaction: (5) 53SO4 2-+ (CH2O)106(NH3)16(H3PO4) = 39CO2 + 67HCO3 -+ 16 NH4 + +53HS -+ 39H2O + HPO4 2-20 This reaction is thought to be of minor importance in freshwater compared to fermentative mechanisms. However, because of the proximity of seawater, we cannot exclude the possibility of SO42-traces far below the surface. NH4 + was observed in samples with salinity > 4, with the highest concentrations (~1.25 mmol L -1 ) at salinity around 15. Based on typical SO4 2concentration in seawater with a salinity 15, we estimate a SO4 2concentration of 12 mmol L -1 in beach surface groundwater, which is sufficient to produce 3.6 mmol L -1 NH4 + by sulfate reduction. This reaction could therefore explain all NH4 + 25 Biogeosciences Discuss., doi:10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License. production in the beach groundwater.
The breakdown of macroalgal deposits derived from wave and tidal action in sediments can also increase N input to beach groundwater (Kelaher and Levinton, 2003;Rossi et al., 2011) and can potentially add NH4 + . At Martinique Beach, algal deposits were not specifically measured but were often observed after storm events. In addition, external contamination from wastewater cannot be completely excluded even if anthropogenic pressure is weak. Moreover, there is no trace of NH4 + 5 contamination in the landward part of the beach and no such evidence is found in inland private wells.

Hot spots of nitrate production along the STE
Hot spots of NOx concentrations (e.g., 7.5 µmol L -1 at 50 cm depth with [DO] < 10% in 2013, 15.21 µmol L -1 at 80 cm depth with [DO] < 30% in 2012) were likely the result of local and sporadic production rather than traces of groundwater-borne NOx, which was likely depleted landward of the most inland sampler. The downward transport of 10 oxygenated seawater over short temporal and spatial scales could be large enough to oxidize NH4 + and produce NO3along the saline circulation cell. Different alternative pathways that involve metal oxides and ammonium can be invoked to support local (and probably sporadic) NOx production. These pathways, which include both biotic and abiotic mechanisms, become important under non-steady-state conditions and are thermodynamically favourable at pH levels (6.5 < pH < 8, data not shown) encountered in the STE (Anschutz et al., 2005;Chaillou et al., 2007;Luther and Popp, 2002;Mortimer et al., 2004). 15 For example, field and experimental studies have indicated that anaerobic oxidation of NH4 + to NO3could be supported by Mn(III/IV) oxides (Hulth et al., 1999;Luther and Popp, 2002;Mortimer et al., 2004). The contribution of these processes to the nutrient load of SGD is largely unknown, although they may occur in the discharge zone (Kroeger and Charette, 2008).
With a maximum groundwater flow rate of 2.41 m 3 m -1 s -1 in beach sediments , the groundwater transit time through the intertidal zone (~25 m) is about 82 days, which is long enough to support tidally driven redox oscillations 20 and subsequent N transformations. The concentration of NO3remained weak (< 6 µmol L -1 ) in the STE probably because of the multiple electron donors that can be used to reduce NO3to N2 under anoxic conditions, i.e., DOC, Fe 2+ , NH4 + , H2S, and FeS.

Nutrients transport along the flow path
The non-conservative behaviour of nutrients within the STE makes it difficult to estimate the export of nutrients to the coastal ocean. As pointed out by the recent review of Moore (2010), robust measurements of nutrient fluxes are needed on a site-specific scale to obtain accurate regional and global estimates. In non-conservative systems, however, the determination of appropriate nutrient end-member concentrations for flux calculations is not straightforward. Beck et al. (2007) previously  5 highlighted the need to closely scrutinize the biogeochemical processes in the STE to refine nutrient export to coastal areas.
Here, inorganic and organic N inventories (in µmol) were explored along the groundwater flow path to the beach face. Based on these inventories, the nitrogen fluxes out of the STE were estimated and compared to the fresh groundwater-borne nutrient fluxes. Fluxes and inventories of the different N species along the groundwater flow path are summarized in Figure   5. 10

Nitrogen inventories
Nutrient inventories were calculated by integrating nutrient concentrations at three distinct sampling locations along the transect and multiplying by the sediment porosity (i.e., 0. 25;Chaillou et al. 2012). Nitrogen inventories were calculated onshore of the STE, and at both the high and low tide marks (Fig. 1C). Inorganic and organic nitrogen inventories are presented in Table 2. 15 In the inland groundwater, fresh groundwater was rich in DON (Table 1). DIN represented only 33% of the TDN, with NOx making up 95% of the inorganic pool (Table 1). In the onshore sample, DON was still the main N species with an inventory of 32.05 µmol (Table 2). However, a shift from NOx to NH4 + occurred from the inland groundwater to the onshore sampler.
NOx became a negligible fraction (NOx inventory = 0.029 µmol) whereas NH4 + was the main inorganic fraction (NH4 + inventory = 28.08 µmol). 20 The inland groundwater clearly acted as a source of nitrogen to the beach groundwater, as has been observed in other STEs, such as in Dor Bay (Mediterranean coast; Weinstein et al., 2011), Cockburn Sound (western Australia; Loveless and Oldham, 2010), and Waquoit Bay (Cape Cod, MA; Talbot et al., 2003;Gonneea and Charette, 2014). However, the groundwater-borne N load was very low in comparison to the above-mentioned sites, where fresh groundwater NOx concentrations as high as 300 µmol L -1 were reported. 25 Biogeosciences Discuss., doi: 10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License. NH4 + is clearly produced along the groundwater flow path from the aquifer to the STE. The TDN pool did not change significantly between fresh inland groundwater and beach groundwater (based on the addition of TDN inventories), At the high tide mark, a strong production of TDN is observed: DIN, NH4 + , and DON increased by 169, 167, and 158%, respectively ( Figure 5). Nitrate hotspots were also observed, as shown in Figure 3A. NH4 + production is larger than the groundwater-borne NOx depletion, and thus could not be supported only by nitrate reduction. The high DON concentrations 5 and the tidal input of oxygen in this zone favored the establishment of oxic / anoxic reactions of mineralization that concomitantly produced high NH4 + and, to a lesser extent, NOx. Based on the previous work of Couturier et al. (2016), the nitrogen source was probably the mineralization of terrestrial organic matter rather than marine organic matter. This zone of strong in situ TDN production, at the high tide mark, altered the groundwater-borne N pool which is discharged to the coastal ocean, by enhancing both DIN and DON. At the low tide mark, less than 15 m further, the N pool decreased to reach 10 similar inventories observed at the onshore sampler. The intertidal zone therefore acts as a filter for in situ N production, whereby beach groundwater undergoes ammonification, nitrification, nitrate reduction, and dilution with low nutrient seawater that is tidally advected into the beach face, and loss by SGD to Martinique Bay. The low nutrient load and rapid recirculation of seawater in this zone limit the establishment of conditions favorable for denitrification, and could favor the export of TDN produced in the STE to the coastal ocean. Our findings show that groundwater-borne TDN, in the form of 15 NO3and DON, is mostly attenuated along the groundwater flow path, but a "new" N pool is produced within the STE.
Nevertheless, if the TDN produced in the STE is released to coastal waters, it can be expected to exacerbate coastal eutrophication.

Nitrogen fluxes delivery to coastal water
At Martinique Beach, fresh inland groundwater is a pathway for DON and DIN produced along the flow path to reach the 20 coastal ocean. Integrating this in situ production is critical to accurately estimate the impact that coastal boreal systems have on regional and global nutrient budgets. The fresh groundwater-borne N fluxes have been calculated as the product of the DON, NOx and NH4 + concentrations of the fresh inland groundwater end-member and the rate of fresh groundwater discharge (Qinland = 1.75x10 -5 m 3 m -1 s -1 ) estimated by Chaillou et al. (2016) at the same site. Based on mean concentrations values (Table 1), the flux of fresh groundwater-derived N was 0.07 mol NH4 + m -1 y -1 , 37.54 mol NOx m -1 y -1 , a 25 Biogeosciences Discuss., doi: 10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License.
corresponding DIN flux of 37.64 mol m -1 y -1 , and 63.57 mol DON m -1 y -1 (Table 3). The estimated groundwater-borne TDN flux is around 102 mol m -1 y -1 , corresponding to an annual N input of ~1700 kg along the 1200 m of the Martinique Beach shoreline. This flux was dominated by DON, and NOx was the dominant inorganic N species. Figure 5 presents a summary of N fluxes along the flow path.
Using the N inventory at the high tide mark as the potential beach groundwater end-member, and assuming that all the in situ 5 production is flushed out of the system by the continental hydraulic gradient and tidal pumping, a maximal beach groundwater N flux has been estimated. Based on the mean Darcy flow measured in the beach (Qbeach = 2.41x10 -5 m 3 m -1 s -1 ; Chaillou et al., 2016), the exported N load thus calculated is 42.8, 1.1 and 43.38 mol m -1 y -1 for NH4 + , NOx, and DON respectively (Table 3). DIN exported to the coastal ocean (43.9 mol m -1 y -1 ) is in the lower range of previous measurements at other sites, such as the Mediterranean coast (France; 530 mol m -1 y -1 ; Weinstein et al., 2011), the Gulf of Mexico (FL, 10 USA; 414 mol m -1 y -1 ; Santos et al., 2009), and the Atlantic Coast (Aquitania Coast, France; 150 mol m -1 y -1 ; Anschutz et al., 2016). It is noteworthy that fewer studies report NH4 + as the main N species exported to the coastal ocean compared to NOx, probably due to anthropogenic pressure. Kroeger et al. (2007) showed high proportions of NH4 + and DON in SGD fluxes to Tampa Bay (FL, USA) and concluded that anthropogenic N could also be exported under these forms. Nevertheless, measurements of DON flux to the coastal ocean are scarce. Kim et al. (2013) reported conservative mixing of DON, with 15 export fluxes of 1.31×10 5 mol d -1 in Hwasun Bay (Jeju Island, Korea) and in the Gulf of Mexico, Santos et al. (2009) estimated that land-derived DON makes up ~52% of the total SGD discharge.
Whatever N species is delivered to the coastal water, the maximum TDN flux at the discharge face of the beach was about 88 mol m -1 y -1 (~1500 kg N y -1 ), an input slightly lower than the fresh groundwater-borne N flux. Whereas the speciation of inorganic N changed during transit, from NOx-rich to NOx-poor groundwater, the production of NH4 + did not compensate for 20 the loss of DON (probably due to mineralization). The net effect of N transformations was to decrease the N load en route to the sea, which means that at Martinique Beach the STE is a sink for terrestrially-derived N as seen in other temperate and Mediterranean sites (Gonneea and Charette, 2014;Loveless and Oldham, 2010;Weinstein et al., 2011).
This input of fresh groundwater N is in the range of fluxes measured by Paytan et al (2006) in various coastal areas in world (i.e., 400-100 000 kg N y -1 ) and in agreement with the calculations recently proposed by Beusen et al. (2013) along boreal 25 Biogeosciences Discuss., doi:10.5194/bg-2016-535, 2016 Manuscript under review for journal Biogeosciences Published: 23 December 2016 c Author(s) 2016. CC-BY 3.0 License. and subarctic coasts (i.e., 1000 -10000 kg N y -1 ). However, the calculated N flux is negligible in comparison to the estimated 18.9 Tg of DIN delivered by world rivers to coastal seas (Seitzinger et al., 2010), and, in particular, by the St.
Lawrence River (i.e., 0.4 Tg y -1 ; Savenkoff et al., 2001;Thibodeau et al., 2010). This calculation gives a good estimate of the impact of fresh groundwater N input in pristine regions without much anthropogenic pressure. At a local scale, however, because rivers are absent in the archipelago, this fresh groundwater N input probably plays a key role in the nutrient budget 5 of Martinique Bay.

Conclusion
This study highlights the role of the STE in processing groundwater-derived N. In the boreal microtidal STE of Martinique Beach, chemical pathways and sources of N species were studied. N is mobilized within the STE and we observed a shift in the relative proportions of NO3and NH4 + in coastal groundwater due to the removal of NO3and the addition of NH4 + within 10 the STE. Nitrate loss along the flow path could be attributed to the mineralization of OC, since DOC concentrations were elevated in the STE or to alternative reduction pathways due the transient system with temporal variations in redox conditions (i.e., oxygen input). While the input of NO3represents 32% (37.54 mol m -1 y -1 ) of the fresh groundwater input of TDN to the STE, NO3fluxes exported to the coastal environment only represent 1% of the total TDN export flux. Thus NH4 + dominated the TDN load to Martinique Bay. Even if in situ NH4 + production was important in the Martinique Beach 15 STE, these coastal aquifer sediments represent a sink for N in spring when aquifer recharge is maximal. This study shows the export of N in groundwater discharge to the coastal ocean from a system without anthropogenic pressure. The export of NH4 + and DON from the STE at Martinique Beach to the coastal bay remains uncertain, since the subsequent nitrification and denitrification of regenerated NH4 + at the sediment-water interface may attenuate groundwater-derived N loads to coastal waters. Moreover, strong variability in N flux was observed within this highly dynamic system and reflects the 20 challenge of accurately estimating groundwater nutrient fluxes to the coastal ocean.   Table 2: Nutrient inventories estimated along the STE. Inventories were calculated at 3 distinct sampling sites as presented 5 in Figure 1C. For dissolved inorganic nitrogen (NH4 + , NOx), the 2013 data set was used. For dissolved organic nitrogen, concentrations measured in 2012 were used.