Isotopic investigation of denitrification in a riparian ecosystem in western France


  • Jean-Christophe Clément,

    Corresponding author
    1. Department of Ecology, Evolution and Natural Resources, Cook College, Rutgers University, New Brunswick, NJ 08901, USA;
    2. UMR 6553–ECOBIO, Université de Rennes I, Avenue Général Leclerc, F-35042 Rennes, France
    Search for more papers by this author
  • Robert M. Holmes,

    1. The Ecosystems Center, Marine Biological Laboratory, Woods Hole, MA 02543, USA; and
    Search for more papers by this author
  • Bruce J. Peterson,

    1. The Ecosystems Center, Marine Biological Laboratory, Woods Hole, MA 02543, USA; and
    Search for more papers by this author
  • Gilles Pinay

    1. UMR 6553–ECOBIO, Université de Rennes I, Avenue Général Leclerc, F-35042 Rennes, France
    Search for more papers by this author
    • §

      Present address: CEFE–CNRS, 1919 route de Mende, 34293 Montpellier Cedex 5, France.

Jean-Christophe Clément, Department of Ecology, Evolution and Natural Resources, Cook College, Rutgers University, New Brunswick, NJ 08901, USA (fax +732 932 8746; e-mail


  • 1Nitrogen (N) loss from agricultural fields and urban areas to stream and groundwaters is a world-wide environmental problem. Excessive nitrogen loading is partly responsible for eutrophication of fresh water and estuarine ecosystems, while elevated nitrate in drinking water has consequences for human health. Under certain conditions, riparian zones improve water quality by removing groundwater nitrate before it enters adjacent stream ecosystems. Nitrate decline along riparian flow paths is most often attributed to denitrification activity and vegetation uptake, but spatio-temporal distributions and rates are notoriously difficult to establish.
  • 2We used natural δ15N techniques in two riparian wetlands with differing vegetation to distinguish between the two processes responsible for reducing nitrate fluxes. We collected groundwater and above-ground vegetation samples along riparian transects where hydrology and groundwater chemistry had been investigated previously.
  • 3By measuring the natural abundance distribution of nitrogen isotopes in both the groundwater nitrate and riparian plant tissues along the transects, we attempted to determine to what extent the groundwater nitrate decline (from c. 15 to < 1 mg N L−1) observed in these two riparian sites with contrasting vegetation resulted from denitrification and/or plant uptake.
  • 4Denitrifying bacteria preferentially use the lighter isotope and hence tend to increase δ15N-NO3. In most groundwater samples we observed a significant increase of δ15N-NO3 (from +5 to +28‰) as nitrate concentrations declined, which demonstrated that denitrification was predominantly responsible for nitrate retention even when the water table was low.
  • 5This isotopic approach provided evidence of seasonal variation in the occurrence of denitrification, and helped to delimit the area where denitrification was most active. δ15N of overlying vegetation along the riparian transects was higher (from +1·7 to +14·2‰) than the typical range for terrestrial plants, and was related to the isotopic composition of nitrate in underlying groundwater when the water table was high. Thus, in this case, both plant uptake and denitrification contributed to the observed nitrate decline. However, given that the roots were limited to the upper 50 cm of soil, direct uptake of groundwater nitrate by riparian vegetation was only important when the water table was high.
  • 6Measurement of δ15N in plants may be a simple and powerful means of identifying buffer zones where denitrification is actively processing allochtonous nitrate in riparian ecosystems and their surrounding catchments.
  • 7Synthesis and applications. The isotopic approach described in this paper is a useful diagnostic tool for easily identifying actual denitrification locations where groundwater nitrate removal is taking place. It should allow investigation at a landscape scale of the spatio-temporal patterns of biogeochemical hot spots where denitrification rates are disproportionately high relative to the surrounding area. This could provide a sound basis for landscape management and restoration in the context of diffuse nitrogen pollution control.


Anthropogenic nitrogen (N) fixation is a major contributor to available nitrogen in the biosphere (Vitousek 1994; Galloway et al. 1995). Much of this nitrogen is used for fertilizer, some of which is not assimilated by growing crops but instead leaches into groundwater. Nitrogen-enriched groundwater eventually discharges into streams and contributes to the eutrophication of downstream lakes, estuaries and the coastal ocean (Vitousek et al. 1997). Nitrate loads in wetlands can also impact plant species richness by enhancing colonization and dominance by invasive species (Green & Galatowitsch 2002). However, numerous studies have demonstrated that groundwater nitrate concentrations may decrease substantially as water moves through riparian ecosystems before being discharged into streams (Lowrance et al. 1984; Peterjohn & Correll 1984; Pinay & Décamps 1988; Fustec et al. 1991; Groffman, Gold & Simmons 1992; Jordan, Correll & Weller 1993; Pinay, Roques & Fabre 1993; Lowrance, Vellidis & Hubbard 1995; Hill 1996; Hedin et al. 1998; Sabater et al. 2003). As a result, there is considerable interest in exploiting the nitrogen ‘filtration’ capacity of riparian ecosystems in order to improve surface water quality. However, there is still much uncertainty about the mechanisms and controls of nitrogen retention in riparian ecosystems, thus hindering their potential use and effective management.

Decline in the concentrations of nitrate–nitrogen (NO3-N) along riparian flow paths is most often attributed to denitrification (the reduction of inline image to N2O and N2), plant uptake and microbial immobilization. However, denitrification is notoriously difficult to quantify and even its detection can be challenging (Lowrance 1992; Simmons, Gold & Groffman 1992; Nelson, Gold & Groffman 1995; Verchot, Franklin & Gilliam 1997). The commonly used microcosm-scale measurements of denitrification are difficult to interpret at larger scales, because chamber effects may result in estimates that differ substantially from those in the field (Groffman, Gold & Simmons 1992; Groffman et al. 1996; Holmes et al. 1996; Gold et al. 1998). More elaborate mesocosm measurements may come closer to mimicking field conditions (Gold et al. 1998; Jacinthe et al. 1998; Addy et al. 1999) but riparian sediment samples are still isolated from hydrological exchange and other field conditions such as water table fluctuation, plant uptake and carbon exudates from intact roots. Given these factors, the lateral and vertical distributions of denitrification along riparian flow paths, and their quantitative significance for NO3-N decline compared with other processes, are difficult to assess using these methods alone.

Often the relative importance of different biological processes to groundwater NO3-N removal within a riparian wetland is first estimated with a mass–balance approach. This technique is difficult and subject to considerable error. The fate of nitrogen imported to a defined area is determined by measuring the subsequent changes in nitrogen standing stocks of potential sinks (e.g. microbial biomass and plant tissue). The difference between nitrogen inputs and losses due to measured assimilation is attributed to denitrification. Accurate measures of nitrogen assimilation are difficult, and proper spatial dimensions are often unclear. Furthermore, the approach yields little information regarding the temporal and spatial distribution of denitrification activity or plant uptake. Finally, riparian zones are often hydrologically complex, with numerous groundwater flow systems converging in near-stream environments (Hill 1990; Mulholland 1992, 1993; Fisher et al. 1998), while changes in water regimes can have critical effects on nitrogen retention capacities (Pinay, Clément & Naiman 2002). This makes identification of discrete flow paths even more difficult and complicates mass–balance approaches, because mixing of nitrate-enriched shallow groundwater with low-nitrate deeper groundwater results in nitrate concentration decline without nitrate retention (Altman & Parizek 1995; Pinay et al. 1998). Nevertheless, recent studies have shown that a detailed instrumentation comprising piezometer nests can help to understand the groundwater flow system, which may influence nitrate attenuation in riparian wetlands (Cey et al. 1999; Devito et al. 2000; Burt et al. 2002; Clément et al. 2003).

Natural abundance N-isotope ratios (expressed as δ15N) provide insights into the nitrate removal processes occurring in groundwater (Mariotti, Landreau & Simon 1988; Böttcher et al. 1990; Brandes, McClain & Pimentel 1996; McMahon & Böhlke 1996; Blicher-Mathiesen, McCarty & Nielsen 1998; Brandes et al. 1998; Mengis et al. 1999). There are two stable isotopes of nitrogen, 14N and 15N. The most common, 14N, accounts for approximately 99·63% of atmospheric nitrogen. Although the nitrogen isotopic composition of the standard (atmospheric N2) is constant, other materials have variable isotopic composition because some processes discriminate (i.e. fractionate) between nitrogen isotopes. The lighter isotope often reacts more rapidly in biogeochemical cycles than the heavy one; therefore processes involved in the nitrogen cycle can also affect the ratio between the light and heavy isotopes in environmental nitrogen pools. Among these processes, microbial denitrification significantly alters the nitrogen isotope ratio, resulting in the progressive enrichment of the remaining NO3 pool with the heavier isotope (Mariotti et al. 1981; Böhlke & Denver 1995; Mengis et al. 1999). In contrast to denitrification, nitrate uptake by terrestrial vegetation appears to fractionate minimally or not at all (Fig. 1; Mariotti et al. 1982; Fry 1991). Recent studies have used natural abundance δ15N techniques to assess denitrification in soil–stream interface (Ostrom et al. 2002) but the method has yet to be used in riparian zones with the intent of distinguishing among various processes (chiefly denitrification and plant uptake) responsible for reducing nitrate fluxes.

Figure 1.

Conceptual diagram illustrating how the natural amount of groundwater δ15N-NO3 and δ15N of vegetation tissues can be used to assess the relative importance of denitrification and plant uptake as nitrate retention mechanisms along riparian flow paths.

We hypothesized that if plant uptake alone is responsible for nitrate retention along a riparian flow path, isotopic composition of the remaining nitrate would remain essentially unchanged. If both denitrification and plant uptake contribute to nitrate retention along riparian flow paths, isotopic composition of residual nitrate would become progressively enriched and the overlying vegetation would reflect the isotopic composition of its increasingly enriched nitrogen source. Alternatively, if denitrification was the only process involved, the isotopic composition of residual nitrate would become progressively enriched while plant isotopic composition would remain unchanged with a different δ15N, its nitrogen source not being influenced by denitrification activities (Fig. 1). Hence, this approach is potentially useful for determining whether groundwater nitrate retention along subsurface flow paths in riparian wetlands is due to denitrification and/or plant uptake. It should be noted that the same pattern can be produced by mixing along the flow path between high NO3 concentrations–low δ15N-NO3 water from upland and deep upwelling water with low NO3–high δ15N-NO3. However, when such a phenomenon is suspected and hydrological data are missing, dilution can still be distinguished from actual fractionation using a δ15N concentration diagram and calculations (Mariotti, Landreau & Simon 1988).

The goals of this research on nitrogen cycling in a riparian ecosystem in western France were to: (i) study the temporal and spatial variation of 15N natural abundances in groundwater nitrate and vegetation; (ii) demonstrate whether denitrification occurred and influenced groundwater nitrate concentrations; and (iii) investigate the relationships between groundwater nitrate retention, water table fluctuations, plant uptake and the spatial distribution of denitrification. We selected two riparian zones differing in their vegetation cover (i.e. a grassland riparian site and a shrub riparian site). Using a grid network of piezometers, the groundwater flow system and geology had been previously investigated and well documented in a parallel study to understand clearly the nitrate behaviour along the riparian groundwater flow paths (Clément et al. 2003). Groundwater nitrate concentrations and isotopic ratios along groundwater flow paths were measured at each site. We also evaluated ambient denitrification activity by a microcosm method and measured δ15N natural abundance in vegetation along the riparian flow paths.


study site

The study was conducted in a riparian ecosystem along a fourth-order stream (Petit Hermitage Stream) in western France (48·3°N, 1·3°W). The region has a mild oceanic climate and the landscape is predominantly agricultural, with maize and wheat being the primary crops. Fertilizer application rates are high, about 200 kg N ha−1 year−1, which have resulted in groundwater nitrate concentrations ranging from 10 to 20 mg N L−1. The upland–riparian boundary was characterized by a steep 2–3 m drop in elevation from surrounding fields into the riparian ecosystem. The whole riparian zone was about 80 m wide from the upland–riparian interface to the stream, and was periodically flooded via over-bank inundation. A relict channel, roughly parallel to the stream and 30 m from the hill slope, flowed during high water periods (Fig. 2).

Figure 2.

Map and cross-section of the study site in Brittany, western France. Topographical lines (dashed line) represent land surface in metres above sea level as indicated in boxes. The bank of the stream is 20 m above the sea level.

Two distinct types of riparian zone were present along the c. 300 m length of the study site (Fig. 2). The upstream zone (‘shrub riparian site’) was characterized by abundant shrubs but no mature trees. The primary species in this zone are young willows (Salix sp.) with herbaceous cover such as Phalaris arundinacea L., Agrostis stolonifera L., Symphytum officinale L., Holcus lanatus L., Glyceria fluitans (L.) R. Br., Urtica dioica L., Ranunculus repens L. and Gallium palustre L. The downstream zone (‘grassland riparian site’) was cleared long ago of all trees and is now a wet meadow mainly characterized by Juncus effusus L., Polygonum amphibium L., Urtica dioica, Glyceria fluitans, Holcus lanatus and Ranunculus repens. Based on field observations, the root zone was limited to the upper 50 cm at both sites.

hydrological context

The geological substratum of the study area is made of ‘cornéennes’ and mica-schists (Brioverien Schist). The soils are fine silty-clay loam, mixed, mesic, Typic Haplaquoll (Soil Survey Staff 1988). The two study sites presented a similar soil profile. Below a litter layer, an organic O horizon, from 0 to −5 cm, rapidly evolved into a brown-grey transitional mineral hydromorphic A horizon, with ferrous oxidation spots along the root channels (from −5 to −25 cm, 9·6% organic matter (OM), pH 4·8, 75% silt plus clay), followed by a grey pseudogley (from −25 to −50 cm), with ferrous oxidation spots (3·9% OM, pH 5·2, 80% silt plus clay). Some roots could be found up to −50 cm deep. Beyond this depth, the gley horizon was re-oxidized until −200 cm deep (2·7% OM, pH 5·5, 76% silt plus clay). Then, a reduced blue-grey horizon was present and reached the geological substratum made of schist. The texture was loamy-clay in the first 100 cm, then loamy-sand in the deeper layers. The structure became denser with depth, reducing the permeability. Bulk densities were measured in the first 25 cm and ranged between 0·45 and 0·70 g cm−3.

Twelve piezometers were installed in each of the two types of riparian zone. Transects were arranged perpendicular to the stream channel, in three transects of four piezometers each, from the upland–riparian interface towards the stream. Transects were spaced about 10 m apart with an average distance of 3 m between piezometers. Piezometers were constructed from a 2·5 cm-diameter PVC pipe, inserted to depths c. 2 m below the ground surface, with the lowest 50 cm perforated to allow water to be collected from that horizon. All piezometers were surveyed using a geographical positioning system (GPS), and topography and water table elevation are referenced to mean sea level.

Hydrology within the riparian area has been investigated intensively and detailed in another report (Clément et al. 2003). Briefly, in order to categorize the regional aquifer and to determine whether mixing of different groundwater sources might confound data interpretation, some additional deeper piezometers (between 7 and 9 m deep) were installed at the upland–riparian interface (Fig. 2) and also within the riparian area. Water samples were also taken from the stream, the relict channel and the three farm wells (from 7 to 30 m deep) in the surrounding catchment (DW1, DW2 and DW3; Fig. 2). Water level measurements in each piezometer, as well as in the stream and the relict channel, were made at least once a month. Further investigations on water table level fluctuations were carried out using both vertical and lateral profiles of the water table from uplands to the riparian zones and information from topographical maps. Hydraulic conductivities were performed using the ‘slug test’ method (Fetter 1994). Finally, nitrate and chloride contents were measured in water samples from the various piezometers, the stream, the relict channel and the farm wells to help identify different water sources. Checking the hydrological connectivity along the riparian transects had been carried out previously using a chloride tracer experiment (Clément et al. 2003). Chloride-enriched groundwater was dropped into an upslope piezometer and the plume appearance was surveyed in the down-gradient piezometers using a conductivity probe. Groundwater flow paths were found to be parallel to the slope, which was confirmed by the water table survey described below.

sampling and analytical procedures

Groundwater samples were collected on four dates: April 1998, August 1998, February 1999 and June 1999, corresponding to different water table levels ranging from the high water period (i.e. April 1998) to the low water period (i.e. August 1998). On each sampling date, water samples were collected from each riparian piezometer in each of the three transects of the grassland and shrub riparian sites. However, NO3-N isotopic analyses were carried out only for a single transect in each of the two zones, so we will focus on those two transects: one transect in the wet grassland zone, the other in the shrub zone. The other transects of each riparian zone were used as spatial replications of NO3-N concentrations. On the same four dates, we collected vegetation in the vicinity of piezometers from which groundwater 15N-NO3 samples were withdrawn in order to measure 15N in plant tissues. Vegetation sampling involved harvesting 10–15 leaves of five different plant species that were present along the transects in both riparian sites, to account for possible interspecific variability in 15N uptake. The sampled species were Glyceria fluitans, Holcus lanatus, Symphytum officinale, Urtica dioica and Ranunculus repens.

For the two intensively sampled transects, all water samples were analysed for nitrate (NO3-N), ammonium (NH4-N), total dissolved nitrogen (TDN), chloride and δ15N-NO3. Piezometers were sampled using a peristaltic pump, and one volume was displaced before sampling water was pumped from the piezometer. Water samples collected from the surrounding deep wells (Fig. 2) and the stream were also analysed for nitrate, ammonium, total dissolved nitrogen, chloride and δ15N-NO3. All water samples were stored in a cooler in the field and filtered (Whatman GF/C) upon return to the laboratory. Samples were analysed within 24 h of collection, for NO3-N using a Technicon auto analyser (Technicon, Tarrytown, NY, USA) including Cd-reduction to nitrite (Wood, Armstrong & Richards 1967) and for ammonium using the phenyl-hypochlorite method (Solorzano 1969). TDN concentrations were measured as nitrate after mineralization in a basic solution (NaOH) with potassium persulphate (K2S2O8) as catalyser (Koroleff 1983). Dissolved organic nitrogen (DON) is calculated as the difference between TDN and the total amount of inorganic nitrogen forms [DON = TDN – (NO3 + NH4)]. Chloride was determined by a titrimetric method using a silver nitrate electrode with a Sherwood Chloride Analyser 926 (Sherwood, Cambridge, UK) (Clarke 1950). Water samples to be analysed for isotopic compositions of nitrate were processed using diffusion methods (Sigman et al. 1997; Holmes et al. 1998). Analytical deviation of 15N-NO3 was assessed using standards (4 µm and 1000 µm NO3) as well as sample replicates, and was estimated to be ±3·26%.

All 15N samples were analysed by mass spectrometry after combustion of the sample to produce N2. The isotopic composition of nitrate was measured using a dual-inlet Finnigan MAT ‘Delta S’ isotope ratio mass spectrometer (Bremen, Germany), and vegetation samples were analysed using a continuous-flow PDZ Europa ‘20–20’ (Northwich, UK), at the Stable Isotope Laboratory, Marine Biological Laboratory, Woods Hole, MA. Nitrogen isotopic compositions were reported using ‘delta’ notation, where δ15N =[(RSA/RST) − 1] × 103 and R =15N/14N, expressed as ‰ deviation of the sample (SA) from the standard (ST), which is N2 in atmospheric air (δ15Nair = 0‰).

denitrification activity

Quantification of in situ denitrification in riparian soils (DNT) was assayed by a static core acetylene inhibition method that blocks denitrification at the nitrous oxide (N2O) step (Yoshinari & Knowles 1976). Each riparian grid of piezometers was divided into three zones (upper, mid and lower) lateral to the hill slope. The ‘upper’ zone corresponded to the first piezometers of the transects (nearest the upland–riparian interface), the ‘mid’ zone ranged from the second to the third piezometers, and the ‘lower’ zone was in the vicinity of the most down-gradient piezometers. Three soil samples were collected seasonally, from autumn 1998 to summer 1999, from each of the three zones of each riparian site and at three different depths (i.e. 0–25 cm, 25–50 cm and below 50 cm). Three intact cores (length 10 cm, diameter 3 cm) were then capped with rubber serum stoppers and amended with acetone-free acetylene to bring core atmosphere concentration to 10 kPa (10% v/v) acetylene and 90 kPa air. The denitrification rate was calculated as the rate of N2O accumulation in the headspace after 2 h of incubation at field temperature. Headspace samples were removed from all cores and analysed via gas chromatography (GC Chrompack CP 9001, Bergen op Zoom, Holland). Estimates of annual nitrogen loss to denitrification were calculated by averaging seasonal measured rates. DNT were then studied using Kruskal–Wallis non-parametric anova of ranks (Clément, Pinay & Marmonier 2002). Differences were considered statistically significant at the P < 0·05 level. All DNT statistical analyses were performed on Statistica for Windows (Stat Soft, Tulsa, OK).



The groundwater level in the riparian zone followed the stream hydrological regime, with high water levels in late winter and early spring and low water levels in summer and early autumn (Fig. 3; Burt et al. 2002). In April 1998, the water table level was above the ground level in the lower part of the gradient because the whole riparian zone was flooded via over-bank inundation. Water table slopes and the chloride tracer experiment confirmed that groundwater flow lines were generally perpendicular to the river during the experimental periods. Observations of water levels along the transects at both sites showed that the saturated zone reached the upper soil horizons in the high water periods (April 1998, February and June 1999) but remained below about 1 m during the low water periods. Based on the intensive hydrological survey and the chloride and nitrate patterns, three main groundwater reservoirs were detected (Clément et al. 2003). First, a deep water table from 4 m below the soil surface was characterized by low concentrations in nitrate (i.e. DW1; Table 1). Secondly, DW2 and DW3 characterized the subsurface high nitrate, high chloride groundwater flowing from upland to the riparian zone (Fig. 2). Finally, a shallow water layer fluctuated between ground surface and 4 m deep, in which all the riparian piezometers were placed. This last zone received the upland flow from the hill slope and was characterized by intermediate nitrate concentrations (Fig. 4). Some groundwater exchanges occurred between the deep aquifer and the shallow compartment during the recharge period (2–3 months year−1) and could involve dilution, especially for the two last wells of each riparian transect. However, because of low hydraulic conductivities (average value 1·98 cm h−1 ± 1·32 SD) the dilution was very limited. Regardless, given the eventual mixing that occurred in the last piezometers of the riparian transects, we took special care to analyse the natural abundance of nitrogen isotope ratios avoiding the recharge period.

Figure 3.

Water table profiles along a riparian transect during the experimental period in April (diamonds), June (squares), August (triangles) and February (circles).The dashed line represents the ground surface. For the complete hydrological study, see Clément et al. (2003).

Table 1.  Characterization of the regional aquifer. All the groundwater farm wells are within 400 m of the study site (Fig. 1). Values are means (± SD)
Well locationDepth (m)Nitrate (mg N L−1 )Chloride (mg Cl L−1 )δ15NO3-N (‰)
DW130–32 0·29 ± 0·1238·2 ± 3·13·2
DW211 9·71 ± 1·2642·1 ± 5·25·1
DW3 7–812·79 ± 2·1139·3 ± 3·19·7
Stream water  4·63 ± 1·2636·3 ± 3·38·2 ± 0·7
Figure 4.

Groundwater nitrate concentrations in riparian transects along Petit Hermitage Stream, western France in April (diamonds), June (squares), August (triangles) and February (circles).

patterns of nitrate concentrations

Nitrate concentrations declined greatly as groundwater flowed through the riparian zone of Petit Hermitage Stream. On average, groundwater nitrate concentrations were reduced by 95% within 9 m of travel in the shrub and the grassland riparian sites (Fig. 4). After 6 m of travel, mean groundwater NO3-N concentrations along the wet meadow transects were already 2·16 mg N L−1 ± 2·55 SD (86% reduction), while shrub concentrations were still 10·25 mg N L−1 ± 3·55 SD (36% reduction). Furthermore, at both sites nitrate concentrations were always higher in the winter period (i.e. February). Wells DW2 and DW3, located in the surrounding catchment and 7–11 m deep, showed high groundwater nitrate concentrations (9·71 and 12·79 mg N L−1, respectively; Table 1 and Fig. 4),while in the deepest well (DW1; 30 m deep) concentrations were much lower (0·29 mg N L−1 ). Finally, the stream water had rather constant nitrate concentrations, with an average of 4·63 mg N L−1 during the experimental period. Ammonium concentrations were constantly low for all samples (i.e. ≤ 1 mg N L−1 , average 0·25 mg N L−1 ± 0·04 SD), while DON remained essentially unchanged and also low as TDN concentrations were nearly equal to nitrate values.

ambient denitrification measurements

At any given depth (i.e. 0–25 cm, 25–50 cm and below 50 cm) and in any given zone (i.e. upper, mid and lower), no significant seasonal pattern of DNT was detected in the different sites (P > 0·05; Clément, Pinay & Marmonier 2002). Therefore, annual average activities were compared. Annual in situ DNT rates were significantly higher (from 247 to 689 µg N2O-N g−1dry soil day−1; P < 0·01) in the upper organic horizon than in the lower ones. Nevertheless, despite the steep gradient of DNT rates measured along the soil profiles, significant activities (from 186 to 294 µg N2O-N kg−1dry soil day−1) were measured even in the deepest soil samples (Table 2). In the shallow soil layer, the grassland riparian site generally had higher DNT rates than the shrub site. Furthermore, DNT activities were relatively constant along the slope in the wet meadow, while denitrification in the shrubby zone showed a sharp increase from the upper to the lower part.

Table 2.  Annual averages (± SD, n= 12) of in situ denitrification activities measured as the accumulation of N2O-N after 2 h of incubation in the presence of acetylene along the shrub and the grassland transects at three different depths. ‘Upper’ is close to the first piezometer of the transects, while ‘lower’ is close to the last piezometer. For further information see Clément, Pinay & Marmonier (2002)
DepthsUpper µg N2O-N kg−1 day−1Mid µg N2O-N kg−1 day−1Lower µg N2O-N kg−1 day−1
0–25 cm247 ± 71689 ± 173502 ± 172598 ± 164637 ± 200637 ± 205
25–50 cm180 ± 27389 ± 109175 ± 27268 ± 66313 ± 126257 ± 72
Below 50 cm227 ± 44226 ± 55195 ± 24294 ± 93186 ± 45284 ± 96

isotopic data

The deep groundwater sampled in the farm wells had δ15N-NO3 values varying from 3·2‰ for DW1 (i.e. the deepest farm well) to 9·7‰ for DW3 (Table 1). The mean value of δ15N-NO3 found in the groundwater entering the two riparian sites was 7·22‰ ± 1·75 SD (Fig. 5). Further downstream the residual inline image pool became enriched in 15N along the transects. However, differences between study sites were noticeable. Mean wet meadow δ15N-NO3 along the transect increased from 7·83‰ ± 2·47 SD at the upland–riparian interface to 13·10‰ ± 4·56 SD in the second piezometer and to 9·87‰ ± 2·02 SD in the third one, before it declined to 4·36‰ ± 4·44 SD in the last piezometer. Mean shrub zone δ15N-NO3 slightly increased from 6·60‰ ± 0·62 SD at the upland–riparian interface to 7·28‰ ± 0·59 SD in the second piezometer, and then increased to 15·37‰ ± 10·94 SD with a maximum of 28‰. In addition, δ15N-nitrate exhibited a seasonal pattern in both riparian sites: February samples presented the lowest values while the highest ones were found in August.

Figure 5.

δ15N-NO3 in groundwater piezometers of the riparian shrub and grassland sites in April (diamonds), June (squares), August (triangles) and February (circles). Each δ15N-NO3 value represents a measurement from a single sample. Analytical deviation is estimated about ±3·26% (SD). There are two missing values for samples collected in June 1999 due to instrumental malfunction.

Three groups were apparent when groundwater nitrate concentrations were plotted against δ15N-NO3 (Fig. 6). The first group (A) comprised samples with low nitrate concentrations and low δ15N-NO3, and corresponded mostly to down slope areas. The second group (B) comprised samples collected along the groundwater flow path from the hill slope–wetland interface. Their nitrate concentrations were inversely correlated to their δ15N-NO3. The third group (C) comprised stream water samples and the farm well closest to the stream (DW2). The deepest hill slope well (DW1) had both a nitrate concentration and δ15N-NO3 similar to those of group A, while the other hill slope well (DW3) was located within the second group (B).

Figure 6.

Relationship between nitrate concentration and δ15N-NO3 in riparian groundwater samples from the Petit Hermitage Stream study site. All sampling dates are included and a distinction is made according to the position of each sample along the riparian transects. Three groups of samples are identified and discussed in the text.

Assuming that denitrification led to the observed nitrate decline, we calculated the fractionation factor (ɛ) from the slope of the regression of ln[NO3-N] vs. δ15N-NO3 (Fig. 7; Mariotti et al. 1981; Fustec et al. 1991). This fractionation factor (ɛ) is the instantaneous difference in isotopic composition of reactant and product of denitrification, nitrate and N2 or N2O, respectively. For this analysis, we used samples corresponding to group B (Fig. 6). We excluded samples in group A because of potential problems associated with mixing and dilution. The slope of the regression of ln[NO3-N] vs. δ15N-NO3 (i.e. the fractionation factor) was −8·38‰ (Fig. 7).

Figure 7.

Plot of ln[inline image–Ν] vs. δ15N-NO3, the slope of which gives ɛ, the fractionation factor (−8·38‰). Each δ15N-NO3 value represents a single groundwater sample from group B (Fig. 6).

Plant uptake δ15N in plants sampled near the piezometers were plotted against the groundwater δ15N-NO3 measured in the corresponding riparian piezometers (Fig. 8). Almost all δ15N values measured in plants were enriched compared with typical values for terrestrial vegetation, and their δ15N values spanned +1·7 to +14·2‰ (Table 3). For the high water period (February) samples showed similar δ15N both in vegetation and groundwater samples along the riparian transects. In the drier sampling periods (i.e. June and August) higher δ15N values were found in the groundwater NO3 than in their corresponding plant samples, and no significant differences were detected among the latter (Table 3).

Figure 8.

Relationship between δ15N-NO3 in riparian groundwater and the δ15N isotopic composition of overlying vegetation in April (diamonds), June (squares), August (triangles) and February (circles).

Table 3.  δ15N ranges found in the leaves of the different riparian plant species sampled along the shrub and grassland transects
Plant speciesRange of δ15N measured (‰)
Glyceria fluitans8·2–9·3
Holcus lanatus5·0–9·2
Symphytum officinale3·4–7·6
Urtica dioica1·7–8·6
Ranunculus repens4·1–14·2


groundwater nitrate elimination along the flow paths

As is frequently observed in riparian ecosystems, groundwater nitrate concentrations decreased along flow paths through both of our riparian sites (Fig. 4). Given the very low ammonium and total dissolved nitrogen concentrations measured along the riparian flow paths, the groundwater nitrate decline was not due to conversion to ammonium or dissolved organic nitrogen, but instead resulted from either abiotic (dilution) or biotic processes such as denitrification and/or plant uptake. Although a previous hydrological investigation by Clément et al. (2003) showed that dilution occurrence could not be totally ignored for the last two piezometers of the transects, it was likely to be limited in space and time during our experiments because the sampling dates for this 15N study were chosen during periods with reduced contribution of the deep aquifer, and thus lower dilution probability.

In situ denitrification measurements demonstrated that this microbiological process actually occurred at each riparian site (Table 2). Despite the lack of seasonal pattern, high DNT rates were measured in the upper and middle zones of the grassland site, which corresponded to a sharp decrease of the groundwater nitrate concentrations (Fig. 4). Likewise, the DNT activities in the shrub site, which were higher from the middle to the lower zone, could be related to a sharp decrease of groundwater nitrate concentrations at 9 m within the riparian zone. In accordance with other studies of deep denitrification activity (Groffman, Gold & Simmons 1992; Lowrance 1992), these high denitrification rates were mostly limited to the upper soil horizon (i.e. 0–25 cm) and significantly decreased with depth (Table 2). Nevertheless, DNT activity was measured below 50 cm. This observation is in accordance with some other studies that have also found significant denitrification potentials in soils in contact with groundwater (Smith & Duff 1988; Obenhuber & Lowrance 1991). However, it is unlikely that denitrification is responsible for all of the nitrate elimination, particularly when the groundwater table is far from the soil surface (i.e. August 1998) but also during high water periods, because the contact between the most active denitrifying soil layer and the subsurface water is restricted to only 25–50 cm of the soil profile.

δ15n-no3 and denitrification

In our riparian piezometers along Petit Hermitage Stream, we found that δ15N-NO3 varied from c. 5‰ up to c. 25‰ and, in most cases, nitrate concentration and δ15N-NO3 were inversely related (Fig. 6). These data clearly emphasize the importance of denitrification in the elimination of nitrate at depth (i.e. 1·5–2 m deep). Based on the hypothesis that denitrification contributed to the nitrate decline along the riparian flow paths, we predicted that δ15N-groundwater NO3 would increase as groundwater nitrate concentration declined (Fig. 1). This pattern is well illustrated by the significant enrichment in 15N associated with declining NO3-N concentrations.

The high nitrate concentration measured in the groundwater at the upland–riparian interface had low δ15N-NO3 values (i.e. 7·22‰ ± 1·75 SD, Fig. 5). As nitrate concentrations declined along the flow path (up to c. 2 mg N l−1), δ15N-NO3 values generally increased (Fig. 6). The elevated δ15N-NO3 in the riparian zone (Fig. 5) cannot be due to mixing with deep groundwater because its δ15N-NO3 values were low (DW1; Table 1). Therefore, the elevated δ15N-NO3 values in the riparian zone are clear evidence that denitrification was contributing to the observed nitrate decline up to a depth of 75 cm.

This conclusion is also supported by the calculated fractionation factor ɛ=−8·38‰, which is close to values reported for denitrification occurring in groundwater on Cape Cod, MA (Smith, Howes & Duff 1991), in a sandy aquifer in Germany (Böttcher et al. 1990), in shallow alluvial groundwater in France (Fustec et al. 1991), in an aquifer flowing through chalk rocks in northern France (Mariotti, Landreau & Simon 1988) and in forested soils in Japan (Koba et al. 1997). This significant increase of δ15N-NO3 also reveals that there is a constant input of upland nitrate at the wetland interface. This corresponds to high δ15N-NO3 samples (group B) taken along the first few metres of the riparian flow path, where the residual highly enriched nitrate is further denitrified (Fig. 6). The low δ15N-NO3 samples (group A), which were mostly taken along the lower part of the riparian transects, could be the result of several alternative processes. The first possibility is that another water source, such as the deep regional groundwater with a low nitrate concentration and relatively low δ15N-NO3 value (e.g. DW1; Table 1), was upwelling to some extent in the vicinity of the last piezometers of the riparian transects, and mixing with the denitrified upland flow. However, this hypothesis is probably not relevant in the present study because the chloride tracer experiment and water table elevations did not support such mixing of water sources. A more likely explanation is that the low δ15N-NO3 values measured in these low nitrate concentration riparian samples (group A) derived from the alternation of intermittent nitrate leaching from the overlying unsaturated nitrifying soil horizon and complete denitrification in the saturated horizon.

seasonal patterns

The seasonal variations in the natural abundance of δ15N-NO3 along each transect demonstrate that the relative importance of denitrification to groundwater nitrate removal also varies seasonally (Fig. 5). It illustrates the value of the isotopic approach compared with the classical denitrification measurement based on the acetylene block method. Indeed, while in situ denitrification measurements did not exhibit any significant seasonal pattern (Clément, Pinay & Marmonier 2002), the 15N natural abundance survey showed that groundwater nitrate was more likely to be enriched/processed by denitrification in summer (August) than in spring (April) or late winter (February; Fig. 5). Moreover, in terms of spatial δ15N-NO3 trends, a distinction could be made between the two riparian zones. Along the shrub riparian transect, the δ15N-groundwater NO3 increased after 3 m of flow under the riparian zone, indicating the location of the ‘active denitrification zone’, i.e. the zone where denitrification was most probably processing the groundwater nitrate.

In the wet meadow, δ15N-groundwater NO3 trends showed that the active zone was located closer to the upland–riparian zone interface, i.e. within the first 3–6 m, and was more limited in space. Indeed, the δ15N-NO3 generally decreased beyond this point, although δ15N changes between the second and the third piezometer were not clear, as is illustrated by the standard deviations calculated for these samples (Fig. 5). A narrow active zone was also reported by Ostrom et al. (2002). Nitrate concentrations in the third piezometer (6 m) were almost totally depleted in April and June, and therefore δ15N-NO3 at this location and beyond along the flow path were most probably the result of a mixture between remaining denitrified allochtonous groundwater nitrate (with high δ15N-NO3) and nitrate originating from nitrification of autochthonous nitrogen (with low δ15N-NO3) and their subsequent leaching from the upper soil horizons. On the other hand, samples from February and August still yielded significant amounts of groundwater nitrate after 6 m as well as the expected trend for denitrification in August with a regular increase of δ15N-NO3 up to the third piezometer. Finally, the decrease of δ15N-NO3 observed after 9 m for that date also corresponded with a significant decrease of N-NO3 concentration and further supports the eventual mixture with low δ15N-NO3 from autochthonous nitrate. These data show that other processes could be involved in the variations of δ15N-NO3 along the groundwater flow paths.

An alternative hypothesis is that the nitrate source is available during limited periods. This lack of constant nitrate input would alter the δ15N-NO3 signal because both isotopes would eventually be used. These different δ15N-NO3 patterns concerning two closely located riparian transects also illustrate the spatial heterogeneity of a buffering system that may occur even within a similar hydrogeomorphic context.

uptake vs. denitrification

In addition to denitrification, uptake of groundwater nitrate by riparian vegetation could also have contributed to the observed nitrate decline. As hypothesized, if uptake by riparian vegetation and denitrification were both contributing to the nitrate concentration decline, the isotopic composition of the overlying vegetation would reflect the isotopic composition of the riparian groundwater. We found that there was a significant 15N enrichment of the riparian plants tissues (from +1·7 to +14·2‰) compared with the typical isotopic composition of terrestrial vegetation (Fry 1991; Gebauer & Schulze 1991; Nadelhoffer & Fry 1994). However, this enrichment was apparent regardless of season, species, site or position along the transect (Fig. 8). This pattern demonstrates that the riparian vegetation assimilated the residual 15N-enriched nitrogen resulting from fractionation during the denitrification process, and that plant uptake and denitrification both contributed to nitrate retention at shallow depth (i.e. the upper 25–50 cm).

Therefore, uptake by riparian vegetation made a direct contribution to nitrate retention at shallow depths in the riparian sites, at least during the high water period. Indeed, stronger, although not significant, relationships were found between δ15N-groundwater NO3 and δ15N of the surrounding vegetation during the high water periods (i.e. February and April; Fig. 8), when the groundwater table level was within the root zone (i.e. the upper 50 cm), than in August when the groundwater table was located below the root zone. This is counter to the commonly accepted model for temperate zone riparian ecosystems, which predicts that nitrate uptake by vegetation is a dominant nitrate retention process during summer (Haycock, Pinay & Walker 1993). During low water periods the soil is more aerated and plants can therefore change their nitrate source from groundwater, which becomes inaccessible to soil nitrate resulting from organic matter mineralization. This nitrogen soil pool is isotopically lighter due to fractionation during nitrification and could explain the lower δ15N of plant tissues in summer. During high water periods the soil is waterlogged and plants may have difficulty assimilating soil nitrate produced by nitrification, which is likely to be prevented under anaerobic conditions; hence, groundwater becomes the primary source of nitrate. In other words, during low water periods vegetation uptake and groundwater nitrate are spatially disconnected and their respective 15N/14N ratios tend to differ (e.g. in August). During high water periods, denitrification in the surface groundwater and vegetation uptake are linked and δ15N-groundwater NO3 are better correlated to δ15N-vegetation (e.g. in February).


Nutrient concentration data and classical ambient denitrification measurements are inconclusive in determining the respective roles of denitrification and plant uptake in groundwater nitrate elimination. The use of the natural abundance distribution of nitrogen isotopes allowed further investigation of the mechanisms of nitrogen elimination. We have demonstrated that denitrification was the primary mechanism responsible for the observed nitrate concentration decline in deep layers along riparian groundwater flow paths, even during low water table periods. This in situ approach facilitates the detection of deep denitrification as even relatively low denitrification rates can be inferred by an inverse relationship between nitrate concentration and δ15N-NO3. Moreover, this method avoids concerns commonly associated with denitrification measurements in microcosm or mesocosm experiments, particularly the question of whether denitrification rates measured in isolated sediment slurries adequately represent actual conditions occurring in the field. In addition, the δ15N-NO3 method gave better insight into the spatio-temporal variations of denitrification activity within the riparian zones as a means of buffering upland nitrate fluxes. Measurements of δ15N in riparian plant tissues demonstrated the shared contribution of denitrification and plant uptake in the decline of groundwater nitrate, especially during high water periods.

The 15N natural abundance surveys in both groundwater and riparian vegetation have improved our understanding of denitrification and groundwater nitrate removal within a riparian ecosystem under a temperate climate. Although the natural abundance distributions of nitrogen isotopes do not always unambiguously identify nitrogen cycling processes, locations and rates, they do greatly constrain the range of viable explanations for nutrient concentration changes and thus improve our understanding of riparian ecosystem functioning. In terms of further use and development of this isotopic approach, the overall high 15N enrichment found in the plants located in a riparian zone where denitrification is occurring suggests that the 15N/14N ratio measurement in plant tissues could easily be used to identify actual denitrification locations as opposed to potential ones. This method should allow investigation at a landscape scale of the spatio-temporal patterns of biogeochemical hot spots where denitrification rates are disproportionately high relative to the surrounding area. We also suggest that the isotopic approach would benefit studies on the importance of denitrification in regulating nitrogen fluxes in landscape types such as field margins (hedgerows) and ditch networks.


This study was supported by a European Union project called Nitrogen Control in Agricultural Landscapes (NICOLAS, grant no. ENV4-CT97-0395), by a visiting scientist fellowship from the University of Rennes to R. M. Holmes, and by NSF-9524740 and NSF-9818199. The authors warmly thank Zaffreen Pinay, Nathalie Josselin, Luc Brient, Olivier Troccaz, Saïd Nassur, Marie-Paul Briand, Gwenaelle Desury and Florence Rouaud for their valuable help with laboratory and/or field work. We also thank the two anonymous referees whose comments improved this article.