JEQ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 27 October 2006
Published in J Environ Qual 35:2113-2122 (2006)
DOI: 10.2134/jeq2006.0091
© 2006 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Montreuil, O.
Right arrow Articles by Merot, P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Montreuil, O.
Right arrow Articles by Merot, P.
Agricola
Right arrow Articles by Montreuil, O.
Right arrow Articles by Merot, P.
Related Collections
Right arrow Wetlands and Aquatic Processes
Right arrow Ecosystem Management
Right arrow Watershed-Scale Studies
Right arrow Nutrient Management

TECHNICAL REPORTS

Landscape and Watershed Processes

Nitrogen Removal in Valley Bottom Wetlands

Assessment in Headwater Catchments Distributed throughout a Large Basin

Olivier Montreuil and Philippe Merot*

INRA-Agrocampus Rennes, UMR Sol-Agronomie-Spatialisation, 65 rue de Saint-Brieuc, CS 84215, 35042 Rennes Cedex, France

* Corresponding author (Philippe.merot{at}rennes.inra.fr)

Received for publication March 6, 2006.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSION
 REFERENCES
 
Although the reduction of nutrient loading between uplands and streams is sometimes considered evidence of the effect of wetlands acting as buffer zones, the influence of valley bottom wetlands (VBWs) on NO3– loading has seldom been assessed at the catchment scale. The objective of this study was to quantify the impact of VBWs on NO3– concentrations in streams in the Brittany region of France. We analyzed the spatial variation in NO3–N concentrations in 18 headwater catchments located in a 400-km2 basin, with varying topographic, climatic, and agricultural intensity conditions. Approximately every 10 d, water was sampled during the high flow season. We investigated the relationships between the mean NO3– concentration and different characteristics of the catchments: (i) the amount of effective rainfall, i.e., the combined effect of precipitation and actual evapotranspiration on discharge and chemical dilution, (ii) the intensity of farming, i.e., the area used for farming in the catchments and the surplus of the agricultural N budget, and (iii) the relative area of VBWs. Although the first two characteristics were the main factors controlling N concentration variability, a step-by-step regression allowed us to attribute a significant part of the NO3– concentration decrease to the increase of VBW area in each catchment. For an increase of VBW area from 11 to 16%, the NO3–N concentration decreased from 5.3 to 4.2 mg L–1. Therefore in this basin, VBWs reduced the NO3– concentrations in streams with sources in agricultural fields by 30%. This work demonstrates the contribution of natural VBWs to NO3– removal at the catchment scale compared to other sources of variation, which is a current need for integrating water quality criteria into wetland management.

Abbreviations: Ci, CORPEN index (agricultural nitrogen surplus) • TNI, total nitrogen input • VBW, valley bottom wetland • NO3–Ncd, nitrate concentration corrected for dilution • NO3-NcN, hypothetical NO3–N concentration for a zero agricultural N surplus


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSION
 REFERENCES
 
THE REDUCTION OF NUTRIENT LOADING between uplands and streams is sometimes considered evidence of the effect of wetlands acting as buffer zones (Fisher and Acreman, 2004). Nitrogen removal has currently been described in numerous types of wetlands. In valley bottom wetlands (VBWs), N removal assessment is often performed at a local scale (Haycock and Pinay, 1993; Hill, 1996; Haycock et al., 1997; Hefting, 2003; Machefert and Dise, 2004), generally showing a decrease in NO3– concentration due to denitrification, vegetation uptake, or dilution. By contrast, the influence of VBWs on NO3– loading has seldom been assessed at the catchment scale (Johnston et al., 1990; Jansson et al., 1998; King et al., 2005). Nevertheless, some authors have studied the influence of scale on the relation between water quality and landscape features (Johnson et al., 1997; Spruill, 2000, 2004; Gove et al., 2001; Sponseller et al., 2001; Strayer et al., 2003; Buck et al., 2004; Burt and Pinay, 2005). We often have the paradox of efficient wetlands with low NO3– concentration at a local scale close to streams that, in contrast, exhibit high amounts of NO3– downstream (Burt, 2005). Therefore, we are still faced with the challenge of quantifying the impact of VBWs on nutrient removal at the catchment scale to support the conservation, restoration, or creation of wetlands (Merot et al., 2000, 2006; Kennedy, 2001; Trepel and Palmeri, 2002; Viaud et al., 2004).

The present study aims to quantify the effect of VBWs on NO3– removal in headwater catchments throughout a large agricultural basin exhibiting various topographic, climatic, and farming intensity conditions.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 CONCLUSION
 REFERENCES
 
The Study Area
The study was performed in the Scorff River Basin (480 km2), in the Brittany region of France (48° N, 3°12' W). The Scorff is a fifth-order river (Strahler, 1957) that runs into the Atlantic Ocean (Fig. 1). The basin is mainly underlain by granite (80%) and schist (20%), with a stream slope varying from 0.0070 to 0.0015 m m–1, according to these respective substrates. The maximum height above sea level is 270 m. Due to the impervious substratum, groundwater bodies are shallow, particularly in the valleys. The upslope domain is occupied by well-drained soils, whereas the valleys are mainly covered by waterlogged soils (glossaqualf and fluvent) corresponding to VBWs.


Figure 1
View larger version (62K):
[in this window]
[in a new window]
  		Fig. 1. Geographic location of the Scorff basin within the Brittany region and delimitation of the 18 selected catchments (S1 to S18).

 
The region has a mild oceanic climate, with a mean annual temperature of 11°C. The mean annual rainfall varies from 950 mm south of the basin to 1100 mm north of the basin. In parallel, the actual evapotranspiration was roughly estimated as up to 750 mm in the south and 650 mm in the north (Guyot et al., 1975). Therefore, the mean annual runoff of the tributaries varies from 200 mm in the south to 450 mm in the north.

Agriculture covered 60% of the basin, and most of the remaining area was forested (30%). In 2002, 46% of the agricultural area was covered by pasture, 29% by maize (Zea mays L.) and 22% by cereals (Hubert-Moy et al., 2003). A total of 171 kg N ha–1 yr–1 was applied as fertilizer, with 100 kg N ha–1 yr–1 of organic compounds and 71 kg N ha–1 yr–1 of inorganic compounds (Tachez, 2005).

With an average value of 6.5 mg L–1 between January and May 2005, the NO3–N concentration of the Scorff River was close to the regional mean concentration in streams (6.5 mg L–1 in 2004). In the Scorff River, this concentration has been decreasing slightly since 2000 after a long period of degradation beginning in the 1970s, with a concentration of around 1.3 to 2.3 mg L–1 in 1974 (Merot, unpublished data, 1975).

Eighteen catchments, denoted S1 to S18 (Fig. 1), and covering a total area of 163.8 km2, were selected among the catchments surveyed since 2001 for monthly NO3– concentrations by the Scorff River Authority. They were chosen with similar areas and had a maximum stream order of three (Strahler, 1957), in the absence of any tidal influence, while exhibiting the greatest possible diversity in landscapes and water quality.

Chemical and Hydrological Measurements at Catchment Outlet
The catchment, as do many catchments, displays two contrasted hydrologic behaviors during the hydrologic year, one encountered in winter and early spring (the high flow period), and a second one encountered in late spring and in summer (the low flow period). This has major consequences on water quality and wetland impact (Merot and Durand, 1995). We only sampled during the wet period of the year for this large catchment, and therefore our results may not apply to the entire hydrologic year. Water was sampled from January through May 2005 (a period of high flow) in all of the 18 catchments during inter-storm periods approximately every 10 d. Water was sampled in the middle of the stream channel by suction with a syringe and then filtered through a disposable capsule filter (Hydrophilic PVDF 0.45 µm, Millipore Millex-HV). Samples were maintained at 4°C before analysis. Concentrations of NO3–N, Cl–, and SO4–S were measured using ion chromatography (Dionex, Sunnyvale, CA) with a precision of 3, 5, and 5%, respectively. We also used chemical data collected between 2000 and 2004 by the Scorff River Authority.

Regional maps of precipitation and evapotranspiration allowed us to observe the climatic gradient from south to north in the Scorff Basin, while showing that the discharge of the northern tributary was twice that of the southern tributary. Nevertheless, due to the lack of sufficiently dense networks of rainfall and evapotranspiration measurements at the scale of the 18 catchments, we needed to use other methods to evaluate the effective rainfall and specific discharge gradients that partially control the concentration of dissolved chemical elements. To check the gradient, we first measured the discharge simultaneously at the 18 catchment outlets, on the same day and during baseflow, with an electromagnetic current flow meter (Nautilus C2000, OTT) and using the velocity-area method. Then, we assumed that the Cl– concentration was an indicator of the degree of solutes diluted by water, i.e., the gradient of specific discharge (Neal et al., 2004). The higher the Cl– concentration, the lower the specific discharge. The concentration in tributary waters depends on the dilution due to the difference in discharge, which is itself a consequence of the gradient of effective rainfall at the basin scale. The concentration of Cl–, considered a nonreactive element (Altman and Parizek, 1995), has often been used as a dilution indicator for studying the relationship of water chemistry to elevation (Barbier, 2005), response time in catchments (Kirchner et al., 2000), or discharge in a delta (de Cabo et al., 2003). Here, we assumed that the Cl– concentration gradient was an indicator of the gradient of discharge or so-called effective rainfall between catchments. This point will be more deeply considered in the discussion section.

Methods for Characterizing Catchments
Topography, Hydrography, and Delimitation of Valley Bottom Wetlands
The topographic characteristics of the catchments (Table 1) were based on a Digital Elevation Model and computed with the dedicated software MNTSurf (Aurousseau and Squividant, 1995). The spatial characteristics were computed in raster mode and then vectorized to be used in GIS software (ArcGIS 8.3, ESRI) for comparison with land cover characteristics. The land cover characteristics extracted from satellite images acquired in 2002 were used to distinguish forests, crops (maize, cereals, legumes), and grassland areas. These characteristics were calculated for each catchment and expressed as a percentage of the catchment area.


View this table:
[in this window]
[in a new window]
  		Table 1. Characteristics of catchments: topography, land-use, wetland dimensions, hydrology, and chemical data measured and calculated. All analyses are performed by the Unité Mixte de Recherche Sol Agronomie Spatialisation. Institut National de la Recherche Agronomique.

 
Due to the lack of spatial ground-based characterization of VBWs on the Scorff Basin, we applied the method proposed by Merot et al. (2003, 2006) for predicting wetland delineation in small catchments. Valley bottom wetlands were defined in two steps. The first step predicted the potential VBWs, i.e., wetlands derived from catchment geomorphological features. The second step extracted the existing VBWs, i.e., wetlands unmodified by anthropic activity among the set of potential VBWs. The potential VBWs (Fig. 2) were defined using a topographic index, taking into account the local slope and upslope drainage area, following an approach first proposed by Beven and Kirkby (1979).


Figure 2
View larger version (44K):
[in this window]
[in a new window]
  		Fig. 2. Potential wetlands mapped from a topographic index.

 
Land cover for existing VBWs in this catchment was mainly represented by wet meadows, small woodlands, or in some cases, peatlands or marshlands. We assumed that when part of a potential VBW was cultivated, this part no longer acted as a wetland and as a buffer zone because it had generally been drained before cultivation. Therefore, the existing VBW area corresponded to the previous potential VBW area minus the cultivated area included in the potential VBW area (Fig. 3).


Figure 3
View larger version (61K):
[in this window]
[in a new window]
  		Fig. 3. Existing wetlands derived from potential wetlands using land-use data.

 
Input and Loss of Agricultural Nitrogen
Calculations were made first for every farm, then results for every farm inside a catchment were added to give the value for each catchment. Information came from questionnaires on the fertilization plan performed by the local farmer advisory service (Chambre d'Agriculture du Morbihan). The fertilization plan accounted for organic and mineral N input (including import and export of manure between farms as currently implemented to reduce the organic surplus on some farms). For each farm, f, we calculated the total nitrogen input (TNI) spread on the available agricultural area, AAA (TNIf, kg N haAAA–1 yr–1), and an estimation of the agricultural N surplus known as the CORPEN index (Cif, kg N haAAA–1 yr–1) (CORPEN, 1988; Benoît, 1992). The CORPEN index was the difference between major sources of N in the soil and an estimation of the N plant uptake and export (or import). The major sources of N considered were the mineral input, the easy mineralizable fraction of organic input of the year n, and the slightly mineralizable fraction of organic input of the year n-1. The Cif was negative for a N deficit at the farm scale, and positive for a surplus. Before pooling the N budgets at the catchment scale, Cif was forced to zero because a within-farm deficit could not compensate for a N surplus in another farm located elsewhere in the catchment. Finally, the TNI and the Ci were computed at the catchment scale (and therefore noted TNIc and Cic) using Eq. [1] and [2], with Cif ≥ 0;

Formula 1[1]

Formula 2[2]
where CA was the catchment area, TNIc was expressed in kg N haCA–1 yr–1, and Cic expressed in kg N haCA–1 yr–1.

Data Analysis and Statistical Method
Each catchment was characterized by its topographical parameters, including valley bottom characteristics, input and surplus of agricultural N, as well as NO3–N and Cl– mean concentrations (Table 1). To assess the buffering role of the VBWs, we first performed a step-by-step regression between chemical species concentrations and catchment characteristics. As a preliminary, we checked the independence of the explicative variables.

Parametrical statistical tests were used for the analysis because of the normality of the data distribution (calculated with the Kolmogorov-Smirnov test). The Pearson product moment test was used to determine if differences in NO3– concentration were correlated with agricultural, hydrological, and landscape characteristics, particularly with VBWs. Differences were considered significant at p < 0.05. The step-by-step regression was finally checked by a multiple linear regression. Prediction quality was estimated by a Student test between measured and calculated values after checking for normality. Similarity was considered significant at p > 0.80. Prediction quality was also tested using the average error and quadratic average error calculated from the difference between measured and calculated values. All of these statistical tests were performed with SigmaStat 3.0 software.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 CONCLUSION
 REFERENCES
 
Statistical Characteristics of the Nitrate-Nitrogen Concentration Data
Figure 4 shows the statistical characteristics of the NO3–N concentration data. The average outlet NO3–N concentration between January and May 2005 was 6.5 mg L–1. The minimum value was 2.5 mg L–1 for S14 and the maximum was 9.8 mg L–1 for S15. Due to the short period of sampling in 2005, we needed to check its representativity. Comparison of the 2005 average NO3–N concentrations with the 2000 through 2004 average concentrations for the different catchments indicated a similar gradient between catchments (Table 2a; r = 0.94; p < 0.01), but with a 14% higher average concentration in 2005. Moreover, between January and May 2005, the range of NO3–N average concentrations between catchments reached 3.3 mg L–1, whereas it was 4.7 mg L–1 in 2000 through 2004 for the same winter and spring months. Higher variation range and lower average concentrations can be explained by interannual climatic differences, particularly considering the especially wet hydrological year 2000 through 2001. Nevertheless, as shown by the similar NO3–N concentration gradients for both periods, the 2005 gradient was assumed to reflect permanent and stable differences between catchments.


Figure 4
View larger version (24K):
[in this window]
[in a new window]
  		Fig. 4. Box plot of average NO3–N concentrations sampled every 10 d from January through May 2005 for each subcatchment and during baseflow (the boundary of the box closest to zero indicates the 25th percentile, a line within the box marks the median, the boundary of the box farthest from zero indicates the 75th percentile, whiskers above and below the box indicate the 90th and 10th percentiles, and other points are outlying points).

 

View this table:
[in this window]
[in a new window]
  		Table 2. Regression analysis between different datasets and characteristics of the catchments. (r, correlation coefficient; n, number of sites; p, probability of being wrong).

 
Agricultural Characteristics of Catchments
Agriculture accounts for 44.6 to 84.6% of the catchment area, with an average of 68.2%. The TNI expressed for the whole catchment area was on average 121 kg N ha–1 yr–1 and was between 92 and 144 kg N ha–1 yr–1. The TNI was strongly correlated with the proportion of agricultural area (Table 2b; r = 0.92; p < 0.01) in each catchment because there was little variation in the N load per unit of available agricultural area. On the other hand, the N surplus estimated from the Ci was more variable because it also included the type of crop production. The Ci was independent of topographic variables and land cover. The mean N surplus was approximately 32 kg N ha–1 yr–1 and ranged between 10.5 and 58.1 kg N ha–1 yr–1.

Spatial Variability of Valley Bottom Wetland Area
Relative potential VBW area (VBWpot) varied from 16.6 (S14) to 25.1% (S7) of the total catchment area, with an average of 20.5%. The VBWpot exhibited a significant negative correlation with forested area (Table 2c; r = –0.47; p = 0.05). The correlations of VBWpot with land cover data were explained by the relationship between these two variables and the topography. Moreover, VBWpot was correlated with average catchment slope (Table 2d; r = –0.70; p < 0.01). Although the relative forested area did not show a correlation with average slope, the Pearson product moment was close to significance (Table 2e; r = 0.43; p = 0.07). This result indicated that VBWpot was not an independent variable, and thus did not represent VBW characteristics alone.

The relative area of existing VBW (VBWexist) varied from 10.8 (S6) to 16% (S9), with an average of 13.2%. On average, the VBWexist represented 70% of the VBWpot. The VBWexist did not show any significant correlation with other landscape variables. This variable was therefore considered independent.

Step-by-step Regression between Valley Bottom Wetland Characteristics and Nitrate-Nitrogen Concentration
As a first step, we investigated the overall relationship between the mean NO3– concentration and VBW characteristics. The statistical analysis revealed a positive correlation between NO3– concentration and VBWpot (Table 2f; r = 0.56; p = 0.02). This correlation, which was in contradiction with the hypothetical buffer role of VBWs, was likely due to the relationship between VBWpot and landscape variables. In contrast, statistical analysis failed to reveal any significant correlation between NO3– concentration and VBWexist (Table 2g; r = 0.10; p = 0.69).

Nitrate concentration was significantly correlated with agricultural characteristics, such as relative agricultural area, AA (Table 2h; r = 0.61; p < 0.01), or Ci (Table 2i; r = 0.74; p < 0.01). The agricultural activity partially explained the spatial variation of NO3– concentration. However, NO3– concentration was also correlated with Cl– concentration (Table 2j; r = 0.63; p < 0.01) and sulfate concentration (Table 2k; r = 0.62; p < 0.01). Chloride concentration was independent of agricultural characteristics, with no correlation observed with either relative AA (Table 2l; r = 0.12; p = 0.64) or Ci (Table 2m; r = 0.15; p = 0.61).

In a second step, we assessed the influence of the degree of dilution on NO3– concentration. We attempted to describe the bulk effective rainfall gradient given in the introduction as representing a general trend in this basin. First, we measured streamflow at the 18 catchment outlets and analyzed the variations of specific discharge. For this analysis, S14 was excluded from the sampling set because it showed an unusually low stream flow due to local high leakage. The specific streamflow varied from 0.085 to 0.165 L s–1 ha–1 and was positively correlated with the mean elevation (Table 2n; r = 0.68, p < 0.01). The spatial variation of specific stream flow confirmed a south-north gradient of effective rainfall dilution that increased by a factor of two. Second, we checked the factors involved in the variation of Cl– concentration. Figure 5 shows a significant negative correlation of Cl– concentration with specific flow (r = –0.64; p < 0.01), and with mean catchment elevation (r = –0.90; p < 0.01). The correlation reflects the gradient at the basin scale, quantified by the twofold increase in Cl– concentration and the decrease in specific flow from north to south in the basin.


Figure 5
View larger version (26K):
[in this window]
[in a new window]
  		Fig. 5. Variation of the average chloride concentration between January and May 2005 as a function of specific flow and average catchment elevation.

 
Because the mean NO3– concentration of the 18 catchments was correlated with the mean Cl– concentration (Table 2j), we assumed this correlation was essentially due to the degree of dilution depending on the difference of discharge between catchments. To normalize each catchment to the same dilution condition, chosen as the mean dilution for all the tributaries, we corrected the NO3–N concentration values using Eq. [3]:

Formula 3[3]
where NO3–Ncd is the nitrate concentration corrected for dilution, NO3–N is the temporal average of measured nitrate concentration for the catchment, Cl is the temporal average of measured Cl– concentration for the catchment, and ClScorff is the Scorff Basin Cl– concentration (temporal and spatial average for the whole catchment).

In a third step, we analyzed the relationships between NO3–N concentration after correction for dilution effects, VBW, and agricultural characteristics. No significant correlation was observed with VBWpot (Table 2o; r = 0.43; p = 0.08) or VBWexist (Table 2p; r = –0.04; p = 0.88). The two correlation coefficients had a negative value. We then assessed the effect of agricultural intensity on NO3–Ncd. By contrast, we found a moderately significant relationship between agricultural characteristics and NO3–Ncd. The NO3–Ncd was correlated (Fig. 6) with the three agricultural characteristics, i.e., relative AA (r = 0.71; p < 0.01), TNI (r = 0.65; p = 0.01) and N surplus (Ci: r = 0.68; p < 0.01). The relationship with the N surplus (Fig. 6c) exhibited a clear trend, with a homogeneous distribution of the samples except for one catchment (S13). We assumed that this catchment can be excluded from the sampling set due to its erratic behavior. Piezometric observations not included here showed a different behavior of this sampling site compared to the others. Removing this catchment from the statistical analysis led to a strong increase of correlation significance (r = 0.88; p < 0.01).


Figure 6
View larger version (23K):
[in this window]
[in a new window]
  		Fig. 6. The NO3–N concentration corrected for rainfall impact, plotted as a function of agricultural characteristics: (a) relative agricultural area; (b) total nitrogen input, and (c) nitrogen surplus (the solid line, dashed line, and dotted line represent, respectively, the fitted regression line, the confidence interval (95%), and the prediction interval (95%)).

 
To analyze the role of wetlands as buffers, we modified the NO3–Ncd concentrations in the same way as for dilution to reduce the variation between catchments due to differences in agricultural inputs. We calculated the hypothetical NO3–N concentration in the stream (NO3-NcN) for a zero agricultural N surplus (Ci) for each catchment using Eq. [4]:

Formula 4[4]
where 0.062 is the slope of the regression line fitted between N surplus and NO3–Ncd concentration adjusted for dilution, without S13 (Fig. 6c).

The final step was to analyze the relationship between NO3–NcN and wetland characteristics. No significant correlation was observed with VBWpot (Table 2q; r = –0.18; p = 0.56), but NO3–NcN was significantly correlated with VBWexist (Fig. 7; r = –0.77; p < 0.01).


Figure 7
View larger version (19K):
[in this window]
[in a new window]
  		Fig. 7. The NO3–N concentration corrected for agriculture and rainfall impact, plotted as a function of existing wetlands area. The solid line, dashed line, and dotted line represent, respectively, the fitted regression line, the confidence interval (95%), and the prediction interval (95%).

 
The regression line fitted in Fig. 7 expresses the relationship between NO3–NcN concentration and VBWexist, without S13:

Formula 5[5]

For similar rainfall, evapotranspiration, and farming pressure, we observed that when VBWexist varied from 11.5 to 16% (i.e., a relative increase of 40%), the NO3–NcN concentration decreased from 5.3 to 4.2 mg L–1.

Multiple Regression between Corrected Nitrate Concentration and Characteristics of Subcatchments
Finally, from this step-by-step regression, we assessed the nitrate removal efficiency of wetlands under the conditions of this study with the following formula:

Formula 6[6]

To check the step-by-step regression, we computed a multiple regression between NO3–N concentration corrected for dilution (plotted as a dependent variable), and all landscape variables (plotted as independent variables). As explained previously, NO3–N concentrations can be predicted from a linear combination of relative existing wetland area and agricultural N surplus (r2 adjusted = 0.89; p < 0.01):

Formula 7[7]

The coefficients of influence for these two variables were equivalent to those calculated previously. These two variables contributed to predicting an NO3–Ncd concentration with a p value lower than 0.01 (significant contribution for p < 0.05).

Prediction quality was first estimated by a Student t test between NO3–Ncd measured and NO3–Ncd calculated. The normality (P > 0.05) and equal variance (P = 0.81) of the two variables was estimated. Result of t test indicated no statistical differences between the two variables (P = 0.96; {alpha} = 0.80). The mean root quadratic error was about 0.26 mg L–1 and the bias or mean error was –0.02 mg L–1.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
CONCLUSION
 REFERENCES
 
Impact of Valley Bottom Wetlands
The analysis allowed us to measure the influence of variations in the area of VBWs on the NO3–N concentration in the Scorff tributaries. If we extrapolate the curve given in Fig. 7, we can calculate a hypothetical NO3–N concentration in the absence of wetland, yielding a value close to 7.5 mg L–1. From this, we can infer the overall role of VBWs. Based on the mean NO3–N concentrations in the 18 catchments, we estimated a 30% reduction in NO3–N concentrations due to wetlands. This average decrease converged with the estimate obtained from hydrochemical catchment modeling in the region (Durand et al., 1999; Basset-Mens et al., 2006; Durand et al., 2006). Such a strong role is also consistent with local estimation of denitrification in Brittany. It is in particular due to the mild temperature (Clement et al., 2002, 2003). We noted also that the Scorff catchment had particularly well preserved wetlands close to the stream network, as shown by the comparison between potential and existing wetlands. A question could be raised concerning the impact of the length of contact between wetlands and the potential source of pollutants, as compared with the influence of their areal extent. This length of contact was often considered as more important than the areal extent (Bidois, 1993; Skaggs et al., 1994; Beaujouan et al., 2001, 2002; Sabater et al., 2003). However, this variable was not easily derived from GIS, so it had not been studied in this case.

Physical and Agricultural Factors
The present analysis showed that the impact of the wetlands was not of primary importance, and that it was masked by dilution effects and agricultural factors. The hidden role of wetlands can also be explained by the small relative variation of their surface area (from 11 to 16%) within catchments. Even if there was a 40% variation of the area covered by wetland, this only produced a variation of 5% in the catchment surface area.

We could also question the relevance of the indicators chosen for dilution and agricultural factors. The use of Cl– as a dilution tracer is questionable in areas close to the sea and in agricultural regions. The main sources of Cl– are atmospheric deposition (Hutton, 1976) and agricultural inputs of KCl in organic and mineral fertilizers (Martin, 2003), whereas inputs from the parent rock are insignificant (Lockwood et al., 1994). In the Central Massif, Barbier (2005) showed an excellent exponential relation between elevation (over a gradient of more than 1000 m) and Cl– concentration for sampling sites not influenced by agriculture. He deduced from data that an increase of 300 m in elevation leads to a twofold decrease in Cl– content in natural water, due to the combined gradients of precipitation and evapotranspiration. In the present study, we found a gradient with the same order of magnitude. Barbier (2005) also considered that the effect of elevation was dominant compared with the effect of distance to the sea. Many authors have used Cl– to provide a signature of dilution at the scale of a wetland or small catchment (Altman and Parizek, 1995; Mengis et al., 1999; Sabater et al., 2003). However, Cl– has not often been used to indicate dilution at a scale corresponding to the Scorff Basin. The present study showed that, in the absence of spatialized rainfall and evapotranspiration data, we can assess the climatic (effective rainfall) gradient from the average content of Cl–. In this way, we were able to quantify and correct for the influence of this spatial dilution on NO3–N concentrations.

The relative AA appeared to be a good index of the effect of agricultural pressure on NO3–N content in the Scorff Basin, where the agricultural production systems were relatively homogeneous. Whereas we can significantly improve the correlation by using the agricultural N surplus measured by means of an agronomic budget or the Ci, the effect was less marked than we might expect. Nevertheless, in Fig. 6c we observe that for a agricultural N surplus equal to zero, the concentration, based on the regression line (without the catchment S13), reaches 4.7 mg L–1. This concentration seems too high compared with the value of 1.3 to 2.3 mg L–1 measured during the less intensive agricultural period. Four reasons may explain this high value. Although the Ci was pooled over the catchment, it was first calculated at the overall farm scale. This implied that the calculation produced a compensation of surpluses and deficits located at the field level, whereas no compensation occurred in the real world. In this case, the Ci underestimated the agricultural N surplus. This underestimation was more important for heterogeneous agricultural practices. Second, the agricultural data were collected by surveys with farmers. In this case, we observed a bias that led to an underestimation of the surplus by the farmers. The third reason was that outputs at the catchment level could be out of equilibrium with the agronomic budget, due to the response time in catchments (Molénat et al., 2002, 2005) when the recent application of Best Management Practices tended to lower the N inputs. Lastly, some other processes like atmospheric N deposition, soil mineralization, or N plant fixation were not taken into account and could also explain the high NO3–N values. Nevertheless, we assumed that the differences were the same for all the 18 catchments.

The primary interest of this work was that it gave an order of magnitude for the impact of natural VBWs on nitrate removal at the catchment scale. With the implementation of the European Water Framework Directive, such benchmarks are currently needed in Europe as guidelines for integrating water quality criteria into wetland management.


   CONCLUSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSION
REFERENCES
 
The aim of this study was to assess the buffering effect of VBWs on NO3–N concentrations at the scale of a 400-km2 basin. By comparing different catchments included in this basin, we were able to attribute part of the decrease in NO3–N concentrations to the areal extent of VBWs in each catchment. For a combination of N surplus and runoff in this catchment, an increase in VBW from 11.5 to 16% decreased the NO3–N concentration from 5.3 to 4.2 mg L–1. Therefore, the overall impact of VBWs in this catchment reduced the N concentration in streams with sources in agricultural fields by 30%.

The variability of natural and anthropic factors often masks the impact of landscape buffer structures at this large scale. We have first to precisely assess these factors such as precipitation, geomorphology, and land use and land cover, coupled with the variation of the agricultural N budget at the catchment scale. In this case study, the two main factors controlling N variability were the amount of effective rainfall, i.e., the combination of precipitation and actual evapotranspiration on discharge and chemical dilution, and the intensity of farming, i.e., the area used for farming in the catchment and the agricultural N surplus.


   ACKNOWLEDGMENTS
 
The authors would like to express their thanks to Christophe Tachez (chambre d'agriculture du Morbihan) for the agricultural data, Thierry Mounier (syndicat de bassin du Scorff) for chemical and geographic data, Laurence Carteaux and Beatrice Trinkler for their analytical assistance, and Nicolas Jeannot (INRA, U3E) and Nicolas Gillet for their help for field work.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSION
 REFERENCES
 




This article has been cited by other articles:


Home page
 J. Environ. Qual.Home page
D. Moreno-Mateos, U. Mander, F. A. Comin, C. Pedrocchi, and E. Uuemaa
Relationships between Landscape Pattern, Wetland Characteristics, and Water Quality in Agricultural Catchments
J. Environ. Qual., October 23, 2008; 37(6): 2170 - 2180.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Montreuil, O.
Right arrow Articles by Merot, P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Montreuil, O.
Right arrow Articles by Merot, P.
Agricola
Right arrow Articles by Montreuil, O.
Right arrow Articles by Merot, P.
Related Collections
Right arrow Wetlands and Aquatic Processes
Right arrow Ecosystem Management
Right arrow Watershed-Scale Studies
Right arrow Nutrient Management

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Vadose Zone Journal
Soil Science Society of America Journal Journal of Plant Registrations The Plant Genome