Nitrogen retention in natural Mediterranean wetland-streams affected by agricultural runoff

Nitrogen retention efficiency in natural Mediterranean wetland-streams affected by agricultural runoff was quantified and the effect of the temporal variability and hydrological/chemical loading was examined from March 2007 to June 2008 in two wetland-streams located in Southeast Spain. Nitrate-N (NO−3 -N), ammonium-N (NH + 4 -N), total nitrogen-N (TN-N), total organic nitrogen-N (TON-N) and chloride (Cl ) concentrations were analyzed to calculate nitrogen retention efficiencies. These wetland-streams consistently reduced water nitrogen concentration throughout the year with higher values for NO −3 -N (72.3%), even though the mean value of inflow NO −3 -N concentrations was above 20 mg l−1. Additionally, they usually acted as sinks for TONN (8.4%), but as sources for NH +4 -N. Over the entire study period, the Taray and Parra wetland-streams were capable of removing on average 1.6 and 0.8 kg NO −3 -N a day −1, respectively. Retention efficiencies were not affected by temperature variation. NO−3 -N retention efficiency followed a seasonal pattern with the highest retention values in summer (June–September). The temporal variability for NO −3 -N retention efficiency was positively and negatively explained by the hydrologic retention and the inflow NO −3 -N concentration ( R2 adj=0.815,p <0.01), respectively. No significant regression model was found for TON-N and NH +4 -N. Finally, the conservation of these Mediterranean wetlandstreams may help to not only improve the surface water quality in agricultural catchments, but to also achieve good ecological status for surface waters, this being the Water Framework Directive’s ultimate purpose. Correspondence to: V. Garćıa-Garćıa (viquigar@um.es)


Introduction
Nitrogen is an essential nutrient for aquatic ecosystem functioning.Its variation influences community structure, microbial activity and primary production (Pringle, 1990;Peterson et al., 2001;Dodds et al., 2002).In recent years however, nitrogen (N) concentrations have increased in many areas as a result of human activities and have important negative effects on natural ecosystems (Townsend et al., 2003;Niyogi et al., 2004).Therefore, a great deal of attention has been paid to the movement (fluxes) and transformation of N, especially in streams (Peterson et al., 2001;Kemp and Dodds, 2002;Gücker and Boëchat, 2004).
Agricultural runoff is an important source of non point pollution of aquatic ecosystems, causing eutrophication through nutrient load enrichment (Peterjohn and Correll, 1984;Mitsch et al., 2005).Unlike point source pollution, diffuse pollution is less easily controlled and its reduction can only be achieved by appropriate land management techniques.
Over the last few decades, much interest has been manifested in specific natural systems, such as riparian zones which are able to reduce or buffer the flux of N from terrestrial to aquatic ecosystems (Lowrance et al., 1984;Groffman et al., 1992;Sabater et al., 2003).In general, wetlands can improve water quality through physical, chemical and biological processes that remove N from water (Howard-Williams, 1985).This is possible because they have zones of high primary productivity in surface environments and decomposition in sediments that create coupled aerobic and anaerobic transformations of N molecules that pass through them (Bowden, 1987).The role of wetlands in removing N from runoff surface waters is globally recognized (Lowrance et al., 1984;Fisher and Acreman, 2004), but the extreme variability of biological and hydrological processes make it difficult to predict the efficiency of N retention of the different types of wetlands.
V. García-García et al.: N retention in natural Mediterranean wetland-streams Nitrogen retention efficiency in constructed wetlands has been extensively studied for wetlands to be used in conjunction with agricultural drainage and wastewater treatment systems (Spieles and Mitsch, 2000).However, few studies have analyzed nutrient retention efficiencies in natural wetlands (Jordan et al., 2003;Vellidis et al., 2003;Fisher and Acreman, 2004;Knox et al., 2008), despite some studies demonstrating their utility in water quality control on the catchment scale (Mitsch, 1992;Mitsch et al., 2005;Chavan et al., 2008).Indeed, the European Framework Directive (2000/60/EC) emphasizes the role of wetlands as significant elements of the hydrological networks required to obtain a "good water status" for surface and ground waters (Wetlands Horizontal Guidance, 2003).
In the Southeast Iberian Peninsula (Spain), the presence of small wetland-streams is a typical feature of the Mediterranean landscape of sedimentary catchments (Gómez et al., 2005).These wetlands, which are associated with stream drainage systems, intercept the runoff waters originating from the agricultural catchments in which they are located.This spatial arrangement converts wetland-streams into natural tools to control non point pollution.These agricultural areas are typically fertilized with N inorganic salts (KNO 3 ) with a load of 250-300 kg N ha −1 year −1 .
Apart from the studies on N retention in Mediterranean streams (Martí and Sabater, 1996;Sabater et al., 2000;Martí et al., 2004;Von Schiller et al., 2008), there are virtually no studies related with Mediterranean wetlands.
Unlike temperate wetlands, a feature of Mediterranean wetland-streams and other arid and semi-arid aquatic systems is the hydrological intermittency (Gasith and Resh, 1999;Acuña et al., 2005) which strongly influences the structure and functioning of aquatic ecosystems, including N dynamic (Bernal et al., 2005;Von Schiller et al., 2008;Gómez et al., 2009).Moreover, the N concentration and water discharge in aquatic systems affected by agricultural runoff inputs likely show wide temporal fluctuations, mainly due to crop irrigation practices.Over longer time scales, the nature and extent of N input into wetlands will likely affect attenuation processes.By way of example, riparian zones that have been subject to long-term nitrate inputs may have attenuation capacities that differ from non nitrate enriched areas (Groffman et al., 1992).Many studies reported a negative effect of high N concentrations and discharges on N retention efficiency in wetlands (Emmett et al., 1994;Spieles and Mitsch, 2000;Knox et al., 2008).In fact, high N concentrations may have a saturation effect on N microbial and plant uptake (Sabater et al., 2003;Bernot and Dodds, 2005).On the other hand, high water discharges provide short retention times and low surface areas for N exchange per unit volume of water (Peterson et al., 2001;Pinay et al., 2002).
The two objectives of this study were to quantify the N (NO − 3 -N, NH + 4 -N, TN-N and TON-N) retention efficiency in Mediterranean wetland-streams affected by agricultural runoff and to examine the effect of the temporal variability and hydrological/chemical loading on N retention.
An understanding of the N retention capacity of the Mediterranean wetland-streams receiving agricultural runoff is important for several reasons: it may help to determine the key factors driving N retention in these systems; it allows better predictions of how N retention in wetland-streams will vary in response to fluctuations of hydrologic/chemical loading and it allows researchers and managers to design better management plans to control non point pollution in agricultural catchments.

Study site
The study was carried out in two natural wetland-streams, the Taray and Parra wetlands, located in the Murcia Region in Southeast Spain (Fig. 1).The climate of the study area is semiarid Mediterranean with temperate winters and hot, dry summers.Average annual precipitation is 300 mm and the average annual temperature is close to 18 • C.
Wetland-streams are situated at the outlet of small catchments (the mean altitudes are 207 and 172 m over sea level for the Taray and Parra wetlands, respectively) and collect runoff waters from agricultural lands and natural surrounding areas (Fig. 1, Table 1).Surface water flows through the Taray and Parra wetlands and finally, water leaves them via an intermittent channel that flows into the Salada and Parra streams, respectively (Fig. 1).Both wetland-streams are intermittent with periods of low discharge (usually during summer) or even drought periods.Geomorphological features and discharge data are shown in Tables 1 and 2, respectively.
Hydrol.Earth Syst.Sci., 13,[2359][2360][2361][2362][2363][2364][2365][2366][2367][2368][2369][2370][2371]2009 www.hydrol-earth-syst-sci.net/13/2359/2009/The wetland-streams' catchments are characterized by impermeable sedimentary marls (from the Miocene) with a considerable gypsum content (calcium sulfate) and halite (sodium chloride).As a result of this lithology, water conductivity is very high (Table 2) and wetland-stream sediments have a considerable clay and silt content.Natural vegetation in catchments is scarce and dominated by Mediterranean shrubs, including species like Stippa tenacissima, Lygeum spartum and Thymus hyemalis.Wetland-stream plant communities are composed of helophitic species like Phragmites australis and Juncus maritimus, and halophytic species like Suaeda vera, Arthrocnemum macrostachyum and Sarcocornia fruticosa, in the lower flooded areas.P. australis is located in the upper-part of the wetlands with a plant cover that ranges from 47.2% to 58.7% for the Taray and Parra wetlands, respectively.J. maritimus only appears in small patches in the lower part of the Taray wetland.With the exception of small patches of Vaucheria dichotoma, aquatic macrophytes are absent.Periphyton communities are frequent on fine substrates.

Methods
To determine the wetland-stream retention efficiencies for NO − 3 -N, NH + 4 -N, TN-N and TON-N, four sampling transects were located on each wetland, perpendicularly to the water flow direction and with a separation of approximately 100 m (Fig. 2).Sampling transects were opened through vegetation areas to reach the surface water.Surface water samples were collected once a month from the different transects, from March 2007 to March 2008 (13 sampling dates) in the Taray wetland and from April 2007 to June 2008 (15 sampling dates) in the Parra wetland (Fig. 2).The four transects of Parra wetland were dry from July to September 2007, while in the Taray wetland surface water only disappeared in the transect 3 during August and September 2007 (Fig. 2).
Surface water samples were collected with plastic syringes (100 ml) as the water was so shallow (Table 1), and were stored in previously acid-washed polyethylene bottles (500 ml) under dark and cold conditions until they were analyzed at the laboratory.The number of samples per transect varied between 1 and 4, depending on the water sheet width (Fig. 2).The total number of samples per sampling date ranged from 7 to 11 and from 10 to 13 for the Taray and Parra wetlands, respectively.Air and water temperatures, salinity and conductivity (conductivity meter Tetracon 325; WTW, Munich, Germany) and the presence of macrophytes species or periphyton communities were also recorded at each transect.
The outlet discharge was estimated for both wetlandstreams as the product of the average water velocity (current meter MiniAir2; Schiltknecht Co, Zürich, Switzerland) and the cross-sectional area at the wetland outlets (Transect 4, Fig. 2).It was not possible to measure the inlet discharge because of the diffuse surface water inputs to the wetlands.The precipitation data were obtained from the two nearest thermo-pluviometric stations to the studied wetlands (SIAM; Servicio de Información Agrometeorológica, Región de Murcia), Fortuna station that was located approximately 0.4 km from Taray wetland and Abanilla station that was located approximately 5.5 km from Parra wetland.

Chemical analyses
Water samples were analyzed for N dissolved forms within 24 h of collection.They were filtered through glass-fiber filters (Whatman GF/C, 1.2 µm nominal pore size; Whatman International Ltd., Maidstone, England).NO − 3 -N concentration was measured by a colorimetric method following cadmium reduction to nitrite-N (NO − 2 -N) (Wood et al., 1967).NO − 2 -N concentration was analyzed by diazotization (Strickland and Parsons, 1972).NO − 3 -N concentration was estimated by subtracting the NO − 2 -N concentration obtained by diazotization.NH + 4 -N concentration was measured by the phenyl-hypochlorite colorimetric method (Solorzano, 1969).Dissolved inorganic nitrogen (DIN) was calculated as the sum of the NO − 3 -N, NO − 2 -N and NH + 4 -N concentrations.Total nitrogen concentration (TN-N) was measured on unfiltered and frozen samples.These samples were digested to NO − 3 -N using potassium persulfate (D'Elia, 1977) and were analyzed by cadmium reduction using an automated ion analyzer (EasyChem Plus, Systea Analytical Technologies, Italy).TON-N concentration was estimated by subtracting the DIN concentration from the TN-N concentration.Chloride concentration (Cl − ) was analyzed within 48 h of collection by the silver nitrate volumetric method (APHA, 1985).

Retention calculations
Chloride was used to calculate N retention in the wetlandstreams (e.g.Simmons et al., 1992;Sabater et al., 2003).As a passive tracer, Cl − undergoes dispersion, dilution and diffusion, but is not significantly removed from solutions and consequently, its movements largely track water flow.Thus, the variations in Cl − concentration allow the detection of possible dilution (by lateral or subsurface water inputs) or solute concentration (by evapotranspiration) that also affects N forms.
An input-output nutrient budget for a wetland depends on a hydrological budget which in simple terms we assumed for the studied wetland-streams as SW in =E+SW out , where SW in is the inflow surface water, E is the evapotranspiration and SW out is the outflow surface water.With the exception of evapotranspiration as a water output, we assumed no hydrological inputs and outputs through the studied wetlandstreams.Piezometric levels and subsurface Cl − concentration data (not showed in this paper) together with the surface water Cl − concentrations suggested that groundwater inputs (shallow subsurface flow sources) and outputs (surface water infiltration) through the stream-wetlands were negligible.Only in two occasions (August and September 2007, in the Taray wetland), surface water was infiltrated (T3) but few meters downwetland it went back into the wetland surface (before T4).
Thus, retention efficiency (%R) was calculated for the different N forms (NO − 3 -N, NH + 4 -N, TN-N and TON-N) on each sampling date by considering, Eq. (1) (Trudell et al., 1986): N/Cl − in and N/Cl − out are the concentration ratios of both solutes in the inlet (T1) and outlet (T4) of both Hydrol.Earth Syst.Sci., 13,[2359][2360][2361][2362][2363][2364][2365][2366][2367][2368][2369][2370][2371]2009 www.hydrol-earth-syst-sci.net/13/2359/2009/ wetland-streams, respectively.Although to estimate N retention were only used N and Cl − data registered in T1 and T4, data from the rest of sampling transects (T2 and T3) were used to check any possible water input and to control the applicability of the used equations (described above and below).%R is the percentage of the N removed by the wetlands in relation to the inflow of N. A positive retention value indicates that the inflow N/Cl − ratio was higher than the outflow N/Cl − ratio.Under this circumstance wetland-streams were N sinks.On the contrary, a negative retention value indicates that the outflow N/Cl − ratio was higher than the inflow N/Cl − ratio and wetland-streams were N sources.The outflow N load (mg N day −1 ) was calculated as the product of outflow N concentration (mg l −1 ) by outlet discharge (l s −1 ).Under the previously described assumption, the percentage of retention (%R) was applied to the outflow N load to estimate the inflow N load (mg N day −1 ).The N net removal was calculated as follows: Finally, the net hydrologic retention for each sampling date in both wetland-streams was calculated by considering, Eq. ( 3), used by Stanley and Ward (1997): The net hydrologic retention was estimated as an indirect measurement of the water residence time inside the wetland.
Positive values of the net hydrologic retention (<1) indicate that the discharge diminishes as surface water flows through the wetland and as a consequence, the water velocity diminishes and the water residence time increases.The net hydrologic retention is 1 when the wetland is dry.If the discharge does not change through the wetland, the net hydrologic retention is 0. A negative value indicates that the outlet discharge is higher than the inlet discharge as result of surface or subsurface water inputs.Because it was not possible to measure the inlet discharge, due to the diffuse water inputs to the wetlands, this was calculated for each sampling date as follows:

Statistical analyses
The coefficient of variation (CV) for inflow N concentrations and retention efficiencies was used as an indicator of their temporal variability throughout study period.
The relationship between N retention efficiency and the physical, chemical and hydrological parameters was evaluated using Spearman correlations with the SPSS software rel.15.0.1 for Windows (SPSS Incorporated, Chicago, Illinois).Seasonal differences in N retention were analyzed using analysis of variance (one-way ANOVA), followed by proportion, in the range 0-1.The transformed variable (y) was fitted to the multiple regression analysis.Arcsin transformation is usually applied to binomial data in order to approximate normal variance.However, our purpose here was to bind the limits of prediction by multiple regression into the bounds 0-1 (0-100%) as the result of transforming back to original units is constrained to the range 0-1.Not transforming the variable may result on the unrealistic case of multiple regression predicting >100% retention, in some cases.The multiple linear regression analyses were performed with R rel.2.6.0 for Windows (R Development Core Team, Vienna, Austria).

Inflow water characterization
Figure 3 shows the variation of accumulated total precipitation between consecutive sampling dates and the outlet discharge in both wetland-streams during the study period.Temporal variability of the accumulated total precipitation was high and the maximum values were registered mainly in months of spring and fall (March, April and October).The outlet discharge also differed vastly between study months (CV=87.7%and 68.2% in the Taray and Parra wetlands, respectively) but their highest and lowest values did not always correspond with increases or decreases in the precipitation, respectively.Despite the high temporal variability of the outlet discharges, the mean values in both wetland-streams were similar (Table 2).
Table 2 compiles the physicochemical characterization of the inflow water in the wetland-streams during the study period.Although the mean value for inflow TN-N concentration was higher in the Parra than in the Taray wetland (Table 2), the relative contribution of N forms in the inflow water was similar in both wetlands (90.4%, 9.5%, 0.1% and 92.6%, 7.3%, 0.1% as NO − 3 -N, TON-N and NH + 4 -N, respectively).The highest variability in the range of inflow N concentrations throughout the study period corresponded to the Parra wetland.

The inflow NO −
3 -N concentrations in the Taray wetland were consistently similar throughout the study period (CV=8.6%,n=13), while a higher temporal variability was noted for the Parra wetland (CV=37.1%,n=12) (Fig. 4).This difference between both wetland-streams was mainly influenced by the sharp increase of inflow NO − 3 -N concentration (30-43 mg l −1 ) registered from March to June 2008 in the Parra wetland (Fig. 4).The inflow TON-N and NH + 4 -N concentrations varied considerably among the study months (Fig. 4).The CV values for TON-N were 58.0%(n=13) and 131.3% (n=12) in the Taray and Parra wetlands respectively and were 83.6% (n=13) and 117.2% (n=12) for NH + 4 -N.The mean value for inflow Cl − concentration was high and similar in both wetland-streams (Table 2).As NO − 3 -N, inflow Cl − concentrations were similar throughout the study period in the Taray wetland (CV=6.7%,n=13), while they showed a greater temporal variability in the Parra wetland (CV=17.6,n=12) (Fig. 4).On the other hand, decreases of the inflow Cl − concentration generally coincided with increases of the inflow NO − 3 -N concentration in the Parra wetland (Fig. 4).The mean value of the net hydrologic retention was higher in the Taray wetland than in the Parra wetland (Table 2), while their temporal variability was higher in the Parra wetland (CV=84%, n=12) than in the Taray wetland (CV=14%, n=11).

Nitrogen retention efficiencies
When all the sampling data from both wetland-streams were considered, the mean retention efficiency for TN-N was 70.1% (median value = 82.9%,n=25) and it was higher in the Taray than in the Parra wetland (Table 3).Both wetlandstreams showed the highest retention efficiency for NO − 3 -N, followed by TON-N and NH + 4 -N (Table 3).The mean retention efficiency for NO − 3 -N was 72.3% (median value = 84%, n=25), ranging from 31.7% to 100%.However, the mean retention efficiency and net removal for NO − 3 -N was consistently higher in the Taray wetland than in the Parra wetland (Table 3).
The retention efficiency for TON-N was low with a mean value of 8.4% (median value = 56.1%,n=25) and ranged from −437% to 99.5%.The mean retention efficiency and the net removal were significantly higher in the Taray wetland than in Parra wetland (Table 3).There was not removal of TON-N from the water of both wetland-streams on 6 of the 25 sampling dates, as show the existence of negative retention values (Fig. 5).On these occasions, the TON-N/Cl − ratio was higher at the outlet than at the inlet of both wetlandstreams, denoting TON-N exportation.Ammonium-N was not removed from water, but was exported instead on the majority of the sampling dates (13 of 25) (Fig. 5).The mean retention efficiency was −96.7% (median value = −3.2%,n=25), ranging from −1537.5% to 96.0%.As same as for TON-N, the mean retention efficiency was only positive in the Taray wetland and the net removal was also higher in this wetland (Table 3).

Temporal variability of N retention efficiencies
The temporal variability of the retention efficiencies for NO − 3 -N was higher in the Parra wetland than in the Taray wetland; CV=42.7% (n=12) and CV=8.2% (n=13), respectively (Fig. 5).Retention efficiencies for NO − 3 -N increased during the summer (June-September) in both wetlandstreams (Fig. 5).However, differences among seasons were only statistically significant in the Taray wetland (one-way ANOVA, F =29.9, p <0.05).The scarcity data during summer in the Parra wetland (drought period) could be the reason of the absent of statistical significance for this wetland.The maximum NO − 3 -N retention values (99.9% and 96.0%) were recorded in August and October in the Taray and Parra wetlands, respectively (Fig. 5).wetland-streams, the mean retention efficiency for NO − 3 -N (72.3%) was higher than that found in these aforementioned studies, even though the mean inflow concentration for NO − 3 -N was above 20 mg l −1 .Besides, other studies performed in constructed wetlands generally show lower retention efficiencies for NO − 3 -N than our results (Spieles and Mitsch, 2000;Braskerud, 2002;Mitsch et al., 2005).By considering both the annual mean inflow load of NO − 3 -N and the annual mean retention efficiency, the Taray and Parra wetlandstreams were capable of removing mean values of 1.6 and 0.8 kg of NO − 3 -N a day −1 , respectively.Denitrification, biological uptake and microbial immobilization are the main mechanisms for NO − 3 -N removal in wetlands (Reddy and Patrick, 1984;Bowden, 1987;Groffman et al., 1992).These processes are influenced by the hydrologic conditions of wetlands (De Laune et al., 1981;Bowden, 1987;Pinay et al., 2007).In the studied wetlandstreams, NO − 3 -N retention efficiency was negatively correlated with the inlet discharge and positively correlated with net hydrologic retention, thus suggesting that longer water residence times allow a longer time for NO − 3 -N removal from surface water.Nutrient retention in wetlands is governed not only by changes in the hydrographs, but also by both the flow-through (velocity) and water residence time rates (Howard-Williams, 1985).If water moves through a wetland at a quicker rate than that of N retention processes (denitrification or biological uptake), then considerable flow-through of N will take place.Peverly (1982) found that wetlands only retained nutrients when flow-through rates were low, while Stanley and Ward (1997) observed that net retention for all the N forms was strongly correlated with hydrological retention in the Talladega Wetland Ecosystem (TWE, Alabama, USA).
On the other hand, several authors have reported that denitrification may be potentially important in aquatic systems dominated by fine sediments, high NO − 3 -N and organic carbon availability, a low redox potential of sediments, and warm water temperature (Faulkner and Richardson, 1989;García-Ruiz et al., 1998;Inwood et al., 2007;Pinay et al., 2007).Unlike organic matter (and N) accumulation, which conserves N within the wetland, denitrification represents a permanent N loss from the system.Natural wetland sediments are chemically reduced and frequently contain ample organic carbon.Therefore, denitrification in wetlands is generally limited by nitrate availability (Ambus and Lowrance, 1991).Nonetheless, this is not the case of the wetlands affected by agricultural inputs.Therefore, although denitrification was not estimated in the studied wetland-streams, this process is proposed to be an important pathway for NO − 3 -N loss because its occurrence is consistent with their environmental characteristics (high NO − 3 -N availability, high water temperature, anoxic-black sediments and high hydrologic retention).
The retention capacity of wetlands varies seasonally, particularly in temperate regions where biological activity diminishes in winter (Howard-Williams, 1985;Groffman et al., 1992).In fact, studies performed in these regions show that NO − 3 -N retention efficiency is controlled mainly by the temperature (Spieles and Mitsch, 2000;Chavan et al., 2008).In the studied wetland-streams, NO − 3 -N retention efficiency tended to increase in summer months although significant differences among seasons were only observed in the Taray wetland (the lack of statistical significance for Parra wetland may be explained by the absence of data from July to September 2007, during the drought period).However, in contrast with the previously mentioned studies, we did not find correlation between NO − 3 -N retention efficiency and temperature in the studied wetland-streams.One suggestion to explain the lack of correlation between both variables is that the warm temperate climate of the study area enables the continuous operation of the essential biogeochemical processes involved in NO − 3 -N removal.This lack of correlation reinforces the fact that Mediterranean wetland-streams can significantly remove N input.We attributed the increase of NO − 3 -N retention efficiency during the summer months to optimum hydrological conditions, as high net hydrologic retention rates, that favour N proccesing in wetlands (mainly biological uptake and denitrification).The increases of Cl − concentrations during the summer months support the idea of evapotranspiration as the responsible factor of the hydrologic retention increases in the studied wetland-streams.
Wetland-streams acted as sinks for TON-N during most of the study period with net removal mean values of 153.9 and 34.4 g a day −1 for Taray and Parra wetlands, respectively.However, they were also sources for TON-N in some occasions.Similar results were obtained by other authors.For example, Jordan et al. (2003) reported TON-N retention efficiencies ranging from −15.0% to 39.0%.
TON-N retention in the studied wetland-streams, as same as in other wetlands, could be greater than the values obtained by input-output balance.Leaching and decomposition of autochthonous particulate organic matter is an additional source of organic N and decreases net TON-N retention.Decomposition of litter is probably the major source of TON-N in our wetland-streams, as other studies reported (Howard-Williams, 1985;Bowden, 1987;Chapman et al., 2001).In fact, some of these studies show that TON-N concentrations are generally higher in summer and fall and suggest increases relate to the autochthonous litter decomposition or to primary production.In contrast, no seasonal pattern was observed in our wetlands study.Bernal et al. (2005) also reported the absence of such a pattern in TON-N retention for an intermittent Mediterranean stream.
As same as previous studies (Braskerud, 2002), TON-N retention efficiency was positively correlated with the inflow TON-N concentration.In addition, it was positively correlated with the net hydrologic retention, probably because the sedimentation of the organic matter associated with soil particles and the processing rates of TON-N to inorganic forms were higher under greater residence time of water within the wetlands (Jordan et al., 2003).
The studied wetland-streams were usually net sources of NH + 4 -N over the study period.However, when wetlandstreams occasionally retained NH + 4 -N, their retention values were relatively high in comparison with those of previous studies.For example, Braskerud (2002) showed a mean retention value of 1.0% in small constructed wetlands that treat agricultural non-point source pollution.We suggest that litter decomposition and mineralization are the main autochthonous sources of NH + 4 -N in wetland-streams.Once wetland vegetation has died, a large and complex series of nutrient transformations emerges, all of which are associated with the leaching of detritus and simultaneous decomposition (Howard-Williams, 1985).Several studies have demonstrated that plant detritus processing may be an important source of nutrients (Howarth and Fisher, 1976;MacLean and Wein, 1978).Kinetic mineralization of TON-N probably proceeds more rapidly than nitrification, thus NH + 4 -N concentration increases in surface water (Kadlec and Knight, 1996;Braskerud, 2002).
On the other hand, NH + 4 -N is more sensitive than NO − 3 -N to slight changes of local conditions (chemical, physical and biological variables) (Hill, 1996;Butturini and Sabater, 1998;Gücker and Boëchat, 2004), which also change as flow discharge does (Fisher et al., 1998;Von Schiller et al., 2008).Furthermore, NH + 4 -N reacts abiotically via adsorption/desorption reactions, and displays processing lengths that reflect the nature of the sediments and the chemical environment (Triska et al., 1994).Both properties are spatially heterogeneous in wetlands, and this variability increases as flow discharge decreases (Gücker and Boëchat, 2004), which also occurs close to wetland-stream outlets.Thus, slight changes in the sediment redox potential may not only affect the exchange of NH + 4 -N at the water-sediment interface, but may also influence the NH + 4 -N concentration in surface water (De Laune et al., 1981;Bowden, 1987).The fact that NH + 4 -N retention efficiency was lower than that for NO − 3 -N, and that it was even exported from wetland-streams, is consistent with this idea.
The temporal variability of the NH + 4 -N retention was very high in this study and was only positively correlated with the inflow NH + 4 -N concentration.However, Sabater et al. (2000) showed that 83.0% of the seasonal variation in the NH + 4 -N retention efficiency in a Mediterranean stream without riparian vegetation is explained by water temperature.The lack of correlation between other environmental factors and NH + 4 -N retention in the studied wetland-streams may be explained by the high sensitivity of NH + 4 -N concentration to slight changes of the local conditions (sediment redox potential, organic matter content, etc.) as we previously suggested.

Influence of the hydrologic retention and the inflow N concentration on the NO − 3 -N retention efficiency
The main factors controlling the NO − 3 -N retention efficiency in the studied wetland-streams are the hydrologic retention and the inflow NO − 3 -N concentration.We hypothesized that a higher hydrologic retention increases NO − 3 -N retention efficiency through an increase of biological processing rates (as biological uptake and denitrification).
Net hydrologic retention was used as an indirect measurement of the water residence time in wetland-streams.This factor often influences the N retention in aquatic systems because a longer contact time between surface water and Hydrol.Earth Syst.Sci., 13,[2359][2360][2361][2362][2363][2364][2365][2366][2367][2368][2369][2370][2371]2009 www.hydrol-earth-syst-sci.net/13/2359/2009/ sediment implies that the total amount of processed N increases (Peterson et al., 2001;Gücker and Boëchat, 2004).Inflow NO − 3 -N concentration was the second factor controlling the NO − 3 -N retention in the studied wetlandstreams.Other studies in both, riparian buffers and natural/constructed wetlands, report a similar relationship between both variables (e.g.Spieles and Mitsch, 2000;Sabater et al., 2003).In addition, these authors suggest a saturation effect by a high NO − 3 -N load which exceeds the buffering capacity of these systems.Although the inflow NO − 3 -N concentrations registered during the study period were high, they never exceeded the loading capacity of the wetlandstreams, as high NO − 3 -N retention rates indicated.Because almost 100% of NO − 3 -N retention efficiency is obtained with even high inflow NO − 3 -N concentrations (e.g.99.5 % with 21.4 mg l −1 ), the retention efficiencies could not increase at lower inflow concentrations.Lower inflow NO − 3 -N concentrations than 10.3 mg l −1 would be necessary to know the wetland-stream response to low N concentrations.

Conclusions
Our results emphasize the high efficiency of Mediterranean wetland-streams as N sinks all year round.This feature is influenced by low water discharges and probably, by the warm climate, both of which are key factors that make Mediterranean wetland-streams especially interesting compared to temperate wetlands.Wetland-streams consistently reduced NO − 3 -N concentration, the dominant N form, throughout the year.They usually acted as sinks for TON-N and as sources for NH + 4 -N.Hydrological retention and inflow NO − 3 -N concentrations were the main factors explaining the variability in NO − 3 -N removal efficiency.However factors explaining TON-N and NH + 4 -N retention were not found.NO − 3 -N retention showed a seasonal pattern but not directly associated with the temperature, but with the hydrological retention.The highest NO − 3 -N retention effienciencies were detected during summer when evapotranspiration increased and as a consequence, wetland discharge decreased.Stream-wetlands showed a high effiency in N removal even at high N concentrations (21.4 mg l −1 ).Our study in the Taray and Parra wetlands clearly demonstrates the crucial role of Mediterranean wetland-streams in the control of the N flux from agricultural landscapes to aquatic ecosystems located downstream.
The studied wetland-streams consistently reduce the N load, in such a way that the water leaving the wetlands is always of better quality than that entering them.In some countries, surface flow wetlands are highly valued for their high nutrient retention potential and their unique biodiversity.However, despite the high efficiency of the Mediterranean wetland-streams to improve surface water quality, they are often desiccated for new agricultural purposes.Presently, there are an increasing number of activities aimed at restoring these sites as multifunctional landscape entities.In fact, there are studies which focus on identifying the most suitable areas for the restoration of surface flow wetlands to improve the water quality of a given catchment (Mitsch, 1992).The wide distribution and strategic location of the Mediterranean wetland-streams in upstream reaches of basins makes them more interesting as special preservation ecosystems.Our results highlight the conservation interest of Mediterranean wetland-streams for two reasons, to protect wetland biodiversity and to improve the surface water quality in agricultural catchments in accordance with WFD's objective (2000/60/EC).

Fig. 1 .
Fig. 1.Location of the studied wetland-streams and their catchments.T1, T2, T3 and T4 represent the four transects on each wetland where samples were collected.

Fig. 2 .
Fig. 2. Location of the four sampled transects in the studied wetland-streams.In each transect (black lines) the number of samples (s.) collected in the different sampling dates (s.d.) are shown.0 samples mean that the transect was dry.

Fig. 3 .
Fig. 3. Accumulated total precipitation (between consecutive sampling dates) and outlet discharge registered during the study period in the (a) Taray and (b) Parra wetland-streams.

Table 1 .
Surface area, land uses at wetland-stream catchments and geomorphological features recorded in the wetland-streams during the study period.

Table 2 .
Mean, median, 10th and 90th percentile values for solute concentrations, conductivity and temperature of inflow water to wetlandstreams.The inlet/outlet discharges and the net hydrologic retention values are also shown.Inlet discharge values were estimated by Eq. (4) (method section).n=13 and n=12 in the Taray and Parra wetlands, respectively.* = (n=11).

Table 3 .
Concentration, load, net removal, and retention efficiency for TN-N, NO − 3 -N, TON-N, and NH + 4 -N registered at inflows and outflows of the wetland-streams.Values are the mean ± standard deviation based on the data collected over the study period (n=13 and n=12 in the Taray and Parra wetlands, respectively).* = (n=11).

Table 4 .
Results of Spearman correlations between the retention efficiencies (%R) of the different nitrogen forms and the environmental factors by considering the dataset registered during the study period in both wetland-streams.Significant at the 0.01 probability level.