Transport of pesticides and artificial tracers in vertical-flow lab-scale wetlands


  • Romy Durst,

    1. Laboratory of Hydrology and Geochemistry of Strasbourg (LHyGeS), University of Strasbourg/ENGEES, UMR 7517 CNRS, Strasbourg, France
    2. Institute of Hydrology, University of Freiburg, Freiburg, Germany
    Search for more papers by this author
  • Gwenaël Imfeld,

    1. Laboratory of Hydrology and Geochemistry of Strasbourg (LHyGeS), University of Strasbourg/ENGEES, UMR 7517 CNRS, Strasbourg, France
    Search for more papers by this author
  • Jens Lange

    Corresponding author
    • Institute of Hydrology, University of Freiburg, Freiburg, Germany
    Search for more papers by this author

Corresponding author: G. Imfeld, Laboratory of Hydrology and Geochemistry of Strasbourg (LHyGeS), University of Strasbourg/ENGEES, CNRS, 1, rue Blessig, F-67084 Strasbourg Cedex, France. (


[1] Wetland systems can be hydrologically connected to a shallow aquifer and intercept upward flow of pesticide-contaminated water during groundwater discharge. However, pesticide transport and attenuation through wetland sediments (WSs) intercepting contaminated water is rarely evaluated quantitatively. The use of artificial tracers to evaluate pesticide transport and associated risks is a fairly new approach that requires evaluation and validation. Here we evaluate during 84 days the transport of two pesticides (i.e., isoproturon (IPU) and metalaxyl (MTX)) and three tracers (i.e., bromide (Br), uranine (UR), and sulforhodamine B (SRB)) in upward vertical-flow vegetated and nonvegetated lab-scale wetlands. The lab-scale wetlands were filled with outdoor WSs and were continuously supplied with tracers and the pesticide-contaminated water. The transport of IPU and UR was characterized by high solute recovery (approximately 80%) and low retardation compared to Br. The detection of desmethylisoproturon in the wetlands indicated IPU degradation. SRB showed larger retardation (>3) and lower recovery (approximately 60%) compared to Br, indicating that sorption controlled SRB transport. MTX was moderately retarded (approximately 1.5), and its load attenuation in the wetland reached 40%. In the vegetated wetland, preferential flow along the roots decreased interactions between solutes and sediments, resulting in larger pesticide and tracer recovery. Our results show that UR and IPU have similar transport characteristics under the tested subsurface-flow conditions, whereas SRB may serve as a proxy for less mobile and more persistent pesticides. Since UR and SRB are not significantly affected by degradation, their use as proxies for fast degrading pollutants may be limited. We anticipate our results to be a starting point for considering artificial tracers for investigating pesticide transport in environments at groundwater/surface-water interfaces.

1. Introduction

[2] Pesticide loss may reach 30% during application (i.e., via spray drift) and 90% following application (i.e., via volatilization, infiltration, runoff) and eventually impacts surface-water and groundwater [Aubertot et al., 2005]. Wetland systems, or similar organic-rich environments at groundwater/surface-water interfaces, can be hydrologically connected to a shallow aquifer and intercept pesticide-contaminated water during groundwater discharge, in particular during extended flood periods. Pesticides in wetland systems intercepting contaminated water have been investigated under field [Reichenberger et al., 2007; Maillard et al., 2011] and laboratory conditions [Prochaska et al., 2007]. In constructed wetlands, sorption, volatilization, hydrolytic and photolytic oxidation, biological degradation, bioaccumulation, and sedimentation may contribute to attenuation of organic contaminants [Gregoire at al., 2009; Imfeld et al., 2009]. Pesticide attenuation in wetlands largely depends on the pesticide properties [Margoum et al., 2006], wetland sediment (WS) and biogeochemical characteristics [De Wilde et al., 2009a; Si et al., 2011], and the hydrological regime [Stearman et al., 2003]. However, little is known about the transport and attenuation of widely used pesticides in wetlands and similar groundwater/surface-water interfaces intercepting contaminated water.

[3] Isoproturon (3-(4-isopropylphenyl)-1,1-dimethylurea); IPU) is a phenylurea herbicide used to control grasses and broadleaves in cereal production and is commonly detected in European rivers and groundwater [Loos et al., 2009; Sørensen et al., 2003]. IPU is mobile, moderately sorptive [De Wilde et al., 2009b], and subject to microbial degradation under oxic [Sørensen et al., 2003] and anoxic conditions [Larsen et al., 2001]. The main degradation products of IPU are the desmethylisoproturon (N-(4-isopropylphenyl)-N′-methylurea) and the didemethylisoproturon (N-(4-isopropylphenyl)urea), which may pose ecotoxicological risks [Sørensen et al., 2003]. Metalaxyl (methyl N-(methoxyacetyl)-N-(2,6-xylyl)-dl-alaninate; MTX) is a widely used phenylamide fungicide for vine-growing. MTX is mobile, moderately sorptive, and subject to microbial degradation in aquatic ecosystems [Massoud et al., 2008; De Wilde et al., 2009a]. The MTX degradation can lead to the formation of the acid degradation product N-(2,6-dimethylphenyl)-N-(methoxyacetyl)alanine, whereas N-methoxyacetyl-2,6-dimethyl-aniline is formed by N dealkylation either directly from MTX or from N-(2,6-dimethylphenyl)-N-(methoxyacetyl)alanine [Pesaro et al., 2004]. Though negative effects on terrestrial rhizospheric microorganisms have been shown for a fungicide mixture including MTX [e.g., Ahemad and Khan, 2011], there is little knowledge about its fate in aquatic ecosystems.

[4] Artificial hydrological tracers have been used to investigate surface and subsurface transport processes, including retention and mixing in wetlands [e.g., Gooseff et al., 2008] and hyporheic exchanges [e.g., Ward et al., 2010] in studies limited to several hours or days. Bromide (Br) is typically used to evaluate wetland hydraulic properties [e.g., Maloszewski et al., 2006]. Environmentally harmless and low-cost fluorescent dye tracers, such as uranine (UR) and sulforhodamine B (SRB), have been recently used to evaluate the transport of pesticides [Passeport et al., 2010; Lange et al., 2011] and pharmaceuticals [Kunkel and Radke, 2011]. Although biodegradation of UR and SRB has not been reported so far, UR is sensitive to photodegradation, mobile, and less sorptive onto negatively charged surfaces, whereas SRB is more stable to photolysis and subject to sorption onto heterogenerously charged surfaces [Kasnavia et al., 1999; Sabatini, 2000].

[5] In this study, we evaluate and compare the transport of pesticides, i.e., IPU and MTX, and artificial tracers in lab-scale wetlands. The vertical-flow wetlands were designed to study the upward discharge of contaminated water into environments at groundwater/surface-water interfaces. The transport and attenuation of pesticides and tracers at surface-subsurface interfaces were thoroughly characterized during 84 days in a vegetated and a nonvegetated wetland. To the best of our knowledge, this represents the first quantitative study on the joint transport of IPU, MTX, and artificial tracers.

2. Material and Methods

2.1. Chemicals

[6] The physicochemical properties of UR, SRB, IPU, and MTX are provided in Table 1. Br was obtained as sodium bromide from Merck (Darmstadt, Germany). UR and SRB were supplied by Sicomet (Flörsheim, Germany) and Chroma (Münster, Germany). Analytical grade (>99.6%) IPU and MTX were obtained from Sigma-Aldrich (Darmstadt, Germany). Stock solutions of dye tracers and pesticides were prepared in ultrapure water and stored at 4°C and −20°C, respectively.

Table 1. Physicochemical Properties and Molecular Structure of Chemicals
  1. a

    Source: Leibundgut et al. [2009].

  2. b

    Source: Gaspar [1987].

  3. c

    Source: European Commission, Health and Consumer Protection Directorate-General [2002].

  4. d

    Source: European Commission, Health and Consumer Protection Directorate-General [2010].

  5. e

    Source: Available on Pesticide Properties DataBase online (

  6. f

    Source: De Wilde et al. [2007, 2009b].

  7. g

    Source: KOC for UR and SRB based on Kd values taken from Sabatini [2000].

  8. h

    Source: Available on Merck Millipore (

  9. Source: Blatchley et al. [1992].

Molecular formula  BraC20H10O5Na2bC27H30O7N2S2Na2bC12H18N2OcC15H21NO4d
Molar mass g mol−1102.89a376.15b604.67b206.28c279.33d
Photolytic decayDT50photolysisdaysStablea0.5b34a48c6.5d
Hydrolytic decayDT50hydrolysisdaysStableaStablebStableb1560c200d
Decay in soilDT50soildaysStablea  23ef42f
Organic carbon-water partitioningKOCmL g−1 0–62g147–498g36f47f
Octanol-water-partitioning coefficientlog POW  −0.67h−2.02f2.5e1.65c
Henry's law constantHccPa m3 mol−12.21 × 10−5h  1.46 × 10−5c1.6 × 10−5d
Molecular structure   

2.2. Storm-Water Wetland and Field Sediment Collection

[7] Sediment and water used in the lab-scale wetlands were collected in a storm-water wetland located in Eichstetten, southwest Germany (48°0.5′38″N, 7°43′50″E). The wetland was previously characterized by Lange et al. [2011]. WS cores were collected on 16 June 2011 using metal cylinders (ø: 15 cm, height: 50 cm, depth: 10 cm) to collect the upper 10 cm of the WS representative of the biogeochemically active interface between surface water and deeper sediment. One sediment core was collected with a common reed plant (Phragmites australis, Cav.), and another was collected without vegetation. Water was collected from the storm-water wetland and supplied to the lab-scale wetlands. Physicochemical characteristics of WS were as follows (%): coarse sand (200–1000 µm) 3.8, fine sand (50–200 µm) 17.0, coarse silt (20–50 µm) 40.1, fine silt (2–20 µm) 32.7, and clay (<2 µm) 6.4. Sediment bulk density was 0.95±0.02 g cm−3 (mean±SD, n=3), and hydraulic conductivity of sandy silt textures was <10−6 m s−1. The composition of sediment was as follows (%): SiO2 47.4, Al2O3 6.8, MgO 2.2, CaO 17.3, Fe2O3 2.7, MnO 0.1, Na2O 1.0, K2O 1.5, P2O5 0.2, TiO2 0.4, and organic carbon 5.2.

2.3. Lab-Scale Wetlands and Experimental Setup

[8] The upward vertical-flow lab-scale wetlands consisted of borosilicate glass columns (inner diameter: 15 cm, height: 65 cm; Figure 1), which were successively filled, from the bottom to the top, with a 5 cm layer of fine gravel (ø: 1–2 mm), a 27.5 cm layer of medium sand (ø: 400–630 µm), a 8 cm layer of WS, and a 10 cm top layer of medium sand (ø: 400–630 µm). The columns were designed to investigate the influence of upward discharge of the aerobic contaminated water into environments at groundwater/surface-water interfaces. In the upward configuration of the vertical-flow wetland, the contaminated water flows through the fully saturated sediment profile. This configuration enables larger root-water contact and lower residence time variation compared to the downward vertical-flow configuration.

Figure 1.

Scheme of the vegetated lab-scale wetland. Black spots show the position of dissolved oxygen measurement spots across the wetland.

[9] The collected WS cores were homogenized by gentle mixing before packing the lab-scale wetlands in order to obtain similar biogeochemical conditions along the WS profile. Gravel and sand physicochemical characteristics are provided in Table S1 (supporting information). The total wetland volume was 8.94 L, and the pore volume was 4.32 L (48%). The mean bulk density of the filling material was 0.75 g cm−3, and organic content was 0.98%. Dissolved oxygen concentrations were measured in situ using noninvasive oxygen sensors (Presens, Germany) mounted at discrete depths across the lab-scale wetlands (see Figure 1). The water flux was maintained at 0.33 mL min−1 throughout the experiment using a high precision pump (Ismatec, Wetheim-Mondfeld, Germany). The nominal residence time in the wetlands was 218 h. Effluent water accumulated as supernatant (3 cm) before reaching the wetland outflows. Viton (Rotilabo®, Carl-Roth, Karlsruhe, Germany) tubing and stoppers were used to limit sorption of the analytes. The room temperature was maintained at 20±0.5°C, and daily photoperiods (8–16 h) were controlled with light-emitting diode lamps (Philips® GreenPower) for plant growth throughout the experiment. The wetlands were covered with aluminum to limit algal growth and photolytic decay of pesticides and tracers.

[10] The experiment was conducted during 84 days (corresponding to 9.3 pore volumes). After 28 days (3.5 pore volumes), steady-state conditions were reached, and a Dirac pulse injection of 200 mg of Br in each lab-scale wetland for 42 days (4.6 pore volumes) was performed in each lab-scale wetland. Thereafter, 100 µg L−1 of SRB (1.72 × 10−7 mol L−1), 40 µg L−1 (1.94 × 10−7 mol L−1) of IPU, and 50 µg L−1 (1.79 × 10−7 mol L−1) of MTX were spiked and supplied to the lab-scale wetlands. Only 10 µg L−1 of UR (3.01 × 10−8 mol L−1) was spiked due to its higher fluorescence intensity compared to SRB. Water supplied to the wetland was spiked every 5 days with fresh stock solutions of tracers and pesticides. No degradation of fluorescent tracers and pesticides was observed in the stock solutions. The continuous injection of tracers and pesticides ended after 36 days (3.9 pore volumes), and the wetlands were then supplied with unspiked wetland water for 44 days (4.8 pore volumes).

2.4. Flow Measurement and Sampling Procedure

[11] Supernatant water at the wetland outflows was collected using glass syringes at each sampling time. Effluent water volumes and compound concentrations were measured and related to the inflow to establish mass balances and estimate the fluxes. Samples for Br, UR, and SRB analyses were collected every 6–24, 24–48, and 48–72 h, respectively, according to the temporal change in concentrations. Water samples were filtered using 0.45 µm PTFE (Teflon) membranes for Br analysis and using 0.7 µm glass microfiber for UR, SRB, and pesticide analyses. Samples were immediately stored in 100 mL brown glass bottles (water samples) or plastic tubes (filter membranes) and placed at 4°C (fluorescent tracer analysis) or −20°C (pesticide analysis). At the end of the experiment, the wetlands were drained, and sediment samples were collected for tracer and pesticide analyses.

2.5. Chemical Analysis

[12] Electrical conductivity (EC), pH, and dissolved oxygen concentration were directly measured in wetland water using WTW Multi 350i sensors (WTW, Weilheim, Germany). Concentrations of carbon species (total inorganic carbon, total organic carbon, dissolved inorganic carbon (DIC), dissolved organic carbon (DOC)), total suspended solids (TSS), major ions (Na+, K+, Mg2+, Ca2+, Cl, NO3−, SO42−), and total phosphorous and metals (Si, Al, Fe, Mn, Sr, Ba, Ni, Zn, Cu) were determined by FR EN ISO and ISO standards and laboratory procedures (detailed procedures are provided in Table S2 (supporting information)). Br was quantified by ion chromatography according to EN ISO 10304-1:2009-07 (Dionex DX 500, dl: 0.018 mg L−1 Br), whereas UR and SRB were quantified by luminescence spectrometry (Perkin Elmer LS 50 B) according to the methods described by Leibundgut et al. [2009]. Detection limits in wetland samples were 0.05 (UR) and 0.1 (SRB) µg L−1. Particle- and sediment-associated tracers were analyzed after treatment with an ammonia-Titriplex III solution, as previously described [Wernli, 2011].

[13] IPU, MTX, and the two IPU degradation products, didemethylisoproturon and desmethylisoproturon, were analyzed according to NF XPT 90-210. Filtered water samples were solid-phase extracted before analyzing the subsequent extracts using liquid chromatography coupled to tandem mass spectrometry. Quantification and detection limits for pesticides and IPU degradation products were 0.02 and 0.007 µg L−1 in water, 0.01 and 0.05 µg g−1 in TSS, and 0.003 and 0.017 µg g−1 in sediment, respectively. Extraction efficiencies, obtained by spiking with surrogates, were 90% in water and ranged from 50% to 70% in TSS and sediment samples. RDS (Relative Standard Deviation) was 25% in water and 50% in TSS and sediment samples.

2.6. Data Analysis

[14] Hydraulic characteristics of the lab-scale wetlands were derived based on the moment analysis of the residence time distribution (RTD) of Br breakthrough curves (BTCs) [Holland et al., 2004; Lange et al., 2011]. The zeroth absolute moment of the RTD is equivalent to the total tracer mass recovery R (%). The first and second central moments yield the mean residence time τ (h) and the variance of the RTD around τ2), respectively. Other transport parameters were the first arrival of the target compound t1 (h), the mean and maximum pore water velocities vmean and vmax (cm h−1), and the absolute or normalized (COUT CIN−1) maximum tracer concentration Cmax (mg L−1).

[15] Transport parameters of tracers and pesticides, such as the dispersion coefficient D (cm2 h−1), the retardation factor Rd, the phase-partitioning coefficient β, the mass transfer coefficient ω, and the liquid-phase degradation coefficient µL (h−1), were estimated using the CXTFIT model code [Mao and Ren, 2004], assuming conservative transport. Degradation on the solid phase was not considered. Fitting an equilibrium convection-dispersion equation (CDE) enabled to estimate D in the simulation of Br BTCs [Toride et al., 1999]. Transport parameters for UR, SRB, IPU, and MTX were estimated based on the nonequilibrium CDE, with v and D values set to values derived from the Br BTCs [Toride et al., 1999]. Based on the modeled retardation (Rd), the soil-water-partitioning coefficient Kd (mL g−1) and the organic-carbon-partitioning coefficient KOC (mL g−1) were determined (equations (1) and (2)).

display math(1)

where Rd is the retardation coefficient, ρb is the average bulk density of the medium (g cm−3), Kd is the adsorption coefficient (mL g−1), and θ is the average volumetric water content of the medium.

display math(2)

where KOC is the organic-carbon-partitioning coefficient (mL g−1), and fOC is the average mass fraction of organic carbon content of the medium.

3. Results

3.1. Water Balance and Biogeochemical Development

[16] Hydrological data are provided in Figure 2, and detailed hydrochemical data are shown in Table S3 (supporting information). While in the nonvegetated wetland 94.2% of the total injected water volume was recovered in the effluent, only 91.1% was recovered in the vegetated wetland. Assuming that water consumptive biogeochemical processes in the wetland matrix were negligible, the evaporative water loss was derived from the negative water balance of the nonvegetated setup, which accounted for 5.8% (1.93 L). The difference between the water balances for the vegetated and nonvegetated wetlands was accounted for the plant transpiration effect (3.1% or 1.03 L). Overall, the flux of effluent water during dye tracer and pesticide transport gradually decreased by 2% and 5% in the nonvegetated and vegetated wetlands, respectively (Figure 2). When pesticides and tracers were supplied to the wetlands, the effluent water volume gradually increased in the vegetated wetland, whereas the effluent volume again gradually decreased when uncontaminated water was supplied. This effect was not observed in the nonvegetated wetland. Globally, physicochemical conditions did not significantly change during the experiments. pH values ranged from 7.3 to 8.0 (vegetated) and from 7.5 to 8.2 (nonvegetated), which did not differ from the influent pH values. Values of EC were (mean±SD) 802±44 (vegetated) and 730±50 (nonvegetated) µS cm−1.

Figure 2.

Water balance of lab-scale wetlands during the tracer and pesticide transport experiments.

[17] Two redox zones could be distinguished in the lab-scale wetlands. In the lower part (50.5–20 cm below the surface), dissolved oxygen concentration (mean±SD) was 7.4±1.0 mg L−1 (vegetated) and 7.3±0.5 (nonvegetated) mg L−1, which underscored the prevalence of oxic conditions. In contrast, dissolved oxygen concentration was lower than 0.1 mg L−1 in the upper part (0–20 cm below surface), and black precipitates could be observed, likely iron monosulfite-type minerals.

3.2. Transport of Artificial Tracers and Pesticides in the Aqueous Phase

3.2.1. Bromide BTCs

[18] The mass recovery rate of the Dirac pulse injection of Br exceeded 94% in both wetlands. Br transport was characterized by sharper concentration peak and lower curve skewness in the vegetated compared to the nonvegetated wetland (Figure 3). The hydraulic retention time (HRT) was 13.4 and 13.9 days in the vegetated and the nonvegetated wetland, respectively. Overall, Br transport was faster in the vegetated wetland, as underscored by larger mean pore water velocity (Δ[vmean]=0.005 cm h−1), shorter HRT (Δ[τ]=12 h), and earlier tracer detection (Δ[t1]=9 h). In contrast, the maximum concentration was larger in the nonvegetated wetland (Δ[Cmax]=2.7 mg L−1; Table 2). Modeling of Br transport in the wetlands also emphasized a larger dispersion coefficient D in the vegetated wetland. Transport-related equilibrium effects prevailed, with 98.4% (vegetated) and 99.6% (nonvegetated) of the total pore water being mobile (Table S4 (supporting information)).

Figure 3.

Observed and modeled Br BTCs in the lab-scale wetlands.

Table 2. Observed and Estimated Transport Parameters of Dissolved Artificial Tracers and Pesticides
ParameterSymbolUnitVegetated WetlandNonvegetated Wetland
  1. a

    Absolute maximum concentration of Br in mg L−1.

  2. b

    The longitudinal dispersion is identical or all compounds in a wetland because diffusion is assumed to be a negligible part of the hydrodynamic dispersion.

  3. c

    Transport type 1: mobile and persistent, type 2: mobile and nonpersistent, type 3: moderately mobile and persistent.

First arrivalt1h1019616896144110118192144144
Maximum velocityvmaxcm h−10.500.530.310.530.350.460.430.260.350.35
Mean pore water velocityvmeancm h−10.157    0.152    
Maximum concentrationCmaxC C0−150.2a0.810.640.800.4447.5b0.840.600.830.44
Mean hydraulic residence timeτh321    333    
Solute recoveryR%94.281.765.676.639.598.878.655.979.737.5
Longitudinal dispersionbDcm2 h−10.70770.70770.70770.70770.70770.60220.60220.60220.60220.6022
Retardation factorRd 10.95453.8211.1411.36411.0015.9521.2721.379
Soil-water partitioningKdcm3 g−1  1.8050.0900.233  3.1690.1740.243
Organic carbon-water partitioningKOCcm3 g−1  184924  3231825
Degradation coefficientµh−101.8 × 10−302.5 × 10−31.9 × 10−203.6 × 10−301.9 × 10−31.9 × 10−2
Transport Typec
    1312 1312

3.2.2. Dye Tracers and Pesticides

[19] Transport characteristics of tracers and pesticides are provided in Figures 4 and 5a, as well as in Tables 2, 3, and S4 (supporting information). The transport characteristics of both tracers and pesticides differed between the vegetated and nonvegetated wetlands. UR and IPU were detected after 100 h in the vegetated wetland, and only after 118 (UR) and 144 (IPU) h in the nonvegetated wetland. BTCs of UR and IPU were similar, with maximum relative concentrations Cmax ≈ 0.8 COUT CIN−1 and solute recovery R ≈ 80%. Didemethylisoproturon could not be detected in the wetlands. Desmethylisoproturon was detected when the IPU flux was larger than 10 µg d−1 (>20 µg L−1). In contrast, IPU concentrations in the dissolved phase were below the detection limit when the IPU flux was lower than 5 µg d−1 (<2 µg L−1; Figure 5a). The daily flux of desmethylisoproturon was three orders of magnitude lower than that of IPU.

Figure 4.

Observed and modeled concentrations of UR, IPU, SRB, and MTX in the vegetated and nonvegetated lab-scale wetlands. Uncertainties of tracer concentrations are lower than 1% and are not shown.

Figure 5.

(a) Dissolved and (b) particle-associated flux of MTX, IPU, and desmethylisoproturon (desmethyl IPU) in the lab-scale wetlands. (left) Unhatched columns: vegetated wetland and (right) hatched columns: nonvegetated wetland.

Table 3. Water Balance and Recovery of Artificial Tracers and Pesticides in Dissolved, Particle-Associated, and Sediment-Bound Phasea
 Water (L)Br (mg)UR (µg)SRB (µg)IPU (µg)Desmethylisoproturon (µg)MTX (µg)
Injected Mass33.220016816806720840
  1. a

    V, vegetated wetland; NV, nonvegetated wetland. Numbers in parentheses are the percentage of total mass.

Dissolved    137.3 (100)132.0 (100)1101.2 (93.8)939.3 (89.1)521.2 (97.4)508.8 (97.4)5.2 (98.1)2.7 (96.4)331.1 (99.6)315.1 (99.8)
Particle-associated      27.5 (2.3)0.5 (0.05)0.7 (0.1)0.4 (0.08)0.1 (1.9)0.1 (3.6)1.2 (0.4)0.5 (0.2)
Sediment-bound      45 (3.8)115 (10.9)13.4 (2.5)13.4 (2.7)    
Recovery (%)  39.537.6

[20] The UR transport modeling pointed at a retardation factor smaller than 1 in the vegetated wetland, indicating that the UR transport was faster than that of Br. Consequently, Kd and KOC values could not be calculated for UR in that case. In the nonvegetated wetland, Rd of UR was approximately 1, and vmax was higher than the vmax detected for Br. The UR recovery was higher in the vegetated (81.7%) than in the nonvegetated wetland (78.6%). KOC and R (Δ 3.1%) of IPU were greater in the nonvegetated wetland. The time of the first detection for MTX was similar to that of IPU in the vegetated wetland, whereas it occurred 48 h later in the nonvegetated wetland (see Figure 4 and Table 2). MTX normalized maximum concentration (Cmax, COUT CIN−1) was 0.44, and solute recovery was lower than 40%. No sorption effect could be observed for UR, whereas sorption could be observed for IPU, MTX, and SRB. Only MTX reached a steady-state concentration in the course of the experiment. MTX concentration decreased 13 days before the end of the injection, and the degradation coefficients µL for modeling MTX in the liquid phase were one order of magnitude larger than those of UR and IPU, which clearly indicates MTX degradation in the wetlands.

[21] SRB was detected only 3 days after UR and IPU (Figure 4 and Table 2). The SRB maximum normalized concentration (Cmax, COUT CIN−1) was 0.6 and solute recovery reached 60%, which are values between those for IPU (UR) and MTX. SRB solute recovery was 10% larger in the vegetated wetland than in the nonvegetated wetland. Assuming no degradation, SRB retardation factor was larger than 3 in both wetlands, which underscores significant sorption-driven retardation, in agreement with previous observations [Sabatini, 2000].

3.3. Particle- and Sediment-Associated Tracers and Pesticides

[22] The mean concentrations of TSS detected in the wetland effluents were larger in the vegetated wetland (55.5±34.2 mg L−1) and exceeded those at the inlet (11.1–11.5 mg L−1) in both wetlands. Total carbon concentrations were larger in the vegetated wetland (61.2±26.8 mg L−1), and mostly consisted of DIC. DOC concentrations were 2.2 mg L−1 at the inlet and 4.5±1.2 and 5.2±1.5 mg L−1 at the outlet of the vegetated and nonvegetated wetlands, respectively.

[23] Particle-associated transport of IPU, desmethylisoproturon, MTX, and SRB occurred in both wetlands (Table 3 and Figures 5b and S3 (supporting information)), whereas it could not be observed for UR. Dissolved and particle-associated fluxes of IPU, desmethylisoproturon, and MTX were positively correlated in both the vegetated and the nonvegetated wetland (p<0.05). The portion of particle-associated transport in the total transported loads was, for the vegetated and nonvegetated wetlands, respectively, 0.12%±0.08% and 0.15%±0.31% for IPU, 0.12%±0.33% and 0.23%±0.44% for desmethylisoproturon, and 0.60%±0.44% and 0.62%±0.79% for MTX. Particle-associated transport of SRB accounted for 2.3, and only 0.05% of the total recovered SRB load in the vegetated and nonvegetated wetlands, respectively (see Table 3 and Figure S3 (supporting information)).

[24] Only SRB and IPU could be detected in the wetland sediments collected at the end of the experiment (Table 3). The SRB load recovered in WSs accounted for 3.8% and 10.9% of the total injected load in the vegetated and nonvegetated wetlands, respectively. The IPU load recovered in sediment accounted for 2.5% and 2.7% in the vegetated and nonvegetated wetlands, respectively. IPU and MTX could not be detected in the vegetation.

4. Discussion

[25] In this study, the transport and attenuation of IPU and MTX were evaluated quantitatively during 84 days along with artificial tracers in the lab-scale wetland experiments. The experiments were designed to investigate upward discharge of pesticide-contaminated groundwater into organic-rich environments at groundwater/surface-water interfaces. Our results show that sustained upward vertical-flow transport of IPU and UR largely differed from that of MTX and SRB. Attenuation of solutes in wetlands is primarily controlled by hydrological and biogeochemical characteristics, while several processes can simultaneously and synergistically control the dissolved and particle-associated transport of both pesticides and tracers. In our lab-scale wetlands, sorption and degradation were the major processes controlling IPU and MTX transport, and UR had a transport pattern similar to that of IPU. Since sorption of UR is less important in free surface wetlands and photolytic decay may limit its use, the transport characteristics of SRB were similar to those of IPU [Passeport et al., 2010; Lange et al., 2011].

[26] During our experiment, changes in hydrobiogeochemical properties with concomitant impact on the transport of both pesticides and tracers were particularly relevant in the vegetated wetland. The plant development (see Figure S1 (supporting information)) enhanced evapotranspiration (compare Figure 2 and section 2.1.2) and likely facilitated the transport of solutes along the root channels. This latter phenomenon resulted in gradual reduction of the solute-soil-plant interaction time and higher recovery rates of UR, SRB, and MTX in the vegetated wetland compared to the nonvegetated wetland (Table 3). Tubeworms (Tubificidae) dwelling activities in both wetlands (Figure S2 (supporting information)) might have affected the solute transport by preferential channeling along worm burrows and temporary immobilization in dead-end channels.

[27] Reciprocally, the supplied pesticides may also have impacted the hydrobiogeochemical development of the wetlands. A phase of larger effluent water volumes was observed after the injection of IPU and MTX and was followed by a phase of gradual increase of evapotranspiration in the vegetated wetland (Figure 2). This phenomenon may reflect an initial plant response to pesticide stress, followed by a progressive recovery of plant activities. Fast recovery from IPU-induced inhibition of photosynthetic activity was also previously reported for periphyton [Laviale et al., 2011].

[28] In our study, the transport of solutes in the vegetated and nonvegetated wetland was assessed based on a single pair of columns. Though our observations are derived from a middle term (84 days in our case) and thorough lab-scale experiment, the lack of experimental replicates may limit their significance and generalization to other cases. However, experimental replicates are largely constrained by the sampling intensity and multicompound analysis, which imposes a balance to be found between the experimental feasibility and the number of replicate experiments necessary to validate the observations. In our case, to limit side effects and to permit a quantitative and comparative analysis of the results, sediment cores used for the vegetated and nonvegetated wetlands were collected in parallel, their physical and chemical characteristics were thoroughly analyzed, and their bottom (32.5 cm) and top layers (10 cm) were carefully homogenized. To evaluate in detail the transport of pesticides and artificial tracers, the experiment was run over 84 days, and samples were collected at least three times per week for hydrochemical characterization and to establish a thorough mass balance of solute transport.

[29] While the transport characteristics of pesticides and tracers reflected hydrobiogeochemical changes in the wetlands, biogeochemical characteristics of the vegetated and nonvegetated wetlands also affected the solute transport. Lower retention of artificial tracers and pesticides, in particular for compounds with larger KOC values, was observed in the vegetated wetland due to preferential flow along root channels. Preferential flows likely enhanced the solute transport, while reducing contacts between the solutes and WS and lowering the mechanical filtration effect of the wetland matrix. This phenomenon also enhanced the TSS concentrations as well as the particle-associated transport of tracers and pesticides in the vegetated wetland.

[30] High Br recovery in the wetlands highlighted its usefulness for determining hydraulic characteristics, although Br loss in the vegetated wetland was larger than in the nonvegetated wetland. Xu et al. [2004] showed that Br could be actively taken up by reed grass at Cl/Br ratios lower than 2, whereas Br uptake was reduced at larger Cl/Br ratios [Xu et al., 2004]. In this study, Cl/Br ratios reached 1.5, which indicates Br loss by plant uptake in the vegetated wetland. In addition, heterogeneous distribution of Br may be caused by transpiration and preferential flows along plant roots, which resulted in a sharper concentration peak and eventually influenced the mean pore velocity.

[31] The evaluation of IPU, MTX, UR, and SRB transport characteristics in the wetlands enabled to distinguish three main transport types, which are described as follows (see Table 2). The first transport type encompasses UR and IPU transport and was characterized by high mobility, low dispersion, and low degradation. The steep increase of UR and IPU (Figure 4) ended concurrently to the Br peak (Figure 3). This indicates that UR and IPU were not significantly retarded compared to Br, although sorption and/or degradation processes retained 20% of their initial mass (Table 2). The occurrence of desmethylisoproturon indicated IPU biodegradation. Our results clearly show that the production and transport of desmethylisoproturon was larger in the vegetated wetland (Table 3), which is in agreement with previous studies underscoring that radial oxygen transfer at the rhizophere scale may favor aerobic biodegradation of IPU [Sørensen and Aamand, 2003; Alletto et al., 2006]. The high mobility and recovery of IPU during the upward vertical-flow transport in the wetlands underline its persistence. The absence of significant IPU attenuation in our experiment emphasizes that IPU-contaminated groundwater may discharge to organic-rich wetlands, or similar groundwater/surface-water interfaces, and may reach sensitive surface-water bodies. Although groundwater/subsurface interfaces may be deep enough to enable large attenuation of IPU during its upward transport through the sediment layer, our results suggest that particular attention should be given to the potential IPU mass transport at groundwater/surface-water interfaces. This may be particularly valid when IPU is applied in regions with vulnerable soil and/or climate conditions.

[32] The UR biodegradation in the wetlands cannot be excluded, although it could not be proven in the present study. Traditionally, UR has been treated as a conservative tracer for underground studies because it is considered as stable and nonsorptive at pH values higher than 7 [Käss, 2004]. Similarities between the UR and IPU transport characteristics, such as the similar degradation terms (µL) for IPU and UR, suggest the occurrence of UR degradation in the lab-scale wetlands. This clearly requires further detailed investigation of reactive transport of UR under environmental conditions, with emphasis on UR degradation products.

[33] The second type of transport includes the transport of MTX, which is characterized by high mobility and moderate persistence. Longer time of the first MTX detection and larger modeled Rd and KOC [MTX] values compared to UR suggest that MTX was retarded in the wetlands. MTX degradation in the wetlands was indicated by the absence of concentration tailings, small MTX recovery rates, and a degradation factor µL, which is one order of magnitude larger than those for IPU and UR. The MTX degradation products were not investigated in the present study, which prevents the identification of prevailing degradation pathways. Nevertheless, our results underscore that degradation is a critical process affecting MTX transport, in agreement with a previous study [Massoud et al., 2008]. There was no evidence of an enhanced MTX transport in the vegetated wetland due to strong attenuation processes, likely biodegradation. These overlaid more subtle processes such as preferential flow along plant roots, which was observed during the transport of other solutes.

[34] The third type of transport includes SRB transport, which is characterized by moderate mobility and high persistence. Right-skewed BTCs, longer t1, and elongated tailings clearly underline SRB retardation in the wetlands. SRB recovery and maximum relative concentration were lower than in the case of transport type 1 (UR and IPU), but larger for type 2 (MTX), while retention efficiency ranged between these two transport types. Since SRB degradation has not been reported so far, sorption likely was the main process affecting SRB retardation and retention in the wetlands. SRB sorption is expected to occur onto heterogeneously charged matrices, due to the presence of two electronegatively charged sulfonic acid groups versus the carboxylic groups [Kasnavia et al., 1999]. Hence, SRB sorption may be less selective and is expected to occur on a larger panel of sorbents compared to UR. However, this behavior is expected to be mainly dependent on the sediment mineral composition rather than on the organic content [Kasnavia et al., 1999].

[35] Our results also indicate that particle-associated transport did not occur for UR and increased in the following sequence: IPU<MTX<SRB<desmethylisoproturon. The mass of particle-associated transport of IPU and MTX accounted for less than 1% of the total mass recovery of IPU and MTX, whereas it accounted for 2.3% and 3.6% of the total mass recovery of desmethylisoproturon and SRB, respectively. In general, our results show that subsurface particle-associated transport of the targeted compounds was not a prevailing transport vector under the given experimental conditions. However, particle-associated transport of pesticides may be a prevailing phenomenon in free water surface wetlands, such as storm-water wetlands receiving surface runoff from agricultural surfaces [Maillard et al., 2011]. At the end of our experiment, SRB and IPU in wetland sediments accounted for a large portion of the total mass of solutes that entered the wetland, which is in agreement with their sorption tendency in aqueous solutions (i.e., large Kd and KOC).

5. Conclusions

[36] This study is the first to quantify the joint transport of IPU, MTX, and artificial tracers through wetland systems with respect to changes in hydrobiogeochemical conditions. Although photolytic decay could not be strictly distinguished from hydrolytic decay, and possible degradation hotspots in the wetlands could not be localized, our results indicate that the transport characteristics of MTX, IPU, and SRB largely differ from each other in both the vegetated and nonvegetated wetlands. The transport of MTX, IPU, and SRB mostly differed with respect to Cmax (ranging from 0.44 to 0.83) and mass recovery (ranging from 37.5% to 79.7%). While the retention of UR, SRB, and MTX was generally more effective in the nonvegetated wetland, IPU was more effectively retained in the vegetated wetland, suggesting that rhizospheric processes enhanced the biodegradation process.

[37] The UR and IPU transport displayed similarities, underscoring that UR may serve as a reference tracer for sustained upward vertical-flow transport of IPU. IPU was highly mobile and only partly degraded into desmethylisoproturon, although IPU attenuation reached 20% in the vegetated wetland. Similar transport characteristics for UR and IPU in both wetlands have two implications: first, UR may be subject to degradation in wetlands, and second, IPU may not be significantly attenuated during upward transport in organic-rich wetland or similar groundwater/surface-water interfaces and thus may reach pristine surface-water bodies. In contrast, more than 60% of the MTX load entering the wetlands was attenuated, and no tracers had similar transport characteristics in the wetlands. Although the SRB transport and behavior largely differed from that of other compounds, it may serve as a surrogate for more sorptive and less degradable pesticides in the future (compare experimental KOC values ranging from 184 to 323 (Table 2) and literature values with KOC ranging from 147 to 498 (Table 1)).

[38] The particle-associated transport accounted for less than 4% of the total loads of IPU, MTX, SRB, and UR, which emphasized that the transport mainly occurred in the aqueous phase. Our results also underline that the presence of vegetation in wetlands may enhance preferential flow while lowering interactions between solutes and the solid phase and thus contaminant attenuation.

[39] So far, only conservative tracers, such as Br, have been used to study the hydraulic characteristics of the wetland systems. This study can be anticipated as a starting point for considering nonconservative fluorescent dye tracers as a cost-effective approach for assessing the pesticide transport and associated risks in complex and biogeochemically dynamic environments.


[40] This research has been funded by the PhytoRET project (C.21) of the European INTERREG IV program Upper Rhine. The authors wish to thank Carole Fillinger, Sophie Gangloff, Marie-Pierre Ottermatte, Eric Pernin, René Boutin, Thierry Peronne, and Barbara Herbstritt for the chemical analyses.