Introduction

The intensive use of pesticides in modern agriculture is responsible for global environmental pollution that affects water quality. Between 1998 and 2008, in France alone, 878 water resources were abandoned due to pollution, i.e. nitrates (506), pesticides (183) or both (189) (Direction Générale de la Santé 2012). Diffuse pollution contributes to the contamination of surface water by run-off, leaching, subsurface drainage and spray drift (Margoum et al. 2006). Several techniques have been developed for the removal of pollutants from surface water. Some compounds have been banned due to their excessive occurrence in environmental settings (Schiavon et al. 1995), such as atrazine, which was prohibited in 2003. In addition, agricultural practices have to limit the export of pesticides (Journal Officiel du Conseil de la Communauté Européenne 2006) through optimisation of treatment periods, effective monitoring (monitoring networks) or the use of non-chemical methods (organic farming being the main such method) of reducing pollution. In addition to these global actions, good agro-environmental practices, such as grass cover bands and untreated areas, have been put in place to restrict the export of pesticides. These practices have demonstrated their effectiveness against pathways (run-off, spray drift) of agricultural pollutants (Reichenberger et al. 2007), but they are ineffective against pesticides exported by the flow of water from drainage tiles that bypass these buffer zones.

For drain water, pesticide losses are generally accepted to be approximately 0.5 % of the applied rate (Brown et al. 1995, 2004; Riise et al. 2004) but can reach 5 % (Carter 2000) and even 10 % under certain conditions (Accinelli et al. 2002). In France, more than 10 % of agricultural lands are drained, and in Lorraine, approximately 20 % of agricultural lands are drained (Recensement Agricole 2010). This percentage can reach 70 % in some Lorraine watershed (Chartier, pers. comm.). Water from these drained fields creates hot spots of high pesticide contamination of fresh water in the rivers. Two studies conducted in Lorraine reported peak pesticide concentrations in drainage water ranging from 5.9 to 33.2 μg/L for alachlor, from 5.2 to 118.5 μg/L for bentazon (Dousset et al. 2004) and from 171.6 to 395.3 μg/L for metolachlor (Novak et al. 2001). These values exceed the limits set by the European water framework directive (Journal Officiel du Conseil de la Communauté Européenne 2000).

Therefore, the objective of this study is to evaluate the effectiveness of constructed wetlands, which are often proposed to be good agricultural practices and located between output drains and drainage to reduce export of agricultural pollutants (Schulz and Peall 2001; Rogers and Stringfellow 2009). We evaluated the adsorption and desorption capacities of different matrices found in these rustic remediation zones (RRZ) for drainage water filtration, installed in the grass cover bands of agricultural fields. Six pesticides [2-methyl-4-chlorophenoxyacetic acid (2,4-MCPA), boscalid, isoproturon, napropamide, prochloraz and tebuconazole] used in Lorraine, with various physical and chemical properties, were studied.

Materials and methods

Experimental sites and sampling procedure

The selected substrates were taken from two recent RRZ designed in 2010 at Jallaucourt (Moselle, Lorraine, France) and Ollainville (Vosges, Lorraine, France) (Fig. 1). The RRZ at Jallaucourt consisted of a ditch 10 m long, 4 m wide and 1.5 m deep. A bundle of straw was placed in the middle of the ditch to reduce the rate of water flow and play the role of a filter. The RRZ at Ollainville consisted of a triangular vegetated pond (20.5 × 15.5 × 11 m) with a water depth between 0.6 and 0.8 m. Spontaneous vegetation of Typha latifolia, Callitriche platycarpa and Glyceria notata has developed since the recent construction of this RRZ.

Fig. 1
figure 1

Location map and schematic diagram of the vegetated ditch (Jallaucourt) and pond (Ollainville)

Full size image

Samples were collected within the RRZ at Jallaucourt and Ollainville on 16 and 23 March 2011, respectively, from bare soil (soil), sediment in vegetated zones and straw. The samples were collected in plastic bags and transported to a laboratory. The soils and sediments were air-dried, homogenised and passed through a 2-mm sieve. The straw was air-dried and cut up into 2- to 3-cm fragments. The main physico-chemical properties of the substrates are shown in Table 1. The measurements of particle size (NF X 31–107), pH (NF ISO 10693), organic carbon (NF ISO 10964), total nitrogen (NF ISO 10878), calcium carbonate (NF ISO 10693) and cationic exchange capacity (CEC) (NF X 31–130) were performed at INRA in Arras. Specific surface measurements were conducted at laboratory by adsorption/desorption of nitrogen at 77 K after degassing at 110 °C under a vacuum (10−6 mbar).

Table 1 Physico-chemical characteristics of the studied substrates
Full size table

Selection of pesticides and chemicals

Nearly 70 pesticides were applied to drained plots feeding the two RRZ. These pesticides were analysed by ANSES (Agence Nationale de Sécurité Sanitaire de l'Alimentation, de l'Environnement et du Travail, Laboratoire d'Hydrologie, Nancy, France) in waters flowing into and out of the RRZ. Six pesticides were selected for laboratory on the basis of frequency of use, the applied amounts and the diversity of their physico-chemical properties. The pesticides selected were three herbicides—2,4-MCPA (2-methyl-4chlorophenoxyacetic acid) (MCPA), isoproturon (3-(4-isopropylphenyl)-1,1-dimethylurea) (IPU) and napropamide (N,N-diethyl-2-(1-naphthyloxy)propionamide) (NPP)—and three fungicides—boscalid (2-chloro-N-(4′-chlorobiphenyl-2-yl)nicotinamide) (BCL), prochloraz (N-propyl-N-[2-(2,4,6-trichlorophenoxy)ethyl]imidazole-1-carboxamide) (PCZ) and tebuconazole (1-p-chlorophenyl-4,4-dimethyl-3-(1H-1,2,4-triazol-1-ylmethyl)pentan-3-ol) (TBZ). The main characteristics of these pesticides are given in Table 2. The analytical standards were purchased from Sigma-Aldrich (Seelze, Germany). All are analytical standard grade (>98 %).

Table 2 Main properties of the six studied pesticides (FootPrint 2011)
Full size table

Sorption studies

Sorption of the pesticides to substrates was studied using a batch equilibrium technique. Pesticide dilutions were prepared in 0.01 M CaCl2. The initial concentrations were 0.05, 0.125, 0.25, 0.5, 1, 2 and 4 mg/L for each pesticide. Low concentrations were calculated from the pesticide flows observed in the field, with contact surfaces representative of annual pesticide exposure in the RRZ.

For each concentration, three samples of 2 ± 0.002 g of soil/sediment and 0.2 ± 0.002 g of straw (corresponding to an equivalent soil volume) were shaken in a 40-mL polycarbonate centrifuge tube with 10 mL of the pesticide solution. The mixtures were shaken for 24 h at 20 ± 1 °C with a rotary shaker at 60 rpm. The suspensions were centrifuged for 30 min at 10,400×g, and 1.5 mL of aliquots was sampled for analysis. Blanks and controls were prepared to quantify adsorption of pesticides on the tubes and to check the non-contamination of the matrices, respectively.

Desorption isotherms were developed immediately after adsorption from the initial concentrations at 0.125 and 2 mg/L. After centrifugation, the supernatant was replaced with 10 mL of 0.01 M CaCl2 aqueous solution. The tubes were vibrated to disperse the soil pellets and were shaken for 24 h at 20 °C with a rotary shaker (60 rpm). Four successive desorption steps were performed for the initial concentrations of 2 and 0.125 mg/L.

Adsorption and desorption modelling

The sorbed pesticide concentration (x/m in milligram per kilogram) was calculated from the difference between the initial concentration (C 0 in milligram per litre) and the equilibrium concentration (C eq in milligram per litre), the volume of pesticide solution was added (V in litre) and the mass substrate was given (M):

$$ x/m=\frac{V}{M}\times \left({C}_0-{C}_{\mathrm{eq}}\right) $$

The adsorption data were described using Freundlich Eq. (1) and linear Eq. (2), and the desorption data were described using Freundlich Eq. (3):

$$ x/m={K}_{\mathrm{fads}}\times {C}_{\mathrm{eq}}{}^{\mathrm{nads}} $$
(1)
$$ x/m={K}_{\mathrm{d}}\times {C}_{\mathrm{eq}} $$
(2)
$$ x/m={K}_{\mathrm{fdes}}\times {C}_{\mathrm{eq}}{}^{\mathrm{ndes}} $$
(3)

where K fads and K fdes are empirical adsorption and desorption coefficients (milligram1 − n per litren per kilogram), respectively; n ads and n des are empirical adsorption and desorption coefficients, respectively; and K d is the coefficient of sorption (litre per kilogram). A hysteresis coefficient (H) was calculated (H = n des/n ads) for the described desorption. In general, hysteresis is defined only for H < 0.7 (Barriuso et al. 1994).

Chromatographic analysis

Chromatographic analyses were conducted using the dual RSLC UltiMate 3000, Dionex. Each sample (500 μL) was injected. The sample was flushed onto the online solid-phase extraction (SPE) with acidified water (H3PO4, pH 2.5) at a flow rate of 1 mL/min. The SPE column was the Oasis HLB 25 μm, 2.1 × 20 mm, Waters. After 1 min of flushing, a ten-port switching valve diverted the SPE to the LC system. At 2 min, the SPE was switched back to the SPE pump and flushed with acidified water to precondition the column for the next injection. After switching to the LC system in back flush mode, the analytical pump was used to separate the six analytes under 35 °C on the Synergi Hydro-RP C18 HPLC column (50 × 2.1 mm, 2.5 μm), Phenomenex. The mobile phase at a flow rate 333 μL/min consisted of a gradient of methanol and distilled water at pH 2.5. The gradient started at 50 % methanol and finished at 100 % in 5.3 min. The six pesticides were examined using a diode array detector. Determination of the limits of quantification (LOQ) was based on the standard deviation (SD) of the Y-intercepts of linear regressions and the slope of the calibration curve (S), according to the following formula: LOQ = 10 (SD/S) (Miller and Miller 1988). The calculated LOQ values were 10.3, 55.6, 51.9, 18.9, 10.0 and 15.3 μg/L for IPU, MCPA, PCZ, BCL, NPP and TBZ, respectively.

Statistical analysis

Statistical analyses were conducted using the XLStat software. An analysis of variance (ANOVA) was conducted using Duncan's test to detect significant differences (α = 0.05) between the K f and K foc matrices for the six pesticides studied.

Results and discussion

Sorption modelling

Two sorption models were used to fit the experimental data. The determination coefficient values (r 2) obtained for the linear models ranged from 0.15 to 0.98 for adsorption (data not shown), and those obtained for the Freundlich equation ranged from 0.15 to 1.00 for adsorption. The lowest r 2 values were obtained for MCPA. Indeed, MCPA presents high water solubility, and the common sorption models did not fit (Hiller et al. 2009). For desorption, only the Freundlich equation was used, and the r 2 values ranged from 0.53 to 1.00 for C i of 2 mg/L and from 0.01 to 0.95 for C i of 0.125 mg/L (Tables 3 and 4). The Freundlich equation better described the experimental sorption data of our pesticides (Dousset et al. 2007; De Wilde et al. 2009; Passeport et al. 2011) except for MCPA. This model was therefore selected for use in this study (Fig. 2). Few points corresponding to the lowest concentrations in MCPA, PCZ or BCL were below the LOQ and thus were not taken into account for drawing isotherms.

Table 3 Values of the Freundlich adsorption parameters K fads (milligram1 − n per litren per kilogram) and n ads for the six pesticides studied on substrates and K foc (milligram1 − n per litren per kilogram)
Full size table
Table 4 Values of the Freundlich desorption parameters K fdes (milligram1 − n per litren per kilogram) and n des for pesticides studied on substrates
Full size table
Fig. 2
figure 2

Adsorption (cross symbols) and desorption (triangle symbols) isotherms for all molecules on the straw (Jst) and sediments (Ose). Desorption isotherms from 2 mg/L−1 initial concentration

Full size image

Adsorption isotherms

Presentation of data

The adsorption coefficient values obtained (K fads) ranged from 0.74 to 20.48 mg1 − n Ln kg−1 for IPU and from 16.33 to 442.6 mg1 − n Ln kg−1 for PCZ and for Jallaucourt soil (Jso) and Jallaucourt straw (Jst) substrates, respectively (Table 3, Fig. 2). The observed n ads ranged from 0.14 to 1.12. In general, MCPA had the lowest n ads coefficient values (0.14 < n ads < 0.56), whereas PCZ had the highest coefficient values (0.73 < n ads < 1.12). For each pesticide, the observed n ads values were similar for all of the substrates, which suggest the same adsorption mechanisms. According to Giles et al. (1960), the isotherms observed can be classified as L or S types depending on the pesticides and substrates.

Comparing data

On the basis of adsorption capacity, the pesticides were classified as follows: PCZ >> TBZ-BCL > NPP > MCPA-IPU (Table 3). The relative adsorption capacities could be related to pesticide properties, particularly solubility and K oc (Table 2), as well as the ionic form. Indeed, IPU and MCPA had the lowest K fads values for all substrates (Table 3). The low n ads values for IPU and MCPA indicate a low affinity between the substrates and these compounds, as well as rapid saturation of the sorption sites. The L-type isotherms (n ads < 1) correspond to a decrease in the availability of sorption sites when the pesticide concentration increases (Giles et al. 1960). The values of the Freundlich parameters K fads and n ads obtained were in agreement with those reported by Fouqué-Brouard and Fournier (1996), Nemeth-Konda et al. (2002), Dousset et al. (2007), De Wilde et al. (2009) and Passeport et al. (2011), who reported K fads values ranging from 0.52 to 2.12 mg1 − n Ln kg−1 for soils. The values obtained for MCPA are in agreement with those given by Thorstensen et al. (2001), Paszko (2011) and Hiller et al. (2012), who measured K fads values ranging from 0.37 to 1.50 mg1 − n Ln kg−1. MCPA is a weak acid (pK a = 3.72 at 25 °C). In our study, pH varied from 7.2 to 7.9; as a result, MCPA was predominantly in an anionic form, which limited adsorption due to electrostatic repulsion of negatively charged soil particles (clay, OC) (Thorstensen et al. 2001).

PCZ showed the highest K fads values for all substrates, varying between 16.33 and 442.63 mg1 − n Ln kg−1 for Jso and Jst substrates, respectively. These values are in agreement with those reported by Rütters et al. (1999) and Roy et al. (2000). However, these soils contained lower amounts of organic carbon (0.77–1.42 %) and clay (17.4–28 %). These authors have also demonstrated a positive correlation between K fads and organic carbon content and a negative correlation with pH. These correlations were also observed in our study (Pearson's test, OC% = 0.999 (p = 0.025); pH = −0.999 (p = 0.025)). Prochloraz is a weak base (pK a = 3.8), and the pH of the studied substrates was more than 3 units of pK a. In this condition, the ionic form was of minor importance. However, the pH close to hydrophobic surfaces of colloids can be 1 unit lower (Gaillardon et al. 1991). Thus, the ionic form at the interface could facilitate adsorption (Roy et al. 2000). In such conditions, the adsorption mechanisms involved are both physical and ionic, depending on the pH.

The values of the Freundlich parameters K fads and n ads obtained for NPP, BCL and TBZ were between those obtained for IPU/MCPA and for PCZ. The K fads values ranged from 3.16 to 77.42 mg1 − n Ln kg−1, from 4.75 to 191.76 mg1 − n Ln kg−1 and from 5.77 to 127.85 mg1 − n Ln kg−1 for NPP, BCL and TBZ, respectively. The n ads values varied from 0.75 to 0.78, from 0.79 to 1.11 and from 0.79 to 1.08 for NPP, BCL and TBZ, respectively. Little information is available in the literature on these molecules. The values obtained for NPP are in agreement with those presented by Aguer et al. (2000) and Sadegh-Zadeh et al. (2012).

Effect of matrix substrates

On the basis of retention capacity, the substrates were classified as follows: Jst >> Ollainville sediment (Ose) > Ollainville soil (Oso) > Jso (Table 3). For all pesticides, the K fads values obtained for Jst were significantly greater (ANOVA, Duncan, α = 0.05) than those for other substrates (Table 3, Fig. 2). These results were directly related to the organic carbon content, which was 20 to 30 times higher than that in the soil and sediments. De Wilde et al. (2009) reported the same results for isoproturon and straw; however, those authors reported a lower adsorption (K fads = 7.91 mg1 − n Ln kg−1) than that obtained in this study (20.48 mg1 − n Ln kg−1). Differences in the preparation of the straw could explain these results. Indeed, Rogers and Stringfellow (2009) showed adsorption increasing from 7.6 to 96.2 % when organic substrates were cut more finely (from 5 to 1 cm long).

The differences in K fads values for MCPA, IPU, NPP and BCL were not significant for the other substrates. However, significantly lower K fads values were obtained for the Jso substrate than for the Oso and Ose substrates for PCZ and TBZ, which are the molecules that exhibited the highest adsorption (Table 3). These results could be explained by the higher correlation coefficients for these pesticides between the organic carbon content and the Jso substrate, which contained the lowest amount of organic carbon.

However, while Jst had the highest adsorption coefficient (for all substrates) with respect to its organic carbon content, the nature and quality of the organic matter is also important. Indeed, by standardising the adsorption coefficients by organic carbon content, it is possible to compare the K foc (K foc = K fads/OC% × 100) (Table 3) and, therefore, the nature and quality of the organic carbon in different substrates (Chiou et al. 1998; De Wilde et al. 2009). The K foc values obtained for Jst were significantly lower than those for NPP, TBZ and PCZ (Table 3), which are the pesticides with the highest adsorption capacities. In contrast, Jst had the highest K foc values for MCPA, and no significant differences were detected between the different modalities for IPU. These results could be explained by the lower degree of humification of the organic matter in straw, compared to soils and sediment. However, other parameters also play a role in soil and sediment adsorption, such as clay content, CEC and specific surface, as reported by Passeport et al. (2011) for IPU, metazachlor and epoxiconazole. These authors reported lower K foc values for plants and dead leaves than for soil and sediment, which had the highest K f values.

However, adsorption alone does not dictate the behaviour of pesticides in the environment. Reversibility and the forces involved in these processes influence the fates of pesticides on substrates. Thus, desorption has a very strong influence on their behaviour in the environment.

Desorption isotherms

Presentation of data

The desorbed amounts were highly variable for the different pesticides and substrates. The desorbed amounts varied from a few percentages to more than 70 % of the initial amounts adsorbed. In addition, for the two initial concentrations, no significant desorption was observed for MCPA, nor was significant desorption observed for IPU or PCZ at the initial concentration of 0.125 mg/L (Table 4). For these conditions, the hysteresis coefficient H (n des/n ads) could not be calculated, but the H values were notable for BCL and TBZ. In general, the H coefficient was lowest for the 0.125 mg/L concentration. These results could be explained by the presence of high- and low-energy sorption sites. At lower concentrations, high-energy sites were occupied first, which limited desorption. Hiller et al. (2009) showed a lower desorption of MCPA, ranging from 3.5 to 20.3 %, for initial concentrations of 5.7 and 56.6 mg/L, respectively. The amounts adsorbed at low concentrations are mostly bonded to substrates, which could explain the lower H coefficient values for the initial concentration of 0.125 mg/L.

Comparing data

On the basis of desorption capacity, the pesticides were classified as follows: BCL > TBZ > NPP > IPU > PCZ > MCPA. No significant desorption was observed for MCPA. Alister et al. (2011) and Paszko (2011) reported high desorption for MCPA at higher initial concentrations, ranging from 3.5 to 25 mg/L. However, these studies were conducted under different pH conditions, ranging from 4.3 to 6.0, and the authors explained that MCPA desorption is inversely correlated with pH. In contrast, the pH of our substrates ranged from 7.2 to 7.9, which could explain the low desorption levels observed. PCZ and IPU had the lowest H coefficient values, ranging from 0.16 to 0.44 and from 0.23 to 0.40, respectively, for the 2 mg/L concentration (Table 4, Fig. 2). These low values indicate low sorption reversibility. For IPU, the results obtained are in agreement with those of Fouqué-Brouard and Fournier (1996) and Passeport et al. (2011). For PCZ, Roy et al. (2000) suggested that pesticide diffusion in the soil aggregates was responsible for low desorption. NPP and TBZ had relatively similar H coefficient values, ranging from 0.35 to 0.48 for the NPP and from 0.50 and 0.59 for the TBZ at the 2 mg/L concentration. BCL had the highest H coefficient values, ranging from 0.53 to 0.89 at the 2 mg/L concentration. These values were the highest, and according to Barriuso et al. (1994), no hysteresis can be defined when 0.7 < H < 1. The adsorption of BCL is therefore very reversible, which reduces the amount of the pesticide retained in the long term.

Effect of matrix substrates

The non-desorbable pesticide amounts are reported in Table 5. Although desorption processes were important for the Jst substrates, Jst had the largest amounts of non-desorbable residues. The total amount of non-desorbable pesticide residues was 224.21 mg/kg at the 2 mg/L concentration, or 14 to 28 times greater than for the other substrates. This ratio was lower but still high for the 0.125 mg/L concentration, and the total amount of non-desorbable pesticide residues was 12.10 mg/kg. These results for retention capacity confirm the filter effect of the bundle of straw, even at low concentrations. For the other substrates, the non-desorbable pesticide amounts ranged from 8.15 to 15.79 mg/kg and from 0.80 and 1.03 mg/kg for the 2 and 0.125 mg/L concentrations, respectively (Table 5). The lower values were observed for the Jso substrate, which had the lowest organic carbon content.

Table 5 Non-desorbable amounts of pesticides on different substrates (milligram per kilogram)
Full size table

Conclusions

The results of this study clearly demonstrate the importance of sorption in the operation and effectiveness of an RRZ, which is directly related to the physico-chemical properties of pesticides and substrates. Retention was greater for pesticides with hydrophobic properties (low solubility and high K oc), as shown by the results obtained for PCZ, TBZ and BCL. The organic carbon content and nature of the substrate was found to have a strong effect on sorption. Straw had the highest K fads and the largest amounts of non-desorbable pesticide amounts. The Ose substrate had higher retention than the Oso and Jso substrates, in proportion to their higher organic carbon contents. At the RRZ scale, adsorption processes limit peak pesticide concentrations during the first rainfall events after treatment, and desorption causes pesticides to flow out of the RRZ. These phenomena occur during periods of drainage, i.e. from October to May in Lorraine. The formation of non-desorbable residues in the long term reduces the flow of pesticides into and out of the RRZ and decreases pesticide export to river systems. In addition, the formation of non-desorbable residues increases the residence time of pesticides in the RRZ and subsequently optimises their degradation (biological, chemical and/or physical).

Sorption alone does not determine the effectiveness of an RRZ. Other processes, such as degradation and water flow, need to be taken into account. However, planting vegetation and maintaining organic matrices (straw and dead leaves, etc.) inside RRZs are encouraged to increase their effectiveness. Moreover, plants could increase the efficiency of RRZs due to their absorption capacity and their contributions to the degradation processes. The presence of plants increases the water residence time and pesticide retention, especially during the drainage period.