Pa

JH

a

ARRAA

KPMAT

1

idtiHti

0h

Ecological Engineering 56 (2013) 51– 59

Contents lists available at SciVerse ScienceDirect

Ecological Engineering

journa l h om epage: www.elsev ier .com/ locate /eco leng

esticide de-contamination of surface waters as a wetland ecosystem service ingricultural landscapes

. Tournebize ∗, E. Passeport, C. Chaumont, C. Fesneau, A. Guenne, B. VincentBAN – Research Unit “Hydrosystems and Bioprocesses”, Irstea, 1 rue Pierre-Gilles de Gennes, CS 10030, F 92761 Antony cedex, France

r t i c l e i n f o

rticle history:eceived 10 November 2011eceived in revised form 4 June 2012ccepted 5 June 2012vailable online 18 July 2012

eywords:esticideitigation

rtificial wetlandile drainage

a b s t r a c t

In agricultural landscapes, pesticides can be transported by surface or tile-drained waters to aquaticecosystems. In such situations, artificial wetlands can be placed to receive tile drainage discharge, andbe effective pesticide pollution remediation tools. Artificial wetland implementation and design largelydepend on land availability and suitability, which is particularly critical in Europe where land is scarce.Wetlands can be configured to either divert and treat a portion of stream discharge (i.e., parallel config-uration), or placed to capture all stream discharge (i.e., series configuration) generated by a catchment.

A field experiment was conducted at the outlets of two subsurface drained catchments in France tocompare the pesticide retention efficiency of these two types of artificial wetlands. The two agriculturalcatchments were cultivated for similar crops (winter wheat, barley, rapeseed, and sugar beet) on poorlydrained tile-drained soils. The Aulnoy artificial wetland was situated in-stream, at the outlet of a 36-hawatershed. It included a deep wetland with a 9000-m3 water storage capacity. Its volume and surfacearea to watershed area ratios were 300 m3 per upstream drained hectare (equivalent to 30 mm runoffstorage capacity) and 1.2%, respectively. The Bray wetland was constructed off-stream, in parallel to themain agricultural ditch and associated with an open/close strategy managed by the farmer according tohis pesticide applications. This artificial wetland consisted of three vegetated cells in a series with shallowwater for a total volume of 330 m3 corresponding to 7 m3 ha−1 (0.7 mm runoff storage capacity) and 0.5%of a 46-ha catchment.

Inlet and outlet discharges and pesticide concentrations were continuously monitored to characterizethe seasonality of pesticide export and the removal efficiency of the artificial wetlands. Both artificialwetlands showed positive impacts on water quality (54 and 45% reduction for in- and off-stream configu-

rations, respectively), attributable to distinct retention processes. Significant dilution occurred at Aulnoywhere pesticide concentrations were frequently below detection levels, making it difficult to discernwhether significant degradation or retention took place in the wetland. Conversely, the Bray shallowartificial wetland showed significant pesticide adsorption, desorption and degradation. However, theseprocesses were limited in the Bray wetland due to insufficient retention time. The off-stream constructedwetland appears to be a promising technical solution for reducing land use impact.

aB

ttgl

. Introduction

Pesticide pollution is a major threat to surface water qualityn France. To reduce pesticide pressure, French authorities haveecided on a pesticide reduction plan, aiming for a 50% reduc-ion in pesticides by 2018. In addition, to reduce pesticide fluxesn streams, vegetative filter strips have been widely implemented.

owever, in sub-surface tile drained watersheds, water discharges

hrough sub-surface pipes and drainage ditches that bypass ripar-an zones and the opportunity that riparian practices present to

∗ Corresponding author. Tel.: +33 1 40 96 60 38.E-mail address: julien.tournebize@irstea.fr (J. Tournebize).

wstcbet

925-8574/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.ecoleng.2012.06.001

© 2012 Elsevier B.V. All rights reserved.

ttenuate pollution entering surface waters is lost (Carter, 2000;rown and van Beinum, 2009).

In the case of pesticide transfer, most losses occur during thehree first rainfall-runoff events that follow pesticide applicationo fields (Kladivko et al., 2001; Branger et al., 2009). Due to hetero-eneous agricultural practices at the catchment scale and generalow application rate of most chemicals, pesticide concentrations in

ater are typically well diluted, yet highly variable. From the plotcale to the catchment scale, concentrations can range more thanwo orders of magnitude, from >100 to <1 �g L−1. Annual pesti-

ide exportations from tile drained lands in Europe have reportedlyeen <1% of the amount applied, or about 1 g ha−1 year−1 (Novakt al., 2001; Riise et al., 2004). Pesticide pollution can be mitigatedhrough ecologically engineered practices such as grass or forest

5 cal Eng

bdSstpM(cwsmcp

c1egwciBseawecptp2chawtitbioatRp

2

idsbeilofliote

dctiiTmliwia

3

3

slsrTPlCsti(mhatPamtcatwdsuepa

3

sr(rTDtt

2 J. Tournebize et al. / Ecologi

uffers, artificial (or constructed) wetlands, planted ditches andetention ponds (Gregoire et al., 2009; Reichenberger et al., 2007;tehle et al., 2011). Artificial wetlands appear to be a promisingolution to complement buffer strips and agricultural field prac-ices that reduce risk of loss. Enough research on this topic has beenublished (e.g., see Blankenberg et al., 2007; Gregoire et al., 2009;iller et al., 2002) that it has been reviewed by Reichenberger et al.

2007). The authors noted that constructed or artificial wetlandsould be very effective in reducing pesticide inputs into surfaceaters. They added also that almost all available studies dealt with

trongly sorbing insecticides with a strong tendency to adsorb toacrophytes, suspended particles or bed sediment. Lastly they con-

luded that artificial wetland could be quite area-consuming as aotential drawback.

In terms of efficiency, percentages of pesticide reduction inonstructed wetlands have reportedly ranged from negative to00%. The best performance has been observed under controlledxperimental conditions; small mesocosm experiments (<1 m2)ave the highest concentration reduction (Cheng et al., 2002),hereas large mesocosm experiments (<500 m2) conducted under

ontrolled conditions have shown highly efficient results, rang-ng from 25 to 100% (Hunt et al., 2008; Matamoros et al., 2007;raskerud and Haarstad, 2003; Blankenberg et al., 2007). Yet, fewtudies have assessed artificial wetlands in real conditions (Millert al., 2002; Rose et al., 2006; Schulz et al., 2003). Nevertheless,

targeted performance goal of 50% load reduction in constructedetlands has been suggested as a reasonable design goal (Gregoire

t al., 2009; O’Geen et al., 2010). It is difficult to assess pesti-ide fate and transport in natural aquatic environments. Generally,esticide removal processes include biological degradation, phy-oaccumulation, adsorption and desorption and sedimentation ofarticle-bound chemicals (Passeport et al., 2011a, b; Imfeld et al.,009; Schulz et al., 2003). The review by Stehle et al. (2011) addedoverage by emergent plants (influenced by wetland depths) andydraulic residence time (influenced by volume and inflow vari-bility) are additional factors impacting retention performance inetlands. Furthermore, there are important questions about how

he location, size and design of a wetland at the catchment scalempact pesticide retention (van der Valk and Jolly, 1992). Oppor-unity to study pesticides in actual constructed wetlands are rareecause wetlands change land use from crop production and lim-

ting the extent of that change is a key for stakeholders acceptancef wetlands, as shown in Tournebize et al. (2012). This paper evalu-tes pesticide fate and transport in two wetlands, each constructedo receive discharge from a small working agricultural catchment.esults are used to compare two strategies of artificial wetlandlacements between two similar catchments.

. Concept of in- and off-stream artificial wetland

In tile drainage, the simplest way to reduce pesticide transfers to treat all water discharged from the tile outlet. This meansetaining the outlet water through artificial wetlands. Here, thisolution is called an in-stream artificial wetland (Fig. 1). However,ecause wetlands cannot fully treat drainage water during runoffvents, and because the land available for wetlands is limited inntensive agriculture areas, an alternative solution that requiresess surface area for the artificial wetland may be beneficial. Ourbjective was to intercept and treat a maximum possible pesticideux while diverting the least drainage water into the wetland. Typ-

cally, pesticide transfer is seasonal and dominantly follows periodsf application; the three rainfall-runoff events following applica-ion carry the greatest risk of transferring pesticide loads (Kladivkot al., 2001; Branger et al., 2009). Thus a manually operated gate to

cbwp

ineering 56 (2013) 51– 59

ivert tile flows to a wetland system parallel to the natural creekould be implemented. Operation should be as simple as possible,he system we implemented was to divert flows to the wetland dur-ng a period following pesticide application and to close wetlandnlet during times with drainage flow but no pesticide application.he farmer operates the system and is personally involved in theitigation process. The period when flows are diverted to the wet-

and depends on application date and lasts 1 month. This durations related to an average of three rainfall-runoff events per month

hich occurred generally in 1 month of rainfall (50 mm in averagen temperate climate). Here, this solution is called an off-streamrtificial wetland (Fig. 1).

. Materials and methods

.1. Catchment description

Two research sites in France having similar soil and land-cape characteristics were monitored. The Aulnoy catchment isocated in Aulnoy, France, 70 km North-East of Paris. The studyite is within the Orgeval catchment that is part of a water qualityesearch observatory (Anctil et al., 2009; Billy et al., 2010, 2011).he 36 ha agricultural watershed is situated in the sedimentaryaris Basin. The upper Brie Formation, tertiary siliceous and marlyimestone, is covered by quaternary loess deposits (Bricon andanipelle, 1963), which shows Luvisol (FAO, 2006) soils with alowly permeable layer within the soil’s upper meter. The agricul-ural watershed is managed by two farmers. Dominant crops grownn the region include rotations of cereals such as winter wheatTriticum turgidum), winter barley (Hordeum vulgare), maize (Zeaays), with canola (Brassica napus L.), vegetable crops includingorse bean (Vicia L.), peas (Pisum sativum L.), carrot (Daucus carota),nd lastly sugar beet (Beta vulgaris). The second catchment is theile drained Bray catchment (46 ha) located 225 km South-West ofaris. It has a hydromorphic Gleyic Luvisol soil (FAO, 2006) above

clay-with-flint layer. The farmer occupying the entire catchmentainly grows winter crop as canola (B. vulgaris), winter wheat (T.

urgidum) and barley (H. vulgare). A detailed description of the Brayatchment can be found in Passeport et al. (2010a). The Luvisol soilst both sites belong to the Aqualf suborder in the US taxonomy sys-em. The soil layer induces a perched temporary water table andinter waterlogging. This perched water table need to be artificiallyrained to provide favorable conditions of growth for winter cropseed before November. The agricultural land is subsurface-drainedsing perforated pipe installed 10 m apart and buried 0.8–1 m tonable soil cultivation during winter. The climate is oceanic tem-erate, with a mean annual temperature and precipitation of 12 ◦Cnd 750 mm, respectively.

.2. Artificial wetlands

The in-stream treatment was implemented in the Aulnoy water-hed. The study site comprises a shallow pond (50 cm deep) and aeservoir (3 m deep) used for supplemental irrigation built in 2001Fig. 2). The pond and reservoir surface area are 860 and 3305 m2,espectively, and the total estimated storage volume is 9000 m3.he area of these wetlands is 1.2% that of agricultural catchment.rainage waters arrive at the site via the main collector (diame-

er of 500 mm), then pass through the pond and finally arrive inhe reservoir. Both basins are also supplied by a permanent spring

oming from the Brie aquifer. The pond and reservoir are connectedy a 500-mm culvert. The reservoir is equipped with an irrigationater intake (pumps) and a shaft spillway. During the overfloweriod, reservoir water flows to the Beck of Burgundy River. Ten

J. Tournebize et al. / Ecological Engineering 56 (2013) 51– 59 53

Fig. 1. Sketch of in- and off-stream strategies.

tchm

pcom

t(tlfwgc(ooan

cr

3

3

wmTlaetiol

Fig. 2. In-stream Aulnoy ca

ercent of the pond surface is occupied by macrophytes, mainlyreeping bentgrass (Agrostis stolonifera), watercress (Nasturtiumfficinale), brooklime (Veronica beccabunga), and moneywort (Lysi-achia nummularia). The rest is free surface water.

At Bray, a three-cell ‘in-series’ artificial wetland was constructedo comprise the off-stream treatment wetland in December 2007Fig. 3). The total wetland surface area was 1280 m2, correspondingo less than 0.5% of the catchment surface area. The artificial wet-and water storage capacity was 330 m3 and the water depth variedrom 20 to 80 cm. Vegetation density increased to about 70% of theetland surface by 2010, and was mainly made up of reed sweet-

rass (Glyceria maxima, 53%), tall fescue (Festuca arundinacea, 12%),ommon reed (Phragmites australis, 10%) and reed canarygrassPhalaris aundinacea, 9%). The artificial wetland was constructedn an area that was previously cropped and artificially drained. Theld tile drains were initially cut and blocked in December 2007 but

dditional efforts to render all drains completely ineffective wereecessary in November 2008 and December 2008.

1 Mention of trade names, company names, or commercial products in this publi-ation is solely for the purpose of providing specific information and does not implyecommendation or endorsement by Irstea.

3

bdmotrm

ent and artificial wetland.

.3. Monitoring1

.3.1. The Aulnoy siteInstrumentation was set up in three locations. Two stations

ere placed to monitor flows at two inlets to the system: theain tile drainage outfall, and the flows from the tributary spring.

he third station monitored the outflow discharge from the wet-and. At each station, a calibrated 90-V notch weir was installednd equipped with two water level measurement devices (shaftncoder, SE-200 OTT and pressure transducer, Druck PDCR1830)o provide a system backup. The measurements provided 15-minnterval record at each station. Water level measurements at theutlet also provided a direct measure of the wetland basin’s waterevel.

.3.2. The Bray siteThe water being drained from the Bray catchment could either

e diverted to the treatment wetland or bypass it to directlyischarge to Le Calais creek via an agricultural ditch. Therefore,easurements of discharged from the agricultural catchment were

btained using different devices. A control section was installed inhe agricultural ditch where the water level was related to the flowate by a frequently verified rating curve. An electromagnetic floweter (MAG 8000 Siemens) was set up at the main inlet of artificial

54 J. Tournebize et al. / Ecological Engineering 56 (2013) 51– 59

tershe

wfiflde

aIwpai

3

dcotTacsc(

3

tafspacSal(t

atmcapecw

4

4

5dadiaiWctcsw

4w

laaa

Fig. 3. Off-stream Bray wa

etland. It was placed inside limited diameter PVC pipes. The arti-cial wetland inlet pipe diameter was 200 mm, limiting the inletow rate to approximately 35 L equivalent at 0.8 L s−1 per upstreamrained hectare. A calibrated 90-V notch weir was installed andquipped with a water level sensor at artificial wetland outlet.

For both sites, flow-weighted composite samples were takent the catchment and artificial wetland outlets by means ofSCO 3700 automatic samplers and collected approximately every

eek. Water samples were mainly collected in 500-mL amberolyethylene terephthalate (PET) bottles and stored at −20 ◦C untilnalysis. Discharge and pesticide concentration monitoring begann October 2008 for both sites.

.4. Pesticide applications

On average, 24 different agrichemicals are applied annually,ominated by herbicides (75%), fungicides (23%) and insecti-ides (2%). The chemicals applied encompassed a wide rangef physicochemical properties with average yearly applica-ion of 2 kg ha−1 (herbicides) and 0.3 g ha−1 (fungicides; seeable 1). For instance, weakly sorbing isoproturon, chlortoluronnd metazachlor (Koc sorption coefficient < 173 mL g−1) were asso-iated with large applied doses (>700 g ha−1 year−1), whereastrongly sorbing diflufenican and epoxiconazole (Koc sorptionoefficient > 1000 mL g−1) were applied in smaller quantities<70 g ha−1 year−1) (Fig. 4).

.5. Chemical analysis

We selected some of the pesticides among all those appliedo be representative of a wide range of chemical propertiesnd application amounts. Chemical analyses were treated dif-erently depending the site. Aulnoy samples were sent to aubcontractor, Institut Pasteur de Lille (France). Aclonifen andendimethalin were extracted by liquid-liquid extraction andnalyzed by GC/MS or GC/MS/MS or GC/NPD. Azoxystrobin,hlortoluron, diflufenican, epoxiconazole, iprodione, isoproturon,-metolachlor, napropamide and tebuconazol were analyzed after

solid–liquid extraction using LC–MS-MS. The quantificationimit was 0.02 �g L−1 for both. The Bray samples were filtered0.20 �m PET 20/15 MS Macherey-Nagel CHROMAFIL® syringe fil-ers), underwent a solid-phase microextraction (SPME) and were

0aia

d and artificial wetlands.

nalyzed using a multi-residue gas chromatography–mass spec-rometry (GC–MS, Trace DSQ, Thermo-Fisher Scientific) analytical

ethod. The method’s development, validation and uncertaintyalculations are fully detailed in Passeport et al. (2010b). Thisnalytical method enabled the simultaneous determination of 16esticides in water, including aclonifen, chlortoluron, diflufenican,poxiconazole, isoproturon, metazachlor, napropamide, prosulfo-arb, S-metolachlor and tebuconazol. Limits of quantification (LQ)ere 0.05, 0.1 or 0.5 �g L−1 depending on the chemical.

. Results

.1. Hydrology

Because of a lower than average rainfall in Aulnoy (588 and76 mm for 2008–2009 and 2009–2010, respectively), the volumerained was relatively low, with 120 and 63 mm (for 2008–2009nd 2009–2010) (Fig. 5). Thus, 42,500 m3 of water coming fromrainage area and 9500 m3 from spring inlet passed through Aulnoy

n-stream artificial wetland over 2 years. At Bray, the 2008–2009nd 2009–2010 rainfalls were 595 and 677 mm, respectively, lead-ng to 137 (2008–2009) and 67 mm (2009–2010) of drained water.

ith the open/close strategy, only 58% of the 94,260 m3 were inter-epted in the Bray off-stream artificial wetland. Of the total flowshat entered the wetlands 93% of inflows to the in-stream artifi-ial wetland volume were returned as outflow discharge to thetream, while only 71% of the inflows to the off-stream wetlandere discharged as outflows to the stream.

.2. Pesticide transfer at catchment outlet before artificialetland

In the following, we only consider the applied chemicals ana-yzed and listed in Table 1; however, many other chemicals werepplied on these two sites. All analyzed pesticides combined, theverage annual exported flux varied between 1 and 10 g ha−1. Theverage exportation rate as a function of the quantity applied was

.13%, with large variations according to the quantity applied, thepplication period and chemical’s properties. For instance, at Bray,soproturon (applied in November) exportations were 2.86% of thepplied mass in 2008–2009 and 1.91% in 2009–2010. These values

J. Tournebize et al. / Ecological Engineering 56 (2013) 51– 59 55

Fig. 4. Average amount of pesticides applied on the Aulnoy and Bray watersheds 179 per month, all crops combined.

wet

wa

(h((cKfatao(

4w

tpta

ttconcentrations, non detects were counted as nil.

The second point involves the difference in outflow concentra-tions in the two systems. The mean outflow concentrations were0.17 and 3.53 �g L−1 for in-stream and off-stream, respectively. As

Fig. 5. Daily discharge at drainage outlet for in- and off-stream

ere 3.92 and 0.27% for epoxiconazole (applied in April), and 0.35nd 0.55% for diflufenican (applied in November).

Overall, the exports of the chemicals considered to be mobileKoc < 500 mL g−1) and applied in the autumn are generallyigh. For example, chlortoluron, a herbicide applied in large1000 g ha−1) quantities and presenting a low sorption coefficientKoc = 205 mL g−1) was detected in 43% of the samples. Conversely,hemicals applied in smaller quantities and associated with higheroc values (>1000 mL g−1) were less frequently exported by subsur-

ace drainage, as noted for azoxystrobin. However, some chemicalsre easily transferred in the watershed despite their strong sorp-ion properties. For instance, epoxiconazole (Koc = 1073 mL g−1)nd tebuconazol (Koc = 769 mL g−1) were quantified in 53 and 50%f the samples, respectively, despite their low application rate<180 g ha−1).

.3. Comparison of inflow and outflow concentrations atetlands scale

Fig. 6 presents the sum of the pesticide “in and out” concen-

rations measured weekly in the two artificial wetlands. The firstoint that appears is the different order of magnitude betweenhe two sites. The sampling strategies were similar. The maximumnd the mean concentrations for weekly inflow concentrations at

Fl

land. The opening periods for off-stream wetland are in gray.

he Bray site were 18 and 4.9 �g L−1, whereas on the Aulnoy sitehey were 6 and 0.73 �g L−1, respectively. When summing these

ig. 6. Total pesticide concentration (�g L−1), inflow and outflow of artificial wet-and in in- and off-stream configurations over the 2008–2010 period.

56 J. Tournebize et al. / Ecological EngTa

ble

1Pe

stic

ide

pro

per

ties

(sor

pti

on

Koc

, deg

rad

atio

n

DT 5

0, f

rom

Foot

pri

nt

dat

abas

e

(201

0)),

app

lied

amou

nt,

infl

ow

and

outfl

ow

flu

xes

mea

sure

d

for

the

Au

lnoy

and

Bra

y

arti

fici

al

wet

lan

ds

(AW

s)

for

the

per

iod

2008

–200

9

and

2009

–201

0

per

iod

s.

Pest

icid

es

Ap

pli

cati

on

per

iod

Prop

erti

es

Au

lnoy

(in

-str

eam

con

figu

rati

on)

Bra

y

(off

-str

eam

con

figu

rati

on)

Koc

(mL

g−1)

DT5

0

fiel

d(d

ay)

Ap

pli

ed

amou

nt

atw

ater

shed

scal

e

(kg)

AW

Flu

x

(g)

Ap

pli

ed

amou

nt

atw

ater

shed

scal

e

(kg)

AW

flu

x

(g)

2008

/200

9

2009

/201

0

Inle

t

Ou

tlet

2008

/200

9

2009

/201

0

Inle

t

Perc

(%)a

Ou

tlet

Isop

rotu

ron

(H)

Au

tum

n

122

23

0

0

2.03

1.01

42

2.5

134.

6

(13)

63.4

Met

azac

hlo

r

(H)

Sum

mer

134

6.8

0

0

0

0

0

0

14.1

(36)

4.8

S-m

etol

ach

lor

(H)

Spri

ng

200

21

3.7

3.12

2.62

0.3

(51)

0.04

Ch

lort

olu

ron

(H)

Au

tum

n

205

34

12.6

7.5

8.62

11.3

40

25

353.

9(8

2)

49.4

Ipro

dio

ne

(F)

Spri

ng

373

26

5.6

0

0.43

0.58

Azo

xyst

robi

n

(F)

Spri

ng

423

70

0

1.6

0.06

0Te

buco

naz

ol

(F)

Spri

ng

769

56

0.5

2.43

0.4

0

0

0

1.8

(45)

0.7

Nap

rop

amid

e

(H)

Au

tom

n

885

72

0

0

0

0

7

15

2.7

(74)

1.2

Epox

icon

azol

e

(F)

Spri

ng

1073

120

1.2

0.49

1.21

0.21

0

1

16.6

(22)

4.6

Pros

ulf

ocar

b

(H)

Au

tum

n

1693

9.8

0

0

0

0

0

22

3

(87)

0.8

Pen

dim

eth

alin

(H)

Spri

ng

1744

90

8.2

3.84

3.24

0D

iflu

fen

ican

(H)

Au

tum

n

3186

315

0

0

0

0

0

0.2

3.3

(69)

1.4

Acl

onif

en

(H)

Spri

ng

7126

117

17.8

5.2

0.02

0

0.3

0

2.5

(82)

0.4

H:

her

bici

de;

F:

Fon

gici

de.

aPe

rc

(%)

stan

ds

for

per

cen

tage

of

inte

rcep

ted

flu

x

com

par

ed

to

tota

l exp

orte

d

flu

x

at

wat

ersh

ed

scal

e.

wt

4

sv(wiaarmcilitooac

4

wccwwBca

5

5

ieflTwmd

5

sbcspcflioirc

ineering 56 (2013) 51– 59

ill be discussed, this difference is related to residence time in thewo systems due to mainly storage volume.

.4. Pesticide flux reduction

Artificial wetland inlet loads ranged from 0 to 8.62 g year−1 (in-tream) and 0.3 to 353.9 g year−1 (off-stream), whereas outlet loadsaried from 0 to 11.34 at Aulnoy and 0.04 to 63.4 g year−1 at BrayTable 1). Usually, a larger load than the concentration reductionsas observed because wetland outflow rates were smaller than

nflow rates. Occasionally, larger outlet loads than those measuredt the inlet were observed, such as for chlortoluron at the Aulnoyrtificial wetland. In the in-stream configuration, the mean loadeductions were 54%, with reductions for the individual chemicalsonitored ranging from −32% to 100%. In this case, if mean effi-

iency is calculated by summing and comparing total loads from thenflow and outflow, the mean reduction is 18%, which is relativelyow. In the off stream configuration, internal efficiency (comparingnflow versus outflow) equals 75%, but if the open/close strategy isaken into account, the overall efficiency (comparing the drainageutlet before off-stream artificial wetland and the total catchmentutlet) is reduced to 45%. In both cases, the chemicals with highdsorption coefficients (Koc > 400 mL g−1) present nearly 100% effi-iency, and the mobile chemicals less than 60%.

.5. Interception

In Bray, more than 65% of the pesticide fluxes over the 2 yearsere indeed diverted to the off-stream wetland for aclonifen,

hlortoluron, diflufenican and napropamide, but loads of otherhemicals such as isoproturon, epoxiconazole and prosulfocarbere not as effectively diverted by gating flows to the off streametland (<30%). Overall, 58% of the tile drained flow from theray catchment was diverted to the off-stream wetland, whichomprised 70% of the total observed agrichemical load from thegricultural drainage, over the 2 monitored years.

. Discussion

.1. Water reduction volume

The reduction in volume induced by the wetlands implementeds not inconsiderable and can be amplified by the presence of veg-tation under the effect of evapotranspiration. Leaks or seepageows were also significant, especially at the Bray (off-stream) site.he full water balance including evaporation and seepage lossesere not evaluated here; effects of seepage losses in particularight need to be evaluated at a range of sites if this type of reme-

iation were to be implemented on a broader scale.

.2. Transfer from the drained catchment scale

The measured pesticide behavior is for the most part due topecific hydraulic functioning of agricultural drainage, as discussedy Branger et al. (2009). A chemical’s transfer time in soil profilean be very rapid when the water table is above the tiles and verylow when it is lower. Generally it is confirmed that (i) the mainesticide transfers take place after application, but that (ii) certainhemicals (e.g. iprodione, metazachlor) that are not applied duringow drainage period, can be detected for a longer period of time

n the drainage waters. Setting up an artificial wetland at the drain

utlet should be a complementary solution to other actions such asmplementing best management practices. Even if the dose/exportelation is not linear, reducing doses results in reducing pesti-ide fluxes. This must be placed within the wider perspective of

J. Tournebize et al. / Ecological Engineering 56 (2013) 51– 59 57

F inflow2

ppait<fiwtaw

5

ttccagttfy

vttloaz

tatuamiirOllIimtoitsttlshsppo

ig. 7. Reduction of pesticide fluxes as a percentage calculated by comparing the009–2010 periods.

esticide flux reduction at the catchment scale. The drain outletosition at the approximately 40-ha scale also seems to be welldapted to pesticides. Use of this strategy at a larger scale wouldnvolve diverting stream and river water below where these pollu-ants could have significant ecological impacts. Yet at this scale (i.e.,100 ha) the concentrations measured are sufficiently high for arti-cial wetlands to play a measurable buffer role. A planted artificialetland less than 1 m deep therefore showed better load reduc-

ions because the sorption and biodegradation processes are morective in a shallow vegetated wetland compared to a deep artificialetland as shown by Stehle et al. (2011).

.3. Interception

In the off-stream configuration, the open/close strategy leadso intercepting only a part of the flows theoretically targeted forhe periods presenting a greater risk of transfer. Depending on thelimatic conditions, this result is more or less optimal for pesti-ide flux interception. For isoproturon, the November 2008 climatefter application was rather dry and cold with periods of frost,reatly limiting fluxes in the month that followed. Isoproturonransfers therefore took place in January 2009, outside the periodhat the artificial wetland was open. The open/close rule can there-ore be improved to capture larger pesticide fluxes throughout theear.

The open/close strategy of the off-stream configuration has aery strong impact on the pesticide fluxes intercepted. The opera-ion rules were kept simple to ensure consistent decision making byhe farmer-operator. Yet data from the tile drainage outlet show the

oss of agrichemicals does not always follow the predicted patternf dominant export during the first rainfall-runoff events followingpplication. Certain chemicals such as metazachlor and tebucona-ol appear to have retention-release characteristic resulting from

ob

c

s and outflows, (a) in-stream, (b) off-stream, combined for the 2008–2009 and

he hydraulic functioning of the drain plots. This is also associ-ted with the variability in weather, which can result in pesticideransfer occurring under wet weather when the off-stream config-ration is not open. The operation rule should instead be driven by

drained volume after pesticide application, but this open/closeanagement will require the development of a specific warn-

ng tool for the farmer. This strategy of including the farmern managing these systems improves awareness of the chemicaleduction process and management issues with pesticide transfer.ne advantage in operating the off-stream option is that the wet-

and can be smaller, and a smaller area of productive agriculturaland is sacrificed to reduce environmental impacts of agriculture.n this study, the off-stream option provided a similar efficiencyn flux reduction compared with the in-stream configuration (a

ean 54% and 45% of all chemicals combined for 2 years) butook less land out of production. The land area occupied by theff-stream configuration is reduced to 0.5% versus 1.2% for then-stream configuration. Moreover, the storage volume is substan-ially reduced by 7 m3 per upstream drained hectare (0.7 mm runofftorage capacity) versus 300 m3 (30 mm runoff storage capacity) inhe in-stream configuration. A high storage volume may be difficulto put in place on the land because of topographical restrictions: theand drained is generally situated on a relatively flat plateau. But intream option could provide supplemental irrigation water, whichas some added benefit for the farmer and may even help compen-ate for some of the land taken out of production. The value of 7 m3

er upstream drained hectare seems low since the mean outputesticide concentrations are still high (3.53 �g L−1). Optimizationf the interception strategy and the retention processes would rec-

mmend an intermediate volume per upstream drained hectare, toe defined by modeling, for example.

In contexts in which land area is a limitation, the off-streamonfiguration is a solution to explore, but it requires specifically

5 cal Eng

di

5

drItg(ncucotautavwwtravlbb

risfotTaclrcefliwot

6

tnairaTcs

icrewfnaoiFlcta

A

(NFMc

R

A

B

B

B

B

B

B

B

C

C

F

F

G

H

8 J. Tournebize et al. / Ecologi

eveloped ecological engineering and the farmer’s commitment ints operation (Fig. 7).

.4. Compared efficiency of the 2 options

Associating the quantity applied (Table 1) and concentration, aifference between the inflows and outflows in the two configu-ations clearly appears. The application dose is the main reason.n both cases, the mean concentrations were reduced, althoughhis reduction seems greater for the in-stream configuration. Ineneral, the chemicals with a high adsorption coefficient (Koc)aclonifen, azoxystrobin, pendimethalin and tebuconazol) wereot found in the artificial wetland outlet waters in the in-streamonfiguration and in low quantities (<1 g) in the off-stream config-ration. In both cases, the chemicals with a low Koc (isoproturon,hlortoluron, S-metolachlor and metazachlor) were detected at theutlet and the flow reduction was low, even negative for chlor-oluron and iprodione for the in-stream configuration. Differencesppeared in terms of the processes involved. Yet in this config-ration, many samples showed pesticide concentrations belowhe detection threshold. The in-stream configuration encouraged

high level of dilution, masking other processes. A high storageolume (9000 m3) reduces pesticide concentrations in accordanceith the regulatory 0.5 �g L−1 for the total pesticides in drinkingater. The off-stream configuration reduces pesticide concentra-

ions but much less (3.53 �g L−1 versus 4.9 �g L−1). The study of theesults of the in-stream configuration shows that with a high stor-ge volume (9000 m3) dilution can be an important process. Theegetation in the off-stream configuration (70% of the cover versusess than 10% for the in-stream configuration) plays an indirect roley reducing speeds (Lange et al., 2011) but also a role in stimulatingiological degradation.

Efficiency is the highest for high-Koc chemicals. This results inanking adsorption processes first in the retention of pesticidesn artificial wetlands. However, the re-mobilization mechanismshould not be under-estimated, as shown by the in-stream resultsor chlortoluron and iprodione. This is one of the leading limitationsf artificial wetlands (Passeport et al., 2011a). Pesticide concen-rations should be followed up in sediments over the long term.oday, after 2 years of operation, none of the chemical sedimentnalyses (one per year in October) has presented a pesticide con-entration above the detection limit (1 mg kg−1 dry matter). Theow mean pesticide flux reduction levels in the in-stream configu-ation are partially due to the negative efficiency measured for twohemicals: chlortoluron and iprodione. The first year of follow-up,fficiency was reasonably positive, but the second year the outletuxes were higher than the inlet fluxes, which indicates that there

s a re-mobilization process limiting the efficiency of the artificialetland. The second main limitation of this technique is the toxicity

f pesticide degradation products. Metabolites are not well knowno ensure an actual toxicity reduction back in surface water.

. Conclusions

Both of the experiments presented reflect actual conditions inhe field, including climate variability. The results therefore showot only a range of pesticide dissipation potential through thertificial wetlands, but also provide an actual simulation involv-ng a farmer at the farm scale. Both configurations gave positiveesults, but each within its own limits. The efficiency levels showed

close dependence on each chemical’s sorption properties (Koc).he artificial wetlands are less efficient for mobile chemicals: effi-iency also depends on the climate conditions and varies from oneeason to another. A larger storage volume and depth resulted

I

ineering 56 (2013) 51– 59

n substantial dilution and this apparently reduced the biologi-al degradation of the pesticides. Future studies should examinee-mobilization processes (desorption) in detail. Given the socio-conomic stakes farmers must face, the off-stream configurationith the open/close strategy appears to be a good option to explore

urther. This configuration can strongly reduce the amount of landecessary for a similar level of efficiency. In addition, it can recreate

true wetland system, thereby favoring biodiversity. The in-streamption might work better if, for instance, water storage for re-rrigation of drained land during drought is of interest to farmers.inally, the limitations found in the results prove that artificial wet-ands, although promising, should be associated with a number ofomplementary measures from reduction at the plot level (reduc-ion of the doses applied, adaptation of the treatment strategy) thatlso help limit of pesticide transfers from the catchment.

cknowledgments

Fundings for this research were provided by ArtWET LIFE06/ENV/F/000133) financial supports AESN (Agence de l’Eau Seineormandie), AELB (Agence de l’Eau Loire Bretagne), Région Ile derance (DIM ASTREA) and ONEMA (Office National de l’Eau et desilieux Aquatiques, France). The authors would like to thank the

ooperative farmers E. Gobard and B. Crespin.

eferences

nctil, F., Filion, M., Tournebize, J., 2009. A neural network experiment on the sim-ulation of daily nitrate–nitrogen and suspended sediment fluxes from a smallagricultural catchment. Ecol. Modell. 220, 879–887.

illy, C., Billen, G., Sebilo, M., Birgand, F., Tournebize, J., 2010. Nitrogen isotopiccomposition of leached nitrate and soil organic matter as an indicator of denitri-fication in a sloping drained agricultural plot and adjacent uncultivated riparianbuffer strips. Soil Biol. Biochem. 42, 108–117.

illy, C., Birgand, F., Sebilo, M., Billen, G., Tournebize, J., Kao, C., 2011. Nitratedynamics in artificially drained nested watersheds. Phys. Chem. Earth 36,506–514.

lankenberg, A.G.B., Haarstad, K., Braskerud, B.C., 2007. Pesticide retention in anexperimental wetland treating non-point source pollution from agriculturerunoff. Water Sci. Technol. 55, 37–44.

ranger, F., Tournebize, J., Carluer, N., Kao, C., Braud, I., Vauclin, M., 2009.A simplified modelling approach for pesticide transport in a tile-drained field: The PESTDRAIN model. Agric. Water Manage. 96, 415–428,http://dx.doi.org/10.1016/j.agwat.2008.09.005.

raskerud, B.C., Haarstad, K., 2003. Screening the retention of thirteen pesticides ina small constructed wetland. Water Sci. Technol. 48 (5), 267–274.

ricon, C., Canipelle, A., 1963. Etude expérimentale du Bassin experimental del’Orgeval. Rapport. Bureau de Recherches Géologiques et Miniéres, Paris (inFrench).

rown, C.D., van Beinum, W., 2009. Pesticide transport via sub-surfacedrains in Europe. Environ. Pollut. 157, 3314–3324, http://dx.doi.org/10.1016/j.envpol.2009.06.029.

arter, 2000. Herbicide movement in soils: principles, pathways and processes.Weed Research 40, 113–122.

heng, S., Vidakovic-Cifrek, Z., Grosse, W., Karrenbrock, F., 2002. Xenobioticsremoval from polluted water by a multifunctional constructed wetland. Chemo-sphere 48, 415–418.

AO, 2006. Guidelines for Soil Description. Food and Agriculture Organization of theUnited Nations, Rome, Italy.

ootprint, 2010. PPDB (2009). The Pesticide Properties Database (PPDB) developedby the Agriculture & Environment Research Unit (AERU), University of Hertford-shire, funded by UK National Sources and the EU-funded FOOTPRINT project(FP6-SSP-022704).

regoire, C., Elsaesser, D., Huguenot, D., Lange, J., Lebeau, T., Merli, A., Mose, R., Passe-port, E., Payraudeau, S., Schutz, T., Schulz, R., Tapia-Padilla, G., Tournebize, J.,Trevisan, M., Wanko, A., 2009. Mitigation of agricultural nonpoint-source pesti-cide pollution in artificial wetland ecosystems. Environ. Chem. Lett. 7, 205–231.

unt, J., Anderson, B., Philips, B., Tjeerdema, R., Largay, B., Beretti, M., Bern, A., 2008.Use of toxicity identification evaluations to determine the pesticide mitigation

effectiveness of on-farm vegetated treatment systems. Environ. Pollut. 156 (2),348–358.

mfeld, G., Braeckevelt, M., Kuschk, P., Richnow, H.H., 2009. Monitoring and assessingprocesses of organic chemicals removal in constructed wetlands. Chemosphere74, 349–362.

cal Eng

K

L

M

M

N

O

P

P

P

P

R

R

R

S

S

T

catch-basin: a solution to improve groundwater quality. Irrig. Drain. 61 (Suppl.1), 75–86.

van der Valk, A.G., Jolly, R.W., 1992. Recommendations for research to develop guide-

J. Tournebize et al. / Ecologi

ladivko, E.J., Brown, L.C., Baker, J.L., 2001. Pesticide transport to subsurface tiledrains in humid regions of North America. Crit. Rev. Environ. Sci. Technol. 31,1–62, http://dx.doi.org/10.1080/20016491089163.

ange, J., Schuetz, T., Gregoire, C., Elsasser, D., Schulz, R., Passeport, E., Tournebize, J.,2011. Multi-tracer experiments to characterise contaminant mitigation capac-ities for different types of artificial wetlands. Int. J. Environ. Anal. Chem. 91,768–785.

atamoros, V., Puigagut, J., García, J., Bayona, J.M., 2007. Behavior of selected pri-ority organic pollutants in horizontal subsurface flow constructed wetlands: apreliminary screening. Chemosphere 69 (9), 1374–1380.

iller, P.S., Mitchell, J.K., Cooke, R.A., Engel, B.A., 2002. A wetland to improveagricultural subsurface drainage water quality. Trans. Am. Soc. Agric. Eng. 45,1305–1317.

ovak, S.M., Portal, J.M., Schiavon, M., 2001. Effects of soil type upon metolachlorlosses in subsurface drainage. Chemosphere 42, 235–244.

’Geen, A.T., Budd, R., Gan, J., Maynard, S.J.P., 2010. Mitigating nonpoint sourcepollution in agriculture with constructed and restored wetlands. Advances inAgronomy, vol. 108, pp. 1–76 (Chapter I).

asseport, E., Benoit, P., Bergheaud, V., Coquet, Y., Tournebize, J., 2011a. Selectedpesticides adsorption and desorption in substrates from artificial wetland andforest buffer. Environ. Toxicol. Chem. 30, 1669–1676.

asseport, E., Benoit, P., Bergheaud, V., Coquet, Y., Tournebize, J., 2011b. Epoxi-conazole degradation from artificial wetland and forest buffer substrates underflooded conditions. Chem. Eng. J. 173, 760–765.

asseport, E., Guenne, A., Culhaoglu, T., Moreau, S., Bouye, J.M., Tournebize, J., 2010a.Design of experiments and detailed uncertainty analysis to develop and val-idate a solid-phase microextraction/gas chromatography–mass spectrometrymethod for the simultaneous analysis of 16 pesticides in water. J. Chromatogr.A 1217, 5317–5327.

ineering 56 (2013) 51– 59 59

asseport, E., Tournebize, J., Jankowfsky, S., Promse, B., Chaumont, C., Coquet, Y.,Lange, J., 2010b. Artificial wetland and forest buffer zone: hydraulic and tracercharacterization. Vadose Zone J. 9 (1), 73–84.

eichenberger, S., Bach, M., Skitschak, A., Frede, H.-G., 2007. Mitigation strategies toreduce pesticide inputs into ground- and surface water and their effectiveness:a review. Sci. Total Environ. 384, 1–35.

iise, G., Lundekvam, H., Wu, Q.L., Haugen, L.E., Mulder, J., 2004. Loss of pesticidesfrom agricultural fields in SE Norway – runoff through surface and drainagewater. Environ. Geochem. Health 26, 269–276.

ose, M.T., Sanchez-Bayo, F., Crossan, A.N., Kennedy, I.R., 2006. Pesticide removalfrom cotton farm tailwater by a pilot-scale ponded wetland. Chemosphere 63,1849–1858.

chulz, R., Hahn, C., Bennett, E.R., Dabrowski, J.M., Thiere, G., Peall, S.K.C., 2003.Fate and effects of azinphos-methyl in a flow-through wetland in South Africa.Environ. Sci. Technol. 37, 2139–2144.

tehle, S., Elsaesser, D., Gregoire, C., Imfeld, G., Niehaus, E., Passeport, E., Payraudeau,S., Schafer, R.B., Tournebize, J., Schulz, R., 2011. Pesticide risk mitigationby vegetated treatment systems: a meta-analysis. J. Environ. Quality 40,1068–1080.

ournebize, J., Gramaglia, C., Birmant, F., Bouarfa, S., Chaumont, C., Vincent, B., 2012.Co-design of constructed wetlands to mitigate pesticide pollution in a drained

lines for the use of wetlands to control rural nonpoint source pollution. Ecol. Eng.1, 115–134.