Environmental Pollution 148 (2007) 729e738

www.elsevier.com/locate/envpol

Hydrodynamic and geochemical constraints on pesticide

concentrations in the groundwater of an agricultural
catchment (Brevilles, France)
N. Baran*, C. Mouvet, Ph. Negrel
BRGM, 3 avenue Claude Guillemin, BP 6009, 45060 Orleans Cedex 2, France
Received 29 January 2007; accepted 31 January 2007

We present an integrated approach combining geochemistry and hydrogeology that leads to a better understanding
of the spatial and temporal fluctuations of the pesticide concentrations in groundwater of a pilot agricultural catchment.
Abstract
The monitoring of a spring and seven piezometers in the 3 km2 Brevilles agricultural catchment (France) over five and a half years revealed
considerable spatial and temporal variability in the concentrations of atrazine and its metabolite deethylatrazine (both systematically quantified
at the outlet spring): maximum 0.97 and 2.72 mg L1, mean 0.19 and 0.59 mg L1, respectively. Isoproturon, the pesticide applied in the greatest
amount, was detected in only 10 of the 133 samples. These observations can only partly be explained by land use and intrinsic pesticide properties. Geochemical measurements and tritium dating showed the importance of the stratification of the sandy saturated zone and the buffer
function of the unsaturated limestone. Principal component analysis on 39 monthly data series of atrazine, deethylatrazine, nitrate, chloride
and piezometric levels revealed a temporal structuring of the data possibly reflecting the existence within the aquifer of two different reservoirs
with time-variable contributions.
2007 Elsevier Ltd. All rights reserved.
Keywords: Pesticide; Catchment; Groundwater; Water Framework Directive; Tritium

1. Introduction
Enhanced monitoring of groundwater quality over several
years has revealed a pesticide contamination of aquifers in
North America and Europe (Barbash et al., 2001; EEA,
1999; Gilliom et al., 2006; IFEN, 2004). The observed concentrations are in places higher than the European drinking-water
limit, 0.1 mg L1 per substance except for aldrin, dieldrin and
heptachlor, 0.03 mg L1.
The objective set by the Water Framework Directive (WFD
e 2000/60/EC, OJEC, 2000) is for all groundwater bodies to
achieve the good quantitative and chemical status . at the latest by 2015. The Directive demands that European Union
* Corresponding author. Tel.: 33 (0) 2 3864 3271; fax: 33 (0) 2 3864
3446.
E-mail address: n.baran@brgm.fr (N. Baran).
0269-7491/$ - see front matter 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.envpol.2007.01.033

Member States not only characterize their levels of groundwater contamination, but also that they study the evolutionary
trends of their pollutant concentrations. This means that they
must be in a position to explain possible cases of non-achievement. In France, 99% of the groundwater bodies risk nonachievement due to either nitrate or pesticide contaminations
(Normand and Gravier, 2005). Monitoring groundwater quality for these two parameters is thus particularly relevant.
Several countries have approached the stage of characterization of their groundwater bodies either by using data derived from various measurement networks, as in France
(IFEN, 2004), or by establishing specific sampling and analysis protocols (NAQUA network in Switzerland e OFEFP and
OFEG, 2004; NAWQA network in the United States e USGS,
2006). Pesticide monitoring networks, where they exist, are often less than 10 years old with a fairly low measurement frequency (1e4 analyses per year). Trend interpretations are thus

730

N. Baran et al. / Environmental Pollution 148 (2007) 729e738

difficult and limited. The adopted approach is generally simple, often only considering the frequency at which threshold
values (e.g. potability limit or maximum acceptable concentration limit) are exceeded.
Characterizing an entire groundwater body from observations limited in time and space remains a challenge. Little published data exist concerning intensive monitoring over several
years, whether at the catchment outlet (Morvan et al., 2006;
Rowden et al., 2001) or at observation points spread over a basin (De Guzman et al., 2005; Lapworth et al., 2006). Only
a few studies have been published in which hydrogeological
contexts and hydrodynamic have been considered (Kolpin
et al., 1997; Lapworth and Gooddy, 2006). By taking hydrodynamic into consideration by using the age of the water determined from tritium concentrations, it is possible to show
spatial and temporal trends of herbicide concentrations in
groundwater at both an aquifer scale (Trojan et al., 2003)
and state- or nationwide scale (Kolpin et al., 2004, 1995; Mills
et al., 2005).
Notable changes in the use of pesticides generally result
from the development of regulations. In Europe, the herbicides
atrazine and isoproturon have been classified as priority substances (2455/2001/EC, OJEC, 2001). As of September 2003
all use of atrazine was forbidden in France following restrictions already in force since 1991. In January 2004 the maximum permitted application of isoproturon was reduced from
2500 to 1800 g ha1.
Bearing in mind the gaps in scientific knowledge on the one
hand and the implementation of the WFD on the other, the
aims of our study were to (i) characterize the spatial and temporal variability of groundwater contamination by different
pesticides with varied physical and chemical characteristics,
(ii) identify possible temporal trends, and (iii) seek explanations for the former two points. To attain these objectives,
a 5.5 year monitoring program, in particular of atrazine, isoproturon and their metabolites, but also other geochemical
and hydrodynamic parameters combined in an integrated approach, was conducted in the small (3 km2), well-instrumented
(seven piezometers, one gauging station at the discharging
spring) Brevilles catchment in France (Morvan et al., 2006).
2. Materials and methods
2.1. Site location and hydrogeological context
The main spring is the outlet of a 3 km2 catchment (70 km west of Paris e
49 100 3700 N, 1 410 2000 E e Fig. 1). In 2000, the established atrazine (a herbicide for corn) and deethylatrazine (its main metabolite) contamination of the
spring led the farmers to modify their cultivation practices by halting all use
of atrazine 3 years before the national ban on this product.
The unconfined aquifer is mainly in the 15-m thick Cuise Sands that are
overlain by variably calcareous and clayey Lutetian carbonates. The Lutetian
is more than 30 m thick in the upper part of the basin where it is overlain by
Bartonian limestone. The impermeable substratum of the aquifer is Sparnacian
clay.
Forest covers about 10% of the area, mainly in the upper part of the catchment. The remainder of the catchment is covered by agricultural land. The
catchment contains no farm buildings and no pesticide storage sites or landfill
areas that could cause point pollution. There is also no railway and only

a single secondary paved road. The observed contamination is derived exclusively from diffuse sources.

2.2. Agricultural practices

The gathering of information regarding agricultural practices for over a decade was facilitated by the fact that the agricultural catchment is managed by
only seven farmers, one of whom already covers 40% of the surface area.
Wheat is the main crop (about 50% of the farmland), and corn represents about
15%. The remaining 35% corresponds to peas, oilseed rape, barley or fallow
land, in variable proportions according to the years. The average amount of nitrogen (ammonitrate, urea and manure) applied to the catchment over the years
for which we have information for more than 85% of the catchment (1999e
2003) was 12.84 t (2.14).
Isoproturon and chlortoluron, herbicides for cereals, were the active ingredients used in the largest quantities (a total of 694 and 352 kg, respectively,
during the period 1994e2004). This was not surprising, considering that cereals have represented around 50% of the crop for at least the last decade.
For the considered period (1994e2004), atrazine, the only active ingredient systematically detected at the spring, ranked 15th in total mass applied
(125 kg), which is less than acetochlor (212 kg), the substance used as an atrazine substitute since 2000. Fig. 1 shows the areas of atrazine application declared by the farmers, and the plots of corn cultivation with no declared
atrazine application; nonetheless, these plots have very likely been treated
with atrazine since this herbicide was classically used on corn throughout
the region. The small amount of atrazine applied over the period 1994e
2004 has to be placed in the context of the major change in the agricultural
practices since 2000. The last atrazine applications in April 1999 used
1 kg ha1 and no further applications occurred for the period 2000e2005.

2.3. Piezometer drilling and equipment

Seven piezometers (Pz), located throughout the catchment (Fig. 1), were
drilled and equipped down to the Sparnacian clay at the substratum of the
aquifer in early 2001. The piezometers were screened (0.7 mm slots) over
the full thickness of the Cuise Sands and locally over the lower part of the
limestone to enable tapping the entire aquifer. Pz4 was an exception with
only the deepest part of the Sands being screened (7 m out of 10 m). Piezometric levels were measured at each field visit, i.e. at least once a month.
A monitoring well cluster (Pz17A, Pz17B and Pz17C) was drilled in summer 2005 in a cultivated plot (wheat) close to Pz7. The three drillings of this
cluster, 2 m apart, are of different depths to enable studying the groundwater
stratification. Pz17A taps the upper part of the aquifer from 7.8 to 10.8 meters
below surface (mbs); Pz17B taps the upper part of the Sands from 12 to
15 mbs; Pz17C taps the lower part of the aquifer from 17 to 20 mbs.

2.4. Sampling and analytical methods

Sampling campaigns were conducted until August 2005, monthly for the
seven piezometers from March 2001, and bi-monthly in the spring from October 1999 to August 2004 after which the frequency became monthly. The
piezometers were sampled after pumping three purge volumes to stabilize the
groundwaters physical and chemical parameters, such as pH and conductivity.
The pump in each piezometer was located at the same depth for each sampling
campaign.
Several samples were also taken to quantify tritium activities in the
groundwater (April 2001 and June 2002, only at the spring; February 2003
and August 2005, all piezometers and the spring). Samples for chloride, nitrate
and sulfate determinations were kept, after in situ filtration (0.45 mm), in polyethylene bottles. They were analysed by ion chromatography (accuracy 5e
10%; quantification limit 0.5 mg L1).
Samples for pesticide analysis were collected in 1 L pre-combusted brown
glass containers kept on ice in coolers for transport to the laboratory and then
stored at 4  C until analysis. After spiking with deuterated atrazine-d5 as
surrogate, atrazine (At) and its metabolites (deethylatrazine e DEA and deisopropylatrazine e DIA), isoproturon (IPU) and its metabolites (monodesmethylisoproturon e MDIPU and didesmethylisoproturon e DDIPU), chlortoluron

N. Baran et al. / Environmental Pollution 148 (2007) 729e738

731

Fig. 1. Location of the spring and piezometers (Pz), of plots cultivated at least once with corn and values of the quantities of atrazine applied over the period 1994e
2004. (The size of the circles is proportional to the total quantity applied over the period 1994e2004 whereas each sector corresponds to a year given.)
(CTU), and acetochlor (ACT, a corn herbicide) were extracted by liquideliquid separation. Acetochlor degradates (oxanilic acid, OA, and ethanesulfonic
acid, ESA) were extracted by solid phase extraction after spiking with deuterated 2,4D-d3. Triazines, urea and acetochlor degradates were analysed by liquid chromatography (ThermoHypersil column; 20 mL injected) coupled with
mass spectrometry (DECA XP and Thermo-Finnigan ion trap mass spectrometer equipped with an Electro Spray Ionization source). A gradient of water and acetonitrile as mobile phases (0.75 mL min1) was used. Acetochlor
was analysed by gas chromatography (capillary column: 30 m, 0.25 mm i.d.,
0.25 m film thickness) coupled with mass spectrometry (VARIAN GC 3400
Saturn 2000). The injection of 1 mL of extract was performed in splitless
mode. Pesticide quantification limits were at the most 0.05 mg L1 for all pesticides e a threshold that we were able to lower during the monitoring. In February 2005, liquidesolid extraction with an OASIS HLB (Waters) cartridge
was used, enabling recovery rates of 110% for DIA as against about 30% previously. The extraction yields for the other molecules were higher than 90%
and thus comply with the accepted standards (EPA, 1995). Uncertainties for
pesticide analysis were estimated by the GUM method (ISO, 1995). Expanded
uncertainties, obtained by multiplying the relative combined uncertainty by
a coverage factor (k 2, which is related to a confidence level of 95%) varied
between 5 and 60% according to the concentration and the molecule studied;
they are lower than 20% for almost all the samples for At and DEA in our
study.
When drilling piezometers 2, 3, 5 and 6, solids of the unsaturated zone
were collected and kept in hermetically sealed glass bottles. The water from
these solids was collected by freeze-drying and used for tritium activity

measurements. Sample aliquots (water from solids or groundwater) of

10 mL were mixed with 10 mL of a scintillation compound (Pico-Fluor
LLT) in plastic vials and counted on a low-background liquid scintillation
counter (PACKARD 2250 CA). The detection limit of 10 TU (Tritium Unit)
was empirically defined by analysing IAEA (International Atomic Energy
Agency) standard solutions. Tritium activity below this limit was determined
for certain samples after electrolytic enrichment. Tritium measurements in the
groundwater and in water from solids of the unsaturated zone were used to estimate the residence time of the water (Clark and Fritz, 1997; Kolpin et al.,
2004) and the water velocity in the unsaturated zone, the tritium peak being
attributed to 1960s, the decade with the maximum emissions (Clark and Fritz,
1997). To allow comparison of samples taken and analysed at different dates,
natural radioactive decay was considered.

2.5. Weather data

A national meteorological network rain gauge, installed 3.5 km from the
site in 1993, records the daily rainfall (P) yielding an annual average of
759 mm (1993e2005). Years 1999, 2000 and 2001 (939, 978 and 932 mm, respectively) and years 2003, 2004 and 2005 (535, 634 and 572 mm, respectively) were particularly wet and dry, respectively.
Daily potential evapotranspiration (PET), calculated using the Penmanne
Monteith equation, was taken from the closest national weather station, located
about 40 km southwest of the catchment. Years 2000, 2001, 2002, 2004 and 2005
showed similar values (703, 744, 756, 760 and 703 mm, respectively), whereas

732

N. Baran et al. / Environmental Pollution 148 (2007) 729e738

year 2003 gave a higher value (827 mm). The P and PET data were used to
estimate the number of water infiltration events every year.

3. Results and discussion

3.1. Pesticides
3.1.1. At the catchment outlet (the Brevilles spring)
Atrazine and its main degradate, deethylatrazine, were systematically detected in the spring (Fig. 2), even though the last
atrazine application in the catchment was in April 1999. The
mean concentrations (n 133) and standard deviation were
0.19  0.7 mg L1 and 0.59  0.18 mg L1 for At and DEA, respectively. Maximum concentrations (0.43 and 1.16 mg L1)
were observed on 9th October 2000 and 19th January 2001
for At and DEA, respectively. This continued detection suggests a stock of these compounds in the soil and/or unsaturated
zone that is partly mobilized during each infiltration event. A
second hypothesis to explain the atrazine persistence is that
the groundwater renewal time is longer than the monitoring
period of 5.5 years and that there is limited pesticide degradation within the groundwater body (Albrechtsen et al., 2003).
The atrazine and deethylatrazine concentrations showed
a strong temporal variability, more particularly marked at the
beginning of the monitoring. The values, notably for atrazine,
were more constant after April 2003. A hypothesis to explain
the stabilization of the concentrations is a reduction in the
number of infiltration events after this date. Indeed, a simple
monthly calculation of the hydrological balance, (PePET),
showed that the recharge events were much greater (in both
number and volume) in 2000, 2001 and 2002 than in the following years (Fig. 2).
The time-series of isoproturon and chlortoluron concentrations were very different to that of atrazine (Fig. 2). The two
substituted ureas were only very rarely quantified (once and
twice on 133 samples for IPU and CTU, respectively) throughout the monitoring and a few times detected at low

concentrations that could not be accurately quantified (10

and 3 detections, respectively). These rare detections in comparison to those of atrazine could appear surprising since IPU
and CTU applications in the catchment over the period 1994e
2004 were 5.5 and 2.8 times higher than that of atrazine. However, it is in agreement with earlier studies confirming the rare
detection of these molecules in groundwaters at national scale
(IFEN, 2004), a phenomenon probably due to the molecules
physical and chemical properties and notably to the rapid formation of large amounts of bound residues (Perrin-Ganier
et al., 1996; Pieuchot et al., 1996). The greater risk of At
leaching compared to IPU has been demonstrated by comparison of sorption, volatilization and mineralisation processes of
both molecules (Boivin et al., 2005; Mordaunt et al., 2005). In
addition, Walker et al. (2005) highlighted the reduced leaching
losses with increased delay between application and significant rainfall events. The results of Walker et al. (2005) could
explain why IPU and CTU have only been quantified here during year 2001, the wettest year of the monitoring period with
a very rainy month of March (three times the average rainfall)
closely matching the period of phenylurea applications.
MDIPU and DDIPU were never detected at the spring. The
absence of DDIPU detection could be due to the fact that this
product does not appear in the first stages of biodegradation.
The absence of MDIPU, a metabolite that appears in the first
stage of biodegradation in the surface layers (Alletto et al.,
2006), is more likely to be due to its rapid degradation than
to an absence of production (Sorensen and Aamand, 2001).
After the solideliquid extraction method was adopted, DIA
was detected more frequently (four times out of the nine
samples analysed since February 2005) although the concentrations were very low and close to the 0.025 mg L1 quantification limit. These observations are in agreement with the
results obtained for different groundwater measurement
programs in which DIA is detected less frequently than
DEA or atrazine (Kolpin et al., 2000, 2004; Pucarevic et al.,
2002).

Fig. 2. Time-series of atrazine, deethylatrazine (DEA), isoproturon (IPU), and chlotoluron (CTU) concentrations (mg L1) at the spring and monthly precipitation
minus potential evapotranspiration (mm).

N. Baran et al. / Environmental Pollution 148 (2007) 729e738

3.1.2. Spatial and temporal variability at the

catchment scale
Acetochlor and its two main metabolites, ESA and OA,
have so far not been quantified at any of the eight spatially distributed observation points (including the spring). However,
a risk of metabolite leaching does exist as they have been detected in other aquifers (De Guzman et al., 2005; Kolpin et al.,
2000) and at depth in the soils of the present catchment (Baran
et al., 2004). The absence of detection could result from (i) the
small amounts applied to date over the whole of the catchment, (ii) the location of the treated plots, preferentially in
the upper part of the catchment, thus far from the spring and
in areas where the unsaturated zone is thicker than 15 m,
and (iii) little recharge since 2003, the two last points suggesting that the time lag necessary for the molecules to reach the
groundwater could be longer than the duration of our
monitoring.
IPU was not detected frequently in any of the piezometers
(as observed for the spring), generally less than twice over the
entire monitoring period, except for Pz5 (38 detections for 44
samples, maximum concentration: 0.13 mg L1). The degradation products MDIPU and DDIPU were never detected in any
of the piezometers, whilst CTU was detected only a very few
times and, like IPU, at periods close to the applications, suggesting that a fraction of these pesticides applied in certain
parts of the catchment may rapidly reach the groundwater.
The atrazine and deethylatrazine concentrations in the piezometers varied from not detected (<0.025 mg L1 for atrazine
in Pz4) to several mg L1 (0.97 and 2.72 mg L1, respectively
for atrazine and deethylatrazine in Pz5). Fig. 3 gives whisker
plots of the observed data for each piezometer throughout
the monitoring and shows that the spatial and temporal variabilities of the contamination are large at any given point.
No overall clear logic in the upgradientedowngradient distribution of the contamination was found. The lower At and
DEA concentrations observed for Pz7 and the spring compared to Pz5 can nevertheless be tentatively explained. Since
Pz5 is upgradient from Pz7 and the spring, but also in the
most downgradient plot known to have had a corn crop, recharge water infiltrating plots of no corn cultivation located
between Pz5 and Pz7 and the spring could lower the At and
DEA concentrations in Pz7 and the spring.
The pesticide concentrations were not directly linked with
the amounts applied in the corresponding plot over the last
12 years for any of the piezometers. Pz6, which was in one
of the two plots that received the most atrazine over the period
1994e2004, did not show the highest atrazine or DEA concentrations. Neither did the intensity of contamination at any
given point appear to be directly linked to the thickness of
the unsaturated zone above that point; Pz4 and Pz5, with similar thicknesses of unsaturated zone, showed very different degrees of contamination.
Piezometer 4 appeared to be uncontaminated by both pesticides and nitrate (Fig. 3). It lies 200 m upgradient from the
spring the concentrations of which were close to 0.2 and
0.6 mg L1 for atrazine and deethylatrazine, respectively. No
application of atrazine has been declared for this plot

733

(Fig. 1) and this could explain the fact that the parent molecule
and its degradates were not detected in Pz4. Nevertheless, several plots upstream were treated with atrazine and could have
induced contamination of Pz4 via transfers in the saturated
zone, as was probably the case for Pz5 and Pz8 which were
also in plots untreated with atrazine but downstream of several
treated plots (Fig. 1). Pz4 was in a plot of cultivated land with
nitrate loading comparable with the other plots, unlike applications of pesticides that are variable from one plot to another.
Consequently, the lack of nitrate contamination compared to
other piezometers remains unexplained. We therefore need
to improve our knowledge of the hydrodynamic of the system
so as to better understand the temporal and spatial variabilities
of the pesticide concentrations in the groundwater.
3.2. Hydrodynamic of the system: clues from
groundwater chemistry
Nitrate data, alone or combined with chloride or sulfate and
tritium data, are among the geochemical parameters that may
improve the understanding of aquifer hydrodynamic (Negrel
and Pauwels, 2004). Examination of the chloride vs. sodium
concentrations (Fig. 4a) showed that Pz3 and Pz4 water was
close to the seawater dilution curve (SW) and thus only
slightly modified by human activity, such as the input of mineral fertilizers or organic fertilization (Negrel and Pauwels,
2004). Anthropic impact was, however, perceptible for the
other piezometers where the chloride/sodium ratios moved
away from the SW dilution curve (Fig. 4a).
The low chloride concentrations measured for Pz3 and Pz4,
(median values of 10.3 and 13.8 mg L1, respectively), the sulfate concentrations of the two piezometers being similar to
those of the other piezometers, plus the unconfined character
of the groundwater, suggested that denitrification cannot be invoked to explain the low observed nitrate concentrations for
Pz3 and Pz4 (Pauwels and Talbo, 2004). One hypothesis is
that the solutes have not yet reached the groundwater due to
a slow infiltration rate above these points. Consequently, tritium measurements were made at the spring and all piezometers to estimate the residence time of the water at each
observation point (Clark and Fritz, 1997; Kolpin et al.,
2004). Tritium activities below <0.8 TU correspond to submodern water recharged before 1952 (first thermonuclear
test), high activities (>30 TU) to water with a considerable
component of recharge from the 1960s to 1970s, while intermediate values are a mixture of sub-modern and recent recharge (Clark and Fritz, 1997).
As a first approximation to enable a direct comparison of
all observation points, it was possible to consider that, cumulated over several years, nitrate loadings were relatively similar throughout the cultivated sector because of its small size,
the few farmers concerned, the crop rotations and the inputs
necessary for all crops. Considering the small size of the
catchment, the tritium input function via rainfall could be considered homogeneous on the basin scale, allowing comparison
of nitrate/tritium pairs between different points (Fig. 4b). The
lowest nitrate contamination was clearly in the oldest water,

734

N. Baran et al. / Environmental Pollution 148 (2007) 729e738

Fig. 3. Box-whisker plot of atrazine, DEA, chloride, nitrate, piezometric level and depth of unsaturated zone. Median, 25th and 75th percentiles (1st and 3rd quartiles corresponding to the bottom and the top of the box), minimum and maximum (whiskers) and outliers (any point that falls below lower quartile minus 1.5 times
the distance between the quartiles or above upper quartile plus 1.5 times the distance between the quartiles) are represented.

some older than 1963 (Pz4 and Pz17C) and thus prior to the
agricultural intensification of the 1970s. Pz3 showed a concentration of 4 TU (a value lower than actual concentrations in
precipitations of the Orleans area; unpublished data, BRGM)
that could correspond to a mixture of pre- and post-1953 waters, leading to a weak contamination of the groundwater at
this point. For a same tritium activity (Fig. 4b), the nitrate concentration could nevertheless appear as variable (Pz7 and Pz5).
This offset may result from local differences in agricultural
pressure (total amounts of fertilizer, fertilizer type, organic inputs), which we did not take into account in our first simplistic
hypothesis of homogeneous loadings at the catchment scale.
The spring was close to the curve linking the oldest and youngest waters (Fig. 4b), confirming that it was indeed a mixture
of waters of different age and quality, as illustrated by the results from the monitoring of the piezometers (Figs. 3 and 4a).

Differences in the residence time of the water could merely

be due to differences between the piezometers of the unsaturated zone thicknesses if transfers were homogeneous within
the unsaturated zone at the catchment scale. However, this
was not the case because the water in Pz3 was older than
that in Pz2, which has a thicker unsaturated zone. The notion
of transfer rate in the unsaturated zone therefore also needs to
be examined in more detail.
The tritium profiles for the unsaturated zones of Pz2 and
Pz3 indicated different transfer processes in the unsaturated
zone above the water table of the two piezometers (Fig. 5).
The profile in Pz3 showed a single symmetrical and relatively
broad peak indicating that the main transfer mechanism was
convectionedispersion. Assuming that this peak corresponds
to 1963 (Smith et al., 1970), the average rate would be of
the order of 0.5 m per year, a velocity that explains why the

N. Baran et al. / Environmental Pollution 148 (2007) 729e738

735

Fig. 5. Tritium profiles obtained for the water recovered from the unsaturated
zone of Pz2, Pz3, Pz5 and Pz6.

Fig. 4. Nitrate vs. sodium (a) from the monthly analysis (March 2001 to
August 2005) of groundwater in the piezometers (Pz) and bi-monthly analysis
of the spring (SW: seawater dilution line) and (b) nitrate vs. tritium (measured
value and error) for piezometers and the spring (August 2005).

solutes currently being transferred between the soil and the

water table at Pz3 have yet to reach the groundwater. Pz2,
however, showed several narrower peaks of tritium, of which
the main ones were at 11 and 35 m depth (Fig. 5). This profile
suggested the existence of several infiltration rates, which
could be due to fractures in the limestone of the unsaturated
zone. The infiltration differences in the unsaturated zone
above Pz3 and Pz2, as well as the presence of a forested
area upstream of Pz3 (Fig. 1), could explain the low contamination of Pz3 vs. Pz2.
The tritium profile with several peaks at Pz6 (Fig. 5) again
suggested several infiltration rates due to fracturing; part of the
solutes applied at the surface could thus reach the groundwater
relatively rapidly and cause the relatively high pesticide and
nitrate concentrations that were recorded. Since the profile
for Pz5 is very short (12.4 m) compared to the three others
(38, 35.5 and 22.5 m), it was more difficult to pronounce on

the transfer mechanisms occurring in the unsaturated zone

above the water level of this piezometer. The four tritium profiles illustrated the heterogeneity of the water (and associated
solutes) infiltration rate within the unsaturated zone at the
catchment scale. Such heterogeneity could in part explain
the spatial variability of the contamination.
The groundwater tritium measurements at Pz4, suggesting
water older than 1963, confirmed another characteristic of
the hydrogeological system. Geophysical surveys have revealed the presence of faults in the vicinity of Pz4, which
may have led to its hydraulic isolation (unpublished data).
Coupled with a very slow infiltration in the unsaturated
zone, this hydraulic isolation would result in low contamination at this point. Unfortunately we cannot pronounce on the
hypothesis of slow transfer because no tritium profile was
measured in the unsaturated zone of Pz4. A second hypothesis
to explain the absence of contamination is that of groundwater
stratification resulting from clay enrichment in the Cuise
Sands with depth (Gutierrez, 2002). In addition, Pz4 is
screened only in the lower part of the Sands, which could explain why its sampled waters were old and contamination free.
The hypothesis of groundwater stratification was validated by
the results from the Pz17 well cluster tapping different groundwater levels. The measurements made on 10th May 2005
showed contamination diminishing with depth, both for nitrate

736

N. Baran et al. / Environmental Pollution 148 (2007) 729e738

(67.7, 64.4 and 0.5 mg L1 for Pz17A, Pz17B and Pz17C, respectively) and chloride (36.6, 32.9 and 5.6 mg L1). Tritium
activity also decreased with depth (21  3, 11  2 and
<1 TU) confirming that an older water (whose quality has
been less affected by human activities) exists at the base of
the aquifer. The results equally indicated limited mixing between the upper and deeper parts of the groundwater, the precise thickness of each of these two compartments being
unknown presently. The finding of groundwater stratification
in Pz4 and Pz17, two piezometers located in different parts
of the catchment, suggested this is not a localized phenomenon
but that it might extend over a large part, if not all, of the
groundwater body. This absence of mixing has an effect on
pesticide concentrations along a vertical profile within the saturated zone, as shown by Pz17A, Pz17B and Pz17C where At
and DEA concentrations decreased with depth (0.16, 0.06,
<0.025 mg L1 for At and 0.2, 0.14, <0.025 mg L1 for
DEA, respectively).

3.3. Statistical analysis of the organic and inorganic

parameters: hydrodynamic perspectives
To better understand the relationships between the piezometer and spring observations, principal component analysis
(PCA) was carried out using XLSTAT Software, with the
correlation tests using a threshold value of alpha 0.05;
only the meaningful correlations at this level are discussed below. Thirty-nine variables (At, DEA, nitrate, chloride concentrations, and piezometric level at each of the seven
piezometers corresponding to 5  7 variables, and At, DEA,
nitrate, chloride concentrations at the spring) are considered,
each with a series of data for the period April 2001 to October
2005 (40 individuals).
Atrazine concentration and piezometric level correlated
positively for Pz3, negatively for Pz5, Pz8 and Pz7, and
were not correlated for Pz2, Pz4 and Pz6. This variable link
between piezometric level and atrazine concentration possibly
resulted from a variable remobilization of atrazine. This coincides with atrazine inputs that were very heterogeneous spatially and temporally at the catchment scale (Fig. 1).
As regards piezometric levels, Pz4 correlated with no other
piezometer, supporting the hypothesis of its hydraulic isolation. Pz2 and Pz3, the most upstream in the catchment, correlated positively with all the piezometers except Pz4 and Pz7,
the most downstream piezometers. This absence of correlation
between the upstream and downstream piezometers reflected
the systems inertia, which increased with the thickness of
the unsaturated zone. With the climatic context showing a deficiency as of February 2003, the piezometric levels of Pz4 and
Pz7 stopped rising in MarcheApril 2003, whereas those of
Pz2 and Pz3 continued to rise until SeptembereOctober.
There was thus a major phase difference in piezometric evolution between the upstream and downstream parts of the system. The piezometric levels of Pz5, Pz6, Pz7 and Pz8, all
relatively central in the catchment, correlated positively with
one another.

Eight factors were needed to explain more than 80% of the

variance. This is a proof of the complexity of the catchments
hydrodynamic. Projection on the F1eF2 axes (explaining
about 47.4% of the variance e Fig. 6) showed a temporal
structuring on passing from the samples correlated negatively
with factors F1 and F2, through samples correlated positively
with these factors (up to the end of 2003), to end with the samples correlated negatively with factors 1 and 2. The two factors
could correspond to the contribution of two different reservoirs within the aquifer. It was worth noting that the piezometric levels of Pz2 and Pz3, and to a lesser extent Pz5 and Pz8,
showed a strong correlation (0.96, 0.96, 0.78 and 0.74, respectively) with factor 1, in fact the strongest correlation of these
four variables to any factor of the PCA. Factor 2 correlated
strongly with the piezometric levels of Pz7 and Pz6 (0.82
and 0.71, respectively). These observations highlight the importance of recharge on the groundwater quality at the catchment scale.
This conceptual model, considering a contribution of two
reservoirs within the aquifer with time-variable proportions,
is in agreement with the climatic trend, at first very wet and
then very dry as of 2003. The rainfall deficit was reflected
throughout 2003 by a staggered lowering of the piezometric
levels depending on the observation points location in the
catchment. All of the piezometers showed a drop in levels
but this was not synchronous everywhere, since it depended
on the position of the piezometer in the catchment The conceptual model also agrees with the tritium concentrations measured at the spring (5 TU on 27th April 2001, 11 TU on 18th
June 2002, 10 TU on 17th December 2003 and 9 TU on 8th
August 2005, all data corrected on 8th August 2005) which
seemed to indicate a contribution of a more recent water

Fig. 6. Principal component analysis representation (projection on planes 1 and

2) obtained for the 39 variables (see Section 3.3). Each label corresponds to
one of the 40 sampling dates (mm/yy).

N. Baran et al. / Environmental Pollution 148 (2007) 729e738

between the 2001 measurement and those of 2002 and 2003

and, with less certainty, a lessening of its contribution during
the last measurement. Projection onto the F1eF3 axes
(37.4% of the variance e data not shown) also showed a temporal organization in the sample series, which passed from
a negative correlation with F1 to a positive correlation, on
which were superimposed fluctuations linked variably to positive or negative correlations with F3. The contribution of F3
could correspond to a water fraction circulating more rapidly
and to the highly varied chemistry between samples. Rapid
circulation of water and solutes in the unsaturated zone would
be favoured by the fracturing revealed notably by the tritium
profiles and whose existence was also assumed from the rapid
detection of IPU after its applications in 2001.
4. Conclusions
The intensive monitoring of water quality over a long period (5.5 years) in a 3 km2 agricultural catchment (Brevilles)
has made it possible to demonstrate and quantify high spatial
and temporal variabilities in pesticide concentrations and more
particularly as regards atrazine (At) and deethylatrazine
(DEA). The variabilities remained high even though At has
not been used in this catchment for 6 years. Comparison of
the At and DEA time-series and those of the phenylureas
(IPU and CTU) showed very strong differences that were
not related to the quantities applied, but attributed to the fact
that the formation of bound residues is far greater for the latter
than the former.
The spatial variability of the At and DEA concentrations in
the groundwater could be explained only in part by the spatial
and temporal variability of the atrazine applications. The positive or negative correlations in some piezometers between At
or DEA concentrations and piezometric level highlighted the
importance of recharge on the temporal variability of the concentrations, recharge water being contaminated to various degrees by these molecules. However, as this type of correlation
was not found for all the piezometers, other factors must be
sought to explain the temporal variability in groundwater
quality.
The complexity of the systems hydrodynamic has been
shown through the analysis of results from geochemical parameters. The very high inertia of the unsaturated zone, demonstrated by tritium measurements and the absence of marked
anthropogenic effects at certain points of the catchment (Pz3
and Pz4), supported by the nitrate concentrations and sodium/chloride ratios, buffered the system and limited the climatic effects on the groundwater hydrodynamic. Despite this
inertia, it is possible that a proportion (unquantified) of the water, and transported solutes, circulated more rapidly in the
unsaturated zone along a set of fractures and probably also in
the least clayey parts of the saturated sand and in parts of the
saturated limestones.
The spatial distribution and temporal evolution of groundwater contamination were complex. The most important factors involved included the stratification of groundwater
quality, and the spatial and temporal variability of the risk of

737

pesticide leaching associated with, on one hand the different

uses and physical and chemical properties of the molecules
(degradation, sorption and formation of bound residues), and
on the other hand the various geochemical and hydrodynamic
properties of the solids in the soil, the vadose and the saturated
zones.
The Brevilles spring served as an integrating point for all
the processes existing at the catchment scale. The observations
made in the different piezometers enabled framing the observations made at the spring (water that is younger or older, water that is more or less contaminated). However, the question
arose as to the representativeness of observation points such
as piezometers or wells at the scale of the groundwater body.
Although factors that explain the differences in concentrations observed both spatially and over time have been identified, predicting the evolution of the quality of the
groundwater remains a challenge since the different factors
identified superimpose each other. Modelling could be a solution, providing that the way the catchment works is well understood and that additional data relative to certain processes
specific to pesticides (degradation and desorption kinetics,
for example) have been acquired.
The continuation of such integrated monitoring of groundwater quality coupled with the use of different tools (modelling,
isotope geochemistry) will enable a better understanding of the
hydrodynamic of the hydrogeological system and is a very
promising scientific approach. In addition, the response to the
requirements of the Water Framework Directive (where to sample, at what frequency, trend characterization, estimation of the
level of confidence and precision of the results provided by
monitoring networks, etc.) will also require this type of
approach on a detailed study scale.

Acknowledgements
The work was carried out initially under the European 5th
FRDP PEGASE project (contract EVK1-CT1999-00028), then
was supported by the European Union FP6 Integrated AquaTerra Project (project no. GOCE 505428) under the thematic
priority of Sustainable development, global change and ecosystems. Financial aid was also received through BRGM research projects (POLDIF and QUALHYS projects) and an
Agreement (012095) with the Seine-Normandy Water Agency.
The authors would like to thank the farmers and the municipality whose help greatly facilitated the work for this study
and Monsanto for supplying the analytical standards for acetochlor OA and ESA degradates.

References
Albrechtsen, H.-J., Clausen, L., Pedersen, P.G., 2003. Abstracts of the International Symposium Non-Agricultural Use of Pesticides e Environmental
Issues and Alternatives, Copenhagen, Denmark, May 7e9, 2003, pp. 43e44.
Alletto, L., Coquet, Y., Benoit, P., Bergheaud, V., 2006. Effects of temperature
and water content on degradation of isoproturon in three soil profiles. Chemosphere 64, 1053e1061.

738

N. Baran et al. / Environmental Pollution 148 (2007) 729e738

Baran, N., Mouvet, C., Dagnac, T., Jeannot, R., 2004. Infiltration of acetochlor
and two of its metabolites in two contrasting soils. Journal of Environmental Quality 33, 241e249.
Barbash, J.E., Thelin, G.P., Kolpin, D.W., Gilliom, R.J., 2001. Major herbicides in ground water: results from the national water-quality assessment.
Journal of Environmental Quality 30, 831e845.
Boivin, A., Cherrier, R., Schiavon, M., 2005. A comparison of five pesticides
adsorption and desorption processes in thirteen contrasting field soils. Chemosphere 61, 668e676.
Clark, I., Fritz, P., 1997. Environmental Isotopes in Hydrogeology. Lewis Publishers, New York.
De Guzman, N.P., Hendley, P., Gustafson, D.I., van Wesenbeeck, I., Klein, A.J.,
Fuhrman, J.D., Travis, K., Simmons, N.D., Teskey, W.E., Durham, R.B.,
2005. The acetochlor registration partnership state ground water monitoring
program. Journal of Environmental Quality 34, 793e803.
EEA (European Environment Agency), 1999. Groundwater Quality and Quantity in Europe. Technical Report. EEA, Copenhagen, 112 pp.
EPA (Environmental Protection Agency), 1995. Determination of Organic
Compound in Drinking Water by LiquideSolid Extraction and Capillary
Column Gas Chromatography/Mass Spectrometry. Method 525.2. EPA,
60 pp.
Gilliom, R.J., Barbash, J.E., Crawford, C.G., Hamilton, P.A., Martin, J.D.,
Nakagaki, N., Nowell, L.H., Scott, J.C., Stackelberg, P.E., Thelin, G.P.,
Wolock, D.M., 2006. The Quality of Our Nations Waters e Pesticides
in the Nations Streams and Ground Water, 1992e2001. U.S. Geological
Survey Circular 1291. USGS, 172 pp.
Gutierrez, A., 2002. Realisation de pompages dessais dans les sables de Cuise
du bassin de la source des Brevilles, Montreuil-sur-Epte (Val-dOise). Rapport de fin de travaux. Rapport BRGM RP-51534-FR, 37 pp.
IFEN (Institut Francais de lEnvironnement), 2004. Les pesticides dans les
eaux e Sixie`me bilan annuel e donnees 2002. Collection Etudes et travaux, n 42. IFEN, Orleans, ISBN 2-911089-70-7, 32 pp. (Detailed results
on CD-Rom).
ISO (International Organization for Standardization), 1995. Guide 99998,
Guide to the Expression of Uncertainty in Measurement (GUM). ISO,
Geneva.
Kolpin, D.W., Goolsby, D.A.A., Thurman, E.M., 1995. Pesticides in near-surface aquifers: an assessment using highly sensitive analytical methods and
tritium. Journal of Environmental Quality 24, 1125e1132.
Kolpin, D.W., Sneck-Fahrer, D., Hallberg, G.R., Libra, R.D., 1997. Temporal
trends of slected agricultural chemicals in Iowas groundwater, 1982e
1995: are things getting better? Journal of Environmental Quality 26,
1007e1017.
Kolpin, D.W., Thurman, E.M., Linhart, S.M., 2000. Finding minimal herbicide
concentrations in ground water? Try looking for their degradates. Science
of the Total Environment 248, 115e122.
Kolpin, D.W., Schnoebelen, D.J., Thurman, E.M., 2004. Degradates provide
insight to spatial and temporal trends of herbicides in ground water.
Ground Water 4, 601e608.
Lapworth, D.J., Gooddy, D.C., 2006. Source and persistence of pesticides in
a semi-confined chalk aquifer of southeast England. Environmental Pollution 144, 1031e1044.
Lapworth, D.J., Gooddy, D.C., Stuart, M.E., Chilton, P.J., Cachandt, G.,
Knapp, M., Bishop, S., 2006. Pesticides in groundwater: some observations
on temporal and spatial trends. Water and Environment Journal 20, 55e64.
Mills, P.C., Kolpin, D.W., Scribner, E.A., Thurman, E.M., June 2005. Herbicides and degradates in shallow aquifers of Illinois: spatial and temporal
trends. Journal of American Water Resources Association, 537e547.

Mordaunt, C.J., Gevao, B., Jones, K.C., Semple, K.T., 2005. Formation of nonextractable pesticide residues: observations on compound differences,
measurement and regulatory issues. Environmental Pollution 133, 25e34.
Morvan, X., Mouvet, C., Baran, N., Gutierrez, A., 2006. Pesticides in the
groundwater of a spring draining a sandy aquifer: temporal variability of
concentrations and fluxes. Journal of Contaminant Hydrology 87,
176e190.
Negrel, P., Pauwels, H., 2004. Interaction between different groundwaters in
Brittany catchments (France): characterizing multiple sources through
strontium- and sulphur isotope tracing. Water, Air, and Soil Pollution
151, 261e285.
Normand, M., Gravier, A., 2005. Mise en uvre de la DCE e Premie`res synthe`ses des caracteristiques principales et secondaires des masses deau souterraine et de lanalyse du Risque de Non Atteinte du Bon Etat
environnemental en 2015. Pistes de reflexion pour la caracterisation plus
detaillee. BRGM/RP-53924-FR, 105 pp., 34 fig., 41 tabl., 4 ann.
Official Journal of the European Communities, 2000. Directive 2000/60/EC of
the European Parliament and of the Council of 23 October 2000 Establishing a Framework for Community Action in the Field of Water Policy,
72 pp.
Official Journal of the European Communities, 2001. Decision no. 2455/2001/
EC of the European Parliament and of the Council of 20 November 2001
Establishing the List of Priority Substances in the Field of Water Policy
and Amending Directive 2000/60/EC, 5 pp.
OFEFP (Office Federal de lEnvironnement, des Forets et du Paysage) and
OFEG (Office Federal des Eaux et de la Geologie), 2004. NAQUA e
Groundwater Quality in Switzerland 2002/2003. OFEFP, OFEG, Berne,
204 pp. (in French).
Pauwels, H., Talbo, H., 2004. Nitrate concentration in wetlands: assessing the
contribution of deeper groundwater from anions. Water Research 38,
1019e1025.
Perrin-Ganier, C., Breuzin, C., Portal, J.M., Schiavon, M., 1996. Availability
and persistence of isoproturon under field and laboratory conditions. Ecotoxicology and Environmental Safety 35, 226e230.
Pieuchot, M., Perrin-Ganier, C., Portal, J.M., Schiavon, M., 1996. Study of the
mineralization and degradation of isoproturon in three soils. Chemosphere
33, 467e478.
Pucarevic, M., Sovljanski, R., Marjanovic, N., 2002. Atrazine in Ground Water
of Vojvodina Province. Water Research 36, 5120e5126.
Rowden, R.D., Liu, H., Libra, R.D., 2001. Results from the Big Spring basin
water quality monitoring and demonstration projects, IOWA, USA. Hydrogeology Journal 9, 487e497.
Smith, D.B., Wearn, P.L., Richards, H.J., Rowe, P.C., 1970. Water movement
in the undersaturated zone of high and low permeability strata by measuring natural tritium. In: Isotope Hydrology, IAEA Proceedings of a Symposium, Vienna, 9e13 March, pp. 73e87.
Sorensen, S., Aamand, J., 2001. Biodegradation of the phenylurea herbicide
isoproturon and its metabolites in agricultural soils. Biodegradation 12,
69e77.
Trojan, M.D., Maloney, J.S., Stockinger, J.M., Eid, E.P., Lahtinen, M.J., 2003.
Effects of land use on ground water quality in the Anoka Sand Plain Aquifer of Minnesota. Ground Water 41, 482e492.
USGS (US Geological Survey), 2006e. National Water-Quality Assessment
(NAWQA). Available from: <http://water.usgs.gov/nawqa/protocols/
doc_list.html>.
Walker, A., Rodriguez-Cruz, M.S., Mitchell, M.J., 2005. Influence of ageing of
residues on the availability of herbicides for leaching. Environmental Pollution 133, 43e51.