Effects of pesticides on community structure and ecosystem
functions in agricultural streams of three biogeographical
regions in Europe
Ralf Bernhard Schäfera,b,⁎, Thierry Caquetc, Katri Siimesd, Ralf Muellere,
Laurent Lagadicc, Matthias Liessa
aUFZ - Helmholtz Centre for Environmental Research, Dept. System Ecotoxicology, Permoser Straße 15, 04318 Leipzig, Germany
bUniversity Lüneburg, Institute for Ecology and Environmental Chemistry, Scharnhorststraße 1, 21335 Lüneburg, Germany
cUMR 985 INRA-Agrocampus, Ecobiologie et Qualite des Hydrosystemes Continentaux (EQHC),
Equipe Ecotoxicologie et Qualite des Milieux Aquatiques, 65 rue de Saint Brieuc, 35042 Rennes, France
dFinnish Environment Institute (SYKE), Research Programme for Contaminants, Mechelininkatu 34a, 00241 Helsinki, Finland
eEWE AG, Department UT, Laboratory for Environmental Analytics, Bürgerparkstraße 11, 49661 Cloppenburg, Germany
Received 26 February 2007; received in revised form 17 April 2007; accepted 30 April 2007
There is a paucity of large-scale field investigations on the effects of organic toxicants on stream macroinvertebrate community
structure and ecosystem functions. We investigated a total of 29 streams in two study areas of France and Finland for pesticide
exposure, invertebrates and leaf-litter breakdown. To link pesticide exposure and community composition we applied the trait-based
Species At Risk (SPEAR) indicator system.
In the French region, pesticide stress was associated with a decrease in the relative abundance and number of sensitive species in
effects of pesticides were identified by a 2.5-fold reduction of the leaf-litter breakdown rate that was closely correlated with the
structural changes in the contaminated streams. No effects of pesticides were observed in Finnish streams since contamination with
pesticides was very low.
biogeographical regions, also including results of a previous field study in North Germany. Furthermore, change of the community
structure was detectable at a concentration range as low as 1/100 to 1/1000 the acute 48 h-LC50 of Daphnia magna.
Our findings demonstrate that pesticides may influence the structure and function of lotic ecosystems and that the SPEAR
approach can be used as a powerful tool in biomonitoring over large spatial scales.
© 2007 Elsevier B.V. All rights reserved.
Keywords: Macroinvertebrates; Aquatic; Pesticides; Recovery; Leaf-litter breakdown; Species traits
Pesticides represent a relevant stressor for many
aquatic and terrestrial species (Liess et al., 2005b). They
have been shown to potentially affect all groups of
Science of the Total Environment 382 (2007) 272–285
⁎Corresponding author. UFZ - Helmholtz Centre for Environmental
Research, Dept. System Ecotoxicology, Permoser Straße 15, 04318
Leipzig, Germany. Tel.: +49 341 235 2120; fax: +49 341 235 2401.
E-mail address: Ralf.Schaefer@ufz.de (R.B. Schäfer).
0048-9697/$ - see front matter © 2007 Elsevier B.V. All rights reserved.
organisms in aquatic ecosystems: e.g. microorganisms
(DeLorenzo et al., 2001), invertebrates (Castillo et al.,
1994). Although some field studies demonstrated effects
ofheavymetalsontheaquatic communitystructure atthe
regionalscale,there isa paucityofsuchinvestigationsfor
organic toxicants, encompassing more than one stream or
river (Clements et al., 2000; Beasley and Kneale, 2003;
pesticides on important stream ecosystem functions such
as leaf-litter breakdown (Wallace et al., 1997) are still
The establishment of a causal relationship between a
as every sampling site exhibits a unique combination of
environmental variables and species (Liess et al., 2005b).
In addition, confounding factors like the occurrence of
other anthropogenic or natural stressors can mask the
effects of a particular stressor. The use of species traits,
such as generation time or dispersal capacity, represents
an interesting approach towards encompassing both
natural variability and confounding factors (Statzner
affect only certain trait modalities, called response traits,
trait-based approaches may be used to identify the effects
of a specific stressor e.g. pesticides. At the community
level, the relative abundance or number of species with
certain trait modalities would probably decrease thus
making it possible to interpret and/or predict community
Ohe (2005a) developed a trait-based concept with which
to distinguish pesticide effects on freshwater macro-
invertebrates from the influence of other environmental
This concept, called Species At Risk (SPEAR),
classifies macroinvertebrates according to their vulnera-
bility towards pesticides into sensitive species (SPEAR)
ecological and physiological traits. The authors success-
fully employed this approach in a field study on 20
streams in North Germany, where the relative abundance
of SPEAR in a community declined with increasing
pesticide stress. Furthermore, pesticide stress was the
most important explanatory variable for different com-
munity-based SPEAR endpoints (Liess and von der Ohe,
2005a). In another study, Schriever et al. (2007)
demonstrated that the highest correlation between the
fraction of sensitive species and various environmental
parameters was obtained for a modelled indicator of
pesticide exposure, called runoff potential.
use of the SPEAR concept in biomonitoring may be
extended beyond North Germany to different biogeo-
graphical regions (Illies, 1978) and (2) pesticides have
effects not only on the structure but also on the
functioning of aquatic ecosystems. Therefore, we con-
ducted field investigations in two regions of France and
Finland in which the macroinvertebrate communities,
leaf-litter breakdown, pesticides and physico-chemical
characteristics of 29 streams were monitored during the
period of pesticide application. Considering the differ-
between these countries, we were also able to examine
whether the effects of pesticides on non-target organisms
are dependent on usage patterns or if the invertebrate
the first study that comparatively investigates pesticide
effects in different biogeographical regions.
To further evaluate the performance of the SPEAR
approach in large-scale biomonitoring we analyzed its
ability to discriminate reference and contaminated sites
across different biogeographical regions, also including
the sites of the previous field study in North Germany
(Liess and von der Ohe, 2005a).
2.1. Study area and sampling schedule
France and Finland were selected as study countries
because they belong to different biogeographical regions
(Illies, 1978) and exhibit contrasting pesticide use with an
average of approximately 6 and 0.8 kg annually applied
active ingredient per hectare, respectively (EUROSTAT,
2002). This difference partly stems from the lower
prevalence of pests in Finland since the northward
dispersal of many pests is averted by the colder climate.
In France, Brittany in the northwest was chosen as study
area because the local authorities reported frequent
regional and temporal contamination of streams with
sampling sites in first- to third-order lowland streams
(Strahler, 1957) were selected which were expected to
exhibit a gradient in pesticide contamination based upon
Agency for Agriculture and Forestry (DRAF) Bretagne,
personal communication). Since Finnish agriculture is
mainly localized in the southern part of the country, this
region was chosen for the specification of sampling sites.
13 sites were selected in first- to third-order lowland
streams covering different areas of South Finland.
All streams in the two regions of France and Finland
were selected to match the physical properties of those
sampled during a previous field study in North Germany
273R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
(Liess and von der Ohe, 2005a): no drying up in summer;
no dredging in the present or past year; presence of
adjacent fields with vegetable, corn or oil-seed crops;
average stream current velocity ranging between 0.1 and
0.5 m/s; maximum stream depth of 0.8 m. Furthermore,
the sites were checked in field survey and with maps
(France: IGN 1:25,000 maps, Finland: Maanmittauslaitos
1:50,000 maps) to have no waste-water treatment plants,
industrial facilities or mining drainage upstream. Thus,
pollution other than from agricultural sources was
unlikely. The location of all sampling sites is displayed
in Fig. 1.
The sites were sampled before (14–19 April 2005 in
France, 3–9 July 2005 in Finland) and during (19–26
May 2005 in France, 1–6 August 2005 in Finland) the
estimated period of maximum pesticide contamination,
according to the monitoring data from local authorities
Institute (SYKE), personal communication). However,
the timing of pesticide application varies and the sites
may therefore have received pesticide input before the
initiation of sampling. This holds especially for France,
sampling date, possibly leading to pesticide runoff.
If stated below, we also included in the analysis the
results of the study in the German region for April and
May, averaged for the 3 years of study (Liess and von
der Ohe, 2005a). We are aware that the results of non-
randomly chosen, single regions cannot be extrapolated
to the country level. However, for the ease of reading we
refer to the respective region with the countries' name
throughout this paper.
2.2. Physico-chemical and geographical parameters
Concentrations of oxygen, ammonium, nitrite, nitrate
and orthophosphate in the stream water as well as tem-
perature, pH and stream current velocity were measured
Fig. 1. Map of sampling sites and large rivers in Finland (a), France (b) and Germany (c). Sampling streams are not displayed due to scale. Regional
maps were created using ESRI World Basemap Data and the European map was created with R (packages: maps and mapdata).
274 R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
Descriptive statistics of environmental parameters at the study sites in France, Finland and Germany
Mean SD% SD Min.Max.Mean SD % SDMin.Max. MeanSD % SDMin.Max.
Boulder (N20 cm)
Cobble (5–20 cm)
Gravel (1–5 cm)e
Grit (0.1–1 cm)
Sand (0.01–0.1 cm)
Clay and silt (b0.01)f, e
Buffer strip widthf
aMeasured twice in 2005.
bMeasured between 1998 and 2000 and taken from the field study of Liess and von der Ohe (2005a).
cIntercorrelation in France (Spearman's rhoN0.8).
dIntercorrelation in France (Spearman's rho=−0.836).
eIntercorrelation in Finland (Spearman's rho=−0.802).
fIntercorrelation in France (Spearman's rho=−0.817).
R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
as described in Liess and von der Ohe (2005a). Water
conductivity was determined on site with a Multi 340i
device of WTW (Weilheim, Germany). Total water
hardness was measured in the field with an Aquamerck
test (precision: 1 °dH; Merck, Darmstadt, Germany).
Suspended matter was collected in suspended-matter
samplers (Liess et al., 1996), measured biweekly and
converted into a volume (ml) per week. In-stream struc-
ture, depth, width, tailing and buffer strip width were
investigated in a 50-m reach upstream and downstream
from the sampling site.
Previous studies demonstrated that the presence of
agricultural activities positively influenced downstream
habitat quality and partly compensated for the effects of
pesticides (Liess and von der Ohe, 2005a; Schriever et al.,
2007). Therefore we inspected the French and Finnish
streams upstream of each sampling site in field survey or
with maps (France: IGN 1:25,000 maps, Finland:
Maanmittauslaitos 1:50,000 maps) for the presence of
riparian forests. If double-sided riparian forests at least
upstream reach. Modification of these criteria such as
different upstream distances(2 or 4 km) orthe presence of
appreciable effects on the results of the present study. An
overview of the stream characteristics, including the sites
previously investigated in Germany (Liess and von der
Ohe, 2005a) is given in Table 1.
2.3. Pesticide monitoring and chemical analysis
The substances for the screening programs in France
and Finland were selected by (1) identifying potential
compounds based upon the analysis of previous regula-
tory monitoring programs (France: DRAF Bretagne,
Finland: Finnish Environment Institute (SYKE),personal
communication) and (2) ranking them according to their
toxicity, indicated by the 48-h acute median lethal
concentration (LC50) for Daphnia magna as given in
Tomlin (2001). The 10 most toxic pesticides included in
the respective screening program were mainly non-polar
(log KowN4) for Finland and polar to semi-polar (log
arranged to catch runoff-induced exposure because this is
a major input path for pesticides in small streams
(Neumann et al., 2002). They were locally adapted due
to differences in polarity and expected concentration
levels of the compounds.
Characteristics and measurement results of pesticides in French and Finnish streams
Organic phosphorous acid
Organic thiophosphorous acid
Organic thiophosphorous acid
aH = Herbicide, F = Fungicide and I = Insecticide.
bTaken from Tomlin (2001).
cn.d. = not detected.
dTime-weighted average concentration.
276R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
1999b) were deployed and retrieved after heavy rain
events (N10 mm precipitation per day). The water
samples were subsequently solid-phase-extracted using
6 ml Chromabond HR-P columns (Macherey-Nagel,
Düren, Germany). Analytes trapped on the columns were
extracted with 10 ml of 1:1 acetonitrile–ethylacetate and
the extract gently evaporated under nitrogen to 0.3 mL.
Residue analysis was conducted on an Agilent 6890N
(Agilent Technologies Germany, Boeblingen, Germany)
gas chromatograph (GC) linked to an Agilent 5973 mass
selective detector (MSD).
In Finland, continuous water passive sampling was
performed with low-density polyethylene (LDPE) strips
(Booij et al., 2003), which were deployed in each stream
at the beginning of the study and exposed for 28 days.
LDPE strips were extracted by soaking in 500 ml n-
hexane for 48 h. The extract was gently evaporated to
0.3 ml under nitrogen. Residue analysis was performed
on an Agilent 6890N GC linked to a Pegasus III time-of-
flight (TOF) mass spectrometer (Leco, Mönchenglad-
bach, Germany). Time-weighted average (TWA) water
concentrations for the LDPE samplers were calculated
according to distribution coefficients from Booij et al.
(2003). TWA water concentrations were converted to
peak water concentrations by multiplying the TWA
concentrations by a factor of 10 (Schäfer, R.B., Paschke,
A. and Liess, M., unpublished data).
Although the sampling methods differed, we think
that the outcomes are comparable as the results of
passive sampling and runoff-triggered water sampling
correlated very high (Pearson's r=0.995) in another
study on the French streams (Schäfer et al., in
2.4. Calculation of toxicity levels
To compare the toxicity associated with the pesticide
concentrations measured in the different sites, toxic units
(TU) were computed from the peak water concentrations
determined for each site (Liess and von der Ohe, 2005a):
where TU(D. magna)is the maximum toxic unit of the n
pesticides detected at the considered site, Ci is the
concentration (μg/L) of pesticide i and LC50iis the 48 h-
LC50 of pesticide i for D. magna (μg/L) as given in
Tomlin (2001). Although peak water concentrations may
have been underrated due to a delayed response of the
be a conservative measure of pesticide toxicity since (1)
pesticide concentrations usually decrease strongly within
24 h during runoff but the 48 h-LC50 of D. magna was
used for toxicity assessment (Richards and Baker, 1993)
and (2) only the maximum toxic unit was considered
respective site. If no pesticide was found a TU-value of
−5 was assigned to that site, corresponding to the value
found for unpolluted streams in a previous study (Liess
and von der Ohe, 2005a).
2.5. Macroinvertebrate sampling
Four replicate samples (surface ca. 0.12 m2per
sample) of different substrates were taken on each
sampling date with a 500-μm mesh-size Surber Sampler
(Hydro-Bios, Kiel, Germany) and preserved with forma-
lin (ca. 4% vol.). The invertebrates were sorted out,
counted and identified to the lowest possible taxonomic
level, which was genus for most taxa. A list of the taxa
the samples is given in the Supplementary material.
2.6. SPEAR-index calculation and endpoints
The identified taxa were classified into SPEAR and
SPEnotAR according to ecological and physiological
traits as described in Liess and von der Ohe (2005a).
Since life-cycle traits such as emergence time and
voltinism are dependent on the biogeographical region
(e.g. Central and Northern Europe), the classification for
a particular taxon may differ between regions. The
available data and region-dependent classification
information are compiled in a database which is publicly
available and comprises about 1000 macroinvertebrate
taxa (Liess et al., 2006). After classification into SPEAR
and SPEnotAR the relative abundance of taxa which are
potentially sensitive towards pesticides in a community,
was computed for each site and date:
logðxiþ 1Þd y
i andy is: 1 if taxoni isclassified as SPEAR, otherwise 0.
endpoint %SPEAR(PM abundance), where classification of
taxa relies only on physiological sensitivity (P) and
migration ability (M) to exclude biogeographical bias of
this index. %SPEAR(PM abundance)was used to examine
the applicability of the SPEAR concept across different
regions. The endpoint %SPEAR(number), which indicates
277R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
the relative number of sensitive taxa, was computed for
each site and date by:
Finally, the endpoint SPEAR(abundance during/before)was
computed by dividing the %SPEAR(abundance)during the
period of maximum pesticide input in streams (France:
May, Finland: August) by %SPEAR(abundance)before this
period (France: April, Finland: July). Time periods were
estimated on the basis of information from local
authorities and as reported elsewhere (Liess et al.,
1999b; Liess and von der Ohe, 2005a).
2.7. Leaf-litter breakdown
Three grams of air-dried Alnus glutinosa leaves at
abscission was anchored to the streams in coarse (mesh
size: app. 6 mm; polyethylene bag size: 20×20 cm) and
fine (mesh size: 50 μm; nylon cylinder size: 15 cm length,
7.5 cm diameter) enclosures. Leaves in the coarse bags
were accessible to invertebrates whereas leaves in the fine
bags were not and served as control for microbial
degradation and leaching (Gessner and Chauvet, 2002).
Triplicatecoarse andfine bagswere deployedafterthefirst
sampling for approximately 21 days in 12 randomly
selected sites in France and in 8 sites in Finland. To correct
for handling losses three coarse and fine bags were treated
the same way as the others but returned immediately to the
laboratory after a brief deployment in the stream. After
dried to a constant mass at 60 °C (24 to 48 h), reweighed
and averaged for each type of bags for every station. The
remaining leaf massWt(s) in grams for stations after timet
was obtained by summing up the handling-loss corrected
weight of the remaining litter at site s from the coarse
enclosures and the loss due to microbial degradation and
leaching at site s derived from the fine enclosures (for
details see Benfield (1996)). The exponential leaf break-
down rate kswas computed by:
as defined above) (Benfield, 1996).
2.8. Data analysis
The data analysis was performed separately for the
Finnish and French sites if not otherwise indicated. Prior
to analysis, the average values for the two sampling dates
werecalculatedfor all variables thatweremeasuredtwice
at each site, in order to avoid temporal pseudoreplication.
However, we also conducted the analyses for each single
measurement date. The results broadly confirmed the
findings of the analyses performed for the combined data
set and they are therefore not shown. Hierarchical cluster
as similarity measure was performed to check for
collinearity and redundancy among environmental vari-
ables (McGarigal et al., 2000). Environmental variables
that exhibited strong correlation with other variables
(Spearman's rhoN0.8; see Table 1) but were implausible
an explanation of differences in macroinvertebrate
assemblages, on the basis of common ecological
knowledge (Allan, 1995), were removed from the data
set to avoid misspecification of the linear model (Flack
and Chang, 1987; McGarigal et al., 2000).
Multiple linear regression was applied to identify the
We weighted the sites in the regression according to the
total log (x+1) abundance of species (only for SPEAR
metrics). We employed manual model building, defining
models on the basis of expert judgement and automatic
variable included) or the full model (all explanatory
variables included). The statistical procedure was back-
ward and forward entering of variables with Akaike's
InformationCriterionas stepwise modelselectioncriterion
t-test for significance of single variables and analysis of
variance (ANOVA) with F-test for significance of the
complete model. Models with different numbers of
parameters were compared with the F-test. Goodness of
fit was assessed with the adjusted r2(r2for models with
only one explanatory variable). Analysis of covariance
(ANCOVA) witht-testwas appliedtocheckfor significant
differences in slope or intercept for factors in regression.
Model checking included: heteroscedasticity, normal
distribution of residuals and influence of single observa-
applied hierarchical partitioning to determine the relative
importance of independent explanatory variables in the
linear models (Chevan and Sutherland, 1991).
To detect effects of pesticide input on SPEAR
metrics or k in a single country, the respective values
were split according to the TU(D. magna) into sites
potentially receiving (TU(D. magna)N−3.5) and poten-
tially not receiving (TU(D. magna)b−3.5) pesticide input.
Welch's t-test for unequal variances was used to
compare the means of the two groups.
278R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
the respective observations of France, Finland and
Germany were grouped according to their TU(D. magna)as:
– reference (TU(D. magna)b−3.5)
– lightly contaminated (−3.5bTU(D. magna)b−2.25)
– and heavily contaminated sites (TU(D.
The class boundaries (−3.5 and −2.25) were chosen
to make the sample sizes as even as possible for all
countries. However, the use of different class boundaries
(−4 and −2) yielded the same results. To detect
significant differences between means of groups, non-
parametric ANOVA with Kruskal–Wallis test was
conducted, followed by a non-parametric multiple
comparison test of the Behrens–Fischer type (Munzel
All statistical computations and graphics were
created with the open source software package R (R
Development Core Team, 2006) using version 2.4.0 (for
Mac OS X, 10.4.8) with appropriate additional packages
(hier.part, Hmisc, npmc, maps and mapdata).
3.1. Characterization of investigated streams and
Water temperature, pH, hardness and oxygen exhib-
ited only slight variation (up to 34% standard deviation)
among the sampling sites in each study area, while
streambed substrate composition showed the largest
variability (up to 400% standard deviation) (Table 1).
The French and German streams were very similar
concerning most stream characteristics but differed in
the clay and silt content of the substrate. In contrast, the
French and Finnish sites were very similar regarding
substrate composition but mainly differed in water
chemistry characteristics. A total of 94 different taxa
with an average of 27 taxa per stream were identified in
the 13 Finnish streams; the values for the 16 French
streams were 110 and 33, respectively (see Supplemen-
tary material). In the study of Liess and von der Ohe
(2005a) 129 different taxa and an average of 24 taxa per
sampling site were found in the German streams,
applying the same level of taxonomic identification as
in the present study.
In Finland, Asellus aquaticus, Chironomidae spp.,
Dryopoidea spp., Leuctra fusca, Limoniidae spp.,
Oligochaeta spp. and Simuliidae spp. were found in
more than 85% of the samples. The most common taxa
(N85% of samples) in France were Baetis rhodani,
Elmidae spp., Ephemerella ignita, Erpobdella spp.,
Gammarus pulex, Hydropsyche spp. and Oligochaeta
spp.. In the German study, Chironomidae spp., Erpob-
della octoculata, Glossiphonia complanata, Tubificidae
spp., Gammarus pulex and Limnephilus lunatus were
present in more than 85% of the samples.
3.2. Pesticide monitoring
In Finland, only the fungicide trifluraline was
detected; it had a maximum TWA water concentration
of 1.11 ng/L (Table 2), but resulted in a neglectable
TU(D. magna)for the Finnish streams (Fig. 2). In contrast,
all pesticides of the monitoring program were detected
in samples from French streams (Table 2). They were
identified and quantified after the only strong rain event
(20 to 30 mm rainfall between May 12 and 13, 2005
(Meteo France, 2006)) that occurred in the study area
during the sampling period. For the French sampling
sites pesticide concentrations with TU(D. magna)values
up to −0.42 were observed (Fig. 2). For the German
streams TU(D. magna)values up to −0.71 were reported
(Liess and von der Ohe, 2005a).
3.3. Relationship between environmental variables and
For the Finnish streams, no significant linear model
could be established for %SPEAR(abundance), and stream
depth was the only variable to explain %SPEAR(number)
was significantly different for streams with and without
Fig. 2. Box–Whisker plot of Toxic Unit(Daphnia magna)for the sites in
the study areas of Finland (FIN), France (FR) and Germany (GR).
279 R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
undisturbed upstream reaches (Welch's t-test, P=0.439
For the French streams,variabilityin%SPEAR(abundance)
was best explained by TU(D. magna)and the factor undis-
turbed upstream reach (Table 3). The negative relation-
ship between %SPEAR(abundance) and TU(D. magna) is
illustrated in Fig. 3.
Values for %SPEAR(abundance)and %SPEAR(number)
were significantly reduced for streams which received
pesticide input (Welch's t-test, both Pb0.001). Pesti-
cide-impacted streams with undisturbed upstream
reaches had significantly higher %SPEAR(abundance)
and %SPEAR(number) compared to impacted streams
which lacked these reaches (Welch's t-test, P=0.017
3.4. Relationship between leaf-litter decomposition and
The leaf-litter breakdown coefficient k for the Finnish
streams ranged from 0.001 to 0.046 and k was highly
The breakdown rate was significantly different between
Fig. 3. Relation between the benthic invertebrate community structure
regression lines are significant with Pb0.001, r2=0.61 and 0.64 for
French (n=10) and German streams (n=11), respectively. The slopes
and intercept are not significantly different (analysis of covariance,
Summary statistics of linear models to explain SPEAR endpoints and leaf-litter breakdown rate k in French and Finnish streams
Relative importance of explanatory variable in best-fit linear model (%)a
% of filamentous
% of sand
Not explanatory variable
aDetermined in hierarchical partitioning (Chevan and Sutherland, 1991).
br2not adjusted for one explanatory variable.
cNot best-fit model.
280R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
upstream reach could explain a significant part of the
variation in k (Table 3), where k responded positively to
the presence of undisturbed upstream reaches and
negatively to an increase of pesticide stress. However,
the best-fit model for k only comprised the variables %
SPEAR(abundance)and % of sand on the stream bottom
(Table 3). This indicated that pesticide stress had no
direct effect on the leaf-litter breakdown but mediated
through its negative effect on sensitive species. The
relationship between %SPEAR(abundance) and k is
displayed in Fig. 4.
3.5. Comparison of %SPEAR(PM
For Germany, France and Finland, the means of %
SPEAR(PM abundance), a metric excluding traits with
biogeographical variability, were significantly different
when grouped by TU(D. magna)(Kruskal–Wallis test,
=26.32, Pb0.001). All the reference sites in the
three countries (TUb−3.5) exhibited about the same
mean level of %SPEAR(PM abundance)with 46, 52 and
54% for France, Finland and Germany, respectively.
Pairwise comparisons showed that values from refer-
ence sites were significantly different (Pb0.05) from the
values recorded in the highly contaminated sites in
France and Germany (TUN−2.25) (Fig. 5). For all
countries, neither the reference sites nor the highly
contaminated sites were significantly different (Pb0.05)
from the mean %SPEAR(PM abundance)of lightly polluted
streams (−3.5bTUb−2.25). Nevertheless, a clear
decline of the %SPEAR(PM abundance)was visible for
the group of lightly polluted sites compared to reference
sites. Furthermore, when the data from all countries
were pooled, the difference between lightly polluted and
reference sites showed to be significant in multiple
%SPEAR(PM abundance)was lower for France than for
Germany concerning the groups of lightly and highly
contaminated streams (Fig. 5), because of the higher
TU(D. magna)values of the French sites in each group. No
difference was observed in regression analysis (not
shown, but compare Fig. 3).
3.6. Temporal change of %SPEAR(abundance)
For Finland, the average change of the communities
from July to August, SPEAR(abundance during/before), for all
streams was 1±0.06 standard error (s.e.). Hence, no
change in the community structure occurred in the period
of pesticide application. Similarly, French streams that
were not subject to pesticide input (TU(D. magna)b−3.5)
had an average SPEAR(abundance during/before)of 1.01±
0.08 s.e., while stations which received pesticide inputs
(TU(D. magna)N−3.5) had a mean SPEAR(abundance during/before)
of 0.92±0.06 s.e.. However, SPEAR(abundance during/before)
values for the two groups were not significantly different
(Welch's t-test, P=0.215).
The average values of SPEAR(abundance during/before)
for the German streams were 0.92±0.06 s.e. and 0.74±
0.08 s.e. for streams potentially not receiving and
receiving pesticide input, respectively. These values
were significantly different at P=0.048 (Welch's t-test).
The pooled data for community change in the period
of pesticide application (SPEAR(abundance during/before)) of
all regions were significantly different concerning the
Fig. 4. Relation between leaf-litter decomposition and %SPEAR(abundance)
for 11 streams in Brittany, France. Linear regression was significant with
Pb0.001 and r2=0.89.
Fig. 5. Relation between %SPEAR(PM abundance)and Toxic Unit(Daphnia
magna)(TU) for the study sites in France, Finland and Germany with the
Pb0.05). Error bars show standard error.
281 R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
grouping by TU(D.
=6.94, Pb0.031). In multiple comparison tests,
the medium (P=0.032) and highly contaminated sites
(P=0.025) differed significantly from the reference
sites, indicating an acute response of %SPEAR(abundance)
from pre- to pesticide-application period.
magna) (Kruskal–Wallis test,
4.1. Linking pesticide input to community composition
a region of each France and Finland to examine whether
current-use pesticides would affect freshwater macro-
invertebrate communities. Pesticide stress in terms of
TU(D. magna)was almost absent in the Finnish streams, and
the variation in the lotic macroinvertebrate community
could not be attributed to the presence of these con-
agreement with those of a Finnish governmental screening
any field study on Finnish streams reporting pesticide
concentrations at a level potentially toxic to invertebrates.
French may be attributed to the reduced pesticide-
factors, e.g. higher organic carbon content of the Finnish
soils (European Commission Joint Research Centre,
Institute for Environment and Sustainability, personal
communication). We suppose that the results presented
here are representative for the general situation in the
investigation including more compounds, sampling sites
and a longer sampling period could alter this appraisal.
For the French streams, we found a clear relationship
between pesticide stress and community composition as
indicated by %SPEAR(abundance). Among the environmen-
explained the results for %SPEAR(abundance). Furthermore,
taxa after pesticide runoff, although it was not significant.
This may be due to effects on the communities before the
beginning of our study.
The concentrations presumed to cause an effect in the
(Table 2) and are in accordance with those reported
the main cause for the decline in abundances of several
study of Liess and von der Ohe (2005a) demonstrated
decline of SPEAR. Castillo et al. (2006) stated change in
the invertebrate community structure associated with an
to a TU(D. magna)of −0.73. Finally, a review of mesocosm
studies reported effects on the macroinvertebrate commu-
nity above a TU(D. magna)of −2 (van Wijngaarden et al.,
The relationship between the relative abundance of
sensitive taxa and pesticide stress was similar for the
French and German streams (Fig. 3). Concurrently,
geographical and physico-chemical variables varied and
approximately 35% of the taxa found in France were not
recorded in the German streams. Hence, the response of
traits to pesticide stress was not affected by geographical
and taxonomical differences.
The insensitivity of traits to a range of environmental
gradients was also reported by Charvet et al. (2000).
This issue warrants further investigation, especially in
South and non-European regions, because if the dose–
response relationship between traits and pesticide stress
could be extrapolated to a wider geographical scale, it
would constitute a powerful tool for an effect assess-
ment on the continental scale.
4.2. Effects of pesticides on leaf-litter decomposition
In our study on the French streams, we found a
significant decrease in leaf-litter decomposition rates for
leaves of A. glutinosa due to pesticide stress. The
decomposition rates for streams potentially not receiving
pesticide input (0.0588±0.0008 s.e.) are in the range of
values reported for relatively pristine streams in Portugal
during summer (0.051 to 0.064) (Graca et al., 2001).
potentially receiving and sites potentially not receiving
pesticide input (0.42) with the threshold values proposed
by Gessner and Chauvet (2002) confirms the functional
impairment caused by pesticides (ratiob0.5 indicates
severe impairment). Chung et al. (1993) also reported a
ratio of 0.35 of breakdown rates of rhododendron leaves
for a pesticide-treated stream and a reference stream.
(not shown) was closely related to a decrease in leaf-litter
and Cummins (1996)) belong to SPEAR. Maltby et al.
(2002) also reported high positive correlation (Pearson's
r=0.83) between physiological effects of pesticides on a
shredder species and impairment of leaf-litter processing.
282R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
also affected an important ecosystem process, leaf
breakdown, probably mediated through the adverse effects
on number and abundance of SPEAR taxa.
4.3. Undisturbed upstream reaches enhance quality of
The presence or absence of riparian forest parts in the
3-km upstream reach explained a significant part of the
sites. The presence of undisturbed upstream reaches lead
to significantly higher values of %SPEAR(abundance)and
reported the relevance of undisturbed upstream reaches
for recovery from disturbance.
A study on the Suna river (Japan) attributed the
recovery of several invertebrate species after pesticide
spraying to recolonization from unsprayed upstream
reaches within a 5-km distance (Hatakeyama and
Yokoyama, 1997). Liess and von der Ohe (2005a)
demonstrated the recovery of pesticide-impacted com-
munities when riparian forest reaches were available in
modelling with macroinvertebrate composition of 360
streams investigated over a 17-year period in North
Germany, showed that the presence of forest parts in the
1.5 km upstream reach facilitated recovery of the relative
abundance of SPEAR taxa after modelled pesticide
contamination (Schriever et al., 2007).
Two mechanisms could explain this positive impact of
undisturbed upstream reaches. First, undisturbed upstream
reaches may provide recolonization pools from which
species could drift downstream to the impacted reach
(Waters, 1972). Second, input of woody debris and leaf
litter from the riparian forest might supplymore energy for
the downstream reaches and thus increase number and
The latter mechanism should also increase the number and
abundance of sensitive species at slightly or non-
contaminated sites. However, Schriever et al. (2007)
reported no significant difference in %SPEAR(abundance)
for low-contamination streams with and without undis-
turbed upstream reaches in the period before pesticide
application. In the present study also, we did not find
significant differences in abundance or number of SPEAR
taxa for uncontaminated sites having and not having
undisturbed upstream reaches in Finland. Therefore,
current evidence indicates that recovery should mainly be
from undisturbed upstream reaches, although a more
thorough study is still necessary to clarify this issue.
However, regardless of the underlying mechanism,
undisturbed upstream reaches clearly enhance recovery
tool for freshwater conservation in agricultural areas.
4.4. Derivation of an effect threshold for pesticides
For the pooled data ofthe German,French andFinnish
streams, a significant reduction of sensitive taxa was
observed for TU(D. magna) values higher than −3.5.
However, this value is only a rough estimate for a
threshold value because there were hardly any observa-
tions for TU(D. magna)in the interval (−3, −5). Further-
more, regarding the dose–response relationship between
TU(D. magna) and %SPEAR(abundance) in France and
Germany (Fig. 3) it remains open, if an effect threshold
exists or if the relationship is continuously linear up to a
TU(D. magna)of −5. Thus, more field data are needed
especially for low values of TU(D. magna)to clarify this
issue. Nevertheless, our data indicate effects of pesticides
in the TU(D. magna)range of −2 and −3. Two other field
studiesalsoreportedshiftsinthe invertebrate assemblages
Liess and von der Ohe, 2005a). In contrast, we do not
magna)of −2 were found (van Wijngaarden et al., 2005).
However, most studies on stream mesocosms just deal
with a single pulse exposure while in the field repeated
input of many different pesticides frequently occurs (van
Wijngaarden et al., 2005). Recently, a laboratory study
with D. magna showed that a repeated exposure to
dimethoate and pirimicarb significantly increased mortal-
ity (Andersen et al., 2006). In addition, multiple stressors
mayoccurinareaswithintense agriculture (e.g.pesticides
with a different mode of action, chronic ammonium
exposure, eutrophication) and act additively or synergis-
tically, which is commonly not incorporated in mesocosm
studies (Heugens et al., 2001). For example, organophos-
phorous pesticides that were also detected in the French
than-additive responses in combination with various
herbicides (Lydy and Austin, 2005).
We suggest that the community change at TU(D. magna)
values b−2 may have resulted from the long-term
propagation of sublethal effects. This hypothesis is
supported by the fact that the relative abundance of
sensitive taxa exhibited only a small acute response to
reduction in %SPEAR(abundance)). Sublethal effects like
reduced fecundity or delayed emergence are known to
appear up to a TU(D. magna)of −4 (Liess, 2002) and if they
cause a competitive disadvantage for sensitive species this
may result in community change (Fleeger et al., 2003).
283R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
However, this remains open to discussion, as another
explanation would be that the communities of the lightly
contaminated streams may have not recovered from past
impacts (Harding et al., 1998).
Overall, our investigation gives rise to the concern
that the effect threshold for pesticides in the field is
below a TU(D. magna)of −2, which is currently regarded
as protective, for example in the legislation on pesticides
in the European Union (EEC, 1991).
This study showed that the structure and function of
aquatic ecosystems may be impaired by pesticides. We
suggest that effects may also occur below levels that are
commonly thought to be protective. This highlights the
importance of field studies since effects at these levels
have not been observed in artificial systems. It is
noteworthy that no effects were detectable in the Finnish
study area under low pesticide usage.
A very important result for risk managers is that
undisturbed upstream reaches improve the quality of
impaired downstream reaches. This could constitute a
valuable measure for future risk mitigation in addition to
other innovations in agricultural practice.
Furthermore, current risk assessment would take a
great step forward when implementing ecological
knowledge. The use of biological traits in biomonitoring
could be a starting point and may prove superior to
taxonomically based approaches.
The trait-based SPEAR concept was capable of
discriminating between the effects of pesticides and
those of confounding factors and natural variation over
large spatial scales. Thus, the results from the regional
investigations may be extrapolated to other biogeo-
graphical regions in Central and North Europe. How-
ever, more studies are needed in non-European regions
to assess the potential for extrapolation beyond Europe.
For example, the concept could easily be applied to field
observations from North America, as a database on
invertebrate traits is available (Vieira et al., 2006).
Lodenius, Hans-Rudolf Voigt, Esa Tulosa, Peter von der
Ohe, Björn Müller and all other people that contributed
with work and facilities to the field study. We are indebted
to Kari-Matti Vuori and Mikhail Beketov for their support
inspecies determination.We are verygratefultoBranislav
Vrana and Albrecht Paschke who provided us with the
passive samplers. Special thanks to Jochen Mueller and
pesticide measurementsinBrittany.Finally,we wouldlike
to thank Bettina Egert for the chemical analyses. Carola
Schriever, Peter von der Ohe, Mikhail Beketov and two
anonymous reviewers provided valuable suggestions that
improved the manuscript. Financial support was provided
by the European Community (project SSPE-CT-2003-
501997 “HArmonised environmental Indicators for pesti-
cide Risk, HAIR”). The first author received funding
through a scholarship of the “Studienstiftung des
deutschen Volkes e.V.” (Bonn, Germany).
Appendix A. Supplementary data
Supplementary data associated with this article can
be found, in the online version, at doi:10.1016/j.scitotenv.
Akaike H. A new look at the statistical model identification. IEEE
Trans Automat Contr 1974;19:716–23.
Allan JD. Stream ecology. Dordrecht: Kluwer Academic Publishers;
1995. 400 pp.
pirimicarb. Environ Toxicol Chem 2006;25:1187–95.
Beasley G, Kneale PE. Investigating the influence of heavy metals on
macroinvertebrate assemblages using Partial Canonical Correspon-
dence Analysis (pCCA). Hydrol Earth Syst Sci 2003;7: 221–33.
Benfield EF. Leaf breakdown in stream ecosystems. In: Hauer FR,
Lamberti GA, editors. Methods in stream ecology. San Diego:
Academic Press; 1996. p. 579–89.
Berenzen N, Kumke T, Schulz HK, Schulz R. Macroinvertebrate
communitystructure inagriculturalstreams:impact ofrunoff-related
pesticide contamination. Ecotoxicol Environ Saf 2005;60:37–46.
Bond NR, Sabater S, Glaister A, Roberts S, Vanderkruk K.
Colonisation of introduced timber by algae and invertebrates and
its potential role in aquatic ecosystem restoration. Hydrobiologia
Booij K, Hofmans HE, Fischer CV, van Weerlee EM. Temperature-
dependent uptake rates of nonpolar organic compounds by
semipermeable membrane devices and low-density polyethylene
membranes. Environ Sci Technol 2003;37:361–6.
Castillo LE, Martinez E, Ruepert C, Savage C, Gilek M, Pinnock M,
Solis E. Water quality and macroinvertebrate community response
following pesticide applications in a banana plantation Limon
Costa Rica. Sci Total Environ 2006;367:418–32.
Charvet S, Statzner B, Usseglio-Polatera P, Dumont B. Traits of benthic
macroinvertebrates in semi-natural french streams: an initial applica-
tion to biomonitoring in Europe. Freshw Biol 2000;43:277–96.
Chevan A, Sutherland M. Hierarchical partitioning. Am Stat 1991;45:
Chung K, Wallace JB, Grubaugh JW. The impact of insecticide
treatment on abundance, biomass and production of litterbag fauna
in a headwater stream: a study of pretreatment, treatment and
recovery. Limnologica 1993;23:93–106.
284R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285
Clements WH, Carlisle DM, Lazorchak JM, Johnson PC. Heavy
metals structure benthic communities in Colorado mountain
streams. Ecol Appl 2000;10:626–38.
DeLorenzo ME, Scott GI, Ross PE. Toxicity of pesticides to aquatic
microorganisms: a review. Environ Toxicol Chem 2001;20:84–98.
DIREN REA. Les produits phytosanitaires dans les eaux super-
ficielles; 2005. http://www.bretagne.ecologie.gouv.fr/Eau/
EEC. Council Directive of 15 July 1991 concerning the placing of
plant protection products on the market. Luxembourg: Office for
Official Publications of the European Union; 1991, 137 pp.
EUROSTAT. Use of plant protection products in the European Union.
Luxembourg: Office for Official Publications of the European
Union; 2002. 122 pp.
Flack VF, Chang PC. Frequency of selecting noise variables in subset
regression-analysis — a simulation study. Am Stat 1987;41:84–6.
Fleeger JW, Carman KR, Nisbet RM. Indirect effects of contaminants
in aquatic ecosystems. Sci Total Environ 2003;317:207–33.
Frankart C, Eullaffroy P, Vernet G. Comparative effects of four
herbicides on non-photochemical fluorescence quenching in
Lemna minor. Environ Exp Bot 2003;49:159–68.
Gessner MO, Chauvet E. A case for using litter breakdown to assess
functional stream integrity. Ecol Appl 2002;12:498–510.
Graca MAS, Ferreira RCF, Coimbra CN. Litter processing along a
stream gradient: the role of invertebrates and decomposers. J North
Am Benthol Soc 2001;20:408–20.
Grande M, Andersen S, Berge D. Effects of pesticides on fish.
Norvegian J Agric Sci 1994:195–209.
Harding JS, Benfield EF, Bolstad PV, Helfman GS, Jones III EBD.
Stream biodiversity: the ghost of land use past. Proc Natl Acad Sci
U S A 1998;95:14843–7.
Hatakeyama S, Yokoyama N. Correlation between overall pesticide
effects monitored by shrimp mortality test and change in
macrobenthic fauna in a river. Ecotoxicol Environ Saf 1997;36:
Heugens E, Hendriks A, Dekker T, van Straalen NM, Amiraal W. A
review of the effects of multiple stressors on aquatic organisms and
analysis of uncertainty factors for use in risk assessment. Crit Rev
Illies J. Limnofauna Europaea, a compilation of the European
freshwater species with emphasis on their distribution and ecology.
Jena: G. Fischer; 1978.
Liess M. Population response to toxicants is altered by intraspecific
interaction. Environ Toxicol Chem 2002;21:138–42.
Liess M, Schulz R. Linking insecticide contamination and population
response in an agricultural stream. Environ Toxicol Chem 1999a;18:
Liess M, von der Ohe PC. Analyzing effects of pesticides on
invertebrate communities in streams. Environ Toxicol Chem
Liess M, Schulz R, Neumann M. A method for monitoring pesticides
bound to suspended particles in small streams. Chemosphere
Liess M, Schulz R, Liess MH-D, Rother B, Kreuzig R. Determination
of insecticide contamination in agricultural headwater streams.
Water Res 1999b;33:239–47.
Liess M, von der Ohe PC, Schriever CA, Schäfer RB, Beketov MA.
Online database of species at risk (SPEAR database) 2006; http://
Effects of pesticides in the field — EPIF. Brussels: SETAC Press;
2005b. 136 pp.
Lydy MJ,AustinKR. Toxicityassessmentofpesticidemixturestypical
of the Sacramento-San Joaquin Delta using Chironomus tentans.
Archives Environ Contam Toxicol 2005;48:49–55.
Maltby L, Clayton SA, Wood RM, McLoughlin N. Evaluation of the
Gammarus pulex in situ feeding assay as a biomonitor of water
quality: robustness responsiveness and relevance. Environ Toxicol
Maret TR, Cain DJ, MacCoy DE, Short TM. Response of benthic
invertebrate assemblages to metal exposure and bioaccumulation
associated with hard-rock mining in northwestern streams USA.
J North Am Benthol Soc 2003;22:598–620.
McGarigal K, Cushman S, Stafford S. Multivariate statistics for
wildlife and ecology research. New York: Springer; 2000. 284 pp.
Merritt RW, Cummins KW. Trophic relations of macroinvertebrates.
In: Hauer FR, Lamberti GA, editors. Methods in stream ecology.
San Diego: Academic Press; 1996. p. 453–74.
Meteo France. Temps du mois http://www.meteofrance.com/FR/
Munzel U, Hothorn LA. A unified approach to simultaneous rank test
procedures in the unbalanced one-way layout. Biom J 2001;43:
Neumann M, Schulz R, Schäfer K, Müller W, Mannheller W, Liess M.
The significance of entry routes as point and non-point sources of
pesticides in small streams. Water Res 2002;36:835–42.
R Development Core Team. R: A language and environment for
statistical computing, reference index version 2.4.1 www.r-project.
Richards RP, Baker DB. Pesticide concentration patterns in agricul-
tural drainage networks in the Lake Erie basin. Environ Toxicol
Schriever CA, Hansler-Ball M, Holmes C, Maund S, Liess M.
Agricultural intensity and landscape structure: influences on the
macroinvertebrate assemblages of small streams in northern
Germany. Environ Toxicol Chem 2007;26:346–57.
Statzner B, Bady P, Doledec S, Scholl F. Invertebrate traits for the
biomonitoring of large European rivers: an initial assessment of
trait patterns in least impacted river reaches. Freshw Biol
Strahler AN. Quatitative analysis of watershed geomorphology. Trans
Am Geophys Union 1957;38:913–20.
Tomlin CDS. The e-Pesticide Manual (Twelfth Edition) on CD-
Version 2.1. The British Crop Protection Council; 2001.
van Wijngaarden RPA, Brock TCM, van Den Brink PJ. Threshold
levelsfor effects ofinsecticidesin freshwaterecosystems:a review.
Vieira NKM, Poff NL, Carlisle DM, Moulton SR, Koski ML,
Kondratieff BC. A database of lotic invertebrate traits for North
America: U.S. Geological Survey Data Series 187; 2006.
WallaceJB,EggertSL,MeyerJL,Webster JR.Multipletrophic levelsof
a forest stream linked to terrestrial litter inputs. Science 1997;277:
Wallace JB, Whiles MR, Eggert S, Cuffney TF, Lugthart GJ, Chung K.
Long term dynamics of coarse particulate organic matter in three
Appalachian Mountain streams. J North Am Benthol Soc 1995;14:
Waters TF. The drift of stream insects. Annu Rev Entomol 1972;17:
285R.B. Schäfer et al. / Science of the Total Environment 382 (2007) 272–285