Journal of Hydrology 273 (2003) 1–17

www.elsevier.com/locate/jhydrol

Hydrological characterisation of pesticide loads using hydrograph

separation at different scales in a German catchment
K. Müllera,*, M. Deurerb, H. Hartmannc, M. Bachd, M. Spitellerc, H.-G. Freded
a
AgResearch Ltd, Ruakura Research Centre, East Street, Private Bag 3123, Hamilton, New Zealand
b
Department of Soil Science, Technical University of Hannover, Herrenhäuserstr. 2, 30419 Hannover, Germany
c
Department of Ecological Chemistry and Ecotoxicology, University of Kassel, Steinstraße 19, 37213 Witzenhausen, Germany
d
Department of Agricultural Ecology and Natural Resources Management, University of Giessen, Heinrich-Buff-Ring 26-32,
35392 Giessen, Germany
Received 9 October 2001; revised 8 October 2002; accepted 11 October 2002

Abstract
Pesticide pollution of surface waters was studied in a 49.7 km2 catchment in Germany. The loads were differentiated into
point source (PSP) and non-point source pollution (NPSP). The contribution of runoff, interflow, groundwater, drainage and
spray drift to river contamination was defined as NPSP. Pesticides discharged into the river via wastewater treatment plants and
sewer overflows were considered as PSP. Hydrology and pesticide loads were studied on the scale of the entire catchment and in
two sub-catchments not influenced by PSP. River discharges and concentrations of 19 pesticides were measured at four
locations over 447 days. The total load detected in the entire catchment amounted to 3249 g active ingredient (a.i.). The stream
flow data were separated into base and direct flow applying a non-linear reservoir separation algorithm. Single rainfall events
leading to runoff and interflow dominated NPSP in the sub-catchments. For the sub-catchment Rossberg 87% and for
Leidenhofen 71% of the pesticide load could be attributed to direct flow. This corresponded to 14 and 34% time of peak flow
events for Rossberg and Leidenhofen, respectively. For the entire catchment only 15% of the NPSP was transported with direct
flow with 49% attributed to base flow. The difference between the entire and the sub-catchment contribution was attributed to
dilution, hysteresis and merging effects. Applying a digital recursive filter technique for hydrograph separation confirmed the
results of the non-linear reservoir separation algorithm. The impact of the hydrograph separation technique on the hydrological
characterisation of pesticide loads was insignificant. The results emphasise the dynamic nature of pesticide transport and its
complex interaction between size and position of source areas along a stream flow network.
The measured loads in the sub-catchments were scaled up to the entire catchment as an estimate for the total NPSP. The
fraction of the area treated with a pesticide was found not to be an appropriate scaling factor. Research is required to develop
adequate methods to scale detailed measurements of pesticide loads from small scales up to catchments with complex drainage
systems.
q 2003 Elsevier Science B.V. All rights reserved.
Keywords: Pesticides; Surface water contamination; Non-point source pollution; Catchment scale; Base flow separation

* Corresponding author. Tel.: þ 64-7-838-5281; fax: þ64-7-838-5073.

E-mail address: [email protected] (K. Müller).

0022-1694/03/$ - see front matter q 2003 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 2 - 1 6 9 4 ( 0 2 ) 0 0 3 1 5 - 3
2 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17

1. Introduction So far, catchment scale studies dealing with

pesticide pollution have mainly concentrated on the
Pesticide contamination of water resources is a documentation of surface water quality (Donnelly and
major water-quality issue in Western Europe (Albanis Ferrari, 2001; Dubrovsky et al., 1998; Fenelon and
et al., 1998; Barcelo et al., 1996; Carter, 1999). Moore, 1998; Harman-Fetcho et al., 1999) and the
Pesticide residues have been detected in groundwater calculation of total pesticide loads (Bach and Frede,
and surface water. Understanding pesticide dynamics 1996; Isenberg-Schröter et al., 1998; Larson et al.,
in the environment is crucial for any effective 1999). Other studies focused on the contribution of
prevention and management strategy for water different non-point sources to the total load (Ng and
resources. Clegg, 1997; Rawn et al., 1999). Some investigations
Pesticide transport results from many interacting differentiated between point source (PSP) and non-
hydrological, chemical and microbiological pro- point source pollution (NPSP) (Fischer, 1996; Frank
cesses at different spatial scales. Microscale pro- et al., 1982; Müller et al., 2002; Seel et al., 1994). The
cesses such as the kinetics of pesticide sorption importance of hydrology in NPSP is poorly under-
(Brusseau, 1994; Streck et al., 1995) and micro- stood. Whereas nutrient pathways have frequently
biological pesticide degradation (Loll and Moldrup, been analysed with respect to catchment hydrology
2000) have been shown to be important for pesticide (Eidem et al., 1999; Heathwaite and Dils, 2000;
transport on the field scale. Preferential flow of McDowell et al., 2001; Pedersen et al., 1999; Pionke
reactive solutes is a phenomenon that can be caused et al., 1996), pesticide transport was rarely analysed
by a multitude of factors (Wang and Feyen, 1998). It from this point of view. Banchard and Lerch (2000)
is observed from the column (Perret et al., 2000) studied the nitrate and pesticide contamination of 95
over the soil profile (Flury et al., 1994), field (Brown streams throughout northern Missouri over two years.
et al., 1995; Kung et al., 2000; Traub-Eberhard et al., They classified the hydrology of a catchment as one of
1995), hillslope (Noguchi et al., 1999) up to the three key factors for assessing watershed vulnerability
catchment scale (McDonnell, 1991). It takes place in to losses of the analysed chemicals. Fenelon and
the unsaturated (Flury, 1996) as well as in the Moore (1998) investigated the transport pathways for
saturated zone (Kupfersberg and Blöschl, 1995). The nitrate and 82 pesticides into the Sugar Creek by
transport velocity during preferential flow can be so screening 11 wells and two tile drains in the
high that the physico-chemical properties of reactive watershed. Whereas samples from the stratified drift
solutes are practically irrelevant (Kung et al., 2000). aquifers showed hardly any pesticide contamination,
Runoff has to be considered on a hill slope scale shallow groundwater from the alluvial units was
(Frank et al., 1982), bank storages, sinks and identified as a source of pesticide traces detected in
different sources such as discharges by wastewater Sugar Creek at low base flow conditions. Effluents
treatment plants operate at the catchment scale from tile drains were identified as an important
(Blöschl and Sivapalan, 1995). For the contamination pathway for the creek contamination of pesticides.
of surface waters the individual process is less The authors did not measure surface runoff in this
important than the reaction time of a pesticide with study. Ng and Clegg (1997) analysed the contribution
the surrounding soil and plant residue materials. of different flow components for the transport of
Rapid flow processes such as preferential flow or atrazine and metolachlor in the catchment of the
runoff drastically reduce the time available for Nissouri Creek in Ontario, 35 km2. They assumed an
reactions such as sorption or degradation and can average pesticide concentration for each of the 25
lead to direct transfer of pesticides from the soil and stormflow events analysed and calculated the pesti-
plant surfaces into surface waters. The separation of cide loads of the different flow components for an
direct flow from base flow events in a discharge event by multiplying the flow volumes with the
series is a possibility to integrate a variety of rapid average concentration. The results indicated that 75%
flow processes on catchment scale. Little is known of the total detected atrazine load and 65% of the
about the importance of flow processes for the long- estimated metolachlor load was transported by direct
term contamination of surface waters. flow. Squillace and Thurman (1992), applying
 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17 3

a hydrograph separation analysis found in their two nothing if they are to be weighted by pesticide
year study at the Cedar River, Iowa, that about 94% of transfer.
the annual river load of atrazine was transported by This study has two main objectives:
direct flow.
In the present study the contribution of point and To hydrologically classify the non-point source
non-point sources to total pesticide loads for the entire pesticide loads in surface waters by separating
catchment was differentiated and the processes them into base and direct flow events and analyse
underlying NPSP hydrologically characterised by their importance for an entire catchment and two
attributing pesticide loads to the flow components of sub-catchments.
base and direct flow. Pesticides detected in base flow To extrapolate pesticide loads from the sub-
are related to the slow processes of water movement catchments to the entire catchment using scaling
such as subsurface runoff, drainage water and flow factors derived from the percentage of land planted
through the soil matrix. with a crop and treated by a specific pesticide.
Another important question addressed in this study
is how to estimate the amount of applied pesticides
that is transferred via NPSP into surface waters for a 2. Material and methods
given site. One way to answer this question is to study
pesticide contamination in sub-catchments with no 2.1. Study area
PSP and scale the results up to the entire catchment.
However, this approach implies that the sub-catch- The research area is the 49.7 km2 Zwester Ohm
ment represents the average behaviour for pesticide catchment located in a low mountain range in central
transfer of the entire catchment. Little is known how Germany (county Marburg-Biedenkopf, federal state
to define a representative elementary area for Hessen), 10 km south of the city Marburg (Fig. 1).
hydrological processes (Woods et al., 1995), and The Zwester Ohm is a tributary of the river Lahn.

Fig. 1. Location of sampling sites in the catchment area.

4 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17

The catchment area is predominantly used for Table 1

agriculture. The mountainous topography together Area (%) of field crops in the catchment and the sub-catchments
with the silt and clay loam textures of its soils render Crop Area (%)
the catchment susceptible for rapid transfer processes Rossberg Leidenhofen Catchment
of pesticides into surface waters such as preferential SP 3 SP 2 SP 1
flow and runoff. Additionally, the catchment was
chosen because (i) the stream’s flow is unregulated; Arable land 35 50 40
(ii) all farms and households in the catchment area are Winter wheat 15 23 13
connected to the wastewater treatment plant system Winter barley 3 7 10
and, therefore, PSP could easily be identified; and (iii) Rye 2 2 2
Summer barley 2 1 1
the socio-agricultural structure is typical for catch-
Oats 6 4 3
ments in central Germany with many small farms run Rape seed 3 3 6
by part-time farmers. The catchment is 200 –400 m Maize 2 3 2
above the sea level and has an undulating topography. Sugar beets 2 1 1
Geologically Pleistocene loess sediments of different Root Crops 0 6 2
origins lie over redstone, disrupted by volcanic Pasture, fallow 13 15 23
features (Rittweger, 1997). Annual replenishments Forest 50 30 28
to groundwater supplies range between 32 and Settlement 2 5 9
142 mm (Anonymous, 1990b). Total (km2) 1.0 3.6 49.7
Luvisols, cambisols and stagnic gleysols (FAO-
system) are the main soil types, depending on the
The mean annual rainfall is 585 mm and the
parent material and the groundwater level. Soils are
average annual temperature is 8 8C. The study area is
predominantly derived from loess. Based on the
located in the temperate zone without extremes of
SCS-curve number method (US Department of
precipitation. The coldest month is January with an
Agriculture, 1972) the soils were classified into
average temperature of 0.5 8C and the warmest month
hydrologic groups. In the catchment soils of the
is August with 18 8C average temperature. Daily
hydrologic group C are dominating, which are silt
rainfall data were collected at six locations (Fig. 1)
and clay loams with a low infiltration capacity,
and interpolated according to the Thiessen polygon
prone to runoff. The arable area comprises 20 km2
method (Maniak, 1993).
of which about 23% is tile drained.
The actual land use in the catchment (1997/98) was
2.2. Hydrological methods
mapped supported by photos of aerial surveys at a
scale of 1:5000 (Anonymous, 1990a). 63% of the area
2.2.1. River discharges
was used agriculturally, with 23% being fallow land
At all sampling sites (SP 1, SP 2, SP 3 and SP 4),
and pastures (Table 1). Forest covers 28% of the
the water-stage heights h (m) were continuously
watershed and settlement areas account for 9% of the
recorded with water level recorders (R20, Ott,
total catchment. The two sub-catchments Rossberg
Kempten) and then transferred into discharges by
and Leidenhofen cover 10% of the entire catchment.
using sampling-site specific calibration functions. The
They represent the typical land use and soil distri-
discharges of the creeks in the two sub-catchments
bution in the entire catchment (Table 1).
(SP 2 and SP 3) and in the headwaters of the Zwester
The boundaries of the catchment and the two
Ohm (SP 4) were measured at 90-degree V-notch
sub-catchments as well as the flow net of the
weirs. The relationship between measured height h
Zwester Ohm were calculated with the model
and discharge Q (m3 s21) was described by square
TOPAZ 1.2 (Topographic Parameterization (Gar-
root equations as suggested by Dyck and Peschke
brecht and Martz, 1995)). A digital elevation model
(1989):
(625 m2 grids) with a vertical resolution of 5 m
[^ 3 m] (Nöhles, 2000) was the data basis for these pffiffiffiffiffi
8 2:5
Q¼ m 2gh ð1Þ
calculations. 15
 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17 5

where g (m s22) is the acceleration due to gravity and 2.2.2. Hydrograph separation
m (m) a coefficient accounting for friction, velocity Several methods exist to separate a stream
and pressure distributions in the weir (Table 2). At SP hydrograph into the components of base and direct
1 the discharge was indirectly determined by flow (Arnold et al., 1995; Caissie et al., 1996;
measuring the flow velocities at a cross-sectional Chapman, 1999; Nathan and McMahon, 1990). In
area with a flow meter (Flügel C2, Ott, Kempten). this study two algorithms for hydrograph separation
Combining flow velocities and water depths the were applied. Lyne and Hollick (1979) introduced a
mean-section method of Herschy (1995) was applied. recursive digital filter technique that was originally
The fractional discharge Qi;iþ1 in a cross-sectional developed in signal analysis as a hydrograph separ-
area enclosed by the two segments i and i þ 1 is ation method. It therefore has, as manual methods of
calculated as: base flow separation, no true physical basis, but it is
   objective and reproducible. With this algorithm direct
v þ vi hiþ1 þ hi flow, Qdf, is interpreted as a high frequency signal and
Qi;iþ1 ¼ iþ1 ðbiþ1 2 bi Þ ð2Þ
2 2 is separated from base flow, Qbf, the low frequency
signals by recursive digital filtering. Results of the
where v (m s21) is the flow velocity, h the stage height method are comparable to graphic hydrograph
and b (m) the distance between the bank and the separation techniques (Arnold et al., 1995; Muttiah
segments. Integrating the fractional discharges over et al., 1994). Nathan and McMahon (1990) modified
the individual cross-sectional areas gives the total the Lyne and Hollick-algorithm and proposed the
discharge. Having repeated this procedure for many following filter equation:
times with varying stage heights and discharges the
following relationship was derived: Qdf ¼ 0:925Qdfðt21Þ þ 0:9625ðQtf 2 Qtfðt21Þ Þ ð4Þ

subject to Qdf $ 0 for the start time t in (d). Qtfis total

Q ¼ 0:5 2 3:45h þ 6:5h2 ð3Þ
flow. Smoothing depends on the number of filtering
with a resulting r 2 of 0.96 and a standard deviation of steps. Arnold et al. (1995) recommended one filtering
0.05 m3 s21. Assuming Gaussian error propagation step for hydrograph separation.
(Papula, 1998) measurement errors for the discharge The second method applied in this study is a
of 8% and 14.5 to 22% of the total discharge were technique developed by Wittenberg (1999). This
estimated for the indirect and the direct measure- method was considered, because it views an aquifer
ments, respectively. as a non-linear water storage reservoir, which is
particularly important, if the discharging aquifers
consist of different geologic materials or strata. Also it
Table 2 is possible to operationally apply the method to
Parameter of the calibration equations (Eq. (1)) for the weirs in the separate short streamflow time series. The method
catchment
consists of two steps:
Location of na mb rc Average
the weir errord (%) (1) A recession curve representative for the storage
behaviour of the aquifer system was identified in
Sub-catchment Leidenhofen 71 0.78 0.98 20.0 a period that was characterised by low evapo-
(SP 2)
transpiration with no precipitation events. The
Sub-catchment Rossberg 79 0.65 0.97 14.5
(SP 3) following equation was fitted to the streamflow
Zwester Ohm close to its 51 0.48 0.99 22.0 recession data (Wittenberg, 1999):
source (SP 4) !2
Qbf ðt ¼ 0Þ0:5
a
n: number of calibrations. Qbf ðtÞ ¼ Qbf ðt ¼ 0Þ 1 þ t : ð5Þ
b
m: coefficient of the square root equation. a
c
r ¼ Pearson’s correlation coefficient.
d
Calculated according to the error propagation rules by Gauß The temporal resolution t was in daily time steps.
(Papula, 1998). Qbf ðt ¼ 0Þ (m3 s21) is the base flow discharge at
6 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17

the start, t ¼ 0; of the chosen data series and a by automatic samplers (ISCO 3700, ISCO, Lincoln
(m3 – 3b sb) is an adjustable parameter of the and APN 20, Klotzbach Hard- and Software, Gies-
relation. Eq. (5) was fitted to the data with sen). Each sample constituted a composite of 10 ml-
Levenberg – Marquardt’s non-linear least squares aliquots taken at 10 min intervals and was collected in
algorithm adjusting a (Press et al., 1992). a glass bottle that was washed with methanol. To
(2) The data series were separated into ascending prevent degradation, samples were cooled down and
and descending discharge-limbs. Only ascending stored at 4 8C at the site. Every day the samples were
or descending discharge-limbs of a 3-day delivered to the laboratory and stored at 2 20 8C until
duration characterised by a steady trend were analysis. Water samples were analysed for 14
considered (Müller, 2000). This tolerance limit pesticides commonly used in the area (Table 3). In
definition results from (i) the mathematical addition during the spring season samples were also
necessity of at least three data points to analysed for five phenoxy acid herbicides (Table 3)
distinguish between ascending and descending that were exclusively applied in spring.
limbs and (ii) the temporal resolution of the The samples collected during spring were divided
pesticide concentration measurements. The base into two sub-samples of 500 ml, as the extraction and
flow of the descending limbs was directly analysis of the phenoxy acid herbicides were carried
separated from the discharge Qtf ðtÞ using Eq. out separately. The extraction of the pesticides was
(5), and the direct flow was then simply Qtf ðtÞ 2 performed with solid phase extraction according to
Qbf ðtÞ: Neglecting hysteresis Eq. (5) was the method of Janssen et al. (1995) in a modified form
inverted, then also applied to the ascending (Müller, 2000). The recovery efficiencies ranged
discharge-limbs. The computation of the base between 83 and 99% (Table 3). None of the reported
flow recession proceeds backwards in time. The values are corrected for the analytical recovery
base flow at the time t 2 Dt follows from the efficiencies.
base flow at the time t and so on, where in our For the analysis two different procedures were
case Dt was 1 day (Wittenberg, 1999): applied, one for the non-polar and semi-polar
  pesticides (e.g. atrazine) and the other for the five
2t 2
Qbf ðt 2 DtÞ ¼ Q20:5
bf ðtÞ þ : ð6Þ phenoxy acid herbicides. Liquid chromatography was
a carried out with a Gynkotek pump (model P580,
The ascending and descending discharge-limbs Gynkotek, Germering). HPLC was either coupled
of the base flow were finally connected at the directly to diodenarray detection (UV340S, Gynko-
peak of the total discharge hydrograph. tek, Germering) or to mass spectrometric detection
(MS/MS, triple quadrupole mass spectrometer MAT
Daily discharges were then classified with three TSQ 7000, Finnigan, San Jose), respectively. The
categories: limits of detection are presented in Table 3. The
analysis is described in detail elsewhere (Hartmann
1. Dominance of base flow throughout a day, if daily et al., 1998; Müller, 2000).
base flow . 80% of total discharge.
2. Dominance of direct flow throughout a day, if daily 2.3.2. Differentiation between PSP and NPSP
direct flow . 80% of total discharge. The total pesticide load of the entire catchment was
3. No clear dominance of base or direct flow. measured at SP 1 (Fig. 1) and was the result of both
PSP and NPSP. The main source for the PSP, effluents
2.3. Pesticides from the wastewater treatment plant situated in the
headwaters, was directly monitored (SP 4). PSP loads
2.3.1. Measurement of pesticide concentrations from other sources were estimated with the empirical
Throughout the main pesticide application periods model MOMENT 4.0 (Anonymous, 1997). The model
(March 10 –May 31 and September 01 –November was used to approximate loads associated with
30) surface water samples were collected continu- overflow volumes from the combined sewer over-
ously at four sampling sites (Fig. 1) in daily time steps flows, which discharge into the river Zwester Ohm
 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17 7

Table 3
Physico-chemical properties of the pesticides included in the investigation (Huber, 1998), recovery rates (r.r.) and limits of detection (lod)

Common name Chemical name (IUPAC)a KOC (l kg21) DT50 (d) r.r. (%) lod (mg l21)

Atrazine 6-Chloro-N 2-ethyl-N 4-isopropyl-1,3,5-triazine-2,4-diamine 100b 60b 91.0 0.04

Carbetamide (R)-1-(Ethylcarbamoyl)ethyl carbanilate 10c 60c 97.1 0.03
Chlortoluron 3-(3-Chloro-p-tolyl)-1,1-dimethylurea 235 27 98.8 0.01
2,4-D 2,4-Dichlorophenoxy)acetic acid 38 10 91.3 0.04
Dichlorprop-P (R)-2-(2,4-Dichlorophenoxy)propanoic acid 20 12 91.4 0.02
Dimefuron 3-[4-(5-tert-Butyl-2,3-dihydro-2-oxo-1,3,4-oxadiazol-3-yl)-3- 145c 50c 97.0 0.01
chlorophenyl]-1,1-dimethylurea
Diuron 3-(3,4-Dichlorophenyl)-1,1-dimethylurea 800 68 97.9 0.02
Fenpropimorph (^)-cis-4-[3-(4-tert-Butylphenyl)-2-methylpropyl]-2,6- 3400 31 97.2 0.05
dimethylmorpholine
Isoproturon 3-(4-Isopropylphenyl)-1,1-dimethylurea; 3-p-cumenyl-1,1- 85 12 96.1 0.04
dimethylurea
MCPA (4-Chloro-2-methylphenoxy)acetic acid; 4-chloro-o-tolyloxyacetic 55 15 84.3 0.03
acid
Mecoprop-P (R)-2-(4-Chloro-o-tolyloxy)propionic acid 20 9 93.7 0.03
Metamitron 4-Amino-4,5-dihydro-3-methyl-6-phenyl-1,2,4-triazin-5-one; 4- 156 21 93.8 0.05
amino-3-methyl-6-phenyl-1,2,4-triazin-5(4H )-one
Metazachlor 2-Chloro-N-(pyrazol-1-ylmethyl)acet-20 ,60 -xylidide 80 6 97.7 0.01
Metolachlor 2-Chloro-60 -ethyl-N-(2-methoxy-1-methylethyl)acet-o-toluidide 200 39 88.7 0.02
Pirimicarb 2-Dimethylamino-5,6-dimethylpyrimidin-4-yl dimethylcarbamate 53b 7b 92.8 0.03
Simazine 6-Chloro-N 2,N 4-diethyl-1,3,5-triazine-2,4-diamine 201 57 88.2 0.04
2,4,5-T 2,4,5-Trichlorophenoxyacetic acid 80b 30b 87.1 0.02
Terbuthylazine N 2-tert-butyl-6-chloro-N 4-ethyl-1,3,5-triazine-2,4-diamine 250 70 88.5 0.02
Triadimenol (1RS,2RS;1RS,2RS)-1-(4-chlorophenoxy)-3,3-dimethyl-1-(1H- 360 110 98.1 0.02
1,2,4-triazol-1-yl)butan-2-ol
a
Nomenclature of the International Union of Pure and Applied Chemistry.
b
Hornsby (1992).
c
Tomlin (1994).

between SP 4 and SP 1 (Fig. 1). The sum of the as for the sub-catchments and the sampling site in the
directly measured and with simulation estimated PSP headwaters were approximated. Therefore, different
was then subtracted from the measured total pesticide spatial scales can be addressed. The hydrological
load to indirectly derive the NPSP for the entire classification of the pesticide loads was conducted for
catchment. This approach to identify pesticide path- each site, and thus resulted in scale-dependent
ways is discussed in detail elsewhere (Müller et al., measures of base and direct flow components.
2002).
2.3.4. Uncertainty of calculated pesticide loads
2.3.3. Calculation of pesticide loads The uncertainty of the daily pesticide loads Xl
The daily load of a particular pesticide p, Lp ðtÞ (%) is the result of the uncertainty of discharge and
passing a sampling site was calculated by multiplying concentration measurements. The uncertainty of the
the daily stream flow volumes Qtf ðtÞ by the mean daily discharge measurements XQi ranged from 8 to 22%
stream pesticide concentration C p ðtÞ: Concentrations depending on the type of measurement. The error
below the detection limit were set to 0. Therefore, the connected with the aggregation of hourly into daily
calculated load is a minimum of the actual load. The values is assumed to be negligible. Quality
daily loads were summed up over the whole control procedures for the pesticide analysis have
investigation period for the different sampling sites. included the analysis of blanks and replicates as
Total pesticide losses for the entire catchment as well well as the determination of recovery efficiencies.
8 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17

The uncertainty of the extraction method XCi was Der Rat von Sachverständigen für Umweltfragen
estimated to range between 2 and 16% depending (1998) estimated that nationwide, agricultural appli-
on the pesticide (Table 3). The uncertainty of cations accounted for 80% of the total pesticide usage
analysis was considered to be negligible. Following in 1997. Therefore, non-agricultural pesticide appli-
the error propagation of Gauß (Papula, 1998) the cations were neglected in this study. About 4600 kg
uncertainty of the load can then be calculated with active ingredient (a.i.) was applied during the
Eq. (7): investigation period (March 10 1997 –May 31 1998).
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Xl ¼ ^ ðXC2 i þ XQ2 i Þ: ð7Þ
3. Results and discussion
Therefore, the pesticide loads were given with a
maximum most probable uncertainty of 27%.
3.1. Hydrological characterisation
2.3.5. Estimation of pesticide amounts applied
To estimate the amount of pesticides applied, a 3.1.1. River discharges
questionnaire was conducted in the villages of the In 1997 the catchment received 463 mm of rainfall,
catchment in spring 1997, 1998 and fall 1997. which was significantly ðp , 0:05Þ less than the
Different interviewing processes led to information average annual rainfall of 585 mm. Apart from
on 70 and 30% of the total cropped area in the year October and November, which were characterised
1997 and 1998, respectively. From this data pesticide by ten heavy rainfall events (. 5 mm), all months
application rates for specific crops in the catchment were unusually dry. In contrast the following spring
were estimated. The land use in the catchment was was extremely wet with six erosive rainfall events
mapped for both years of the investigation. Estimates (. 10 mm within 24 h or . 5 mm within half an hour,
of pesticide use were inferred from the data on crop Auerswald, 1996) and plenty minor events. The
acreage and from the average application rates of precipitation pattern was mirrored by discharge
pesticides for specific crops. The results from the (Fig. 2). Water required 5 h to flow from the source
inquiries were extrapolated to the total arable land by of the stream to the outlet of the catchment, calculated
statistics analyses (Müller, 2000). Non-agricultural according to Dyck and Peschke (1989). This reflects
applications of pesticides are difficult to quantify, but the direct response of the catchment to rainfall.

Fig. 2. Daily total discharge measured at the outlet of the catchment (SP 1) and catchment precipitation.
 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17 9

Fig. 3. Separation of base flow from total flow for the entire catchments (SP 1) and for the sub-catchment Rossberg (SP 3) according to the
Wittenberg algorithm (Wittenberg, 1999) and the recursive digital filter technique (Nathan and McMahon, 1990).

The annual flow volume at the outlet of the catchment the Zwester Ohm. With regards to pesticide pollution
(SP 1) amounted to 7.33 £ 106 m3. the two sub-catchments seemed to be reasonably
The average flow rates for the sub-catchments were suited as representative units for the entire catchment.
similar if expressed on an area basis (1.76 and They were not influenced by PSP, and together they
0.96 l s21 km22 for Rossberg (SP 3) and Leidenhofen represented the land use in terms of the distribution of
(SP 2), respectively). Therefore, the two sub-catch- agricultural crops in the catchment well. But the
ments behave hydrologically similar. However, the different average flow rates per unit area in the sub-
average flow rate for the entire catchment was with catchments compared to the entire catchment showed
3.82 l s21 km22 markedly higher. In contrast to the that the sub-catchments cannot represent the entire
Zwester Ohm the two creeks draining the sub- catchment hydrologically. More research is needed
catchments are ephemeral channels (Fig. 3). Further- how to define representative areas (Woods et al.,
more, they were not influenced by effluents from the 1995) for reactive transport.
wastewater treatment plant system. About 2% of the
total stream discharge of the Zwester Ohm could be 3.1.2. Hydrograph separation
attributed to the wastewater treatment plant. Further For the non-linear reservoir approach characteristic
4.4% was due to the combined sewer overflow recession curves were chosen from the measured series
emissions. Mainly during spring and summer 1997 of daily stream flow for the entire catchment (SP 1,
human discharges dominated the total discharge of Fig. 4) and the two sub-catchments (SP 2 and SP 3,

Fig. 4. Left: chosen characteristic recession curve (between the two vertical lines) of the daily discharge time series, and right: recession curve
and fitted non-linear reservoir equation for the total catchment (SP 1).
10 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17

Fig. 5. Left: chosen characteristic recession curve (between the two vertical lines) of the daily discharge time series, and right: recession curve
and fitted non-linear reservoir equation for the three catchments (top SP 4, middle SP 3 and bottom SP 2).

Fig. 5). From the discharge series time periods with low comparison in the following sections, the results
evapotranspiration rates and no major precipitation for the recursive digital filter technique are given
events were chosen. Fitting Eq. (5) to the series of in parentheses following the percentage values
discharge recession data the different hydrological calculated applying the Wittenberg algorithm.
behaviour of the three catchments is obvious (Table 4). Table 4
This also theoretically supports the choice of Witten- Parameter fit for the non-linear reservoir equation (Eq. (5)) for the
berg’s hydrograph separation algorithm (Wittenberg, three catchments
1999). A non-linear approach takes complex aquifer
Sampling site na a (m3 – 3bsb) (x 2 )b
compositions within catchments into account.
Wittenberg’s (1999) non-linear reservoir algo- Entire catchment (SP 1) 24 25.45 0.20
rithm and a recursive digital filter technique (Nathan Sub-catchment Leidenhofen (SP 2) 16 38.11 0.95
and McMahon, 1990) were applied to separate base Sub-catchment Rossberg (SP 3) 31 31.77 0.39
flow from the time-series of daily discharge a
n: number of days of the characteristic recession curve.
b
measured at the three sampling sites. For ease of x 2: coefficient of the chi squared test.
 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17 11

The annual base flow accounted for 66% (69%) of separation techniques differed fundamentally: only
the total flow in the entire catchment during the discharge-limbs of at least 3 days characterised by a
entire investigation period of 447 days (Fig. 3). The steady trend were differentiated with the Wittenberg
time-series consisted of 217 days with daily water algorithm. Therefore, sharp and short peaks were
samples. According to the classification outlined in skipped in the separation procedure, e.g. in Rossberg
Section 2.2.2. 66% (81%) of these days were (SP 3) a storm flow event induced by a heavy
dominated by base flow, 6% (2%) by direct flow thunderstorm in August was not separated into base
and 28% (17%) of the time period could not be and direct flow components at all (Fig. 3). Applying
classified. In the Rossberg (SP 3) and Leidenhofen the recursive filter algorithm each event was
(SP 2) sub-catchments the time-series with daily considered. The Wittenberg algorithm led to base
water samples consisted of 162 and 156 days, flow peaks being parallel to the corresponding peaks
respectively. About 25% (36%) and 62% (53%) of of total flow. Percolating rain increased level and
the total flow were on an average due to direct flow head of the groundwater reservoir and explains this
in Rossberg (SP 3) and Leidenhofen (SP 2), behaviour (v. Loewenstern, 1998). In contrast the
respectively. This was significantly ðp , 0:05Þ filter technique has to be regarded as a smoothing
higher than in the entire catchment and again technique that lacks any physical background. For
showed the different hydrological behaviour of the our study the major shortcoming of the non-linear
two sub-catchments. The base flow volume totalled separation method is the crucial importance of the
5313 m3 in Rossberg (SP 3) and in Leidenhofen (SP characteristic recession curve. The slope of the
2) 6780 m3 with 2398 m3 and 16,377 m3 classified characteristic recession curve dictates which peak
as direct flow, respectively (Fig. 3). This emphasises
flow events are considered in the separation process
that rapid water movement in the form of runoff or
and also influences which part of the total flow is
preferential flow can be a scale-dependent local
determined as direct or base flow component (Figs. 4
phenomenon (Heathwaite and Dils, 2000). How can
and 5). Therefore, in practise separating the hydro-
a representative area for the entire catchment be
graphic data with the Wittenberg approach remained
defined in this case? This problem is similar to the
to some extent an arbitrary process due to the
one in soil physics, namely to define a representa-
arbitrary choice of the characteristic recession curve.
tive elementary volume for a measurement of
Only a limited data set of 1.5 hydrological years was
hydraulic conductivity in the presence of
macropores. available for choosing a characteristic recession. In
Flowing distances and reaction times are shorter the case of longer time series an average recession
in smaller catchments. Moreover, they depend on curve could be used to overcome the problem
spatially and temporally highly variable rainfall encountered in our study. Still, there are many
events. Additionally, the flow downstream reflects catchments for which no long-term hydrological
the superposition of several sub-catchments, records exist.
smoothing the importance of localised phenomena Specifically hydrographs of streams in larger
(Fig. 3). This agrees with the results of Squillace catchments have less pronounced peaks due to the
and Thurman (1992) analysing flow components in heterogeneity of soil and land use properties and of
catchments in Iowa. They found that the percentage precipitation events (Fig. 3). Rainfall is not evenly
of direct flow of total stream flow is inversely distributed in larger catchments. This together with
proportional to the size of a catchment. More the different travel distances and times in catch-
research is needed to solve the scale-dependency of ments may lead to superimpositions of peak
hydrologic behaviour, especially when direct flow discharges and complicates hydrograph separation
is an important flow component. in larger catchments such as the entire catchment in
Even though the average results of the hydrograph this study. Moreover, for the reactive transport of
separation methods were comparable, the daily base pesticides different travel times are connected with
flow components differed significantly according to different time periods available for pesticide
the Wilcoxon rank order test ðp , 0:05Þ: The two degradation and sorption.
12 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17

3.2. Pesticides load of 172 g a.i. km22 arable land was estimated.
As a percentage of applied pesticide amounts NPSP
3.2.1. Pesticide loads determined during the entire ranged from 0.16% for metamitron to 0.0004% for
investigation period fenpropimorph. Important losses were detected for
isoproturon, diuron, metazachlor and the phenoxy
The presence of pesticides was unevenly distri-
acid herbicides.
buted throughout the investigation period (Fig. 6).
The number of detected pesticides and the
The seasonal variation in pesticide concentrations in
concentrations were lower in the sub-catchments
surface water showed a marked increase during the
than for the whole catchment. For the sub-catchment
main application periods. A seasonal pattern was also
Rossberg (SP 3) a load of 169 g a.i. km22 arable land
reported by Baker and Richards (1990), Kreuger and
was estimated, while the load accumulated to 19 g a.i.
Törnqvist (1998), Thurman et al. (1991) and Tisseau
km22 arable land for the sub-catchment Leidenhofen
et al. (1996).
(SP 2) during the entire investigation period. In both
The median concentrations varied between
cases the level of pollution was not only characterised
0.05 mg l21 for 2,4-D, metazachlor and dimefurone
by a seasonal pattern as in the entire catchment, but
and 0.27 mg l21 for atrazine and exceeded for
was dominated by a few very heavy rainfall events.
atrazine, carbetamid, chlortoluron, dichlorprop-P,
The impact of individual heavy rainfall events on the
terbuthylazine and triadimenol the drinking water pesticide loading of the two creeks was independent
quality value proposed by the EC. Pirimicarb was of each other. Local variation in both the rainfall
the only pesticide applied, which was not at all events and conditions of the pesticide application
detected. Isoproturon was present in 66% of the differed sufficiently to substantiate the observations.
analysed samples. The highest individual concen- The time of the application (Buttle, 1990), the
tration measured was 23.86 mg l21 for isoproturon amounts and formulations of the pesticides (Wau-
and occurred in October. Pesticide loads are a chope et al., 1990), the soil properties, the actual water
better ecological indicator of pesticide pollution contents at the start of the rainfall events (Flury et al.,
and allow comparison with other catchments. 1994) or the topography may have caused the high
During the entire investigation period the total spatial variability of the detected NPSP. In the
pesticide load determined from 264 water samples Rossberg sub-catchment (SP 3) nearly 46% of the
amounted to 3419 g a.i. and ranged from 0.4 g for total pesticide load was detected in spring 1997 of
2,4-D to 1322 g for isoproturon. A non-point source which 84% arose from a single storm event on 5th of
May 1997 with a rainfall intensity of 31.3 mm h21.
For the same rainfall event no significant pesticide
input was detected for the Leidenhofen sub-catchment
(SP 2). Other studies by Felding (1995) and Ng and
Clegg (1997) also stressed the importance of single
events for NPSP even during longer investigation
periods.

3.2.2. Hydrological classification of pesticide loads

Catchment hydrology has important implications
in the interpretation of the estimated NPSP for the
entire catchment in relation to the measured NPSP in
the two sub-catchments. Of the NPSP detected at SP
1, 64% (69%) could be attributed to the two flow
Fig. 6. Sum of concentrations of all pesticides analysed of the water components (Table 5). On an average base flow was
samples taken from the Zwester Ohm at the outlet of the catchment
(SP 1) during the investigation period (10 March 97–31 May 98).
contaminated with 8605 mg d21 a.i., whereas the
The dashed line delineates the drinking water quality objective for direct flow was four times higher at 31,818 mg d21
the sum of all pesticides set by the EC. a.i. However, for the entire catchment base flow still
 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17 13

Table 5 (SP 2), respectively (Fig. 7). Direct flow contributed

Average direct flow, base flow and total pesticide export via the flow 87% (80%) and 71% (68%) of the pesticides
components for the three catchments determined from daily surface
exported, and occurred during the 14% (6%) and
water samples
34% (11%) time of peak flow events (Table 5).
Direct flow Base flow Not classified In the entire catchment and the sub-catchments
pesticide concentrations associated with direct flow
Entire catchment (SP 1)
were higher than those for the base flow component.
Time (d) 12 144 61
Average flow (m3 d21) 8479 10,263 6815 Both the entire catchment characterised by a complex
Average pesticide 31,818 8605 14,951 flow net and longer travel distances and the smaller
load (mg d21) sub-catchments with practically no branched system
Total pesticide loss (g) 382 1239 912 and short travel distances influenced the level of the
Sub-catchment pesticide concentrations through interactions and
Leidenhofen (SP 2)
Time (d) 53 60 43
Average flow (m3 d21) 309 113 71.8
Average pesticide 0.31 0.11 0.20
concentration (mg l21)
Total pesticide loss (g) 22.3 0.59 8.46
Sub-catchment
Rossberg (SP 3)
Time (d) 22 77 63
Average flow (m3 d21) 109 69 31.4
Average pesticide 2.81 0.38 0.01
concentration (mg l21)
Total pesticide loss (g) 51.4 2.97 4.73

accounted for 49% (59%) of the total export (1239 g),

while direct flow only transported 382 g, 15% (10%).
Although they were hydrologically different, the
two sub-catchments showed a similar pesticide
export pattern. For the sub-catchment Rossberg
92% (83%) and for Leidenhofen 73% (84%) of the
pesticide load could be attributed to one of the flow
components. The mean pesticide concentration of the
direct flow component in the sub-catchments was 2.8
and 0.31 mg l21. The concentrations were not
normally distributed, losses in direct flow varied
markedly from one event to the other and from one
sub-catchment to the other. The timing of a rainfall
event with respect to a preceding pesticide appli-
cation may contribute to this high variability. The
pesticide concentrations associated with base flow
components were 0.38 and 0.11 mg l21 for Rossberg
(SP 3) and Leidenhofen (SP 2), respectively
(Table 5). The mean values of base and direct flow
were significantly ( p , 0.05) different for each of Fig. 7. Percentage of the total pesticide load in the two flow
compartments for the sub-catchments (SP 2 and SP 3) and for the
the sub-catchments. Base flow accounted for 5% entire catchment (SP 1) for the three application periods spring
(3%) and 2% (16%) of the total pesticide export from 1997, fall 1997 and spring 1998, top: pesticide load in direct flow;
the sub-catchments Rossberg (SP 3) and Leidenhofen bottom: pesticide load in base flow.
14 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17

NPSP (Léon et al., 2001). The loads of individual

pesticides for the entire catchment were estimated
using the percentages of the land used for a crop for
each crop specific pesticide as a scaling factor. The
average total NPSP scaled up from the two sub-
catchments (1688 g) favourably agrees with the
estimated NPSP loads at the sampling site SP 1
(1726 g). The use of the arithmetic average is
somewhat arbitrary. But, as it was not known how
representative the sub-catchments were for the entire
catchment nor was the underlying distribution of sub-
catchments of which they were a part, this seemed a
reasonable procedure. The upscaling was only
Fig. 8. Comparison of the average pesticide load measured in the successful for the bulk of all pesticides (Fig. 8).
two sub-catchments and scaled up to the entire catchment to the Comparing the results of individual pesticides the best
estimated NPSP of the catchment (pesticide load SP 1—pesticide matches were found for chlortoluron (extrapolated
load SP 4—loads emitted by the combined sewer overflows).
load equalled 189% of measured NPSP), mecoprop-P
reactions. The pesticide contamination measured at (111%), metolachlor (61%) and isoproturon (248%).
the outlet of the catchment, integrated sources Using the area of a treated crop cannot be
distributed over all parts of upstream catchments. recommended as a suitable scaling factor for NPSP
The contaminated water moved downstream from the of surface waters. The variability of the application
contamination source to the sampling site. The times, rates and methods and the complex hydrology
pesticide concentration was thereby transformed by in the catchment are important factors for NPSP.
mixing with water from other parts of the catchment
with differing pesticide concentrations. Therefore, the
pattern of pesticide contamination over time also 4. Conclusions
depends on the position of the sampling site with
respect to the stream flow net even when soil types, Applying two fundamental different hydrograph
land use and management practises are similar separation techniques showed that the choice of the
throughout the entire catchment. Whether a pesticide separation method is crucial: daily base flow contri-
load was associated with the base or direct flow butions to total flow were significantly different
component on catchment scale, is due to the degree of applying the two algorithms. This should be con-
diluting and mixing processes of water of different sidered in future validations of hydrological models.
origins and with differing loads. Peak concentrations However, the choice of the hydrograph separation
were higher in the sub-catchments while the periods method did not influence the key message of the
with detectable pesticide concentrations were longer study: its impact on the hydrological classification of
for the entire catchment. Baker and Richards (1990) pesticide loads was negligible.
have described the effect of catchment size and how it The results emphasise the dynamic nature and
greatly affected the pattern of pesticide occurrence. scale dependency of pesticide transport. The
interaction between size and position of source
areas along a stream flow net is complex. Storm
3.2.3. Upscaling of NPSP generated flows exported the bulk of pesticides
Although the two sub-catchments were found to be and exhibited the highest concentrations at the
hydrologically different, the results were used for scale of the small sub-catchments. Therefore,
scaling the measured NPSP up to the entire catch- measures to reduce pesticide NPSP should focus
ment. The two sub-catchments are different in terms on controlling losses from critical source areas. The
of topography, soils and land use. Land use was pattern was more complex with base flow con-
hypothesized as key factor for the spatial variability of tributing most pesticide to the total export and
 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17 15

highest concentrations in direct flow for the entire Banchard, P.E., Lerch, R.N., 2000. Watershed vulnerability to
catchment. Dilution and hysteresis effects must be losses of agricultural chemicals: interactions of chemistry,
hydrology, and land-use. Environment Science and Technology
considered when water quality monitoring sites are 34, 3315–3322.
established within a catchment to measure pesticide Barcelo, D., Chiron, S., Fernandez-Alba, A., Valverde, A.,
contamination. The relationships of pesticide con- Aloendurada, M.F., 1996. Monitoring pesticides and metab-
centrations with flow component were very dis- olites in surface waters and groundwater in Spain. In: Meyer,
tinctive for different scales. Using the area of a M.T., (Ed.), Herbicides—Metabolites in Surface and Ground-
water, American Chemical Society, Washington, DC, pp.
treated crop cannot be recommended as a suitable
237 –253.
scaling factor for NPSP of surface waters. The Blöschl, G., Sivapalan, M., 1995. Scale issues in hydrological
variability of the application times, rates and modelling: a review. Hydrological Processes 9, 251–290.
methods throughout the catchment are important Brown, C.D., Hodgkinson, R.A., Rose, D.A., Syers, K., Wilcock-
for NPSP. son, S.J., 1995. Movement of pesticides to surface waters from a
heavy clay soil. Pesticide Science 43, 131–140.
Brusseau, M.L., 1994. Transport of reactive contaminants in
heterogeneous porous media. Reviews of Geophysics 32,
Acknowledgements 285 –313.
Buttle, J.M., 1990. Metolachlor transport in surface runoff. Journal
We thank Jürgen Mank for technical help using the of Environmental Quality 19, 531–538.
Caissie, D., Pollack, T.P., Cunjak, R.A., 1996. Variation in stream
model MOMENT 4.0 and Sabine Martis for technical water chemistry and hydrograph separation in a small drainage
help in the field. The German Ministry for Technol- basin. Journal of Hydrology 178, 137–157.
ogy and Education (BMBF) supported this study Carter, A.D., 1999. Pesticide contamination of water sources and
through funding number 0339568. the monitoring data across the EU. XI Symposium on Pesticide
Chemistry, La Goliardica Pavese, Cremona, Italy, 11– 20.
Chapman, T.A., 1999. A comparison of algorithms for stream flow
recession and baseflow separation. Hydrological Processes 13,
References 701 –714.
Der Rat von Sachverständigen für Umweltfragen, 1998. Flächen-
Albanis, T.A., Hela, D.G., Sakellarides, T.M., Konstantinou, I.K., deckend wirksamer Grundwasserschutz—Ein Schritt zur dauer-
1998. Monitoring of pesticide residues and their metabolites in haften Entwicklung. Sondergutachten, Metzler-Poeschel,
surface and underground waters of Imathia (N. Greece) by Stuttgart, Germany, in German.
solidphase extraction disks and gas chromatography. Journal of Donnelly, C.A., Ferrari, M.J., 2001. Summary of pesticide data
Chromatography, Part A 824, 59–71. from streams and wells in the Potomac River Basin. Open File
Anonymous, 1990a. Orthophoto maps of the whole Hessen Report 97-666, US Geological Survey.
territory. Hessisches Landesvermessungsamt, Wiesbaden. Dubrovsky, N.M., Kratzer, C.R., Brown, L.R., Gronberg, J.M.,
Anonymous, 1990b. Wasserbilanz Mittelhessen-Öffentliche Was- Burow, K.R., 1998. Water quality in the San Joaquin-Tulare
serversorgung-Stand (1989)/90. Schriftenreihe der Hessischen Basins, California, 1992–1995. Circular 1159, US Geological
Landesanstalt für Umwelt, Umweltplanung, Arbeits- und Survey.
Umweltschutz, Heft 110, in German. Dyck, S., Peschke, G., 1989. Grundlagen der Hydrologie, VEB
Anonymous, 1997. Dokumentation des EDV-Programmsystems Verlag für Bauwesen, Berlin, in German.
MOMENT, Vers. 4.0. Brandt-Gerdes-Sitzmann-Wasser- Eidem, J.M., Simpkins, W.W., Burkart, M.R., 1999. Geology,
wirtschaft GmbH, Darmstadt, in German. groundwater flow and water quality in the Walnut Creek
Arnold, J.G., Allen, P.M., Muttiah, R., Bernhardt, G., 1995. watershed. Journal of Environmental Quality 28, 60–69.
Automated base flow separation and recession analysis Felding, G., 1995. Leaching of phenoxyalkanoic acid herbicides
techniques. Ground Water 33, 1010–1018. from farmland. The Science of the Total Environment 168,
Auerswald, K., 1996. Jahresgang der Eintrittswahrscheinlichkeit 11 –18.
erosiver Starkregen in Süddeutschland. Zeitschrift für Kultur- Fenelon, J.M., Moore, R.C., 1998. Transport of agrichemicals to
technik und Landentwicklung 37, 81–84, in German. ground and surface water in a small central Indiana watershed.
Bach, M., Frede, H.G., 1996. Pflanzenschutzmittel in Journal of Environmental Quality 27, 884 –894.
Fließgewässern. Teil 1: Monitoring im Lahngebiet. Deutsche Fischer, P., 1996. Quantifizierung der Eintragspfade für Pflan-
Gewässerkundliche Mitteilungen 40 (3), 110– 114, in German. zenschutzmittel in Fließgewässer. PhD thesis, Justus-Liebig-
Baker, D.B., Richards, R.P., 1990. Transport of soluble pesticides Universität, Giessen, in German, with English abstract.
through drainage networks. In: Kurtz, D.A., (Ed.), Long Range Flury, M., 1996. Experimental evidence of transport of pesticides
Transport of Pesticides, Lewis Publishers, Chelsea, MI, pp. through field soils: a review. Journal of Environmental Quality
241–270. 25, 25–45.
16 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17

Flury, M., Flühler, H., Jury, W.A., Leuenberger, J., 1994. Léon, L.F., Soulis, E.D., Kouwen, N., Farquhar, G.J., 2001.
Susceptibility of soils to preferential flow of water: a field Nonpoint source pollution: a distributed water quality modeling
study. Water Resources Research 30 (7), 1945–1954. approach. Water Resources 35, 997 –1007.
Frank, R., Braun, H., van Hove Holdrinet, M., Sirons, G.J., Ripley, v. Loewenstern, S., 1998. Separierung und Bewertung von
B., 1982. Agricultural and water quality in the Canadian Great Abflusskomponenten für den Stoffaustrag aus Einzugsgebieten
Lakes Basin: V. Pesticide use in 11 agricultural watersheds and mit Tertiärsedimenten (Scheyern, Oberbayern). PhD thesis,
presence in stream water. Journal of Environmental Quality 11, Ludwig-Maximiliams-Universität, München, in German.
497–505. Loll, P., Moldrup, P., 2000. Stochastic analyses of field-scale
Garbrecht, J., Martz, L.W., 1995. TOPAZ—an automated digital pesticide leaching risk as influenced by spatial variability in
landscape analysis tool for topographic evaluation, drainage physical and biochemical parameters. Water Resources
identification, watershed segmentation and subcatchment para- Research 36 (4), 959– 970.
metrization. National Agricultural Water Laboratory, USDA, Lyne, V., Hollick, M., 1979. Stochastic time-variable rainfall–
Durant, OK. runoff modeling. IE Austr. Nat. Conf. Publ., Inst. of Eng,
Harman-Fetcho, J.A., McConnell, L.L., Baker, J.E., 1999. Agri- Canberra, pp. 89 –93.
cultural pesticides in the Patuxent river, a tributary of the Maniak, U., 1993. Hydrologie und Wasserwirtschaft—Eine Ein-
Chesapeake Bay. Journal of Environmental Quality 28, führung für Ingenieure, Springer, Berlin, in German.
928–938. McDonnell, J.J., 1991. Preferential flow as a control of stormflow
Hartmann, H., Burhenne, J., Spiteller, M., 1998. Trace determi- response and water chemistry in a small forested watershed. In:
nation of pesticides in water by coated capillary micro Gish, T.J., Shirmohammadi, A. (Eds.), Preferential Flow,
extraction (CCME) and reversed-phase high performance American Society of Agricultural Engineers.
liquid cromatography. Fresenius Environmental Bulletin 7, McDowell, R., Sharpley, A., Folmar, G., 2001. Phosphorus export
96–102. from an agricultural watershed: linking sources and transport
Heathwaite, A.L., Dils, R.M., 2000. Characterising phosphorus loss
mechanisms. Journal of Environmental Quality 30, 1587–1595.
in surface and subsurface hydrological pathways. The Science
Müller, K., 2000. Diffuse and punktuelle Pflanzenschutzmittel-
of the Total Environment 251– 252, 523–538.
Einträge in ein Fließgewässer: Messungen und Modellierung.
Herschy, R.W., 1995. Streamflow Measurement, E&FN Spon,
PhD thesis, Justus-Liebig University of Giessen, Giessen, in
London.
German, with English abstract.
Hornsby, A.G., 1992. Site-specific pesticide recommendation: the
Müller, K., Bach, M., Hartmann, H., Spiteller, M., Frede, H.-G.,
final step in environmental impact prevention. Weed Technol-
2002. Point- and nonpoint-source pesticide contamination in the
ogy 6, 736–742.
Zwester Ohm catchment, Germany. Journal of Environmental
Huber, A., 1998. Belastung der Oberflächengewässer mit Pflan-
Quality 31, 309–318.
zenschutzmitteln in Deutschland—Modellierung der diffusen
Muttiah, R.S., Allen, P.M., Arnold, J., Srinivasan, R., 1994.
Einträge. PhD thesis, Giessen, in German, with English abstract.
Baseflow Characteristics of Texas, Toxic Substances and
Isenberg-Schröter, M., Kofold, M., König, B., Schramm, T.,
Bedbur, E., Mattheß, G., 1998. Auftreten von Pflanzenschutz- Hydrologic Sciences, American Institute of Hydrology.
mitteln in Oberflächengewässern und im Grundwasser. Grund- Nathan, R.J., McMahon, T.A., 1990. Evaluation of automated
wasser 2, 57 –66, in German. techniques for base flow and recession analyses. Water
Janssen, A., Keuter, M., zu Altenschildesche, D., Ritzkopf, A., Resources Research 26 (7), 1465–1473.
Treder, W., 1995. Chromatographische Pestizidbestimmung Ng, H.Y.F., Clegg, S.B., 1997. Atrazine and metolachlor losses in
entsprechend der Trinkwasserverordnung—Teil I: Anreich- runoff events from an agricultural watershed: the importance of
erung mittels Festphasenextraktion. GIT Fachzeitschrift Labor runoff components. The Science of the Total Environment 193,
5, 417 –424, in German. 215 –228.
Kreuger, J., Törnqvist, L., 1998. Multiple regression analysis of Noguchi, S., Tsuboyama, Y., Sidle, R.C., Hosoda, I., 1999.
pesticide occurrence in streamflow related to pesticide proper- Morphological characteristics of macropores and the
ties and quantities applied. Chemosphere 37, 189–207. distribution of preferential flow pathways in a forested slope
Kung, K.J.S., Steenhuis, T.S., Kladviko, E.J., Gish, T.J., Bubenzer, segment. Soil Science Society of America Journal 63,
G., Helling, C.S., 2000. Impact of preferential flow on the 1413–1423.
transport of adsorbing and non-adsorbing tracers. Soil Science Nöhles, I., 2000. Landnutzungsklassifikation mit multitemporalen
Society of America Journal 64, 1290–1295. Landsat TM-Szenen in einer kleinstrukturierten Agrarregion.
Kupfersberg, H., Blöschl, G., 1995. Estimating aquifer transmissiv- PhD thesis, Justus-Liebig-Universität, Giessen, in German, with
ities—on the value of auxiliary data. Journal of Hydrology 165, English abstract.
85–99. Papula, L., 1998. Mathematische Formelsammlung für Ingenieure
Larson, S.J., Gilliom, R.J., Capel, P.D., 1999. Pesticides in streams und Naturwissenschaftler. Friedr. Vieweg und Sohn, Braun-
of the United States—Initial results from the National Water- schweig/Wiesbaden, in German.
Quality Assessment Program. US Geological Survey Water- Pedersen, M.L., Kronvang, B., Larsen, S.E., Müller-Wohlfeil, D.I.,
Resources Investigations Report 98-4222, US Geological 1999. Catchment-scale analysis of hydrology and nutrient
Survey. pathways. 24th General Assembly Hydrology, Oceans and
 K. Müller et al. / Journal of Hydrology 273 (2003) 1–17 17

Atmosphere, Geophysical Research, European Geophysical linear and nonlinear sorption. Water Resources Research 31 (4),
Society. 811 –822.
Perret, J., Prasher, S.O., Kantzas, A., Hamilton, K., Langford, C., Thurman, E.M., Goolsby, D.A., Meyer, M.T., Koplin, D.W., 1991.
2000. Preferential solute flow in intact soil columns measured Herbicides in surface waters of the Midwestern United States:
by SPECT scanning. Soil Science Society of America Journal the effect of spring flush. Environmental Science and Technol-
64, 469 –477. ogy 25, 1794–1796.
Pionke, H.B., Gburek, W.J., Sharpley, A.N., Schnabel, R.B., Tisseau, M.A., Fauchon, N., Cavard, J., Vandevelde, T., 1996.
1996. Flow and nutrient export patterns for an agricultural Pesticide contamination of water resources: a case study—the
hill-land watershed. Water Resources Research 32, rivers in the Paris Region. Water Science and Technology 34,
1795–1804. 147 –152.
Press, W.H., Teukolsky, S.A., Vetterling, W.T., Flannery, B.P., Tomlin, C., 1994. The Pesticide Manual, The Royal Society of
1992. Numerical recipes in C, Cambridge University Press, Chemistry, Farnham.
Cambridge.
Traub-Eberhard, U., Henschel, K.P., Kördel, W., Klein, W., 1995.
Rawn, D.F.K., Halldorson, T.H.J., Turner, W.N., Woychuk, R.N.,
Influence of different field sites on pesticide movement into
Zakrevsky, J.G., Muir, D.C.G., 1999. A multi-year study of four
subsurface drains. Pesticide Science 43, 121– 129.
herbicides in surface water of a small prairie watershed. Journal
US Department of Agriculture, 1972. National Engineering Hand-
of Environmental Quality 28, 906 –917.
book, USDA, Washington, DC.
Rittweger, H., 1997. Spätquartäre Sedimente im Amöneburger
Wang, Z., Feyen, J., 1998. Susceptibility and predictability of
Becken. PhD thesis, Philipps-Universität Marburg, Marburg, in
German. conditions for preferential flow. Water Resources Research 34
Seel, P., Knepper, T.P., Gabriel, S., Weber, A., Haberer, K., 1994. (9), 2169–2182.
Einträge von Pflanzenschutzmitteln in ein Fließgewässer— Wauchope, R.D., Williams, R.G., Marti, L.R., 1990. Runoff of
Versuch einer Bilanzierung. Vom Wasser 83, 357 – 372, sulfometuron-methyl and cyanazine from small plots: effects of
in German. formulation and grass cover. Journal of Environmental Quality
Squillace, P.J., Thurman, E.M., 1992. Herbicide transport in rivers: 14, 132–136.
importance of hydrology and geochemistry in nonpoint-source Wittenberg, H., 1999. Baseflow recession and recharge as nonlinear
contamination. Environmental Science and Technology 26 (3), storage processes. Hydrological Processes 13, 715 –726.
538–545. Woods, R., Sivapalan, M., Duncan, M., 1995. Investigating the
Streck, T., Poletika, N.N., Jury, W.A., Farmer, W.J., 1995. representative elementary area concept: an approach based on
Description of simazine transport with rate-limited, two-stage, field data. Hydrological Processes 9, 291–312.