Content uploaded by Traugott J Scheytt
Author content
All content in this area was uploaded by Traugott J Scheytt
Content may be subject to copyright.
A New Approach to Calculate EMEA’s Predicted
Environmental Concentration for Human Pharmaceuticals
in Groundwater at Bank Filtration Sites
Beate Müller & Traugott Scheytt &
Matthias Zippel & Stephan Hannappel &
Jutta Klein-Goedicke & Klaus Duscher
Received: 14 January 2010 /Accepted: 20 July 2010
#
Springer Science+Business Media B.V. 2010
Abstract In recent years, human pharmaceutical
substances have been increasingly detected in the
aquatic environment. Specific attention has been
drawn to the occurrence of pharmaceutical substan-
ces at bank filtration sites which are used for
drinking water production. In the course of the
authorisation application for new pharmaceutical
compounds, an environmental risk assessment is
required. Currently, the expected concentration of
the human pharmaceutical compound in groundwa-
ter at bank filtration sites is calculated following the
guideline Pre-Authorisation Evaluation of Medicines
for Human Use issued by the European Medicines
Agency (EMEA 2006). A simple estimation is
applied: The predicted environmental concentration
(PEC
GW
) is the predicted environmental concentra-
tion in surface water (PEC
SW
) multi plied w it h 0 .25.
A new approa ch c onsidering the hydrau lic an d
hydrogeological characteristics of bank filtration
sitesaswellastransportprocessesispresentedin
this study. First, a numerical groun dwater flow
model was developed to simulate the groundwater
flow processes a t bank filtration sites in general.
Flow times were calculated as a function of the
hydraulic and hydrogeological parameters: hydraulic
conducti vi ty, shore -w el l dist anc e , scree n depth and
extraction rate. In a second step, t he PEC
GW
was
calculated based on the compound concentration in
surface water and the modelled groundwater flow
times considering linear sorption and first-order
decay. Sorption and degradation can only be calcu-
lated based on the data provided by the pharmaceu-
tical company in the course of the authorisation
application. The current a pproach following t he
EMEA gu ide line invariably connects t he PEC
GW
with the PEC
SW
without considering sorption and/or
degradation processes. We introduce an approach
that incorporates the hydraulic process bank filtra-
tion and the main transport processes sorption and
degradation. The new approach is compound specific
as well as aquifer, flow and transport specific
resulting in a more realist ic PEC
GW
value compare d
to the old approach.
Water Air Soil Pollut
DOI 10.1007/s11270-010-0568-9
The use of brand names in peer-reviewed papers is for
identification purposes only and does not constitute endorse-
ment by the authors and their employers.
B. Müller (*)
:
T. Scheytt
Hydrogeology Research Group, Department of Applied
Geosciences, Technical University Berlin,
Ackerstraße 71-76,
13355 Berlin, Germany
e-mail: beate@mailbox.tu-berlin.de
M. Zippel
:
S. Hannappel
:
K. Duscher
HYDOR Consult GmbH,
Am Borsigturm 40,
13507 Berlin, Germany
J. Klein-Goedicke
Federal Environment Agency,
Wörlitzer Platz 1,
06844 Dessau, Germany
Keywords Predicted environmental con centration
(PEC)
.
Groundwater flow model
.
Sorption
.
Residence
time
.
Surface water
.
First-order decay
1 Introduction
Since the early 1990s, human pharmaceutical substan-
ces have been detected in the aquatic environment in
surface, ground and drinking water as well as in
sewage effluent (Engelmann and Rohde 2009;Halling-
Sorensen et al. 1998;Heberer2002a). Pharmaceutical
compounds reach the groundwater via different paths
(Fig. 1). A significant amount of pharmaceuticals is
excreted as parent products or their metabolites;
another part is directly disposed into the toilet. In both
cases, pharmaceuticals end up in the sewage system.
Leakages may lead to a direct infiltration of sewage
into the groundwater (Fenz et al. 2005).
In many cases, pharmaceutical compounds are not
completely eliminated in sewage treatment plants
(STPs), and the substances are discharged after
treatment into the surface water (Halling-Sorensen et
al. 1998; Heberer 2002b). During influent aquifer
conditions and at bank filtration sites, pharmaceutical
substances can reach the groundwater. If this ground-
water is used for drinking water purposes, pharma-
ceuticals may be detected in drinking water. Heberer
(2002a) reported pharmaceutically active compounds
in groundwater and drinking water samples from
water works using bank filtrat ion or artificial ground-
water recharge downstream from municipal STPs.
The application of sewage sludge in agriculture, the
irrigation of sewage and the disposal of pharmaceu-
tical substances at landfills are further source s for
(human) pharmaceuticals in groundwater.
During the process of bank filtration, surface water
infiltrates the aquifer and the filtrate is extracted at
groundwater extraction wells for drinking water
production. In Germany, approximately 16% of the
drinking water is produced from bank filtration
(Schmidt et al. 2004). Research studies at bank
filtration sites in the Berlin area show that some
pharmaceutical substances such as carbamazepine or
primidone are not removed during the subsurface
passage (Massmann et al. 2007).
The European Commission has stated in the
Council Directive 2001/83/EC (EC 2001) that an
environmental impact assessment has to be performed
in the course of the mark eting authorisation applica-
tion for new medicinal products for human use. This
finally led to the Guideline on the environmental risk
assessment of medicinal products for human use
published by the European Medicines Agency
(EMEA 2006). However, the studies during the
marketing authorisation application focus in particular
on toxicological issues such as acute, subacute and
chronic toxicity of adverse effects for a new pharma-
ceutical substance on the human body (von Keutz and
Jekat 1998). The guideline states that an environmen-
tal impact should not constitute a criterion for refusal
of a marketing authorisation of medicinal products for
human use (EMEA 2006). Furthermore, for t he
pharmaceutical substances, which already have a
marketing authorisation, an environmental risk assess-
ment is not requested by the regulatory agencies.
The EMEA guideline is based on the comparison
between the predicted environmental concentration
(PEC) and the predicted no-effect concentration
(Liebig et al. 2006). The stepwise assessment includes
the estimation of exposure, the initial prediction of
risk, the substance and compartment-specific refine-
ment and the risk assessment. The first phase (phase
I) comprises the estimation of exposure and includes
the estimation of the PEC
SW
. The trigger for the
performance of a phase II is 0.01 μg/L for the PEC
SW
.
The second phase (phase II) is the environmental risk
assessment which is divided in two parts, so-called tier
A and tier B. Tier A is the initial prediction of risk,
based on a base set of aquatic toxicology and fate data.
Tier B includes a substance and compartment-specific
refinement and risk assessment, based on an extended
data set on emission, fate and effects.
Within the phase II of the environmental risk
assessment following the EMEA guideline, an input
into groundwater is assumed via bank filtration
(EMEA 2006). Therefore, a predicted environmental
concentration in groundwater (PEC
GW
) has to be
calculated. The PEC
GW
is based on the predicted
environmental concentration in surface water (PEC
SW
)
using the simple equation
PEC
GW
¼ 0:25 PEC
SW
ð1Þ
In this equation, the PEC
GW
is invariably connected
to the PEC
SW
without considering groundwater flow
situation or transport behaviour of the pharmaceutical
compound. Thus, the aim of the present study was to
Water Air Soil Pollut
find a compound-specific approach of calculating the
PEC
GW
, based on general hydraulic and hydrogeolog-
ical conditions at bank filtration sites and the limited
sorption and degradation parameters provided by the
pharmaceutical companies in the course of an applica-
tion for marketing authorisation for new pharmaceuti-
cal substances (Zippel et al. 2009).
2 Transport Processes During Bank Filtration
Bank filtration is a specifically designed technique to
produce drinking water. Groundwater extractio n wells
are installed on the shore of surface water bodies, and
an artificial gradient between surface water and
groundwater level induces infiltration of surface water
into the aquifer. The physical and chemical transport
processes in the aquifer improve the quality of the
filtrate (Ray et al. 2002; Hiscock and Grischek 2002).
Due to its nature, the process of bank filtration is
limited to the strip between the shore line and the
extraction well and to sites comprising a porous
aquifer with good hydraulic conductivity. In central
Europe, the majority of river bank filtration plants are
located along the rivers Danube, Elbe and Rhine (Ray
et al. 2002). In Germany, bank filtration is success-
fully applied for more than 100 years, mostly at the
rivers Rhine, Elbe, and in the Berlin area at the rivers
Havel and Spree and several lakes (Ec kert and
Irmscher 2006; Fritz 2003; Lenk et al. 2006).
Fig. 1 Input paths into the aquatic environment for human pharmaceuticals
Water Air Soil Pollut
The occurrence and concentration of pharmaceutical
compounds in groundwater wells depend not only on
the concentrations in river and its fluctuation but also
on the advective transport of the substance and on the
reactions along the flow path. The main attenuation
processes during bank filtration are diffusion, dilution,
dispersion, mixing, biodegradation and sorption occur-
ring within the colmation layer and the aquifer itself
(Hiscock and Grischek 2002).
The ratio between the concentration sorbed onto the
aquifer sediment and the dissolved solute concentration
in water at equilibrium conditions is referred to as a
sorption isotherm. The simplest case of an isotherm with
a linear correlation between the dissolved and sorbed
substance concentration is described by the Henry
isotherm.
c
sorb
¼ K
d
c
w
ð2Þ
c
sorb
concentration of the substance sorbed onto
solid (milligrams per kilogram)
c
w
substance concentration in groundwater
(milligrams per litre)
K
d
Henry sorption coefficient (litres per kilogram)
Sorption depends on the organic carbon content of the
aquifer material. By normalizing the sorption coeffi-
cient with the organic carbon fraction, a significant
reduction of variability is achieved applying the
equation
K
OC
¼
K
d
f
OC
ð3Þ
K
OC
partition coefficient with respect to the organic
carbon fraction (–)
f
OC
organic carbon fraction (weight percent)
Scheytt et al. (2006) have shown that even at low
organic carbon contents in the aquifer sediment,
sorption processes are relev ant for the transport of
pharmaceutical substances. Beside the organic carbon
content of the aquifer material, other parameters such
as the mineral content of the sediment or the
physicochemical properties of the groundwater (pH,
ionic strength) are important for sorption processes.
Retardation is the reduced mobility of a solute in the
aquifer at a velocity less than that of the flowing
groundwater due to sorption processes. The retardation
factor is described by the equation
R
f
¼ 1 þ
1 nðÞ
n
r
S
K
d
ð4Þ
R
f
retardation factor (–)
n porosity (–)
ρ
S
solid density (grams per cubic centimetre)
Since only few data on the fate of human pharmaceuticals
in the aquifer are a vailable, a focus of the investigations
during the past years was the degradation behaviour of
pharmaceutical substances. Several laboratory and field
studies on the degrad ation behaviou r of pharmaceutica l
substances for human use in the saturated and unsaturated
aquifer zone applying first-order kinetics were performed
(Andreozzi et al. 2004; Kunkel and Radke 2008;
Loeffler et al. 2005; Scheytt et al. 2006). Loeffler et
al. (2005) investigated the environmental fate of ten
selected pharmaceuticals in water/sediment systems
including both the analysis of water and sediment.
Due to the lack of sufficient data, degradation of
pharmaceutical substances in groundwater is often
described using a first-order kinetic equation.
c ¼ c
0
e
l t
compound
ð5Þ
l ¼
ln 2
DT
50
ð6Þ
c concentration at production well
(nanograms per litre)
c
0
surface water concentration (nanogr ams
per litre)
λ decay constant (1 per day)
t
compound
residence time for the compound (days)
DT
50
disappearance time of 50% of the
compound; here used as half-life time of
the compound (days)
3 Method
The general approach for t he development of a
calculation tool for PEC
GW
was a stepwise process.
Water Air Soil Pollut
Starting point is the PEC
SW
, the initial pharmaceutical
concentration in surface water. PEC
SW
is calculated by
applying a formula provided in the EMEA guideline,
which incorporates the maximum daily dose consumed
per inhabitant, the percentage of market penetration,
the amount of wastewater per inhabitant per day and a
dilution factor.
The aim of the first phase of the study was to
construct a groundwater flow model, which does not
display one specific site but bank filtration sites in
central Europe in general. The groundwater flow model
has to consider the geometry of bank filtration sites as
well as hydraulic and hydrogeological parameters.
In the second phase, transport processes were
considered. The dat a provided in the application
process include values on K
d
, K
OC
and DT
50
. With
these data, sorption and degradation occurring during
the transport of the pharmaceutical substance from the
shore line to the groundwater extraction well were
incorporated in the approach. The available data on
the environmental behaviour of new pharm aceutical
substances provided in the course of the marketing
authorisation define the limits of incorporating trans-
port processes. Finally, it was necessary to combine
the results of both phases and to develop a user-
friendly computer interface for the estimation of the
PEC
GW
considering the transport time, sorption and
degradation (Fig. 2).
3.1 Groundwater Flow Model for Bank Filtration Sites
Based on the evaluation results of available literature
(Grischek 2003; Massmann et al. 2008a; HYDOR
2004, 2007; Lenk et al. 2006) and unpublished data of
bank filtration sites in central Europe, the variety of
hydraulic and hydrogeological parameters typical for
bank filtration sites was compiled (Table 1). All bank
filtration sites are located in the vicinity of surface
water bodies, and the wells are screened in medium to
highly permeable unconsolidated sediments. There-
fore, geological and hydrogeological properties of the
aquifer and the hydraulic settings are quite compara-
ble at all bank filtration sites. The distance between
surface water body and the extraction wells varies
between 1.5 and 1.00 km, and screen depths reach
from 4 to 70 m. Extraction rates are often difficult to
acquire, but results from the Berlin area show typical
values between 500 and 5,000 m
3
/day per extraction
well (HYDOR 2004, 2007).
The above-mentioned characteristics that are quite
comparable at different bank filtration sites in central
Europe do not apply for groundwater flow conditions.
Note, data on groundwater flow times are generally
rare, although they are essential for describing the
solute transport in groundwater. Here, numerical
groundwater flow models are capable of calculating
these flow times and their variability.
In the present study, a steady-state groundwater
flow model was developed using the program code
VISUAL MODFLOW (Fig. 3). A horizontal discretization
with a 5×5-m grid and a vertical discretization with
20 model aquifer layers with a continuous thickness
of 5 m were applied. The river bed has a width of
30 m and was set as a river boundar y condition in the
groundwater flow model.
The parameters hydraulic conductivity, extraction
rate, depth of filter screen and well-shore distance for
European bank filtration sites, which characterize the
groundwater flow regime, were varied stepw ise
within defined ranges (Table 2).
The following parameters were set to default values:
& Length of the filter screen—10 m
This screen length is a mean value for the water
works in the Berlin area (HYDOR 2004, 2007).
The screen lengths are controlled by the saturated
thickness of the aquifer and the desired pump
design. Schubert (2002) reports screen lengths for
the Lower Rhine Valley of 10 to 15 m.
& Hydraulic conductivity of clogging layer—1×
10
−5
m/s for the river bed area between 0 and
20 m (near to the extraction well) and to 1×
10
−6
m/s for the river bed area between 21 and
30 m
These are a gain mean values from literature (e.g.
Massmann et al. 2008b; Grischek 2003; Doppler
et al. 2007).
& Total porosity—0.35, effective porosity—0.2
& The horizontal K
xx
and vertical K
zz
hydraulic
conductivities are assumed to be constant in each
layer. Their ratio ε =K
xx
/K
zz
is the anisotropy
coefficient. The anisotropy coefficient was set to
a value of 2
In contrast to the diffusive input of pesticides via
the land surface, the input of human pharmaceuticals
from the surface water can be regard ed as a linear
source. The effect of dilution is incorporated in the
PC-based decision matrix.
Water Air Soil Pollut
3.2 Analytical Calculation of PEC
GW
After the development of the groundwater flow model,
it was necessary to incorporate the transport processes
sorption and degradation occurring during the transport
of the pharmaceutical substance from the shore line to
the groundwater extraction well. The environmental
assessment according to the EMEA guideline only
includes the determination of the distribution coeffi-
cients K
d
, occasionally the Freundlich sorption iso-
therm K
F
, the partition coefficient normalized by the
organic fraction K
OC
and the disappearance time DT
50
for the system sediment/water.
Several method descriptions for the determination of
the sorption and degradation parameters are provided
by various organizations, for example, the Organiza-
tion for Economic Co-operation and Development
(OECD), the European Commission and the Interna-
tional Organization for Standardization.
The determination of the DT
50
value is described
in the OECD guideline 308 Aerobic and Anaerobic
Transformation in Aquatic Sediment Systems (OECD
2002). The OECD guideline 308 describes degrada-
tion tests in artificial water/sediment systems under
aerobic and/or anaerobic conditions. DT
50
,DT
75
and
DT
90
values can be derived from the transformation
investigations. The OECD guideline 106 Adsorption/
Desorption using a Batch Equilibrium Method,
published in 2000, describes adsorption and desorp-
tion experiments for the determination of the K
d
and
K
F
describing the mobility of chemical substances in
groundwater and soil (OECD 2000). During batch
tests as described in the OECD guideline 106, the soil
sample is spiked with an aqueous solution of the
chemical substance. Afterwards, the soil sample is
shook for a defined time. Finally, the distribution of
the chemical substance dissolv ed in water and
adsorbed onto soil particles is determined.
In general, the determination of sorption and
degradation parameters depends on the laboratory test
conditions (e.g. pH, sand, clay and organic carbon
fraction). The distribution coefficients are determined
at different pH values and in different soil materials as
well as DT
50
values for aerobic and anaerobic
conditions. K
d
and DT
50
values derived from labora-
tory tests following OECD or other guidelines are
difficult to apply to real bank filtration sit es. However,
if the laboratory tests for K
d
and DT
50
are performed
under conditions similar to those in the natural aquifer
system, these values come quite near to the natural
system. Characteristic physicochemical parameters and
aquifer material composition at four German bank
filtration sites are summarized in Table 3.
In first-order rate laws, the time needed to halve
the compound concentration is known as the half-life
of the reaction and is independent of the initial
Fig. 2 General approach for development of a calculation tool for PEC
GW
Parameter Range
Hydraulic conductivity kF (m/s) 1×10
−2
–1×10
−4
Distance between shoreline and extraction well (m) 1.5–1,200
Depth of filter screen (final depth; m) 4–70
Average extraction rates (m
3
/day) 500–5,000
Groundwater flow times (days) <1–>1,100
Table 1 Ranges of hydro-
geological and hydraulic
parameters at bank filtration
sites in central Europe
(Lenk et al. 2006)
Water Air Soil Pollut
concentration. In a more loose sense, the half-life
concept is also used for other types of reaction rates,
but then the value depends on the initial concentra-
tion. According to the OECD guideline, the half-life
time (t
0.5
or t
1/2
) is the time t aken for 50%
transformation of a test substance when the transfor-
mation can be described by first-order kinetics. It is
independent of the initial concentration. The disap-
pearance time 50 (DT
50
) is the time within which the
initial concentration of the test substance is reduced
by 50%.
Table 4 shows exemplarily a data sheet, which is
provided by the pharmaceutical company in the
course of the marketing authorisation application.
Data are only available on sorption and degradation
processes, the K
d
and K
OC
values for the quantifica-
tion of retardation processes and DT
50
for quantifica-
tion of degradation processes occurring during the
transport in the aquifer.
Based on the results of the groundwater flow
modelling of the first phase, the available sorption
and degradation parameters are used for the analytical
calculation of the PEC
GW
. The parameters listed in
Table 4 are the solely available data which are
provided in the course of the marketing authorisation
for a new pharmaceutical substance for the calculation
of PEC
GW
. The sorption coefficients were used for the
calculation of the pharmaceutical compound retardation
resulting in a longer residence time of the compound in
groundwater compared to the groundwater flow time.
Then, the final pharmaceutical compound concentration
was calculated analytically applying the first-order
decay equation (Eq. 5) and the retarded transport time
of the co mpound. The result is the co mpound
concentration (PEC
GW
) at the extraction well. Because
linear sorption and first-order degradation can be
calculated easily, they were not incorporated into a
transport model. However, the mathematical calculated
results were verified with the transport modelling
results (MT3DMS) and the analyzed pharmaceutical
compound concentrations for different human pharma-
ceuticals at two real bank filtration sites: Flehe at the
river Rhine and Torgau at the river Elbe.
The EMEA guideline sugges ts that substances with
a high K
OC
value (>10,000 L/kg) are retained in the
STP and will not reach the aquifer during the bank
Fig. 3 Sketch of the groundwater flow model displaying a bank filtration site
Table 2 Step sizes for the variables of the groundwater flow modelling
Variables Step sizes for modelling
Distance between shore line and extraction well (m) 1,000/500/300/100/50/25/5
Hydraulic conductivity (m/s) 1×10
−2
/5×10
−3
/3×10
−3
/1×10
−3
/5×10
−4
/1×10
−4
Depth to filter screen (m b.g.l.)
a
80–90/50–60/30–40/20–30/10–20/10–15
Extraction rate (m
3
/day) 5,000/2,000/1,000/750/500/250
a
Top of ground surface was set to a default value of 100 m, which is the upper model boundary
Water Air Soil Pollut
filtration process. It is therefore assumed that these
substances pose no risk for groundwater. The K
OC
value >10,000 L/kg was integrated as a threshold into
the calculation.
4 Results and Discussion
For verification of the groundwater flow model,
groundwater flow times were modelled for the bank
filtration sites Flehe at the Rhine River and Torgau/
Ost at the Elbe River and compared with measured
flow times. Furthermore, the quality of the model was
tested by calculating a multiple regression to obtain
information on the statistical correlation of the
hydraulic and hydrogeological parameters and the
groundwater flow time. Moreover, the analytical
results for human pharmaceuticals analyzed at the
two bank filtration sites were compared with the
calculation results from the approach discussed in this
paper. Finally, the Microsoft Access based application
tool for the calculation of PEC
GW
is introduced.
The results of the groundwater flow model were
compiled in a matrix with a total number of 1,290
data pairs displaying shortest flow times and the
drawdown at the extraction well depending on the
parameter hydraulic conductivity, shore line-well
distance, depth of the screened interval, and extraction
rate. The main statistical parameters are compiled in
Fig. 4.
The median value of the modelled flow times is
110 days, which coincides very well with flow times
from real bank filtration sites. A total number of 464
modelled values revealed groundwater flow times
<50 days. The minimum and maximum values of the
modelled groundwater flow times ar e 0.07 a nd
12,791 days. Groundwater flow times at real bank
filtration sites range between <1 and 1,100 days
(Table 1). It is important to point out that the model
displays all groundwater flow times, which are
hydraulically possible including extreme high and
low values. The mean value of 638 days displays this
high range of the modelled values. Therefore, the
median value gives a more realistic value.
For the two bank filtration sites Torgau/Elbe and
Flehe/Rhine, measured groundwater flow times and
pharmaceutical concentrations in surface water
and groundwater at the extraction well were taken
from Eckert and Irmscher (2004) and Grischek
(2003). The groundwater flow model was adjusted
for both sites according to the information from the
waterworks operator and literature. For the Flehe site,
a groundwater flow time of 30 days was modelled
compared with a flow time of 35 days, according to
information from the wat erworks operator (Eckert and
Irmscher 2004). For the Torgau site, a groundwater
flow time of 210 days was modelled compared to a
medium flow time of >150 days, according to
information in Grischek (2003).
The developed groundwater flow model is based
on the variation and combinati on of four different
parameters groundwater extraction rate, shore-well
distance, depth to filter screen and hydraulic conduc-
tivity. A multiple regression allows to assess the
relationship between one dependent variable (ground-
water flow time) and several independent variables
(filter depth, extraction rate, shore-well distance,
groundwater drawdown) at the same time. It is
suggested that the variation of the groundwater flow
time depends on the variation of several different
variables. The calculation revealed a squared multiple
correlation R
2
of 0.47. The four parameters (see
above) thus contribute to 47% o f variance of the
modelled groundwater flow times. The beta values of
0.63 for the shore-well distance, −0.24 for the
extraction rate and 0.22 for the filter depth are in
agreement with the correlation factors. From a
statistical point of view, a good correlation between
the four parameters and the modelled flow time is
achieved with the groundwater flow model . Neve r-
theless, other parameters also influence the ground-
water flow time, whi ch were not included in the
groundwater flow model, such as river water level,
water viscosity (temperature-related) etc.
Another possibilit y f or the illustration of the
correlation between the model results and the four
parameters is box plots displaying ranges for the
groundwater flow times depending on the individual
parameter. Figure 5 shows the groundwater flow time
as a function of two parameters. At high extraction
rates, the filter depth has only a small influence on the
groundwater flow time (Fig. 5a). At low extraction
rates and inc reasi ng depth to f ilter scr een, th e
groundwater flow times show a higher range (varia-
tion). Short flow times result from short shore-well
distances and low filter depths (Fig. 5b). In other
words, at larger shore-well distances and increasing
depth to filter screen, the modelled groundwater flow
Water Air Soil Pollut
times show a higher range. Finally, Fig. 5c shows that
for the specific boundaries of the model, the extrac-
tion rates have a higher influence on the groundwater
flow time than short shore-well distances. All three
box plots approve the different influence of the
parameters on the groundwater flow time. The shore
well distance seems to have the highest influence on
the flow time followed by the extraction rate and the
filter depth.
Following the statistical evaluation of the model
results, the mathematical calculation approach was
compared with analyzed real-world pharmaceutical
concentrations. Table 5 compares the analyzed and
calculated concentrations for the pharmaceutical sub-
stances carbamazepine (antiepileptic drug) and diclofe-
nac (analgesic, antiarthritic, antirheumatic compound)
at the bank filtration sites Flehe and Torgau.
For the transport modelling as well as the analytical
calculation, the maximum analyzed concentration of
the pharmaceutical substance was chosen as the initial
concentration in surface water (PEC
SW
)inorderto
represent the worst case. The information on distance
between shore line and extraction well and on
groundwater flow time were taken from the results of
the groundwater flow modelling, from information
made available by the waterworks operator or from
literature. The DT
50
and the sorption distribution
coefficient were taken from literature.
For this comparison of actually analyzed concen-
trations and calculated values, diclofenac and carba-
mazepine were c hosen as example compounds
because they occur not only in surface water but also
in groundwater occasionally. Due to the higher
persistency of carbamazepine, documented by the
higher half-life time compared to diclofenac, carba-
mazepine has been detected in groundwater samples
from extraction wells. Likewise, the modelled concen-
trations are also above detection limits for those wells.
For diclofenac, modelled values are below the detec-
tion limit of 2 ng/L in the groundwater at the extraction
wells which coincides very well with analyzed values.
Overall, there are significant differences between
the analyzed and modelled concentrations. These
differences are mainly due to high variations in
concentration in surface water and also due to specific
local conditions with respect to the input history,
geology and hydrogeology. Albeit these restrictions ,
the comparison of the analyzed and calculated
concentration s shows that the appro ach for the
calculation of PEC
SW
by combining groundwater
flow modelling and analytical calculation of retarda-
tion and degradation is a good approximation for the
expected concen trations in groundwater. With this
tool, it will be possible to identify compounds that
have the potential to occur in groundwater even
before they are administered.
Table 3 Groundwater and aquifer sediment characteristics of four German bank filtration sites
Berlin—Tegeler See Berlin—Wannsee Flehe—Rhine Torgau/Ost—Elbe
Aquifer material/
soil texture (clay
content)
Fine- to coarse-grained
medium sand
Fine- to coarse-grained
medium sand
Sandy gravel (Eckert
and Irmscher 2004)
Fine gravel to medium
sand (Grischek 2003)
Natural organic
matter content
aquifer sediment
Total organic carbon
content 0.0–2.1 wt.
% (Massmann et al.
2007)
Total organic carbon
content 0.2–10 wt.%
(Massmann et al.
2008a)
Total carbon content
ranges between
0.02 and 3.46 wt.%
(Eckert 2003,
unpublished results)
Total organic carbon
content ranges
between 0.013 and
0.024 wt.%
(Grischek 2003)
Redox condition
groundwater
Aerobic (upper aquifer) Aerobic to anaerobic
condition (Massmann
et al. 2008a)
Aerobic to denitrifying
(Schmidt et al. 2004)
Denitrifying conditions
(anaerobic conditions)
(Grischek 2003)
Anaerobic (lower aquifer;
Scheytt et al. 2004)
pH value
groundwater
7.4 (mean value;
KWB 2007,
unpublished results)
7.4 (mean value;
KWB 2007,
unpublished results)
∼7.2–7.8 (Eckert and
Irmscher 2004)
6.5–7.4 (Grischek 2003)
Temperature
groundwater (°C)
10.7 (mean value; KWB
2007, unpublished
results)
11.1 (mean value;
KWB 2007,
unpublished results)
13.3–14.2 (Eckert 2003,
unpublished results)
8.0–12.0 (Grischek 2003)
Water Air Soil Pollut
Table 4 Data sheet with sorption and degradation parameters
Pharmaceutical
substance
Sorption Degradation
K
OC
(mL/g) K
d
(mL/g) K
F
Characterization test soil DT
50
water
aerobic
(days)
DT
50
water
anaerobic
(days)
DT
50
sediment
aerobic
(days)
DT
50
sediment
anaerobic
(days)
DT
50
soil
(days)
Characterization
pH % Sand % O.C.
Substance A 25 0.568 5.7 77.5 2.29 0.9 Sandy loam
pH water 7.98
pH sediment 7.11
146 1.812 6.8 48.4 1.24 1.1 Silty clay loam
pH water 8.17
pH sediment 7.19
381 7.235 6.7 22.1 1.90 3.06 Silt loam
pH water 7.09
pH sediment 7.13
151 2.202 7.2 14.5 1.46
K
OC
study acc. OECD 106, April 2007 Water/sediment study valid acc. OECD 308, May 2007
Substance B 55,800 4.3 86 0.5 1 2.8 (whole system) Sand
acc. OECD 106, cannot be validated, study report not available, 2007 acc. OECD 308, cannot be validated, study report not available, 2007
Substance C 2,650 957 1,350 n.s.
a
Sludge 36.1 11 11 Sand
pH 7.6 resp. 6.5
K
OC
study valid acc. OECD 302A, July 2003 Water/sediment study valid acc. OECD 308, April 2006
a
Not specified
Water Air Soil Pollut
Three standard flow times were chosen from the
statistical evaluation of the flow model results
representing the realistic worst case, the worst case
and the median case and correlated with the hydraulic
and hydrogeological conditions at real bank filtration
sites (Lenk et al. 2006). The shortest flow times result
from a combination of small depth to screen, small
well-shore-distance and high extraction rate.
The 5/95 percentile was chosen to represent the worst
case scenario resulting from short shore-well distance
(13 m), high extraction rates of about 3,000 m
3
/day and
high hydraulic conductivity (0.02 m/s). The realistic
worst case is represented by the 20/80 percentile with a
depth to filter screen of 24 m, shore-well distance of
33 m, extraction rate of 2,500 m
3
/day and high
hydraulic conductivity of 0.0069 m/s. Finally, the
median case is characterized by the median values of
the modelled flow times. The three scenarios do not
represent an individual concrete bank filtration site.
Based on these three cases, the groundwater flow time
was modelled with the groundwater flow model.
Table 6 summarizes the results.
The final step was to develop a PC-based, user-
friendly application tool, based on the results of the
groundwater flow model and the calculation routines for
sorption and degradation. The application applies
Microsoft Access and is named “SiMBaFi” (Simulation
Model Bank Filtration). The user has two different
options for the estimation of PEC
GW
(Fig. 6).
Fig. 4 Descriptive statistics of the modelling results (range of
the modelled shortest flow times 0–2,000 days)
a
b
c
2500
2250
2000
1750
1500
1250
1000
750
500
250
0
2000
1750
1500
1250
1000
750
500
250
0
7000
6000
250
extraction rate [m
3
/d]
500
8 074
12 791
750
1000
2000
5000
5000
4000
3000
2000
1000
0
groundwater flow time [d]groundwater flow time [d]groundwater flow time [d]
shore-well distance [m]
depth of filter screen [m]
depth of filter screen [m]
5 10203050
5 10203050
5 25 50 100 300
5
2 142
3 368
shore-well distance [m]
25
50
100
300
250
extraction rate [m
3
/d]
500
3 368
750
1000
2000
5000
Fig. 5 Box plots showing the relationships between ground-
water flow time and hydraulic/hydrogeological parameters.
a Influence of different extraction rates and filter depths.
b Influence of different shore-well distances and filter depths
(maximum flow time 2,000 days). c Influence of different
extraction rates and shore-well distances (maximum flow time
2,500 days)
Water Air Soil Pollut
First, PEC
GW
values can be calculated with three
default flow times representing the worst (0.15 days),
the realistic worst (5 days) and the median (110 days)
case. The calculation routines will only start when a
K
OC
<10,000 L/kg is entered. The K
OC
serves as a
threshold value, and an input is optional. The
calculation with a K
OC
>10,000 L/kg is not feasible
following the EMEA guideline, as substances with an
average K
OC
>10,000 L/k g are regard ed to be
immobile. In the case of an input of a K
OC
>10,000
L/kg, no further calculation will be performed.
It is furthermore possible to modify the results by
choosing the percentage of bank filtrate. The calcu-
lated concentration at a flow time of 5 days (realistic
Table 5 Comparison of calculation, flow model and analysis results
Bank filtration site Flehe/Rhine Flehe/Rhine Torgau/Elbe Torgau/Elbe
Pharmaceutical substance Carbamazepine Diclofenac Diclofenac Carbamazepine
Substance property
Sorption distribution coefficient K
d
(from literature) mL/g 0.131
a
0.572
c
0.572
c
0.131
a
Retardation factor (calculated) 2.30 6.70 6.70 2.30
Half-life time (literature) days 328
b
45
a
45
a
328
b
Lambda (calculated) 1/day 0.002113 0.0154 0.0154 0.002113
Initial concentration surface water
(maximum concentration)
ng/L 200 110 130 340
Model parameter and boundary conditions
Shoreline-well distance m 65 65 300 300
Depth to filter screen m b.g.l. 10–17 10–17 35–55 35–55
Extraction rate m
3
/day 840 840 3,600 3,600
Medium groundwater flow time days 35 35 >150 >150
Hydraulic conductivity k
F
m/s 0.001 0.001 0.002 0.002
Total porosity 0.35 0.35 0.35 0.35
Effective porosity 0.2 0.2 0.2 0.2
Solid density g/cm
3
2.65 2.65 2.65 2.65
Bulk density g/cm
3
2.0 2.0 2.0 2.0
Results
Modelled flow time days 33 33 210 210
Substance concentration at extraction well (analyzed) ng/L 190 b.d.l. b.d.l. 86
Substance concentration at extraction well (calculated) ng/L 177 14 0 164
b.d.l. below detection limit of 2 ng/L
a
Scheytt et al. 2006
b
Loeffler et al. 2005
c
Hanisch et al. 2004
Table 6 Definition of three standard scenarios
Parameter Worst case Realistic worst case Median case
Hydraulic conductivity (k
F
; m/s) 0.02 0.007 0.002
Extraction rate (m
3
/day) 3,000 2,000 1,000
Well-shore distance (m) 10 30 100
Depth of filter screen (metre below surface water level at 95 m) 5–10 10–20 20–30
Modelled time (days) 0.15 5 110
Water Air Soil Pollut
worst case) is referred as the standard case and should
be applied in the course of the authorisation process.
Second, PEC
GW
values can be calculated with
user-defined information on the flow time and/or the
hydrogeological and hydraulic parameter. The user
can either enter a groundwater flow time manually or
can choose values for the hydraulic and hydrogeo-
logical properties as screened interval, extraction rate,
hydraulic conductivity and surface water to well
distance from a dropdown menu. The respective flow
time is then recalled from the database, and finally,
the PEC
GW
values are calculated.
Mandatory input fields (operation):
- date - pharmaceutical compound
- brand name
- reference number - calculation mode
Output fields:
Calculation results of PEC
GW
for default flow times
Optional input of K
OC
No calculation possible
PEC
GW
= 0
Calculation of PEC
GW
Mandatory input fields (compound data):
- PEC
surface water
Optional: input of bank filtrate fraction in %
Standard calculation
Reduction of PEC
GW
due to dilution
Optional input of additional information:
- comment
- institution
- author
Optional output:
- protocol
- data matrix (backup)
Calculation of PEC
GW
at default flow times
User-defined calculation of PEC
GW
Input field:
Groundwater flow time for the distance
between shoreline and extraction well [days]
Output field:
Substance concentration in groundwater for
the indicated flow time based on the input
parameter made for the substance above
Input field:
Depth to filter screen, extraction rate,
hydraulic conductivity, distance between
shoreline and extraction well (based on data
matrix)
Output field:
Option 2: hydraulic and hydrogeological parameters are known
Option 1: groundwater flow time is known
Optional output:
- protocol
- data matrix (backup)
0.15 days
5 days 110 days
Groundwater flow time and substance
concentration in groundwater for the defined
parameter combination based on the input
parameter made for the substance above
- K
d
and DT
50
K
OC
>10 000
mL/kg
K
OC
<10 000
mL/kg
Fig. 6 Flow chart for application of SiMBaFi
Water Air Soil Pollut
5 Conclusions
A new approach to calculate EMEA’spredicted
environmental concentration for pharmaceuticals in
groundwater has been developed. The current approach
of EMEA considers bank filtration and calculates
PEC
GW
merely by multiplying the environmental
concentration of surface water (PEC
SW
) by 0.25. Using
this method, PEC
GW
is invariably connected to PEC
SW
without considering groundwater travel times or
chemical characteristics of the organic compounds.
However, in the course of the marketing authorisation
for new human pharmaceuticals, some limited data on
the characteristics of organic compounds have to be
provided by the applying pharmaceutical companies.
These data include the sorption coefficient K
d
and the
value for the disappearance time of the substance. For
the new approach, these data are utilized to character-
ize the basic transport behaviour of the new com-
pounds.
Although bank filtration is applied in Germany for
more than 100 years, only very little information on
groundwater residence times and the distribution of
residence time could be found. As bank filtration
sites exhibit quite distinctive patterns for geology,
hydrogeology and basic setup at the various sites,
a general groundwa ter flow model re vealed that
the median value for the residence time is
110 days. This might be a sufficiently long
residence time for attenuation processes at bank
filtration sites. However, a total number of 464
modelled values revealed g roundwater flow times
of less than 50 days in dicating surprisingly low
residence time in g roundwater at some locations
and along specific flow lines.
The main advantage of the new proposed
approach is that the PEC
GW
is not any longer
invariably connected to PEC
SW
but is calculated
using the limited albeit available information on the
chemical ch aracteristics of the organic compound s.
The two main processes that can be assessed using
the available data are sorption, i.e. retardation and
degradation/elimination.
Different c hemi cal propert ies lead to different
groundwater transp ort behaviour. In the case of
sorption, this is especially relevant for compounds
with very high or very low sorption coefficients. If the
sorption coefficient is high, the substance may not be
transported far in groundwater due to high retardation.
If the sorpt ion coefficient is very low, sorption can be
neglected and the substance is transported by advec-
tion with groundwater flow velocity.
In the case of degradation/elimination, again very
low or very high degradation exhibits a significant
difference compared to the current method. For sub-
stances which show low degradation, the calcula-
tion following the EMEA approach underestimates
the predicted environmental concentration in
groundwater.
The introduced calculation tool SiMBaFi is a
user-friendly tool for calculation of the predicted
environmental concentration for all compounds with
available data on sorption and degradation. It can be
easily refined to suit more specific demands. Although
the calculation is rather rough, it opens the possibility
for a fast check on the expected concentration of any
organic compound at bank filtration sites.
Acknowledgement This work was funded by the German
Federal Environment Agency (Umweltbundesamt). We would
like to thank the following institutions which provided
hydraulic and hydrogeological data on bank filtration sites as
well as concentration data of pharmaceutical substances in
groundwater: Kompetenzzentrum Wasser Berlin, Stadtwerke
Düsseldorf, Hochschule für Technik und Wirtschaft Dresden
(FH), Fernwasservers orgung Elbaue-Ostharz GmbH Torgau
and Technologiezentrum Wasser (TZW) Karlsruhe.
References
Andreozzi, R., Caprio, V., Ciniglia, C., De Champdore, M., Lo
Giudice, R., Marotta, R., et al. (2004). Antibiotics in the
environment: Occurrence in Italian STPs, fate, and pre-
liminary assessment on algal toxicity of amoxicillin.
Environmental Science & Technology, 38(24), 6832–6838.
Doppler, T., Franssen, H. J. H., Kaiser, H. P., Kuhlman, U., &
Stauffer, F. (2007). Field evidence of a dynamic leakage
coefficient for modelling river–aquifer interactions. Jour-
nal of Hydrology, 347(1–2), 177–187.
EC. (2001). Community code relating to medicinal products for
human use, Directive 2001/83/EC. Brussels: European
Commission, European Parliament and Council.
Eckert, P., & Irmscher, R. (2004). Modelling of hydrogeochem-
ical processes during bank filtration. In: K. Kovar, Z.
Hrkal, & J. Bruthans (Eds.), Proceedings of the interna-
tional conference on finite element models (pp. 193–196).
Karlovy Vary, Czech Republic: Fontainebleau, France,
International Association of Hydrologic Sciences.
Eckert, P., & Irmscher, R. (2006). Over 130 years of experience
with riverbank filtration in Düsseldorf, Germany. Journal
of Water Supply: Research and Technology—AQUA, 55
(4), 283–291.
Water Air Soil Pollut
EMEA. (2006). Guideline on the environmental risk assessment
of medicinal products for human use, Doc. Ref. EMEA/
CHMP/SWP/4447/00. London: European Medicines
Agency.
Engelmann, U., & Rohde, S. (2009). Pharmaceutical substances,
antibiotics and X-ray contrast media in treated waste water
and surface water in the Federal State Saxony (Arzneimittel-
wirkstoffe, Antibiotika und Röntgenkontrastmittel in Abwas-
sereinleitungen und Oberflächengewässern in Sachsen). KA
Korrespondenz Abwasser, Abfall, 56(3), 258–268.
Fenz, R., Blaschke, A. P., Clara, M., Kroiss, H., Mascher, D., &
Zessner, M. (2005). Monitoring of carbamazepine concen-
trations in wastewater and groundwater to quantify sewer
leakage. Water Science and Technology, 52(5), 205–213.
Fritz, B. (2003). Natural processes with a high efficiency: bank
filtration and water recharge (Natürliche Prozesse mit
hoher Effizienz: Uferfiltration und Wasseranreicherung).
wwt,10–11, 10–17.
Grischek, T. (2003). Management of bank filtration sites at the
Elbe River (Zur Bewirtschaftung von Uferfiltratfassungen
an der Elbe). Habilitation, Technische Universität Dresden
Institut für Grundwasserwirtschaft, Dresden.
Halling-Sorensen, B., Nielsen, S. N., Lanzky, P. F., Ingerslev, F.,
Lutzhoft, H. C. H., & Jorgensen, S. E. (1998). Occurrence,
fate and effects of pharmaceutical substances in the
environment—a review. Chemosphere, 36(2), 357–394.
Hanisch, B., Abbas, B., Kratz, W., & Schüürmann, G. (2004).
Human drugs in aquatic ecosystems: approach for the
environmental risk assessment of drug residues (Human-
arzneimittel im aquatischen Ökosystem: Bewertungsansatz
zur Abschätzung des ökotoxikologischen Risikos von
Arzneimittelrückständen). Umwe ltwiss enschaft en und
Schadstoff-Forschung, 16(4), 223–238.
Heberer, T. (2002a). Occurrence, fate, and removal of pharma-
ceutical residues in the aquatic environment: A review of
recent research data. Toxicology Letters, 131(1–2), 5–17.
Heberer, T. (2002b). Tracking persistent pharmaceutical resi-
dues from municipal sewage to drinking water. Journal of
Hydrology, 266(3–4), 175–189.
Hiscock, K. M., & Grischek, T. (2002). Attenuation of
groundwater pollution by bank filtration. Journal of
Hydrology, 266(3–4), 139–144.
HYDOR (2004). Hydrological evaluation of the groundwater
resources of the water works Tegel, Spandau, Tiefwerder,
Kladow and Beelitzhof in Berlin (Hydrologische Berech-
nungen zum Nachweis des Grundwasserdargebotes für die
Wasserwerke Tegel, Spandau, Tiefwerder, Kladow und
Beelitzhof der Berl iner Wasserbetriebe). Unpublished
report, Berliner Wasserbetriebe, Berlin.
HYDOR (2007). Hydrological evaluation of the groundwater
resources of the water works Friedrichshagen, Altglie-
nicke, Johannisthal, Wuhlheide, Kaulsdorf und Buch in
Berlin (Hydrologische Berechnungen zum Nachweis des
Grundwasserdargebotes für die Wasserwerke Friedrich-
shagen, Altglienicke, Johannisthal, Wuhlheide, Kaulsdorf
und Buch der Berliner Wasserbetriebe). Un published
report, Berliner Wasserbetriebe, Berlin.
Kunkel, U., & Radke, M. (2008). Biodegradation of acidic
pharmaceuticals in bed sediments: Insight from a labora-
tory experiment. Environmental Science & Technology, 42
(19), 7273–7279.
Lenk, S., Remmler, F., Skark, C., & Schulte-Ebbert, U. (2006).
Technical concepts and adjusted operational modes for an
optimal adaptation of riverbank filtration to local bound-
ary conditions (Technische Konzepte und abgestimmte
Betriebsweisen zur optimalen Anpassung der Uferfiltra-
tion an lokale Randbedingungen). Schwerte: Institut für
Wasserforschung GmbH Dortmund.
Liebig, M., Moltmann, J. F., & Knacker, T. (2006). Evaluation
of measured and predicted environmental concentrations
of selected human pharmaceuticals and personal care
products. Environmental Science and Pollution Research,
13
(2), 110–119.
Loeffler, D., Rombke, J., Meller, M., & Ternes, T. A. (2005).
Environmental fate of pharmaceuticals in water/sediment
systems. Environmental Science & Technology, 39 (14),
5209–5218.
Massmann, G., Heberer, T., Grützmacher, G., Dünnbier, U.,
Knappe, A., Meyer, H., et al. (2007). Drinking-water
production in urban environments—bank filtration in Berlin
(Trinkwassergewinnung in urbanen Räumen—Erkenntnisse
zur Uferfiltration in Berlin). Grundwasser, 12,232–245.
Massmann, G., Nogeitzig, A., Taute, T., & Pekdeger, A.
(2008a). Seasonal and spatial distribution of redox zones
during lake bank filtration in Berlin, Germany. Environ-
mental Geology, 54,53–65.
Massmann, G., Sültenfuß, J., Dünnbier, U., Knappe, A.,
Taute, T., & Pekdeger, A. (2008 b). Investigation o f
groundwater residence times during bank filtration in
Berlin: A multi-tracer approach. Hydrological Processes,
22, 788–801.
OECD. ( 2000). OECD guidelines for the testing of chemicals/
Section 1: Physical–chemical properties test no. 106:
Adsorption–desorption using a batch equilibrium method.
Paris: Organisation for Eco nomic Co -operation and
Development.
OECD. (2002). OECD guidelines for the testing of chemicals/
Section 3: Degradation and accumulation test no. 308:
Aerobic and anaerobic transformation in aquatic sediment
systems. Paris: Organisation for Economic Co-operation
and Development.
Ray, C., Melin, G., & Linsky, R. B. (Eds.). (2002). Riverbank
filtration—improving source-water quality. Dordrecht:
Kluwer Academic.
Scheytt, T., Mersmann, P., & Heberer, T. (2006). Mobility of
pharmaceuticals carbamazepine, ibuprofen, diclofenac,
ibuprofen, and propyphenazone in miscible-displacement
experiments. Journal of Contaminant Hydrology, 83,53–
69.
Scheytt, T., Mersmann, P., Leidig, M., P ekdeger, A., &
Heberer, T. (2004). Transport of pharmaceutically active
compounds in saturated laboratory columns. Ground
Water, 42(5), 767–773.
Schmidt, C. K., Lange, F. T., & Brauch, H.-J. (2004).
Assessing the impact of different redox conditions and
residence times on the fate of organic micropollutants
during riverbank fi ltration. In R. Gimbel, N. Graham, &
M. R. C ollins (Eds.), Recent progress in slow sand
and alternative biofiltration processes (pp. 5 61–569).
London: IWA.
Schubert, J. (2002). Hydraulic aspects of riverbank filtration—
field studies. Journal of Hydrology, 266, 145–161.
Water Air Soil Pollut
von Keutz, E., & Jekat, F.-W. (1998). Toxicological studies in
the course of a marketing authorization application for
human pharmaceuticals (Toxikologische Untersuchungen
im Rahmen der Zulassung von Humanarzneimitteln). In
Hessische Landesanstalt für Umwelt (Ed.), Arzneimittel in
Gewässern (pp. 65–69). Wiesbaden: Hessische Landesan-
stalt für Umwelt.
Zippel, M., Scheytt, T., Hannappel, S., & Müller, B. (2009).
Mathematical transport simulation of pharmaceutical com-
pounds in groundwater during bank filtration (Mathemati-
sche Simulation des Eintrages von Arzneimitteln aus
Oberflächengewässern in das Grundwasser durch Uferfil-
tration). In Umweltbundesamt (Ed.), Texte 14/2010 (p.
162). Dessau: Umweltbundesamt.
Water Air Soil Pollut
