Global Virtual Conference

GLOBAL VIRTUAL

April, 8. - 12. 2013

C O N F E R E N C E

Analytical Calculation Methods of Riverbank

Filtration
An Overview
Zsuzsanna Vradi Ph.D. Student
Department of Hydraulic and Water Resources Engineering
Budapest University of Technology and Economics
Budapest, Hungary
[email protected]
AbstractThere are several methods of calculating riverbank
filtration (RBF). Analytical, numerical, graphical approaches are
used to estimate the ratio of discharge coming from the surface
water and groundwater. This article is an overview of the most
common analytical methods, their uses and limitations.
Keywords - riverbank filtration; RBF; analytical methods;
streambed loss; Dupuit; Lapsin; Forchheimer;

I.

INTRODUCTION

Riverbank filtration (RBF) is a process when water from a

river or lake enters the aquifer under the riverbed which filters
and cleans it.
It is a very rare phenomenon in the nature because surface
water bodies tend to be found in the deep points of the terrain
or following the deepest line of a valley, draining the
groundwater of the surrounding area which has a relatively
higher position. However natural riverbank processes can
develop in alluvial riverbeds (i.e. Upper-Danube) or in time of
floods [1].
The practical usage of RBF manifests when the entering of the
surface water into the soil is triggered by pumping out the
water from the aquifer through production wells. During the
process the filtered water from the river becomes drinking
water suitable for human consumption.
II.

and there is a thick layer under it, there can be a limited RBF.
In such cases groundwater percolating through the aquifer can
mix with the water coming from the river and we might end up
pumping out water which is not from the river at all. In some
cases the water can contain pollution that nobody would expect
like pollution on the other side of the river [2].
B. Streambed loss
Another important factor of RBF is the permeability of the
streambed. This is due to two main causes:

A layer of worse permeability always develops on the

surface of the riverbed because of the sedimentation of
fine particles, mud (on top of the streambed) and
because of clogging (underneath the streambed) [3].
Even if there was no such layer before the RBF started
or the riverbed was cleaned it would develop or redevelop after the RBF production pumping starts. The
cause of these two kinds of clogging (in which organic
life germinates very soon) is the fine sediment of the
surface water continuously entering the aquifer.
Schubert examined the 300-m-wide streambed of
River Rhein near the water treatment plants of
Dsseldorf and found out that a 20-30-m-wide strip
near the bank was so clogged that no water could enter
the streambed at all. Recharge was coming from
further from the banks and the infiltration (active) zone
reached out as far as the other bank of the river [4]. In
[5] the different types of clogging were studied in
details. According to their findings the riverbed
consisting of mostly gravel and more coarse grains
clogged a lot faster and to a higher degree than a
riverbed of fine-grained sand. They named the
phenomenon as the Clogging-paradox. Their
explanation is that the floating and colloidal sediment
of the surface water can penetrate deeper into the
gravel streambed but once it settles it develops a more
solid, denser clogging. See: Fig. 1

There would be some kind of hydraulic loss when the

water is entering the streambed even if there was no
clogging at all. When the surface water enters the pores

FACTORS INFLUENCING THE PROCESS

A. Quality of geological layer, formation

Obviously the geological parameters of the riverbed are
important factors. There is no RBF in an impermeable, rock
riverbed.
In Hungary the riverbeds are usually made of sandy gravel
with sand underneath. Thickness of the riverbed varies between
a few meters and more than 100 m (i.e. Szigetkz, Hungary). If
the aquifer is very thick RBF may become insignificant
compared to the other percolation processes in the aquifer (i.e.
the flow of the groundwater parallel to the river or groundwater
flow from higher grounds toward the river).
If the river is a so-called partially penetrating river,
meaning that the riverbed intersects only a part of the aquifer

Figure 1. The two types of clogging

1st Global Virtual Conference

http://www.gv-conference.com

SECTION
17. Industrial and Civil Engineering

- 573 -

Global Virtual Conference

GLOBAL VIRTUAL

April, 8. - 12. 2013

C O N F E R E N C E

of the riverbed the magnitude and direction of its

velocity vectors change. Changing of the velocity
vectors requires energy which causes a so-called
entrance head loss. See Fig. 2. This loss is often
neglected in hydrogeology although it is always
present in surface water - groundwater relations, on the
border of the two systems. In the hydraulics of
pipelines it is considered a significant factor and it is
always taken into account [6] [7].

out. Then the two cones of depression (heads) are added as it is

shown on Fig. 5. It can be seen that the two values cancel each
other at the line of the riverbank giving a sum of zero
drawdown there.

Figure 3. Cone of Depression calculated by Dupuit equation

Figure 2. Entrance head loss

III.

CALCULATION METHODS OF RIVERBANK FILTRATION

A. Simplified mathematical description of RBF

Scientists started to study the hydraulic processes behind
RBF not long after the first such systems had been built and
had started operating. They tried to find ways to be able to
calculate its parameters and give a mathematical description so
that such systems could be better designed. Using the available
equipment of his time Forchheimer was the first to give a
formula which has been used ever since. His findings are
shown on Fig. 4 [8].

Figure 4. The Forchheimer mirror-method

He started-up using the Dupuit-Thiem equation of confined

groundwater flow which can be used to calculate the cone of
depression (s - the drawdown in any given point) in and around
a pumping well. [9] [10]. Directly next to the well (or in the
well itself if we neglect the well loss) the drawdown is:

Figure 5. Forchheimer solution

During the mirror-method R (radius of influence) falls out

of the equation and thus the Forchheimer equation is:
(1)
Where r is the radius of the well, R is the radius of
influence (it can be estimated or calculated according to
Sichardt), m is the depth of the aquifer, k is the hydraulic
conductivity. The drawdown in a point which is x-x0 from the
well:

Q: discharge of the well, k: hydraulic conductivity of the

aquifer, m: thickness of the aquifer, x0: distance of the river and
the production well
Fig. 3 shows the cone of depression as calculated by the
Dupuit equation. As mentioned earlier Fig. 4 shows the
Forchheimer solution. In his time it was believed that the cone
of depression around a well rises up to the level of the surface
water. Thus they used the so-called mirror-method. They
assumed a recharge well twice as far from the production well
as the riverbank. The recharge well is imagined to pump the
same discharge into the ground as the production well pumps

1st Global Virtual Conference

http://www.gv-conference.com

(3)
This approach was used for decades despite its
shortcomings when it comes to estimate the parameters and
discharge of RBF. It has been taught in many schools until this
day without giving a warning of these shortcomings.
When using the Dupuit or Forchheimer equation we do not
take well loss into consideration thus the calculated water
levels are only valid directly next to (but outside) of the wells.
Water level in the well will be lower than the ground water
level just outside of it because the water levels are separated
the same way as the surface water and the water table in the
riverbed. Similar hydraulic phenomenon, similar loss.
When we evaluate the results of test pumping or design
production wells, well fields we might need the predicted water
level in the wells and hence we might need to estimate the well
loss. The phenomenon is shown on Fig. 6. While sb (drawdown

SECTION
17. Industrial and Civil Engineering

- 574 -

Global Virtual Conference

GLOBAL VIRTUAL

April, 8. - 12. 2013

C O N F E R E N C E

in well) can be measured we also need sk (drawdown outside

the well).

B. More exact mathematical solutions

As mentioned earlier there is never a direct connection
between the surface water and the aquifer. There will always
be some kind of resistance when the water enters the aquifer.
This fact should be represented in the mathematical description
so that the design of an RBF system and the estimation of its
effects could be calculated better.
There are two parameters which are essential when judging
the real processes. One is the RBF-efficiency; the other is the
vertical conductivity of the streambed. Both can be found out
from data of test pumping carried out on the riverbank. All the
possible solutions use the same equation system which has
become known in the past years as the Lapsin equations [13]
[14].

Figure 6. The well loss

According to Juhsz [11] the following estimation can be

used (using the Dupuit equations in a modified form):

confined environment:

(4)

unconfined environment:

(5)

Further formulae:
(6)

(7)

(according to Sichardt)

(8),

1) Lapsin equations and their usage

We investigate the recharge processes of a production well
installed in the vicinity of a river. We use an analytical
mathematical method which requires some simplifications.
Scheme of the calculation is shown on Fig. 7. As it can be seen
we substitute the fine-grained sedimentation (cake) and the
clogging with only one layer of mo thickness and with ko
hydraulic conductivity. It is possible to use a separate aquifer
of k1 and m1 under the riverbed (which differs from the general
aquifer of parameters k and m).
Three different cones of depression are shown on Fig. 7.
The deepest one (dotted line) develops around the well if there
is no recharge from the river at all. In this case the above
mentioned Dupuit equation can be used to describe the cone
which is symmetrical to the axis of the well. The drawdown at
the line of riverbank is sD.
If there were no loss when the surface water enters the
aquifer the drawdown would be zero at the riverbank and the
cone of depression would be the highest one (also a dotted
line). This is not symmetrical to the axis of the well; it is much
steeper toward the river showing that there is a recharge of
water coming from this direction. The aforementioned mirrormethod of Forchheimer describes this cone [15].

(9)
In all formulae: k[m/s] - hydraulic conductivity, v[m/s] velocity.
The well loss will be sb - sk which can be as high as several
meters in badly designed production wells with high yields. In
these cases despite a large drawdown in the pumping well the
observation wells will show hardly any drawdown. This comes
as no surprise because the drawdown must be very small next
to the pumping well too. Test pumping cannot be evaluated in
this situation.
Simplified mathematical methods are still required today
for quick and easy-to-do estimations which are easy to
understand and are adequate for getting a first impression of an
area. New methods are still developed today which are based
on the Dupuit equations [12].

1st Global Virtual Conference

http://www.gv-conference.com

Figure 7. Cone of depression calculated by Lapsin equation

In reality the cone is somewhere between these two

extremes (dashed line). According to this there is a drawdown
of sL at the line of the riverbank which never exceeds sD of

SECTION
17. Industrial and Civil Engineering

- 575 -

Global Virtual Conference

GLOBAL VIRTUAL

April, 8. - 12. 2013

C O N F E R E N C E

Dupuit. If we knew this curve we could calculate the streambed

loss in three different ways:

The sL=H vertical distance itself. This is the head loss

of the flowing water when it enters the aquifer measured at the line of the riverbank. We could simply
measure this value if there was an observation well at
the line of bank but there is usually not and we usually
do not want to install one because the line of the
riverbank is often not an exact line due to the water
level changes of the river.
The horizontal distance (L) of the line of riverbank
and the intersection of the pressure curve and the water
surface (endpoint of the distance of influence). This is
a theoretical value, it cannot be measured. (It can be
calculated using the Shestakov formula: L = cth(2
b)/ [13] - letters representing the same values as in
the Lapsin equations below.) The L distance has been
considered very important in earlier publications
because it represents a real-world value: if the distance
of the production well and the river was larger with L
then we could calculate the depression as there was no
streambed loss. A river in such distance would have
the same effect as the river in its current position but
with a streambed loss. (A 3-phased seepage develops
in a L-wide strip at the riverbank. There is also some
recharge from the river to the aquifer here too but to a
much lesser extent.)

The values of b vary between 0.05 d-1 10 d-1. A two meter

thick mud layer of 0.1 m/d hydraulic conductivity has b = 0.05
d-1 vertical conductivity. (But only if we neglect the
aforementioned hydraulic entrance head loss. We cannot afford
that so we have to use a smaller b value than that of the one
calculated as b = ko / mo even if we know ko and mo. Hence
providing the 'real' b value requires a thorough model
calibration.)
Actually there are very few recommendations for the value
of b in the literature. In general it can be estimated as ~ 0.1 d-1
[17]. Zime gives an estimation of 0.07 d-1 - 0.27 d-1 based on
his own experiments [18].
There is no correlation between RBF-efficiency and b nor
are they interchangeable. There are cases when b is very small
(the riverbed is clogged) still the RBF-efficiency can be high if
the riverbed fully intersects the aquifer which is relatively thin.
But then the riverbed can be clean and without any clogging
and still having a very low RBF-efficiency if the river is above
a very thick aquifer with high hydraulic conductivity; (River
Saj, Szigetkz) in this case the production well might get
recharge more easily from under the river and the other side of
it.
Both parameters can be calculated using the Lapsin
equations. These apply to two zones as shown on Fig. 7. Zone
A is on the side of the river where the production well is
situated while Zone B is on the other side of the river.

The (sD sL )/sD ratio which we can call streambedconnectivity factor or if we want to measure the
efficiency in % we can call it RBF-efficiency (%). Its
value ranges between 0% and 100% [16].

RBF-efficiency is a better measure when it comes to giving

an overall number to judge the surface water - ground water
interaction. The H and L distances describe the streambed
loss but they depend on the discharge of the production well. In
the case of RBF-efficiency while both sD and sL changes when
the discharge changes but the ratio hardly changes. Analyzing
past test pumps data we can say that:

RBF-efficiency usually peaks at 90 %,

between 50-90 % the site is considered as a good RBF

area,

between 20-50 % there is more water coming from the

background than from the river,

under 20 % there is no recharge coming from the river,

even if the production well is close to it.

The RBF-efficiency can be best used to judge the discharge

of the water coming from the river but it does not enable us to
calculate the exact size or shape of the cone (that would help to
calculate residence time). For this we need the thickness and
hydraulic conductivity of the bottom layer developed on the
surface of the streambed. If we calculate b = ko/mo we have a
value (b) which describes the vertical conductivity of this
bottom layer. This is the value we usually have to provide as a
parameter to modeling software to calculate the recharge from
the river.

1st Global Virtual Conference

http://www.gv-conference.com

(10)

(11)

where: = x + xo, and r = x xo.

The equations were solved with fully penetrating wells in
mind, assuming that T=km transmissivity does not depend the
depression (in theory this applies only to confined
environment). Values of I represent the resistance of riverbed
clogging and the not full intersection of the aquifer. Generally
they can be calculated using Fourier integrals but if we only
want to calculate the cone of depression in a section
perpendicular to the river we can use the following simplified
formulae:
(12)
(13)
unknown values are:

SECTION
17. Industrial and Civil Engineering

(14)
(15)

- 576 -

Global Virtual Conference

GLOBAL VIRTUAL

April, 8. - 12. 2013

C O N F E R E N C E

(16)

(17)
(18)
x = x (in Zone A) and x = x 2c (in Zone B). Finally Ei
is the Theis well function (exponential integral)[19]. Values of
Ei are approximated by infinite series.
It is worth noting that if there is no clogged bottom layer
(mo = 0) and the river intersects the aquifer in full depth then I1
= I2 = 0 thus giving the Forchheimer equation for (10) while
(11) cannot be interpreted (and there is no drawdown on the
other side of the river either).

From the data of the test pumping we can calculate the

vertical conductivity of the streambed (b = ko/mo).
Evaluation of the test pumping can be done by computer
software which enables the determination of the unknown
parameters so that the calculated cone of depression matches
the observed water levels. By doing so, we try to find a
theoretical curve corresponding to the observed values. This
theoretical curve gives us sL - the drawdown at the line of the
riverbank.
C. Graphical and other estimation methods
Sometimes it is worth to estimate the RBF-efficiency
instead of calculating the Lapsin equations. This method is
simpler and makes easier to visualize the results. We create a
graph where the drawdown values are presented corresponding
to the logarithm of the distance. The cone of depression can be
approximated as a line on such a graph. See Fig. 8.

(10) and (11) can be written in a more general form of

s = f (Q , x , k, m, k1 , m1 , ko , mo ).
Thus if we want to calculate the drawdown anywhere
around an RBF production well we need these eight
parameters. At x = xo (at the line of riverbank) we get sL.
During a test pump we know the discharge and we have to
measure the drawdown in observation wells in known
distances. That gives us three unknown parameters (k = k1, ko,
mo) out of eight thus we need at least three s x pairs which
requires at least three observation wells. However due to some
uncertainties it is advisable to have more observation wells
installed in a line between the production well and the river. In
practice we usually install six observation wells. Three between
the production well and the river and three behind the
production well (so all six are installed in Zone A as shown in
Fig. 7). In case of small rivers and thick aquifer it is worth
having an observation well in Zone B on the other side of the
river (if we see a drawdown here we can be sure that there is no
RBF).
On Hungarian fluvial terraces the k = k1 approximation can
be accepted - to reduce the number of unknown values in the
equation. Similarly we can consider m1 (thickness of the
aquifer) known after the geotechnical investigation of the
riverbed. If the impermeable layer cannot be reached (i.e. in the
Szigetkz in Hungary) it is advisable to assume a fictive depth
which can be twice the depth of the screened well. Most
probably the pumping cannot move the ground water in greater
depth than this and whatever happens underneath this distance
that should not have any effect on the RBF process.
To calculate the RBF-efficiency ((sD sL )/sD) we need to
know sD which can be calculated from the Dupuit equation. In
this case the radius of influence (R) is unknown which can be
calculated from empirical formulae (Sichardt, Kusakin,
Kozeni, etc. [20]) or from the data of the observation wells
behind the production well using the R = x exp(2Ts/Q)
formula. From three observation wells (three x s pairs) we
have three values for R of which we calculate the arithmetic
mean.

1st Global Virtual Conference

http://www.gv-conference.com

Figure 8. Graphical method

If there is recharge from the river then the cone of

depression runs higher on the river's side so the line
representing the cone will run higher than that of the mirrored
line of the cone from the background.
If we draw a vertical line at the logarithm of the riverbank's
distance then the line of the measured values intersects it at sL
depth (Lapsin drawdown) and the mirrored line intersects it at
sD depth (Dupuit drawdown). From these values we can
calculate the RBF-efficiency.
We can use this graph to estimate the radius of influence
too because the line on the background side intersects the
surface (where s = 0) at the logarithm of R according to the
Dupuit equation.
The fact itself that the drawdowns are smaller on the river's
side gives us hints about the recharge conditions of a
production well without any analytical or graphical
evaluations. Of course drawdowns depend on the distance from
the well thus we have to calculate the so-called drawdown
efficiency on the river's side and on the background side:
and

LR = xRi sRi

(19)

LB = xBisBi

(20)

LR: drawdown efficiency (river side), xRi: observation

well distance from the production well (river side), sRi:
drawdown (river side), LH: drawdown efficiency

SECTION
17. Industrial and Civil Engineering

- 577 -

Global Virtual Conference

GLOBAL VIRTUAL

April, 8. - 12. 2013

C O N F E R E N C E

(background side), xHi: observation well distance from the

production well (background side), sHi: drawdown (background
side).
The number of the observation wells should be the same on
each side and at least two wells are required. In a usual test
setting we use 3-3 wells, so i = 3. The drawdown efficiency on
the river side (LR) cannot be larger than the 90% of the
background efficiency. If it is larger further evaluations are
unnecessary as there was no RBF during the test pumping.
Finally we have to emphasize that following any
preliminary evaluations of a test pumping the most effective
method of further calculations and protective-area
determinations is the hydraulic modeling.

[2]

[3]

[4]

[5]

B. Wett, H. Jarosch, K. Ingerle: Flood induced infiltration affecting a

bank filtrate well at the River Enns in Austria Journal of Hydrology
266 (3-4), Sep. 2002. pp.:222-234
G. D. Liu, J. T. Li: Seepage laws in aquifer near a partially penetrating
river with an intensive extraction of ground water. in Science in China
Series E-Technological Sciences40 (5), Oct. 1997.
D. C. Mays, J. R. Hunt: Hydrodynamic aspects of particle clogging in
porous media. in Environmental Sciene & Technology 39 (2), Jan.
2005. pp.:577-584
J. Schubert: How does it work? Field Studies on Riverbank Filtration.
in IAWR Conf. (4.), Proceedings of the International Riverbank
Filtration Conference. Nov. 2-4., 2000, Dsseldorf
J. P. Stuyfanzd, M. H. A. Juhsz-Holterman, W. J. de Lange:
Riverbank filtration in the Netherlands: well fields, clogging and
geochemical reactions. in NATO Advanced Research Workshop:
Bratislava, 7-10 Sept. 2004.

1st Global Virtual Conference

http://www.gv-conference.com

[7]

[8]
[9]
[10]
[11]
[12]

[13]

REFERENCES
[1]

[6]

[14]
[15]
[16]
[17]
[18]
[19]

[20]

E. Smela, L. J. Martinez-Miranda: Effect of substrate preparation on

smectic liquid crystal alignment: II. Further results and modeling. J.
Appl. Phys., 77 (5), 1995.
J. M. Burke E. Smela: A novel surface modification technique for
forming porous polymer monoliths in PDMS. Biomicrofluidics, 6 (1)
16506, 2012.
P. Forchheimer: Hydraulics, Teubner Berlin-Leipzig 1930.
J. Dupuit: Etudes Theoriques et Pratiques sur le Mouvement des Eaux,
Dunod. Paris. 1863.
G. Thiem: Hydrologische Methoden (Hydrologic Methods), J.M.
Gebhardt, Leipzig 1906.
J. Juhsz: Szivrgsi tnyez meghatrozsa a ktellenlls
figyelembevtelvel. Hidrolgiai Kzlny, 1966/5
E. I. Anderson: Modeling groundwater-surface water interactions
using the Dupuit approximation. Advances in Water Resources. 2005.
Apr., v. 28, issue 4 pp.: 315-327.
B. M. Sesztakov: Ocenka szoprotivlenija lzsa ruszla vodojemov pri
gidrogeologicseszki raszcsota. Razvedka i ohrana nyedr., No 5, 1964.
M. S. Hantus Pompage dessai dans un puit proximit dune rivire
colmate. Bulletin de BRGM, Sect. III, 3/4.
F. M. Bochever, N. N. Lapsin, A. E. Oradovszkaja: Zascsita podzemnh
vod ot zagrjaznyenyija. NEDRA, Moszkva 1979.
I. Vlgyesi: Mederkapcsolati hatsfok: a parti szrs vztermels fontos
paramtere. Hidrolgiai Kzlny, 73. vf., 5. szm, 1993. pp.:261-264
F. Szkely: Partiszrs kutak szivrgshidraulikai, vdterleti s
vzminsgi mretezse. VITUKI Kzlemnyek 30.
Zima K.(1988): Prognoza mnozstvi a jakosti podzemnh vody jiman v
blizkosti leky. Vodni hospodstvi 2., rada B., Szovjetuni, 1988.
C. V. Theis: The Relation Between the Lowering of the Piezometric
Surface and the Rate and Duration of Discharge of a Well Using
Ground-Water Storage, Transactions, American Geophysical Union,
Vol. 16. p:519-524
J. Juhsz: Hidrogeolgia. Akadmiai Kiad, Budapest, 1987.

SECTION
17. Industrial and Civil Engineering

- 578 -