Seepage Face Height, Water Table Position, and

Well Efficiency at Steady State

by Djaouida Chenaf1 and Robert P. Chapuis2

Abstract
When a fully penetrating well pumps an ideal unconfined aquifer at steady state, the water table usually does
not join the water level in the well. There is a seepage face inside the well, which is a key element in evaluating
the well performance. This problem is analyzed using the finite-element method, solving the complete equations
for saturated and unsaturated flow. The seepage face position is found to be almost independent of the unsaturated
zone properties. The numerical results are used to test the validity of several analytic approximations. Equations
are proposed to predict the seepage face position at the pumping well for any well drawdown, and the water table
position at any distance from the pumping well for any in-well drawdown. Practical hints are provided for install-
ing monitoring wells and evaluating well efficiency.

Introduction supply wells in unconfined aquifers, pump-and-treat

When a fully penetrating ideal well pumps an ideal
unconfined aquifer, there may be a seepage face inside
the well screen and, thus, a vertical discontinuity of the
water table (Sichardt 1928). The seepage face problem for
steady state was studied by Boulton (1951), Boreli
(1955), Hall (1955), Schneebeli (1956), Polubarinova-
Kochina (1962), Heinrich (1964), Brauns (1981), Kovàcs
(1981), and Gefell et al. (1994, 1995). The seepage face
for unsteady state conditions was investigated by Rushton
and Holt (1981), Rushton and Singh (1983), Sakthivadi-
vel and Rushton (1989, 1990), S xen (1990), Lee and Leap
(1997), Ojha and Gopal (1999), and Sxen and Cximen (2001).
The seepage face and the water table position are
key parameters for correctly evaluating the performance
and risks of chemical and biological incrustation of water
1Corresponding author: Department of Civil Engineering, Royal
Military College, P.O. Box 17000, Stn Forces, Kingston, Ontario,
Canada K7K 7B4; (613) 541-6000 ext. 6603; fax (613) 541-6218;
chenaf-d@rmc.ca
2Department of Civil, Geological and Mining Engineering,
École Polytechnique, P.O. Box 6079, Stn CV, Montreal, Quebec,
Canada H3C 3A7; (514) 340-4711 ext. 4427; fax (514) 340-4477;
robert.chapuis@polymtl.ca
Received February 2006, accepted August 2006.
Copyright ª 2007 The Author(s)
Journal compilation ª 2007 National Ground Water Association.
doi: 10.1111/j.1745-6584.2006.00277.x
168 Vol. 45, No. 2—GROUND WATER—March–April 2007 (pages 168–177)
remediation pumping systems combined with either aero-
bic or anaerobic treatment, including recovery of light,
nonaqueous phase liquids from porous media (Waddill
and Parker 1997; van Dijke and van der Zee 1998).
This paper is limited to steady-state conditions. It first
presents a concise but extensive review of previous predic-
tive methods. The predictive capacity of previous methods
is then assessed against a large set of more than 200 numeri-
cal analyses using a finite-element method for both satu-
rated and unsaturated conditions. Finally, the paper presents
new practical equations to predict the minimum height that
the seepage face may have if the water level in the well is
maintained at the base of the aquifer (maximum well draw-
down and flow rate), the seepage face height for any well
drawdown, the position of the water table in the vicinity of
the well, and how to evaluate well efficiency. Practical hints
are also given for installing monitoring wells (MWs) in
order to avoid difficulties in interpreting test data.
Previous Studies
The solution for steady-state pumping of an ideal
unconfined aquifer was developed by Dupuit (1857, 1863),
during the year after Darcy (1856) published his memoirs
containing the filtration law that much later became
Darcy’s law. The Dupuit solution is for a homogeneous
isotropic aquifer of uniform thickness over a horizontal
impervious substratum when the well fully penetrates the explicitly on K. Furthermore, according to Equation 2,
aquifer and is screened over the entire initial saturated this water table would be continuous between h (r = R) =
thickness. It is also assumed that the well is perfect; i.e., H at a large distance and h (r = rw) = hw on the in-well
there are neither skin effects close to the well (head losses surface. When a seepage face appears in the well, then the
due to either too high Reynolds number related to too high water table at r = rw on the out-well surface reaches
water velocities, poor design of the screen/filter pack, and/ a position ha > hw (Figure 1). The seepage face was stud-
or insufficient well development) nor development effects ied by several authors using reduced-scale physical mod-
(increased hydraulic conductivity close to the screen). els (e.g., Hall 1955; Simpson et al. 2003) and numerical
models (e.g., Shamsai and Narasimhan 1991; Clement
The Dupuit Equation et al. 1996; Chenaf and Chapuis 1998; Boufadel et al.
The Dupuit equation relates the constant flow rate, Q 1999; Chapuis et al. 2001).
(steady-state conditions), to the saturated hydraulic The Dupuit equation for the flow rate may be written as:
conductivity, K, and either the saturated thickness bj ( j =
1, 2 .) at radial distance rj from the pumping well or the H 2 2 h2w
Q ¼ pK ð3Þ
drawdown sj defined as sj = H 2 bj, where H is the initial lnðR=rw Þ
saturated thickness:
The maximum flow rate Qmax occurs when hw = 0:
b2 2 b21 ðH2s2 Þ2 2 ðH2s1 Þ2
Q ¼ pK 2 ¼ pK ð1Þ H2
lnðr2 =r1 Þ lnðr2 =r1 Þ Qmax ¼ pK ð4Þ
lnðR=rw Þ
In his proof, Dupuit considered that bj was equal to
the hydraulic head h(rj, z) = hj at any elevation z within This gives a ratio Q/Qmax that, in theory, would be
the aquifer when the elevation datum (z = 0) was taken at independent from R and rw:
the horizontal interface with the impervious substratum.
The equipotential surface at rj was assumed to be a verti- Q H 2 2 h2w ðH2sw Þ2
¼ ¼ 1 2 ð5Þ
cal cylinder, and the slope of the water table was supposed Qmax H2 H2
to be small. Dupuit then assumed that the vertical compo-
nent of the ground water velocity was negligible, i.e., that
Method of Brauns (1981)—Sichardt (1928)
the flow was horizontal and radial. He also assumed that
Combining the Dupuit equation and the equation for
the water table joined the water level in the pumping well.
the maximum entrance gradient of Sichardt (1928), Brauns
For a long time, the original equations of Dupuit have
(1981) obtained Equation 6 when hw = 0 and ha = ha, min ,
been viewed as approximations. Currently, it is known that
which corresponds to the maximum seepage face height:
the flow rate equation is exact when the unsaturated seep-
age (above the water table) is neglected and that the proof 1=2
ha;min =H ¼ ða2 11Þ 2a ð6Þ
for the flow rate does not require all the original Dupuit as-
sumptions (Charny 1951). However, the equation for the in which
water table position is approximate. The true position is
 
higher than that of the Dupuit equation, and the difference rw R
is larger close to the well. The difference becomes negligi- a¼ pffiffiffiffi ln ð7Þ
15H K rw
ble at a radial distance r > 1.5 H. In addition, there may be
a seepage face inside the well (Sichardt 1928; Muskat and Equations 6 and 7 were never verified by any experi-
Wyckoff 1937). The mathematical proof for the flow rate mental or numerical study.
(Charny 1951) was extended by Schneebeli (1956) to dem-
onstrate the existence of the seepage face.
14
The equation for the water table as developed by
Dupuit involves the well external radius rw, the in-well ha = 8.5 m height of
12
seepage face at r = rw
saturated thickness hw or drawdown sw, and the radius of
influence R, where the drawdown is equal to zero. The 10
elevation z (m)

head h at any distance r is then given as:

8
2 lnðr=rw Þ
h ¼ h2w 2
1 ðH 2 h2w Þ ð2Þ 6
lnðR=rw Þ hw = 6.0 m
in the well
4
Note that the radius of influence is an approximation eq. Dupuit
that, in steady state, results from the extrapolation of the 2 free surface
cone of depression to such a distance that the pumping rw = 0.2 m
r = rw
rate is matched by recharge from infiltration and/or a lat- 0
eral recharge boundary. Because R is in a log term, the 0,1 1 10 100

calculation of h using Equation 2 is relatively insensitive radial distance r (m)

to the estimated value of R.
Figure 1. Illustration of the seepage face problem in a fully
According to Equation 2, the shape of the water table penetrating ideal well that pumps an ideal unconfined aquifer.
would depend only on the problem geometry and not
D. Chenaf, R.P. Chapuis GROUND WATER 45, no. 2: 168–177 169
Method of Boulton (1951) presented them in a nondimensional graph that was found
Using a finite-difference method with the method of to have the following equation (Chenaf and Chapuis
relaxation, known also as the Gauss-Siedel method, 1998):
Boulton (1951) proposed:
pKðh2a 2 h2w Þ 1000pKrw2
ha;min 2logðR=rw Þ ¼ 2:63 2 log ð14Þ
¼ 21 ð8Þ Q Q
H ns
Schneebeli (1956) did not specify whether the rela-
in which ns is a coefficient that would be almost indepen- tionship was valid only for the maximum seepage face or
dent of rw/H and close to 3.75. minimum value of ha, ha,min when hw = 0 or for any seep-
Equation 8 was established from a limited number of age face when hw > 0.
cases. It predicts a seepage face height that becomes neg-
ative when R/rw is < 75. To avoid this impossibility, the Method of Polubarinova-Kochina (1962)
comparative study will consider that Equation 8 predicts The relatively complex mathematical method of
that there is no seepage face when R/rw is < 75. Polubarinova-Kochina (1962) was expressed as practical
graphics by Kovàcs (1981).
Method of Gefell et al. (1994) and Kozeny (1953)
Kozeny (1953) developed a graphical method to Method of Heinrich (1964)
determine the maximum height of the seepage face. He Using an approximate analytic solution, Heinrich
assumed that the maximum ground water Darcy velocity (1964) proposed:
at the well screen occurred at a unit gradient (then Darcy
velocity [ K), an assumption that was shown to be incor- ha;min lnðxÞ
¼ ð15Þ
rect (Boulton 1951; Kawecki 1995; Wise and Clement H lnðxR=rw Þ
1995). The equation corresponding to this graphical
method was solved by Gefell et al. (1994). In this paper, in which
the authors have written the solution as: x ¼ 0:5ð1 1 R=rw Þ ð16Þ
2 1=2
ha;min =H ¼ ðb 11Þ 2b ð9Þ In this paper, the predictions of these different meth-
ods are compared to the results of numerical analyses
in which using the finite-element method. Comparisons are made
  for the numerous methods that predict the ratio ha,min/H
rw R
b ¼ ln ð10Þ and the more general method of Schneebeli (1956).
H rw

The mathematical derivation of Equations 9 and 10 is

given in Appendix 1. Numerical Analysis
The numerical analysis used the finite-element code
Method of Hall (1955) Seep/W (Geo-slope International 2003), which was evalu-
After sandbox tests and a numerical analysis using ated in detail by Chapuis et al. (2001). The code considers
finite differences, Hall (1955) proposed: the full characteristic functions of the soil, K(uw) and
h(uw), where uw is the pore water pressure, K(uw) is the
ha;min 1 hydraulic conductivity function, and h(uw) is the volu-
¼ ð11Þ
H 5rw h  R i
metric water content function. The equations of Darcy
11 1 1 0:02 ln
H rw (1856) for flow and Richards (1931) for conservation are
numerically solved as uw-based equations. The numerical
solutions for unconfined aquifers are therefore obtained
Method of Boreli (1955) without needing any of the Dupuit assumptions. The itera-
Boreli (1955) treated eight cases of either partial or tive method used to simulate the seepage face at the well
total penetration with a finite-difference method. His find- is given by Chenaf and Chapuis (1998) and is briefly de-
ing for total penetration was written by Brauns (1981) as: scribed as follows. All simulations are done with a bound-
  ary condition h = H at r = R and a no flow condition at
ha;min A1 2:3
¼ 12 ½0:13B2 0:0123ðB22:3Þ2
ð12Þ the bottom of the aquifer. A first simulation is done with
H B
the following boundary condition along the screen: h = hw
in which from z = 0 to z = hw and a no flow condition for upper
nodes. The first solution is examined. There are a few
A ¼ lnðR=HÞ and B ¼ lnðR=rw Þ ð13Þ upper nodes for which the calculated h is higher than ele-
vation (h > z or uw > 0), a physically impossible situation.
In a second simulation, the boundary conditions at these
Method of Schneebeli (1956) nodes are reviewed to ensure that they are either at zero
Schneebeli (1956) gathered the results of Babbitt and water pressure with water leaving the flow system or unsat-
Caldwell (1948), Boulton (1951), Hall (1955), and Zee urated with water retained by capillarity (no flow). The
et al. (1955) for sandbox tests and numerical analyses. He iterative calculation process stops when all upper nodes
170 D. Chenaf, R.P. Chapuis GROUND WATER 45, no. 2: 168–177
maintain either a zero water pressure (seepage face) or
a negative pressure (above the seepage face).
More than 200 different conditions of steady radial
flow have been analyzed numerically, all of them with
axial symmetry. Each grid was defined in a vertical plane
(r, z), the equations being written for axial symmetry
without using the polar angle in the horizontal plane. For
mathematical consistency, the code provides a pumped
flow rate per unit angle, i.e., in m3/s/rad. The parameters
of this study had the following ranges:
d Initial saturated thickness H from 9 to 20 m
d Radius of the well rw from 0.05 to 2.5 m
d Radius of influence R from 15 to 90 m
d Ratio R/rw from 12 to 900 (this provides a robust test of
the flexibility of the method)
d In-well drawdown sw from 0.1 H to H (empty well)
d More than 20 unsaturated isotropic hydraulic conductivity fun-
ctions, K(uw), six examples of which are shown on Figure 2
d Finite-element grids having between 3000 and 9000 nodes
(grids roughly 50 3 60 to 75 3 120).
In the seepage face zone, and in the first meter above
the water table, grid elements were less than 10 cm high in
order to follow accurately the sharp variation of the K(uw)
function. Close to the well, the first elements extended from
rw to less than 1.3 rw and gradually increased in size with
radial distance. The code has no restrictions for the number
of ‘‘layers’’ to be used; this number may vary with r, and
also the number of ‘‘rows’’ may vary with z, transition ele-
ments being used where appropriate. This capacity enables
one to adjust a grid to a specific problem (e.g., around
a water table) with variable element sizes in both directions.
The code provided by interpolation the position of the seep-
age face, which was known within 6 3 cm or less.
Results of the Numerical Analysis
Pumped Flow Rate
Each numerical analysis calculated a pumping rate
that was slightly higher (1% to 5%) than predicted by the
1.E-02
1.E-03
1.E-04
1.E-05
1.E-06
1.E-07
K (m/s)
Dupuit equation, the small difference depending on the
K(uw) function and problem dimensions (H, R, and rw).
The ratio Q/Qmax was found to be almost independent of
H and R/rw, as predicted by Equation 5. Figure 3 presents
the data for a numerical case where H = 10 m, the
sandbox tests of Hall (1955) where H = 1.22 m, and of
Simpson et al. (2003) where H = 0.90 m. Thus, it
was numerically and experimentally confirmed that the
Dupuit equation is ‘‘incorrect’’ with respect to flow
rate because it neglects unsaturated flow, although the
discrepancy is likely minor enough to be considered
negligible in most practical cases.
Water Table
The calculated position for the water table was always
very close to that predicted by Equation 2 when the radial
distance r was higher than 1.5 times the initial saturated
thickness H as noticed by Boulton (1951). Figure 4 gives
the water table position vs. r and also the values of h(r, zj)
at four elevations zj within the aquifer. Figure 4 clearly
shows that the hydraulic head depends on depth, and this is
why it is not recommended to use the drawdown values
measured in fully penetrating MWs close to the pumping
well (e.g., USDI [U.S. Department of the Interior] 1977;
Kruseman and de Ridder 1990) to calculate K using Equa-
tion 1. Hydraulic heads measured at MWs screened near
the base of the aquifer provide a closer approximation to
the Dupuit curve and are therefore more useful for estimat-
ing K using Equation 1. If the MWs have fully penetrating
screens, the drawdown values to be used in Equation 1 are
those of MWs located at distances r > 1.5 H, where the
equipotentials become almost vertical.
Within the full-length screen of a fully penetrating
MW close to the well, a downward seepage can be mea-
sured (Rushton and Howard 1982). This happens because
the total head h increases with z and, thus, the full-length
screen acts as a vertical drainpipe. As a result, the static
water level in a fully penetrating MW gives only some
average drawdown (the type of average is unknown) for
the full aquifer thickness. The use of this average draw-
down value in Equation 1 has been questioned by van der
Kamp (1985), who suggested that Equation 1 should
1.0
0.8
ratio Q/Qmax

0.6
1.E-08
1.E-09
1.E-10 0.4
K1 K2
1.E-11
K3 K4
1.E-12 0.2 Dupuit, H = 10 m
K5 K6 f.e.m., H = 10 m
1.E-13 Hall, H = 1.22 m
Simpson et al., H = 0.90 m
1.E-14
-50 -40 -30 -20 -10 0 10 0.0
0.0 0.2 0.4 0.6 0.8 1.0
pore water pressure uw (kPa)
ratio sw/H
Figure 2. Examples of curves of K(uw) used in the numerical
analyses. Figure 3. Pumping rate Q vs. drawdown sw in the well.

D. Chenaf, R.P. Chapuis GROUND WATER 45, no. 2: 168–177 171

 9 10
s(well) = 2H/3 = 3 m
h (r,z) depends on z 9 s(well) = H/2 = 4.5 m
s(well) = H/3 = 6 m
8 s(well) = H = 9 m
8
hydraulic head h (m)
7

elevation z (m)
r = 1.5 H maximum at the
6 bottom of the
7 5 seepage face

Dupuit 4
water table
3
6 h (z = 6 m)
h (z = 4 m) 2

hw = 5.0 m h (z = 2 m) 1
5 0
0.1 1 10 100 0.0E+00 5.0E-04 1.0E-03 1.5E-03 2.0E-03
radial distance r (m) horizontal Darcy velocity (m/s)

Figure 4. Comparison of numerical results for the hydraulic Figure 5. Numerical results for the horizontal component of
head h(r, z) vs. the radial distance r at different elevations z the Darcy velocity at r = rw, case of maximum seepage face
within the aquifer, with the numerically computed water height (sw = H), and cases of sw = H/3, H/2, and 2H/3.
table position and the Dupuit equation.

apply when the considered drawdown at a short distance Results for the Seepage Face
is an average established from drawdowns of several short For the same geometric conditions (i.e., rw, R, H), it
screen MWs at this distance. Using a finite-element anal- was found that functions of K(uw) having the same satu-
ysis, Chenaf and Chapuis (1998) found that Equation 1 is rated K value gave almost equal seepage faces, the ha
respected with any r value when the drawdown values are value being slightly increased when the air entry value was
those of MWs having a short screen at the bottom of the increased. The numerical results are examined first for
aquifer, as suggested by Figure 4. With such MWs, the the seepage face when hw = 0 in steady-state conditions.
distance restriction rj > 1.5 H can be removed. However,
deep MWs cannot be used to determine the water table Comparisons to Previous Predictive Methods
position at r < 1.5 H. Additional shallow MWs are needed The numerical results are first compared against the
for this purpose. These practical findings were used by method of Schneebeli (1956) that synthesized several
Chapuis et al. (2005) to monitor a pumping test for tran- experimental and numerical studies and is the most gen-
sient and steady-state conditions. eral method. The numerical results (Figure 6) do not give
The drawdown of deep MWs, when extrapolated to a unique straight line in the nondimensional graph (x, y)
r = rw, gives hw and not ha (Figure 4). Thus, water levels as defined by Schneebeli (1956). The calculated curves
measured in deep MWs should not be used to erroneously make an angle with the straight line of Schneebeli and
conclude that the well has no seepage face. Care must be
exercised when evaluating the performance of the pump-
ing well using step-drawdown tests, and attention must be 4.0
paid to the exact location of each MW in terms of radial R/rw = 900
R/rw = 900*
distance, length, and depth of screen. 3.5 R/rw = 300
R/rw = 150
R/rw = 150*
3.0 R/rw = 90*
Gradient and Velocity along the Seepage Face R/rw = 60
According to the numerical results, the horizontal R/rw = 30
(ha2-hw2) / (Q/πK)

2.5 R/rw = 12
component of the gradient, which is zero above the seep- Schneebeli
Hall, R/rw = 16
age face, is not equal to 1.0 along the seepage face—as 2.0 Simpson et al., R/rw=11
sometimes assumed (Kozeny 1953)—but it increases with
depth to reach a maximum at the bottom of the seepage 1.5
face. This maximum is close to 18 at z = 0 in the example
of Figure 5 when sw = H. The vertical component of the 1.0

gradient is very small above the seepage face and equal to

0.5
one along the seepage face.
The horizontal component of the Darcy velocity 0.0
v(rw, z) was found to reach a maximum at the bottom of 0.1 1 10 100 1000

the seepage face, as already found by Muskat and Wyck-
off (1937). Four sets of results are provided in Figure 5
for a well drawdown sw of H/3, H/2, 2H/3, and the
maximum case of sw = H. Neglecting the variations in the
horizontal velocity would induce a risk of misinterpreting
tracer tests and overestimating longitudinal dispersivity.
172 D. Chenaf, R.P. Chapuis GROUND WATER 45, no. 2: 168–177
103 rw2 / (Q/πK)
Figure 6. Presentation of a few nondimensional numerical
results for the seepage face, using the axes of Schneebeli
(1956). *means that another set of parameters (rw, R, H, and
K) was used.
are usually displaced to the right when the ratio R/rw 20
decreases. Similar angles are obtained with the sandbox Kozeny - Gefell et al.
data of Hall (1955) and Simpson et al. (2003), which Sichardt - Brauns
Hall
come close to the Schneebeli straight line when the well 15 best fit

predicted ha, min (m)

drawdown is large.
Figures 7 and 8 compare numerical results with dif-
ferent predictions that can be made using the previously 10
presented equations for the maximum height of the seep-
age face or the minimum value of ha, ha,min when hw = 0.
It appears that none of the six methods provided good
5
predictions (Figures 7 and 8). Among these, that based on
the calculations of Brauns (1981) gave the best predic-
tions when the maximum seepage face exceeded 7 m. For
0
smaller seepage faces, this method largely underestimated 0 5 10 15 20
the seepage face position. ha, min numerical study (m)
In most cases, a fully penetrating well is operated at
a drawdown sw such as 0.25 < sw/H < 0.75 (Kruseman and Figure 8. Comparison of nondimensional numerical results
de Ridder 1990; Driscoll 1986). Even under ideal theoreti- with the predictive methods of Kozeny (1953), Gefell et al.
cal conditions, the ratio of well drawdown to yield is not lin- (1994), Sichardt (1928), Brauns (1981), and Hall (1955).
ear. Increasing sw/H from 10% to 20% means a large yield
increase, whereas increasing sw/H from 70% to 80% means
much less yield increase (e.g., Figure 3). In addition, when pKh2a;min 1000pKrw2
operating at sw/H > 0.75, it is difficult to maintain the yield y¼ and x ¼ log ð18Þ
Qmax Qmax
with seasonal recharge variations and temperature variation
that affects the water viscosity (e.g., Chapuis 1999); i.e., Note that Figure 10 provides a graphical means to
there is little safety margin for such operating conditions. solve for ha,min without solving for Equations 17 and 18.
The data of Figure 6 are for any well drawdown value. The reader can use Equation 4 to calculate Qmax, solve for
However, when the usual operating conditions are consid- the x-axis value, use the top curve on Figure 10 to esti-
ered, Figure 9 is obtained. Then the straight line relation- mate the resultant y-value, and then solve for ha, min.
ship of Schneebeli (1956) appears as some average for the Equation 19 predicts the seepage face position ha
usual operating conditions (0.25 < sw/H < 0.75). (Figure 1) for the condition hw > 0. Knowing the value of
ha,min from Equations 17 and 18, ha may then be pre-
New Equations for the Seepage Face dicted using:
Equations 17 and 18 predict the maximum seepage   2
face height ha,min for the condition hw = 0. The numerical ha ha;min ha;min hw
¼ 1 12 ð19Þ
results for this condition are not linear (Figure 10) and H H H H
slightly above the straight line of Schneebeli (1956). The
numerical results can be fitted with a third degree equation: The predicted values are very close to the numerical
results as shown in Figure 11.
y ¼ 0:0539x3 2 0:0446x2 2 1:2633x 1 2:9896 ð17Þ
4.0
in which R/rw = 900
R/rw = 900*
3.5 R/rw = 600
20 R/rw = 450
Boreli R/rw = 300
3.0 R/rw = 150
Heinrich
R/rw = 150*
(ha2-hw2) / (Q/πK)

Boulton R/rw = 90*

15 2.5
predicted ha, min (m)

best fit R/rw = 60

R/rw = 30
2.0 R/rw = 12
Schneebeli
10
1.5

1.0
5 hw between 25 and 75% H
0.5

0 0.0
0 5 10 15 20 0.1 1 10 100 1000
ha, min numerical study (m) 103 rw2 / (Q/πK)

Figure 7. Comparison of nondimensional numerical results Figure 9. Comparison of a few nondimensional numerical
with the predictive methods of Boreli (1955), Heinrich results with the predictive method of Schneebeli (1956) for
(1964), and Boulton (1951). the usual operating range of wells in unconfined aquifers.

D. Chenaf, R.P. Chapuis GROUND WATER 45, no. 2: 168–177 173

 5 
y = 0.0539x3 - 0.0446x2 - 1.2633x + 2.9896 a ¼ bðH 2 ha Þ 2 aðhD 2 ha Þ =ða2 b 2 ab2 Þ ð24Þ
R2 = 0.9994
4
ha, min2/ (Qmax/πK) 
f.e.m.
Schneebeli
b ¼ b2 ðH 2 ha Þ 2 a2 ðhD 2 ha Þ =ðab2 2 a2 bÞ ð25Þ
3 Linear (Schneebeli)
Polynomial (f.e.m.)
 0:5
hD ¼ H 2 2ðQ=pKÞlnðR=1:5HÞ ð26Þ
2
An example, given in Figure 12, compares the numer-
1 y = -1.00x + 2.63 ical solution for the water table position and the predicted
R2 = 1.00
position using Equations 17 to 26. More detailed studies
will be required to assess the capacity of these equations to
0
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 adequately predict the water table position.
log (103 rw2 / (Qmax/πK))

Figure 10. Nondimensional numerical results for ha,min

when hw = 0.
Prediction of the Water Table Position
For r > 1.5 H, the water table position can be pre-
dicted by the Dupuit equation. For r < 1.5 H, it can be
predicted using simple geometrical considerations. First,
the descriptive graph must pass through the two extreme
points (r = rw, h = ha) and (r = R, h = H). Second, it must
join the Dupuit curve at (r = 1.5 H, h = hD) where hD is
given by Equation 2. The simplest curve passing through
these three points is a second degree equation such as:
h ¼ ax2 1 bx 1 ha
in which the following parameters are defined as:
x ¼ logðr=rw Þ
a ¼ logðR=rw Þ
b ¼ logð1:5H=rw Þ
Application: Estimating Well Efficiency
A numerical model is used here to qualitatively illus-
trate the influence of either clogging or development
around the screen, as compared to the ideal case. The
parameters are H = 9 m, rw = 0.1 m, R = 90 m, and K =
1024 m/s. Three cases are considered, homogeneous or
ideal, developed, and clogged. For the developed case,
K = 6 3 1024 m/s for 0.10 < r < 0.24 m and K = 2 3 1024
m/s for 0.24 < r < 0.345 m. For the clogged case, K =
1 3 1025 m/s for 0.10 < r < 0.158 m. The hydraulic head
in the lower portion of the aquifer, h (r, z = 2 m), and the
water table position are plotted against the radial distance,
r, in Figure 13. Effective well development lowers all h
values, especially close to the well. It also lowers the
ð20Þ water table and the seepage face height. However, some
clogging of the screen (or a poor design of the screen and
filter pack if any) produces increased h values, especially
ð21Þ close to the well. It also raises the water table and the
seepage face height.
The well efficiency cannot be evaluated directly
ð22Þ from Figure 13. It must be evaluated using a graph of the
corrected drawdown sc vs. the radial distance (Figure 14)
under steady-state conditions. The corrected drawdown sc
ð23Þ
is defined as:

1.0 sc ¼ s 2 ðs2 =2HÞ ð27Þ

best fit line
R/rw = 900
R/rw = 900* and the corrected drawdown at the well is noted swc.
0.9 R/rw = 600
R/rw = 450
R/rw = 300 10
predicted ratio ha/H

R/rw = 900**
0.8 R/rw = 150 predicted
R/rw = 150* f.e.m.
R/rw = 60
9
R/rw = 30
water table position (m)

0.7 R/rw = 12
8

0.6
7

0.5 6

0.4 5
0.4 0.5 0.6 0.7 0.8 0.9 1.0
ratio ha/H, numerical analyses 4
1 10 100
Figure 11. Comparison of the predicted seepage face height radial distance r (m)
(using Equations 17 and 18) vs. the calculated results for any
value of hw . * and ** mean that other sets of parameters (rw, Figure 12. Comparison of the predicted position of a water
R, H, and K) were used. table (using Equations 17 to 26) with the calculated position.

174 D. Chenaf, R.P. Chapuis GROUND WATER 45, no. 2: 168–177

 9.0 show how the well loss can be evaluated using the cor-
8.5
rected drawdowns of MWs having a short screen close to
h (z = 2 m) and water table position, m
the impervious bottom of the unconfined aquifer.
8.0

7.5
7.0
6.5
6.0
5.5
5.0
4.5
4.0
0.1
Figure 13. Influence of skin effect and determination of well
efficiency (solid symbols for the hydraulic head at the bot-
tom of the aquifer, open symbols for the water table, WT).
In order to illustrate field conditions, Figure 14 was
drawn using the drawdowns measured by four MWs hav-
ing short screens close to the aquifer floor. When the
aquifer is homogenous, the corrected drawdown curve
provided by such MWs is a straight line, which by extrap-
olation intersects the vertical line r = rw at sc = swc. This
illustrates an ideal case where the efficiency is 100%. If
the screen is either inadequate or partially clogged, the
straight line of corrected drawdown crosses the vertical
line r = rw at sc(rw) < swc: the ratio sc(rw)/swc gives the
efficiency of the well. If the well has been correctly de-
signed and successfully developed, the straight line of
corrected drawdown crosses the horizontal line sc = swc
at r(swc) > rw, which indicates a fully efficient well act-
ing with an effective radius r(sw) that is higher than the
physical rw.
Figures 13 and 14 illustrate why the seepage face
should not be considered as a well loss, as correctly
pointed out by Gefell et al. (1994). Figures 13 and 14 also
0.01
0.0
0.5 r = rw
corrected drawdown sc (m)
Discussion and Conclusion
The findings presented in this paper are for steady
state only. They were obtained using the finite-element
method to solve the problem in the saturated and unsatu-
rated zones. According to the numerical results for more
h, homogeneous than 200 conditions, for an ideal well in an ideal uncon-
h, developed
h, clogged fined aquifer, the pumping rate is slightly higher (differ-
WT, homogeneous ence < 5%) than the value given by the Dupuit equation.
WT, developed
WTL, clogged
The water table position, which is higher than that pre-
dicted by the Dupuit equation close to the well, is almost
1 10 100 identical at radial distances greater than 1.5 times the ini-
radial distance r (m) tial saturated thickness. This confirms the usual condi-
tions of textbooks relative to the position of fully
screened MWs to be used with the Dupuit equation to
obtain the saturated K value for an ideal aquifer, even
when the well is not ideal and is affected by well losses.
However, the numerical results indicate that there is no
distance restriction when all MWs have short screens in
the lower third of the saturated thickness. The corrected
drawdowns (Equation 27) of these MWs can be used to
calculate K, estimate R by extrapolation at large dis-
tances, and estimate the well efficiency by extrapolation
at short distances. However, these MWs cannot provide
the water table position close to the well, which only shal-
low MWs can provide.
Six simplified methods and one nondimensional
method (Schneebeli 1956) were available to predict the
seepage face height, ha,min, when hw = 0. The numerical
results confirm that these methods did not consider all
parameters of the seepage face problem. Usually, their
predictions do not compare well with the numerical results.
However, the method of Schneebeli seems to be fair for
the usual operating conditions of wells in homogeneous
unconfined aquifers.
According to the numerical results, the height of the
seepage face is almost independent of the unsaturated
K(uw) function. Two predictive equations have been pro-
posed to predict the height of the seepage face at steady
radial distance r (m)
state. The first nondimensional predictive equation is for
0.1 1 10 100 the maximum seepage face when there is no saturated
thickness in the well (hw = 0). The second predictive
equation is for the general case hw > 0. The values
r(scw) rw predicted by the two equations are very close to the

1.0
numerical values. A good agreement was found for the
1.5 sandbox experiments of Hall (1955) and Simpson et al.
2.0 (2003). This is probably due to the fact that the capillary
sc(rw) fringe effects, which may be more important in a sandbox
2.5
than in the field, were small in the models of Hall (1955)
3.0 where K ’ 4 3 1023 m/s and those of Simpson et al.
homogeneous
3.5 (2003) where K ’ 8 3 1024 m/s.
swc developped The predictive equations depend on the pumping rate
4.0
clogged Q, the well radius rw, the radius of influence R, and the
4.5 saturated hydraulic conductivity K. Since the Dupuit
equation gives an excellent approximation for the flow
Figure 14. Curves of corrected drawdown sc vs. log r to
evaluate the well efficiency under steady-state conditions. rate, the values of K and r(sw)—the effective well radius,
which must be considered instead of rw—can be determined
D. Chenaf, R.P. Chapuis GROUND WATER 45, no. 2: 168–177 175
using a plot of corrected drawdown vs. the log of dis-
tance for several MWs having short screens at the bottom
of the aquifer, without any distance restriction. The same
plot can be extrapolated to estimate the radius of influ-
ence, R.
The results presented herein will be useful for evalu-
ating the performance of pumping wells in unconfined
aquifers, either for drinking water or for ground water
remediation.
Acknowledgments
The research work was subsidized by the Academic
Research Program at the Royal Military College, and
the Natural Sciences and Engineering Research Council
of Canada at École Polytechnique, Montreal. We thank
Philippe Pasquier and Marie-Claude Cormier, who ran
several numerical analyses in the Civil Engineering
Department of the Royal Military College, Kingston,
during work terms of their geological engineering coop
program, and John Molson for reading and commenting
on the article. We thank Michael Gefell, Philippe Renard,
and an anonymous reviewer for their helpful comments.
References
Babbitt, H.E., and D.H. Caldwell. 1948. The free surface
around, and interference between, gravity wells. Bulletin
Series No. 374. Champaign-Urbana, Illinois: University of
Illinois, Engineering Experimental Station.
Boreli, M. 1955. Free-surface flow towards partially penetrating
wells. Transactions, American Geophysical Union 36, no.
4: 664–672.
Boufadel, M.C., M.T. Suidan, A.D. Venosa, and M.T. Bowers.
1999. Steady seepage in trenches and dams: Effect of capillary
flow. Journal of Hydraulic Engineering 125, no. 3: 286–294.
Boulton, N.S. 1951. The flow pattern near a gravity well in a uni-
form water bearing medium. Journal of the Institution of
Civil Engineers (London) 36, no. 10: 534–550.
Brauns, J. 1981. Drawdown capacity of groundwater wells. In
Proceedings of the 10th International Conference on Soil
Mechanics and Foundation Engineering, vol. 1, 391–396.
Stockholm, Sweden: Balkema.
Chapuis, R.P. 1999. Guide d’Interpre´tation des Essais de Pom-
page. Québec, Canada: Les Publications du Québec.
Chapuis, R.P., D. Chenaf, N. Acevedo, D. Marcotte, and M.
Chouteau. 2005. Unusual drawdown curves for a pumping
test in an unconfined aquifer at Lachenaie, Quebec: Field
data and numerical modeling. Canadian Geotechnical
Journal 42, no. 4: 1133–1144.
Chapuis, R.P., D. Chenaf, B. Bussière, M. Aubertin, and
R. Crespo. 2001. A user’s approach to assess numerical
codes for saturated and unsaturated seepage conditions.
Canadian Geotechnical Journal 38, no. 5: 1113–1126.
Charny, I.A. 1951. A rigorous derivation of Dupuit‘s formula
for unconfined seepage with seepage face (in Russian).
Proceedings of the Academy of Sciences, USSR 79, no. 6:
937–940.
Chenaf, D., and R.P. Chapuis. 1998. Étude numérique du pom-
page en régime permanent dans un aquifère à nappe libre.
In Proceedings of the 51st Canadian Geotechnical Con-
ference, 523–528. Edmonton, Canada. Alliston, Ontario:
Canadian Geotechnical Society.
Clement, T.P., W.R. Wise, F.J. Molz, and M. Wen. 1996. A com-
parison of modeling approaches for steady state unconfined
flow. Journal of Hydrology 181, no. 1–4: 189–209.
176 D. Chenaf, R.P. Chapuis GROUND WATER 45, no. 2: 168–177
Darcy, H. 1856. Les Fontaines Publiques de la Ville de Dijon.
Paris, France: Victor Dalmont.
Driscoll, F.G. 1986. Groundwater and Wells, 2nd ed. St. Paul,
Minnesota: Johnson Division.
Dupuit, J. 1863. E´tudes The´oriques et Pratiques sur le Mouve-
ment des Eaux dans les Canaux De´couverts et à Travers les
Terrains Perme´ables, 2nd ed. Paris, France: Dunod.
Dupuit, J. 1857. Mémoire sur le mouvement des eaux dans les
terrains perméables. Mémoire déposé en 1857, rapporté
aux Comptes Rendus des séances de l’Académie des
Sciences, Paris, tome LII, séance du 3 juin 1861.
Gefell, M.J., G.M. Thomas, and S.J. Rossello. 1995. Maximum
water-table drawdown at a fully penetrating pumping well:
Reply. Ground Water 33, no. 3: 502.
Gefell, M.J., G.M. Thomas, and S.J. Rossello. 1994. Maximum
water-table drawdown at a fully penetrating pumping well.
Ground Water 32, no. 3: 411–419.
Geo-slope International. 2003. SEEP/W User’s Guide, Version 5.
Calgary, Canada: Geo-slope International.
Hall, H.P. 1955. An investigation of steady flow toward a grav-
ity well. La Houille Blanche 10, no. 1: 8–35.
Heinrich, G. 1964. EineNäherung für die Freie Spiegel-fläche
beim Vollkommenen Brunnen. Österreichische Wasserwirt-
schaft 16, no. 12: 15–20.
Kawecki, M.W. 1995. Maximum water-table drawdown at a fully
penetrating pumping well: Discussion. Ground Water 33,
no. 3: 498–499.
Kovàcs, G. 1981. Seepage Hydraulics. Amsterdam, The Nether-
lands: Elsevier.
Kozeny, J. 1953. Hydraulik. Wien, Austria: Springer-Verlag.
Kruseman, G.P., and N.A. de Ridder. 1990. Analysis and Evalu-
ation of Pumping Test Data, Pub. 47, Wageningen, The
Netherlands: International Institute for Land Reclamation
and Improvement.
Lee, K.K., and D.I. Leap. 1997. Simulation of a free-surface and
seepage face using boundary-fitted coordinate system
method. Journal of Hydrology 196, no. 1–4: 297–309.
Muskat, M., and R.D. Wyckoff. 1937. The Flow of Homoge-
neous Fluids through Porous Media. New York: McGraw
Hill.
Ojha, C.S.P., and V. Gopal. 1999. Seepage face modeling for
large-diameter well in unconfined aquifer. Journal of Hydro-
logic Engineering 4, no. 3: 275–279.
Polubarinova-Kochina, P.Y. 1962. Theory of Ground-Water Move-
ment. Translated from Russian by J.M. DeWiest. Princeton,
New Jersey: Princeton University Press.
Richards, L.A. 1931. Capillary conduction of liquids through
porous medium. Physics 1, no. 5: 318–333.
Rushton, K.R., and S.M. Holt. 1981. Estimating aquifer para-
meters for large-diameter wells. Ground Water 19, no. 5:
505–509.
Rushton, K.R., and K.W.F. Howard. 1982. The unreliability of
open observation boreholes in unconfined aquifer pumping
tests. Ground Water 20, no. 5: 546–550.
Rushton, K.R., and V.S. Singh. 1983. Drawdown in large-diame-
ter wells due to decreasing abstraction rates. Ground Water
21, no. 6: 670–677.
Sakthivadivel, R., and K.R. Rushton. 1990. Numerical analysis
of large diameter wells with a seepage face—Reply. Jour-
nal of Hydrology 119, no. 1–4: 398–399.
Sakthivadivel, R., and K.R. Rushton. 1989. Numerical analysis
of large diameter wells with a seepage face. Journal of
Hydrology 107, no. 1–4: 43–55.
Schneebeli, G. 1956. Sur l’hydraulique des puits. In Symposia
Darcy, Publication No. 41, tome 2, 10–27. Gentbrugge,
Belgium: Association Internationale d’Hydrologie Scienti-
fique.
xen, Z. 1990. Numerical analysis of large diameter wells with
S
a seepage face—Comment. Journal of Hydrology 119, no.
1–4: 393–398.
xen, Z., and M. Cximen. 2001. Seepage face modeling for large-
S
diameter well in unconfined aquifer: Discussion. Journal of
Hydrologic Engineering 6, no. 4: 363.
Shamsai, A., and T.N. Narasimhan. 1991. A numerical investi- a unit gradient (i = 1) along the seepage face. Thus, ac-
gation of free surface-seepage face relationship under cording to Darcy’s law:
steady state flow conditions. Water Resources Research 27,
no. 3: 409–421. Q ¼ AvDarcy ¼ ð2prw ha;min ÞðK 3 IÞ ð1:1Þ
Sichardt, W. 1928. Das Fassungsvermögen von Rohrbrunnen
und seine Bedeutung für die Grundwasserabsenkung, in-
sbesondere für größere Absenkungstiefen. Berlin, Germany:
The pumping rate may be obtained using the Dupuit
Julius Springer. equation as:
Simpson, M.J., T.P. Clement, and T.A. Gallop. 2003. Laboratory
and numerical investigation of flow and transport near H 2 2 h2a;min
a seepage-face boundary. Ground Water, 41, no. 5: 690–700. Q ¼ pK ð1:2Þ
lnðR=rw Þ
USDI. 1977. Ground Water Manual. Washington, D.C.: U.S.
Department of the Interior. Superintendent of Documents,
Equating the two expressions for the pumping rate
U.S. Government Printing Office.
van der Kamp, G. 1985. Brief quantitative guidelines for the gives:
design and analysis of pumping tests. In Hydrology in the  
Service of Man. Memoirs of 18th Congress IAH, 197–206. ha;min =H ¼ h=2rw lnðR=rw Þ 1 2 ðha;min = HÞ2 ð1:3Þ
Cambridge.
van Dijke, M.I.J., and S.E.A.T.M. van der Zee. 1998. Analysis This can be simplified by defining:
of oil lens removal by extraction through a seepage face.
Computational Geosciences 2, no. 1: 47–72. x ¼ ha;min =H ð1:4Þ
Waddill, D.W., and J.C. Parker. 1997. Recovery of light, non-
aqueous phase liquid from porous media: Laboratory
experiments and model validation. Journal of Contaminant b ¼ rw lnðR=rw Þ=H ð1:5Þ
Hydrology 27, no. 1–2: 127–155.
Wise, W.R., and T.P. Clement. 1995. Maximum water-table
drawdown at a fully penetrating pumping well: Discussion. which gives for Equation 1.3
Ground Water 33, no. 3: 499–501.
Zee, C.-H., D.F. Peterson, and R.O. Bock. 1955. Flow into x ¼ ð1=2bÞð1 2 x2 Þ ð1:6Þ
a well by electrical and membrane analogy. Proceedings
ASCE 81, October: 817.1–817.21.
x2 1 2bx 2 1 ¼ 0 ð1:7Þ

Appendix 1 Only the positive root is physically admissible

0:5
x ¼ ðb2 11Þ 2 b or ð1:8Þ
Derivations of Equations 9 and 10
When the well drawdown is maximum, Kozeny h i
0:5
(1953) assumed that flow into the empty well occurs at x ¼ b ð12b22 Þ 2 1 ð1:9Þ

D. Chenaf, R.P. Chapuis GROUND WATER 45, no. 2: 168–177 177