Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Identification of aquifer and well parameters

from step-drawdown tests

Cem B. Avci & Emin Ciftci & A. Ufuk Sahin

Abstract A new analysis technique has been proposed for necessary to evaluate the efficiency, specific capacity and
interpreting transient step-drawdown test data. The pro- safe yield of pumping wells. Jacob (1947) introduced the
posed method is based on taking the derivative of the concept of a step-drawdown test as a single well-perform-
drawdown with respect to time for the entire pumping test ance test in order to assess the impact of the well discharge
period to eliminate the time-independent well-loss terms. on the drawdown measured in the pumping well. The test
The derivative function is subsequently integrated to provides information on the yield of an extraction well
obtain the time-dependent aquifer drawdown as a con- which is important in determining the optimum pump and
tinuous function. The well-loss parameters are then the depth of pumping (Bouwer 1978). In a step-drawdown
obtained with higher accuracy once the aquifer behavior test the pumping rate is increased in a step-wise manner
is identified. The proposed method is applicable for during successive periods of time (Bear 1979).
analyzing data obtained not only from ideal confined Step-drawdown tests can be conducted either under
aquifers but also from other aquifer types (i.e. unconfined) steady state or transient conditions. The Thiem equation
and non-ideal aquifers (i.e. heterogeneous). The technique (Thiem 1906) can be employed for evaluating the
was tested for synthetically generated and field data; the formation loss for the steady-state case (Todd 1980). The
proposed approach was noted to provide accurate aquifer specific capacity of a well, defined as the ratio of its
and well-loss parameter estimates. The results of the discharge to the steady-state drawdown, is correlated with
proposed method were compared with those of some of the aquifer transmissivity on the basis of the Thiem
the existing methods for analyzing step-drawdown test equation (Kruseman and de Ridder 1990). The correlation
data and were found to be more reliable and robust. between the specific capacity and the transmissivity for
different aquifer types was studied by a number of
Keyword Step-drawdown test . Well loss . Derivative investigators (Razack and Huntley 1991; Huntley et al.
analysis . Groundwater hydraulics . Hydraulic testing 1992; El-Naqa 1994; Mace 1997; Verbovsek 2008). Due
to the fact that pumps are usually operated intermittently,
the steady-state step-drawdown tests requiring long periods
of time may not be appropriate (Helweg 1994). In the
Introduction unsteady-state step-drawdown tests, the productivity of the
well varies not only with the discharge but also with time.
Well-performance tests are conducted in order to estimate The Theis well function (Theis 1935) or the Cooper-Jacob
the energy losses in the aquifer and the pumping well that approximation (Cooper and Jacob 1946) can be utilized
develop during groundwater extraction. The drawdown in a for assessing the formation loss (Batu 1998). The typical
pumping well consists of the head loss due to the laminar duration of each step may vary between 30–120 min
flow as well as the head loss resulting from the turbulent (Kruseman and de Ridder 1990); the test is recommended
flow of water through the well screens and the pump intake to be conducted for at least four steps to ensure accuracy
(Batu 1998). The estimation of these head losses is (Clark 1977). Pumping rates are taken to be constant
during each step testing period as a common practice;
Received: 14 October 2009 / Accepted: 25 May 2010 however, Singh (2002) suggested that a variable discharge
Published online: 12 June 2010 test can appropriately replace the conventional step-draw-
© Springer-Verlag 2010
down test, thus allowing the pumping discharge to be
changed more frequently without waiting for the 30–120-
min step period.
C. B. Avci ()) : E. Ciftci : A. U. Sahin Jacob (1947) proposed the following formulation for
Civil Engineering Department, the drawdown in the pumping well under transient
Bogazici University,
34342, Istanbul, Turkey conditions;
e-mail: avci@boun.edu.tr
Tel.: +90-2123596410
Fax: +90-2122872457 sw ðtÞ ¼ Bðrew ; t; T ; S ÞQ þ CQ2 ð1Þ

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2


1592
Table 1 The step-drawdown data for the ideal confined aquifer
Predefined well-loss parameters: C=1E-6, p=2.4
Step 1 2 3 4 5 6
3
Discharge rate (m /day) 100 200 300 400 500 600
Time from beginning of step (min) Drawdown (m)
2 0.814 2.107 3.733 5.719 8.097 10.895
4 0.869 2.165 3.792 5.779 8.157 10.956
6 0.901 2.200 3.828 5.816 8.195 10.994
8 0.924 2.225 3.855 5.843 8.223 11.022
10 0.942 2.245 3.876 5.866 8.245 11.046
15 0.974 2.283 3.917 5.908 8.290 11.091
20 0.997 2.311 3.948 5.941 8.324 11.127
25 1.015 2.334 3.973 5.969 8.353 11.157
30 1.029 2.353 3.995 5.992 8.378 11.183
35 1.041 2.369 4.014 6.013 8.401 11.207
40 1.052 2.384 4.031 6.032 8.421 11.229
45 1.061 2.397 4.047 6.050 8.440 11.249
50 1.070 2.409 4.061 6.066 8.458 11.268
55 1.077 2.420 4.075 6.081 8.474 11.285
60 1.084 2.431 4.087 6.095 8.490 11.302

where B(rew, t, T, S) is a continuous function defining the Birsoy and Summers (1980) and Kawecki (1995) pre-
aquifer properties, C represents the nonlinear well-loss sented alternative graphical interpretation methods to
coefficient, Q is the discharge rate, t denotes the pumping determine aquifer well losses. Labadie and Helweg
time, rew, is the effective well radius, T and S are the (1975) performed a least-squares curve-fitting analysis
transmissivity and the storativity of the aquifer, respec- for determining the values of B, C and p. The method
tively. The total drawdown sw(t) consists of the aquifer proposed by Yeh (1989) to determine the well-loss
loss B(rew, t, T, S)Q and the well loss CQ2. parameters was based on a nonlinear least-squares
Rorabaugh (1953) generalized Jacob’s formulation by approach and the finite-difference method of Newton.
introducing CQp as the nonlinear well-loss term, where p Miller and Weber (1983) derived analytical expressions
could essentially be within the range of 2.4–2.8. Lennox for the evaluation of these parameters. Gupta (1989), Avci
(1966) suggested that the well-loss exponent p may be as (1992) and Helweg (1994) proposed alternative numerical
high as 3.5. Sheahan (1971) proposed the range of 1–4 for techniques for analyzing step-drawdown data. Shekhar
p. Singh (2002) indicated that a well-loss exponent equal (2006) suggested fitting a trend line on the drawdown
to 2 should be considered. Rorabaugh (1953) developed a versus discharge plot to approximate the equation of
graphical trial-and-error approach for the evaluation of the Rorabaugh (1953) and so to assess the well-loss param-
well-loss components. Sheahan (1971) developed a eters. Jha et al. (2004) introduced the use of a genetic
method for direct analysis of step-drawdown data using algorithm for the estimation of loss parameters. Jha et al.
type curves. Bierschenk (1963), Eden and Hazel (1973), (2006) conveyed that both the traditional gradient-based

Table 2 The loss parameters calculated for each iteration step for the ideal confined aquifer test case
Iteration number p C Bo RMSE (m) Iteration number p C Bo RMSE (m)
1 1.100 2.23E-02 −0.0280 0.3136 16 2.550 3.75E-07 0.0078 0.0287
2 1.600 2.33E-04 0.0038 0.1804 17 2.500 5.20E-07 0.0077 0.0195
3 2.100 7.32E-06 0.0067 0.0618 18 2.450 7.22E-07 0.0076 0.0103
4 2.600 2.70E-07 0.0078 0.0378 19 2.400 1.00E-06 0.0075 0.0028
5 3.100 1.05E-08 0.0085 0.1191 20 2.350 1.39E-06 0.0074 0.0100
6 3.050 1.45E-08 0.0084 0.1117 21 2.355 1.35E-06 0.0074 0.0090
7 3.000 2.00E-08 0.0084 0.1041 22 2.360 1.30E-06 0.0074 0.0081
8 2.950 2.77E-08 0.0083 0.0964 23 2.365 1.26E-06 0.0074 0.0072
9 2.900 3.83E-08 0.0083 0.0885 24 2.370 1.22E-06 0.0074 0.0063
10 2.850 5.30E-08 0.0082 0.0805 25 2.375 1.18E-06 0.0074 0.0054
11 2.800 7.34E-08 0.0081 0.0723 26 2.380 1.14E-06 0.0074 0.0046
12 2.750 1.02E-07 0.0081 0.0639 27 2.385 1.11E-06 0.0075 0.0039
13 2.700 1.41E-07 0.0080 0.0554 28 2.390 1.07E-06 0.0075 0.0033
14 2.650 1.95E-07 0.0079 0.0467 29 2.395 1.04E-06 0.0075 0.0029
15 2.600 2.70E-07 0.0078 0.0378 30 2.400 1.00E-06 0.0075 0.0028

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2


1593
Table 3 Estimated parameters for the ideal confined-aquifer test case
Parameters Labadie and Helweg (1975) Birsoy and Summers (1980) Gupta (1989) Present method Predefined values
p 2.289 2.399 2.400 2.400 2.400
C 2.318E-06 9.963E-07 1.000E-06 1.003E-06 1.000E-06
T (m2/day) – 99.98 100.02 100.1 100

optimization techniques and the non-traditional optimiza- drawdown data with respect to time and then integrating
tion technique (genetic algorithm) provide an efficient and the derivative curve back to establish the B(t) function in
reliable tool for identifying loss parameters through step- order to get a more complete picture of the aquifer
drawdown data. behavior. The origin of the derivative analysis goes back
A review of these solution techniques shows that the to the field of petroleum engineering (Wong et al. 1986;
majority of the proposed methods assume a priori Ehlig-Economides 1988; Ostrowski and Kloska 1988;
behavior of the aquifer such as a linear drawdown Bourdet et al. 1989; Parks and Bentley 1996). Today,
behavior with respect to the logarithm of time in each derivative analysis of aquifer testing data has been
test step (confined aquifer assumption) or a stabilized accepted and is extensively applied. Singh (2001) pro-
drawdown level developing at the end of each pumping posed a method which makes use of the temporal
step. This assumption leads to the utilization of Cooper derivative of drawdowns for the evaluation of confined
and Jacob’s approximation (1946). Singh (2002) took into aquifer parameters utilizing the early drawdown values.
account the nonlinear behavior of drawdown by utilizing The use of the drawdown derivative as a diagnostic plot
the Theis well function rather than the Cooper-Jacob together with its advantages and disadvantages has been
approximation. As indicated by Karami and Younger discussed in detail by Renard et al. (2009).
(2002), one of the drawbacks of the existing methods
was that the tested aquifer was assumed to be homoge-
neous even though it might be highly heterogeneous. Methodology
Karami and Younger (2002) pointed out that ignoring the
heterogeneous behavior of the aquifer would lead to Assessment of the aquifer behavior
misinterpretation of the aquifer as well as the well-loss In a step-drawdown test, the drawdown observed in the
parameters. first test step, s1, can be calculated by the following
A method has been proposed with the aim to establish equation proposed by Rorabaugh (1953);
the aquifer behavior given as B(t) in Eq. 1 for the whole 0
duration of the step-drawdown testing period followed by sðt Þ ¼ BðtÞQ1 þ CQ1p t1  t  t1 ð2Þ
the estimation of the well-loss parameters. The estimated
B(t) curve provides a larger database compared to those where B(t) is the continuous hydraulic function defining
0

provided by the traditional step-drawdown test analysis the aquifer behavior under stressed conditions, t 1 and t 1
methods for identifying the aquifer parameters. The stand for the times of the first and the last drawdown
method is based on calculating the derivative of the measurements in the first test step, respectively, and CQ1p
represents the nonlinear well-loss component. With an
increase in the discharge rate of Q2–Q1 for the second test
step period, the drawdown for the second pumping period
can be calculated using the principle of superposition
(Bear 1979) as :

sðt Þ ¼ Bðt  t 1 ÞðQ2  Q1 Þ þ BðtÞQ1 þ CQ2 p


0
ð3Þ
t2  t  t2

where t 2′ and t 2 stand for the times of the first and the last
drawdown measurements in the second test step, respec-
tively. Equation 3 can be generalized to calculate the
drawdown for the ith test step to be:
X
i1
sðtÞ ¼ BðtÞQ1 þ CQi p þ ½Bðt  t k ÞðQkþ1  Qk Þ ð4Þ
k¼1
0
t1  t  ti i ¼ 2; 3 ; . . . ; N
Fig. 1 The calculated B(t)Q1 curve for the ideal confined aquifer

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2


1594
Table 4 The drawdown data from the confined sandstone aquifer (Clark 1977)
Step 1 2 3 4 5 6
3
Discharge rate (m /day) 1,306 1,693 2,423 3,261 4,095 5,016
Time from beginning of step (min) Drawdown (m)
7 3.117 5.669 8.557 12.149 15.952 20.741
8 3.345 5.705 8.592 12.184 16.022 20.811
9 3.486 5.740 8.672 12.219 16.022 20.882
10 3.521 5.740 8.672 12.325 16.093 20.917
12 3.592 5.810 8.663 12.360 16.198 20.952
14 3.627 5.810 8.698 12.395 16.268 21.022
16 3.733 5.824 8.733 12.430 16.304 21.128
18 3.768 5.845 8.839 12.430 16.374 21.163
20 3.836 5.810 8.874 12.501 16.409 21.198
25 3.873 5.824 8.874 12.508 16.586 21.304
30 4.014 5.824 8.979 12.606 16.621 21.375
35 3.803 5.881 8.979 12.712 16.691 21.480
40 4.043 5.591 8.994 12.747 16.726 21.551
45 4.261 5.591 9.050 12.783 16.776 21.619
50 4.261 6.092 9.050 12.818 16.797 21.656
55 4.190 6.092 9.120 12.853 16.902 21.660a
60 4.120 6.176 9.120 12.853 16.938 21.663
70 4.120 6.162 9.155 12.888 16.973 21.691
80 4.226 6.176 9.191 12.923 17.079 21.762
90 4.226 6.169 9.191 12.994 17.079 21.832
100 4.226 6.169 9.226 12.994 17.114 21.903
120 4.402 6.176 9.261 13.099 17.219 22.008
150 4.402 6.374 9.367 13.205 17.325 22.184
180 4.683 6.514 9.578 13.240 17.395 22.325
a
Interpolated data

where N is the total number of the test steps, t i′ and t i The Bo term can be substituted into the general
denote the times at which the first and the last drawdown drawdown expression described in Eq. 4 to give;
measurements were taken in the ith test step, respectively.
From Eq. 2, it is possible to obtain the variation B(t)
function with time as follows: iP
1
sðt Þ ¼ B ðt ÞQ1 þ Bo Qi þ CQi p þ ½B ðt  t k ÞðQkþ1  Qk Þ
k¼1
@Bðt Þ 1 @sðt Þ 0 0
ti  t  ti i ¼ 2; 3; . . . ; N
¼ t1  t  t1 ð5Þ
@t Q1 @t
ð8Þ
and the B(t) curve within the first test step can be obtained
as; By taking the time derivative of Eq. 8 first and then the
Z integral of the result obtained, it is possible to construct
1 t
@sðt Þ 0 the complete B*(t) curve for the duration of all test steps
Bðt Þ ¼ Bo þ dt t1  t  t1 ð6Þ
Q1 t1
0 @t as;

where Bo is an unknown constant representing the value of


the B function at the time of the 0first drawdown measure-
ment of the first test step (at t= t 1 ). If the last term on the @B ðtÞ 1 @sðtÞ
¼
right hand side of the above equation is replaced by a new @t Q1 @t
function, B*(t), the following identity can be obtained; i1 
X 
ðQkþ1  Qk Þ @ 
 ðB ðt  t k ÞÞ ð9Þ
Bðt Þ ¼ Bo þ B ðt Þ ð7Þ k¼1
Q1 @t

Table 5 Estimated parameters for the confined sandstone aquifer (Clark 1977)
Parameters Labadie and Helweg (1975) Gupta (1989) Avci (1992) Singh (2002) Present method
p 2.100 2.709 2.005 2.191 2.540
C 8.14E-08 2.72E-10 1.43E-07 2.63E-08 1.74E-08
T (m2/day) – 250.9 280.7 252.4 261.6
RMSE (m) – 0.094 0.090 0.067 0.035

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2


1595

Fig. 2 Calculated vs. observed drawdown along 1:1 line (Clark Fig. 3 The calculated B(t)Q1 curve (Clark 1977)
1977)

and where sij and Bij* represent the observed drawdown and
the value of the B*(t) function at the jth time level of the ith
Z t
@B ðt Þ 0 test step, respectively. N and m stand for the number of
B ðt Þ ¼ B ðt i1 Þ þ dt t i  t  t i test steps and the number of time levels in each test step,
ti
0 @t
ð10Þ respectively. An iterative linear least-squares optimization
procedure is performed using an algorithm similar to the
i ¼ 2; 3; . . . ; N one developed by Labadie and Helweg (1975). The p
parameter was reported to take values between 1 and 4
A central difference formulation and the composite (Sheahan 1971). The optimization algorithm follows the
trapezoidal integration rule can be employed for the logic that the objective function has a minimum for the
evaluation of the derivative and the integral operations interval defined for p. As an initial guess, p is chosen to be
given in Eqs. 9 and 10, respectively. 1.1, and for this given p value, corresponding C and Bo
values minimizing the objective function are obtained
through the solution of the linear system constructed by
Assessment of the well-loss parameters taking the partial derivatives of the objective function with
Once the B*(t) curve is constructed, the unknown respect to C and Bo and equating them to zero as described
parameters Bo, C and p in Eq. 8 can be obtained through in the following equations;
the minimization of the following objective function ξ
which measures the squared error between the observed m  
@x P
B1j þ Bo þ CQ1p1  Q1j1
s
drawdown (left-hand side of Eq. 8) and the estimated @Bo ¼
drawdown (right-hand side of Eq. 8); j¼1
 
PN P
m P h  ðQkþ1 Qk Þ i sij
i1
m  2 þ Bij QQ1i þ Bo þ CQi p1 þ Bikj 
P Qi Qi
B1j þ Bo þ CQ1p1  Q1j1
s
x¼ i¼2 j¼1 k¼1
j¼1
 2 ¼0
N P
P m P h  ðQkþ1 Qk Þ i sij
i1
ð12Þ
þ Bij QQ1i þ Bo þ CQi p1 þ Bikj Qi  Qi
i¼2 j¼1 k¼1
and
ð11Þ
m 
P 
@x
B1j þ Bo þ CQ1p1  Q1j1 Q1p1
s
@C ¼
j¼1
 
P
N P
m P h  ðQkþ1 Qk Þ i sij
i1
þ Bij QQ1i þ Bo þ CQi p1 þ Bikj Qi  Qi Qi p1
Table 6 The predefined parameters for the hypothetical homoge- i¼2 j¼1 k¼1
neous unconfined aquifer
¼0 ð13Þ
Initial saturated aquifer thickness, b 10 m
Horizontal hydraulic conductivity, Kr 5×10−5 m/s
Vertical hydraulic conductivity, Kz 5×10−5 m/s
Specific storage, Ss 10−5 1/m
Specific yield, Sy 0.1 Using the parameters attained in each iteration step, the
Pumping rate, Q 100 m3/day drawdown values are re-calculated for assessing the root

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2


1596
Table 7 The synthetically generated step-drawdown data for the hypothetical homogeneous unconfined aquifer
Predefined well-loss parameters: C=1E-7, p=3
Step 1 2 3 4 5 6
3
Discharge rate (m /day) 100 125 150 175 200 225
Time from beginning of step (min) Drawdown (m)
2 1.533 2.272 2.942 3.662 4.447 5.311
4 1.620 2.296 2.967 3.686 4.471 5.334
6 1.633 2.303 2.972 3.691 4.477 5.340
8 1.640 2.306 2.976 3.695 4.481 5.344
10 1.647 2.311 2.981 3.700 4.485 5.349
20 1.679 2.332 3.000 3.719 4.505 5.369
30 1.705 2.351 3.017 3.737 4.523 5.387
40 1.728 2.367 3.034 3.753 4.539 5.403
50 1.749 2.384 3.049 3.768 4.555 5.419
60 1.768 2.399 3.063 3.783 4.570 5.435
70 1.785 2.414 3.077 3.797 4.584 5.449
80 1.801 2.427 3.090 3.810 4.598 5.463
90 1.815 2.439 3.102 3.822 4.610 5.476

mean squared error (RMSE) between the observed and present analysis: 0.0001). A standalone FORTRAN-
calculated drawdown as follows based program (mathematical formula translation sys-
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi tem) was developed for the execution of the explained
1X n
2 optimization procedure.
RMSE ¼ sio  sic ð14Þ The B(t) curve can provide significant information on
n i¼1 the aquifer being stressed during the step-drawdown test.
By examining the shape of the curve, the aquifer type can
where n denotes the number of observed drawdowns, so be assessed in the same manner as assessing a typical
and sc represent the observed and the calculated draw- time-drawdown shape from an arbitrary monitoring well.
down, respectively. For the next iteration step, the p Once the aquifer type (whether confined, unconfined or
value is increased by an amount of Δp whose initial non ideal) is identified based on the time-drawdown
value is chosen to be 0.5. When the RMSE value behavior, conventional solution techniques can be imple-
obtained in an iteration step is greater than that in the mented with the use of B(t) values to obtain the aquifer
previous step, the search direction is reversed and Δp is parameters such as the aquifer transmissivity. For instance,
set to 0.1Δp. The iteration procedure is carried on until if the B(t) curve obtained for the whole aggregate data set
the difference between the p values from successive behaves like the drawdown-time curve from a confined
iteration steps is smaller than an acceptable error (for aquifer, the traditional Theis (1935) or Cooper-Jacob
(1946) methods can be employed for the evaluation of
the aquifer parameters.

Results and discussion


In order to test the validity of the proposed method,
synthetically generated step-drawdown data sets from
hypothetical confined, unconfined and confined heteroge-
neous aquifers were investigated. In addition, field data
sets obtained from a confined sandstone aquifer (Clark
1977) and an unconfined basin-fill aquifer (Maurer 2006)
were assessed as part of the validity testing.

Confined aquifer tests


The proposed method was first implemented on hypo-
Fig. 4 The calculated B(t)Q1 curve for the hypothetical homoge- thetical drawdown data generated for a homogeneous
neous unconfined aquifer confined aquifer. The data were generated using the Theis

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2


1597
Table 8 Estimated parameters for the hypothetical homogeneous unconfined aquifer test case
Parameters Labadie and Helweg Birsoy and Summers Gupta Avci Present Predefined
(1975) (1980) (1989) (1992) method values
p 2.032 2.794 2.648 2.802 2.993 3
C 4.01E-05 3.21E-05 8.58E-07 3.52E-07 1.04E-07 1.00E-07
T (m2/day) – 76.5 74.1 77.0 44.2 43.2

(1935) equation. The Theis well function was calculated The step-drawdown data reported by Clark (1977)
via using the first 1,000 terms of the series expansion for given in Table 4 were utilized as the second confined
the first-order exponential integral (Abramowitz and aquifer data set to be analyzed with the proposed method.
Stegun 1965). The calculated drawdown values were The test was reportedly carried out in a confined sandstone
taken to be the values of the B(t)Q1 function; these values aquifer where six pumping steps were conducted to collect
were substituted into Eqs. 2 and 4 along with the the data (Clark 1977).
predefined loss parameters and the discharge rates to The results from the proposed method are shown in
generate the data shown in Table 1. The loss parameters Table 5 together with the results obtained from existing
were selected to be within the ranges defined in previous analysis methods. The comparison between the observed
investigations (Walton 1962; Sheahan 1971) drawdown and the calculated drawdown based on the
The convergence of the parameter values obtained at estimated B(t)Q1 curve and the well-loss parameters, is
each iteration step is presented in Table 2. Table 3 presented in Fig. 2. The results of the present investigation
summarizes the well loss and aquifer parameters calcu- fall within the range of well-loss coefficients estimated by
lated with both the present method and the methods other investigators. The RMSE measuring the discrepancy
developed by Labadie and Helweg (1975), Birsoy and between the observed and the calculated drawdown is
Summers (1980) and Gupta (1989). Figure 1 demonstrates seen to be lower for the present method when compared
the calculated B(t)Q1 curve utilized in the evaluation of with the methods of Gupta (1989), Avci(1992) and Singh
aquifer parameters. As seen in Fig. 1, the dimensionless (2002). The aquifer parameters were computed by utiliz-
time was sufficiently small (u<0.01), hence the trans- ing the B(t)Q1 values obtained as part of the analysis as
missivity of the aquifer was estimated via the traditional shown in Fig. 3. It can be seen that the aquifer does indeed
Cooper and Jacob (1946) method. The results indicate that display confined characteristics throughout the entire step-
the proposed method works accurately to retrieve the well- drawdown testing period. A straight line can be fitted
loss parameters as well as aquifer drawdown behavior B(t) through the drawdown behavior estimated by the present
for ideal confined aquifer test data. method of analysis for the entire testing period. The
drawdown data reflect a straight line behavior starting
from the 10th minute of the test until the end of the
Table 9 The drawdown data from the unconfined basin-fill aquifer pumping period.
(Maurer 2006)
Step 1 2 3
Discharge rate (m3/day) 828 1,656 2,484 Unconfined aquifer tests
Time from beginning of step (min) Drawdown (m) The proposed method was implemented for a hypothetical
homogeneous unconfined aquifer data set generated by
4 3.935 9.284 15.572
5 4.008 9.394 15.697 WTAQ (Barlow and Moench 1999), a FORTRAN-based
6 4.045 9.507 15.840 program for analyzing confined or water table aquifers.
7 4.112 9.580 15.917
8 4.161 9.632 15.990
9 4.191 9.683 16.054
10 4.240 9.738 16.157
15 4.380 10.031 16.398 Table 10 Estimated parameters for the unconfined basin-fill aquifer
20 4.456 10.183 16.517 (Maurer 2006)
30 4.587 10.317 16.636 Parameters Labadie and Gupta Avci Present
45 4.779 10.464 16.730 Helweg (1975) (1989) (1992) method
60 4.883 10.540 16.935
75 4.968 10.671 17.136 p 2.35 2.73 2.29 2.75
90 4.993 10.723 17.285 C 3.17E-08 1.07E- 2.45E- 3.48E-10
105 5.032 10.799 17.383 09 08
120 5.060 10.906 17.563 T (m2/day) – 181.6 166.0 50.8
150 5.124 10.955 17.700 RMSE 0.210 0.098 0.052
180 5.157 11.034 17.846 (m)

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2


1598

Fig. 5 The calculated B(t)Q1 curves for the unconfined basin-fill Fig. 6 Calculated vs. observed drawdown along 1:1 line (Maurer
aquifer (Maurer 2006) 2006)

The specified parameters required for the generation of drawdown analysis methods overestimate the transmissiv-
synthetic data are presented in Table 6. The generated data ity of the aquifer compared to the proposed method as in
were then substituted into Eqs. 2 and 4 to obtain the the case for the analysis of the synthetically generated data
synthetic step-drawdown test data given in Table 7. set.
Once the B(t)Q1 curve was constructed, Neuman’s
semilogarithmic method (1975) was utilized on the late
time data to estimate the transmissivity of the aquifer as
illustrated in Fig. 4. The delayed yield phase represented
by the initial line segment (up to 102 min) and the late-
time phase through which the regression line was passed
can be observed in Fig. 4. Table 8 summarizes the
estimations of the well-loss parameters as well as the
aquifer transmissivity. The nonlinear behavior of the B(t)
Q1 curve with respect to the logarithm of time causes the
step-drawdown test methods relying on traditional con-
fined aquifer-based analysis techniques to suffer from lack
of accuracy. On the other hand, the results obtained with
the proposed method can be seen to be in good agreement
with the actual predefined values. The unconfined aquifer
test case demonstrates that the traditional step-drawdown
analysis methods over-estimate the transmissivity of the
aquifer, whereas the proposed method provides a better
estimate of this parameter by utilizing the late-time
segment (not the delayed yield segment) of the calculated
B(t)Q1 curve.
The unconfined aquifer data set (Table 9) obtained
from pumping in a 90-m thick unconfined basin-fill
aquifer in Carson Valley, Nevada (Maurer 2006) was
tested with the proposed method. The estimated parame-
ters are tabulated in Table 10. The calculated B(t)Q1 curve
shown in Fig. 5 clearly displays the unconfined nature of
the aquifer. As shown in Fig. 6, the estimated and the
observed drawdowns match well, giving the lowest RMSE
among the compared methods (Table 10). The unconfined- Fig. 7 a Log-transmissivity (m2/day) and b log-storativity fields
aquifer-field-data analysis shows that the traditional step- for the hypothetical heterogeneous confined aquifer

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2


1599
Table 11 The step-drawdown data for the hypothetical heterogeneous confined aquifer
Predefined well-loss parameters: C=1E-6, p=2.8
Step 1 2 3 4 5 6
3
Discharge rate (m /day) 50 75 100 125 150 175
Time from beginning of step (min) Drawdown (m)
2 5.939 10.893 15.270 19.748 24.386 29.209
4 6.421 11.150 15.524 20.002 24.640 29.464
6 6.603 11.255 15.627 20.106 24.744 29.568
8 6.730 11.333 15.703 20.182 24.821 29.645
10 6.836 11.400 15.767 20.246 24.886 29.710
15 7.038 11.531 15.894 20.374 25.014 29.840
20 7.197 11.638 15.998 20.479 25.120 29.946
25 7.327 11.727 16.085 20.566 25.209 30.036
30 7.434 11.803 16.160 20.641 25.285 30.113
35 7.524 11.869 16.224 20.706 25.351 30.180
40 7.600 11.926 16.280 20.763 25.409 30.239
45 7.665 11.976 16.329 20.814 25.460 30.291
50 7.722 12.020 16.373 20.859 25.506 30.338
55 7.772 12.060 16.412 20.899 25.547 30.380
60 7.815 12.095 16.448 20.936 25.585 30.419

Heterogeneous aquifer test demonstrated in Fig. 8 where drawdown data do not


A step drawdown simulation was performed for a hypo- behave linearly with the logarithmic time scale.
thetical heterogeneous confined aquifer by means of a
PMWIN-MODFLOW (Chiang and Kinzelbach 2001)
numerical model. The field generator in PMWIN-MOD-
FLOW was used to generate the transmissivity and Conclusions
storativity fields. In the light of previous investigations
(Freeze 1975; Delhomme 1979; Clifton and Neuman A new analysis technique has been proposed for interpret-
1982; Hoeksema and Kitanidis 1985; Dagan 1989; ing transient step-drawdown test data. The proposed
Hantush and Marino 1990), the aquifer transmissivity method is based on taking the derivative of the drawdown
and storativity were assumed to be lognormally distrib- with respect to time for the entire pumping test period,
uted. The log-transmissivity mean was determined to be eliminating the constant well-loss terms. The derivative
2 m2/day and the log-transmissivity variance was defined function is subsequently integrated to obtain the time-
as 2 m2/day; the mean and the variance of the log dependent aquifer drawdown as a continuous function.
storativity were chosen to be −4 and 2, respectively, and a The method was tested on synthetically generated data as
correlation length of 10 m was specified for both well as field data for confined and unconfined aquifers.
lognormal fields. The transmissivity and storativity fields Since the logic of the method is based on the
are shown in Fig. 7 and the generated step-drawdown data establishment of the actual response of the aquifer without
are presented in Table 11. any preconditioned behavior pattern, the method is seen to
The results summarized in Table 12 reveal that the be able to provide information about the aquifer being
present method provides the most accurate solution for the tested (whether it is confined or unconfined). The
well-loss parameters for the given heterogeneous aquifer. conducted tests showed that the proposed method allows
The calculated B(t)Q1 values shown in Fig. 8 are observed the generation of the aquifer function B(t) for the entire
to be very close to the actual ones generated by the duration of the step-drawdown test which in turn provides
numerical model. The heterogeneous behavior is also a greater set of data to identify the aquifer parameters. The

Table 12 Estimated parameters for the hypothetical heterogeneous confined-aquifer test case
Parameters Labadie and Helweg (1975) Birsoy and Summers (1980) Gupta (1989) Present method Predefined values
p 1.738 3.147 3.354 2.780 2.800
C 6.151E-04 1.378E-07 3.521E-08 1.129E-06 1.000E-06

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2


1600
Freeze RA (1975) A stochastic conceptual analysis of one-dimen-
sional groundwater flow in nonuniform homogeneous media.
Water Resour Res 11:725–741
Gupta AD (1989) On analysis of step-drawdown data. Ground
Water 27(6):874–881
Hantush MM, Marino MA (1990) Temporal and spatial variability
of hydraulic heads in finite heterogeneous aquifers: Numerical
modelling. In: Kovar K (ed) Calibration and reliability in
groundwater modelling. IAHS Publication no. 195, IAHS,
Oxfordshire, UK
Helweg OJ (1994) A general solution to the step-drawdown test.
Ground Water 32(3):363–366
Hoeksema RJ, Kitanidis PK (1985) Analysis of the spatial structure
of properties of selected aquifers. Water Resour Res 21:563–572
Huntley D, Nommensen R, Steffey D (1992) The use of specific
Fig. 8 The actual and the calculated B(t)Q1 curves for the capacity to assess transmissivity in fractured-rock aquifers.
hypothetical heterogeneous confined aquifer Ground Water 30(3):396–402
Jacob CE (1947) Drawdown test to determine effective radius of
artesian well. Trans ASCE 112:1047–1070
well-loss parameters can then be obtained with higher Jha MK, Nanda G, Samuel MP (2004) Determining hydraulic
accuracy once the aquifer behavior is identified. characteristics of production wells using genetic algorithm.
Water Resour Manage 18:353–377
Jha MK, Kumar A, Nanda G, Bhatt G (2006) Evaluation of
traditional and nontraditional techniques for determining well
parameters from step-drawdown test data. J Hydrol Eng 11
References (6):617–630
Karami GH, Younger PL (2002) Analysing step-drawdown tests in
Abramowitz M, Stegun IA (1965) Handbook of mathematical heterogeneous aquifers. Q J Eng Geol Hydrogeol 35(3):295–303
functions. US Government Printing Office, Washington, DC Kawecki MW (1995) Meaningful interpretation of step-drawdown
Avci CB (1992) Parameter estimation for step-drawdown tests. tests. Ground Water 33(1):23–32
Ground Water 30(3):338–342 Kruseman GP, de Ridder NA (1990) Analysis and evaluation of
Barlow PM, Moench AF (1999) WTAQ: a computer program for pumping test data, 2nd edn. Int. Institute for Land Reclamation
calculating drawdowns and estimating hydraulic properties for and Improvement, Wageningen, The Netherlands
confined and water-table aquifers. US Geol Surv Sci Invest Rep Labadie JW, Helweg OJ (1975) Step-drawdown test analysis by
99–4225, 74 pp computer. Ground Water 13(5):438–444
Batu V (1998) Aquifer hydraulics: a comprehensive guide to Lennox DH (1966) Analysis and application of step-drawdown
hydrogeologic data analysis. Wiley, New York tests. J Hydraul Div ASCE 92(HY6):25–47
Bear J (1979) Hydraulics of groundwater. McGraw-Hill, New York Mace RE (1997) Determination of transmissivity from specific
Bierschenk WH (1963) Determining well efficiency by multiple capacity tests in a karst aquifer. Ground Water 35(5):738–742
step-drawdown tests. Int Assoc Sci Hydrol 64:493–507 Maurer D (2006) Carson Valley-Aquifer tests. USGS-Nevada
Birsoy YK, Summers WK (1980) Determination of aquifer Water Science Center. http://nevada.usgs.gov/water/AquiferTests/
parameters from step tests and intermittent pumping data. carsonvalley_Sclapham.cfm?studyname=carsonvalley_
Ground Water 18(2):137–146 Sclapham. Last cited April 2010
Bourdet DJ, Ayoub A, Pirard YM (1989) Use of pressure derivative Miller TM, Weber WJ (1983) Rapid solution of the nonlinear step-
in well test interpretation. SPE Form Eval 4(2):293–302 drawdown equation. Ground Water 21(5):584–588
Bouwer H (1978) Groundwater hydrology. McGraw-Hill, New Neuman SP (1975) Analysis of pumping test data from anisotropic
York unconfined aquifer considering delayed gravity response. Water
Chiang W-H, Kinzelbach W (2001) 3D-Groundwater modeling with Resour Res 11(2):329–342
PMWIN a simulation system for modeling groundwater flow Ostrowski LP, Kloska MB (1988) Use of pressure derivatives in
and pollution. Springer, New York analysis of slug test or DST flow period data. SPE Paper 18595
Clark L (1977) The analysis and planning of step-drawdown tests. SPE, Richardson, TX
Q J Eng Geol Hydrogeol 10(2):125–143 Parks KP, Bentley LR (1996) Derivative-assisted evaluation of well
Clifton PM, Neuman SP (1982) Effects of kriging and inverse yields in a heterogeneous aquifer. Can Geotech J 33(3):458–469
modeling on conditional simulation of the Avra Valley aquifer Razack M, Huntley D (1991) Assessing transmissivity from specific
in southern Arizona. Water Resour Res 18:1215–1234 capacity in a large and heterogeneous alluvial aquifer. Ground
Cooper HHJ, Jacob CE (1946) A generalized graphical method for Water 29(6):856–861
evaluating formation constants and summarizing well field Renard P, Glenz D, Mejias M (2009) Understanding diagnostic
history. Trans Am Geophys Union 27:526–534 plots for well-test interpretation. Hydrogeol J 17:589–600
Dagan G (1989) Flow and transport in porous formations. Springer, Rorabaugh MI (1953) Graphical and theoretical analysis of step-
Heidelberg drawdown test of artesian well. ASCE Proc., Hydraulic
Delhomme JP (1979) Spatial variability and uncertainty in ground- Division, ASCE, Washington DC, vol 9, pp 1–14
water flow parameters: a geostatistical approach. Water Resour Sheahan NT (1971) Type-curve solution of step-drawdown test.
Res 15:269–280 Ground Water 9(1):25–29
Eden RN, Hazel CP (1973) Computer and graphical analysis of Shekhar S (2006) An approach to interpretation of step drawdown
variable discharge pumping tests of wells. Civil Eng Trans 15 tests. Hydrogeol J 14(6):1018–1027
(1–2):5–10 Singh SK (2001) Confined aquifer parameters from temporal
Ehlig-Economides C (1988) Use of pressure derivative for derivative of drawdowns. J Hydraul Eng ASCE 127(6):466–470
diagnosing pressure-transient behaviour. J Petrol Technol 40 Singh SK (2002) Well loss estimation: variable pumping replacing
(10):1280–1282 step drawdown test. J Hydraul Eng ASCE 128(3):343–348
El-Naqa A (1994) Estimation of transmissivity from specific Theis CV (1935) The relation between the lowering of the piezometric
capacity data in fractured carbonate rock aquifer, central Jordan. surface and the rate and duration of discharge of a well using
Environ Geol 23:73–80 groundwater storage. Trans Am Geophys Union 2:519–524

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2


1601
Thiem G (1906) Hydrologic methods (in German). Gebhardt, Walton WC (1962) Selected analytical methods for well and aquifer
Leipzig, Germany evaluation. Illinois State Water Surv Bull 49:1–81
Todd DK (1980) Groundwater hydrology. Wiley, New York Wong DW, Harrington AG, Cinco-Ley H (1986) Application of the
Verbovsek T (2008) Estimation of transmissivity and hydraulic pressure-derivative function in the pressure-transient testing of
conductivity from specific capacity and specific capacity fractured wells. SPE Form Eval 1(3):470–480
index in dolomite aquifers. J Hydrol Eng ASCE 13(9):817– Yeh H-D (1989) Step-drawdown data analysis. J Hydraul Eng
823 ASCE 115(10):1426–1432

Hydrogeology Journal (2010) 18: 1591 – 1601 DOI 10.1007/s10040-010-0620-2

You might also like