Professional Documents
Culture Documents
available at www.sciencedirect.com
KEYWORDS
Groundwater;
Pumping test;
Sensitivity analysis;
Parameter estimation
model;
Leaky aquifer;
Unconfined aquifer
Introduction
Groundwater hydrologists often conduct pumping tests to
obtain aquifer parameters, such as hydraulic conductivity
and storage coefficient, which are necessary information
for quantitative groundwater studies. Theis (1935) obtained
the solution for unsteady groundwater flow toward a pumping well in a confined aquifer by analogy to the problem of
* Corresponding author. Fax: +886 3 5726050.
E-mail address: hdyeh@mail.nctu.edu.tw (H.-D. Yeh).
0022-1694/$ - see front matter 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2006.12.007
407
408
r
Q
W u;
1
4pT
B
where s is drawdown, r is the distance between pumping
well and observing well, u is dimensionless variable and it
is defined as r2S/4Tt, K 0 is the vertical conductivity of leaky
confining bed, b 0 is thickness of aquitard,pr/B
= L is called
exp y
W u;
dy
2
B
y
4y
u
s
r20 < y 2
Ksrw, and rD = r/rw. Notice that rw represents the well radius. The symbol en is the root of
p
11
en tanen
rbw p=c
where r = S/Sy, c = a1bSy/Kz, and a1 is a fitting parameter for
drainage from the unsaturated zone and has units of inverse
time (1/T). A large value of a1 effectively eliminates this
parameter from the solution.
oPi
DPi
12
where O is the output function of the system (i.e., the aquifer drawdown) and Pi is the ith input parameter of the system. However, the values of the parametric sensitivity for
various parameters are useless for making comparison if
the unit and/or the order of magnitude of the parameters
are different. Thus, the normalized sensitivity is used and
defined as (Kabala, 2001)
Si;t
oO
oO
Pi
oPi =P i
oPi
13
where Si,t is the normalized sensitivity of ith input parameter at time t. Note that O is a function of Pi and t. The partial derivative of this equation may be approximated by a
forward differencing formula as
oO OPi DPi OP i
oPi
DPi
14
and
rrn sinrn y 2 r2n cosrn 0;
10
n
X
Ohi Phi 2
15
i1
where n is the total time step and Ohi and P hi are respectively the observed and predicted drawdowns at time step
i. Based on Eq. (15), Simulated Annealing method can deter-
Table 1
No.
Time (min)
Drawdown (m)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
0.017
0.050
0.100
0.250
0.750
1.000
1.500
2.000
2.500
3.500
4.500
5.000
10.000
20.000
50.000
100.000
200.000
500.000
700.000
1000.000
0.013
0.099
0.203
0.380
0.621
0.687
0.781
0.847
0.899
0.977
1.035
1.059
1.215
1.365
1.539
1.640
1.702
1.728
1.730
1.730
Q = 3000 m3/day, R = 30 m.
409
cate that the parameters T and S have influence on the drawdown right at the beginning of pumping. In addition, the
influence of S is larger than that of T at early pumping period.
This result to some extend reflects the physical behavior of
parameters T and S during the pumping.
The time-drawdown data set 1 of an unconfined aquifer,
generated by Neumans model (1974), for pumping starting
from 1 to 176,360 s (49 h) in an unconfined aquifer are listed
in Table 2. The thickness of the aquifer, b, is 10 m, pumping
rate Q is 3000 m3/day, and the distance between the pumping well and observation well R is 10 m. The radial hydraulic
conductivity Kr, vertical hydraulic conductivity Kz, storage
coefficient S, and specific yield Sy are set to 1 103 m/s,
1 104 m/s, 1 104, and 1 101, respectively. The
time-drawdown data and related normalized sensitivities
are plotted in Figure 2. Similar to Figure 1, the temporal distribution of each normalized sensitivity reflects the temporal change of the drawdown in response to the relative
change of each parameter, and all aquifer parameters affect the drawdown at different periods. The normalized
sensitivity of parameter S starts from 1 to 10 s, Kz ranges
from 1 to 1000 s, and Sy appears from 80 s to the end of
pumping. The drawdown is most sensitive to the parameter
Kr except at the early period of the pumping and the influence of Kr on the drawdown increases at the beginning and
through the end of the pumping.
The normalized sensitivity of S begins with highest value
and drops quickly after the start of the pumping. The normalized sensitivity of Kz reaches its highest value in a range
between 10 and 1000 s, implying that the slow decline of
the water table is attributed to the contribution of the Kz
at the moderate pumping time. The drawdown stops
increasing when the normalized sensitivity of Kz approaches
its maximum. The temporal distribution of Krs normalized
sensitivity, displaying three segments during the pumping
period, is similar to the drawdown curve. The second segment appears at 10 s and vanishes at 1000 s (16.67 min). Figure 2 shows that the drawdown increases in the third
segment along with the decrease of Kzs normalized sensitivity, clearly indicating rapid decrease of vertical drainage.
The sensitivity curve demonstrates that the aquifer parameter Sy does not have any contribution in response to the
pumping at the beginning of the test and starts to react at
about 80 s (1.33 min).
The time-drawdown data set 2 listed in Table 3 is generated by Moenchs model (1997). The pumping starts from
0.6 to 600,000 s (16.67 hours). The thickness of the aquifer,
b, is 10 m, pumping rate Q is 1000 m3/day, and the distance
between the pumping well and observation well R is 10 m.
The parameters Kr, Kz, S, Sy, and well radius rw, are set to
1 103 m/s, 1 104 m/s, 1 104, 1 101, and 1 m
respectively. The time-drawdown data and related normalized sensitivities are plotted in Figure 3. The upper panel of
Figure 3 shows the same plot without the normalized sensitivity of parameter Kr because the magnitude of Krs normalized sensitivity is very large at the late time of
pumping. Removing Krs normalized sensitivity is helpful
to recognize the small change of other parameters normalized sensitivities at the early time of pumping. The normalized sensitivity of parameter rw varies from 2 to 2000 s, S
changes from 0.6 to 1000 s, Kz ranges from 100 to
10,000 s, and Sy appears from 100 s toward the end of pump-
410
Figure 1
Table 2
The time-drawdown data and the normalized sensitivities of the leaky aquifer parameters.
No.
Time (s)
Drawdown (m)
No.
Time (s)
Drawdown (m)
No.
Time (s)
Drawdown (m)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
30
44
58
74
0.22
0.31
0.36
0.38
0.40
0.41
0.41
0.42
0.42
0.42
0.43
0.43
0.43
0.43
0.43
0.43
0.43
0.43
0.43
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
87
120
149
176
212
272
332
393
472
600
792
967
1143
1350
1723
2154
2632
3215
4385
0.44
0.44
0.44
0.44
0.45
0.45
0.46
0.46
0.47
0.48
0.49
0.50
0.52
0.53
0.55
0.58
0.61
0.64
0.70
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
6000
8000
9354
11,429
14,925
18,235
22,274
25,882
32,696
41,295
47,195
59,224
69,279
81,302
95,126
118,168
151,775
176,360
0.76
0.83
0.86
0.91
0.98
1.03
1.09
1.13
1.19
1.25
1.29
1.35
1.40
1.44
1.48
1.54
1.61
1.65
Figure 2
Table 3
411
The time-drawdown data and the normalized sensitivities of the unconfined aquifer parameters (Neumans model).
No.
Time (s)
Drawdown (m)
No.
Time (s)
Drawdown (m)
No.
Time (s)
Drawdown (m)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
0.6
1
2
3
4
5
6
7
8
9
10
12
14
16
18
21
0.0001
0.0003
0.0008
0.0014
0.0020
0.0028
0.0034
0.0040
0.0047
0.0055
0.0064
0.0075
0.0087
0.0101
0.0116
0.0134
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
24
27
31
36
41
47
54
63
72
82
95
125
189
249
497
655
0.0155
0.0178
0.0204
0.0234
0.0268
0.0306
0.0348
0.0396
0.0449
0.0507
0.0572
0.0722
0.0993
0.1199
0.1720
0.1892
33
34
35
36
37
38
39
40
41
42
43
44
45
46
1138
1722
1977
2992
5970
11,912
18,029
35,973
62,514
94,619
124,732
188,789
328,078
600,000
0.2130
0.2251
0.2290
0.2424
0.2741
0.3189
0.3507
0.4088
0.4577
0.4950
0.5200
0.5578
0.6084
0.6638
412
Figure 3
Time (min)
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
0.10
0.25
0.75
1.00
1.50
2.00
2.50
3.50
4.50
5.00
10.00
20.00
50.00
100.00
200.00
500.00
700.00
1000.00
Estimated parameters
T (m2/day)
S 104
L 102
1000.53
1000.32
1000.52
999.93
1000.02
999.96
999.98
999.99
999.99
999.95
1000.06
1000.02
1000.01
1000.02
1000.02
1000.04
1000.06
1000.05
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.12
2.52
2.74
3.06
3.00
3.03
3.01
3.00
3.01
3.01
3.00
3.00
3.00
3.00
3.00
3.00
3.00
3.00
Figure 4
case.
413
The field time-drawdown data and the estimated parameters for a leaky aquifer using different number of observations
Number of observations
Time (min)
Drawdown (m)
1
2
3
4
5
6
7
8
9
10
11
12
0.2
0.5
1
2
5
10
20
50
100
200
500
1000
0.536
0.838
1.094
1.298
1.609
1.798
1.972
2.109
2.167
2.195
2.198
2.198
4
5
6
7
8
9
10
11
12
Estimated values
T (m2/day)
S 104
L 102
1060.40
1182.30
1182.70
1203.80
1211.33
1222.18
1232.32
1236.93
1239.28
1.12
1.05
1.04
1.03
1.02
1.00
0.99
0.98
0.98
15.70
1.61
6.76
5.85
5.61
5.32
5.09
4.99
4.93
414
Table 6
Number of observations
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Time (s)
4
5
6
7
8
9
10
11
12
13
14
15
30
44
58
74
87
Estimated parameters
Kr (m/s) 103
Kz (m/s) 104
S 104
Sy 101
0.997
1.000
1.000
0.997
1.000
0.999
1.000
0.998
1.000
1.000
1.000
0.998
1.000
0.998
1.000
1.000
1.000
1.006
0.999
0.999
1.010
0.998
1.000
1.000
1.000
0.995
0.997
0.998
0.998
1.000
0.997
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
0.612
0.616
1.190
0.444
1.570
0.933
0.972
0.712
2.010
1.140
1.220
0.816
1.040
0.987
1.010
0.993
1.000
2.01 101 and the largest relative errors are 101% when
using 12 observation data. The identification results of Sy
did not approach the target value until the number of observation is over 20, i.e., about 80 s. Therefore, the on-line
PEM may not obtain accurate results of Sy if the time-drawdown data is too short to cover the response period of Sy.
Similar to Figure 4, the curve of estimated Sy versus time
displayed in Figure 5 shows dramatic fluctuation in the early
period and converges to a constant value after about 80 s.
Figure 2 demonstrates that the on-line PEM can successfully
identify the aquifer parameters when Sy just starts to affect
the drawdown. Therefore, the on-line estimation based on
Number of observations used in the data analysis and the estimated parameters based on the synthetic data set 2
Number of observations
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
415
Time (s)
7
8
9
10
12
14
16
18
21
24
27
31
36
41
47
54
63
72
82
95
125
189
249
Estimated parameters
Kr (m/s) 103
Kz (m/s) 104
S 104
Sy 101
rw (m)
1.66
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.01
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.63
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.01
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.67
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.01
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
0.10
0.53
2.91
0.20
0.67
1.17
0.48
0.56
0.69
0.65
0.77
0.85
1.23
0.59
0.96
1.03
0.90
1.00
0.99
1.03
1.00
1.00
1.00
1.01
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
416
Table 8
Number of observations
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Time (min)
0.15
0.22
0.32
0.47
0.68
1.00
1.47
2.15
3.17
4.75
6.75
10.10
14.90
21.90
31.90
46.90
67.90
99.90
151.00
221.00
325.00
492.00
675.00
1050.00
1470.00
2190.00
3100.00
4330.00
Figure 7
Estimated parameters
Kr 103 (m/s)
Kz 105 (m/s)
S 103
Sy 101
0.65
0.73
0.91
0.88
0.96
0.51
0.32
0.22
0.24
0.30
0.26
0.25
0.44
1.01
1.39
1.74
1.92
1.96
1.97
1.92
1.82
1.70
1.60
1.54
1.50
1.47
1.46
1.45
0.10
1.05
0.98
1.18
1.19
1.89
2.15
2.23
2.41
2.78
2.60
2.20
2.34
1.51
0.74
0.48
0.39
0.37
0.36
0.38
0.42
0.46
0.49
0.51
0.52
0.53
0.54
0.54
7.18
7.29
7.45
7.17
7.16
5.61
4.28
3.32
3.52
4.31
3.83
3.64
5.64
8.14
7.00
6.21
5.65
5.52
5.49
5.63
5.98
6.39
6.71
6.94
7.10
7.19
7.26
7.29
3.00
3.00
3.00
3.00
3.00
3.00
3.00
3.00
2.91
3.00
3.00
2.95
3.00
1.49
0.56
0.27
0.18
0.16
0.16
0.18
0.24
0.33
0.43
0.52
0.59
0.64
0.68
0.69
Concluding remarks
The sensitivity analysis is used to investigate the influence
period of aquifer parameters in both leaky and unconfined
417
aquifers. The influences of parameters L and Sy on the drawdown are shown to have time lag in response to pumping in
the leaky and unconfined aquifers, respectively. An on-line
parameter estimation model is applied to estimate the
parameters based on the data obtained from synthetic and
field pumping tests for both leaky and unconfined aquifers.
The results indicate that the on-line estimation can be terminated when the parameters are stabilized and their corresponding normalized sensitivities start to react the
pumping. In the synthetic cases, the termination time of
the on-line estimation is consistent with the influence period of the parameter which has longest time lag from the
beginning of the pumping. This phenomenon indicates that
the on-line estimation can be terminated if all identified
parameters tend to be stabilized, i.e., the drawdown already reacts to the effect of aquifer parameters. In the field
cases, the results indicate that the on-line parameter estimation model can save 90% pumping time in the leaky aquifer and 77% pumping time in the unconfined aquifer. Note
that the small fluctuation in the estimated parameters at
the late period of pumping and a longer on-line estimation
time than that of the synthetic case occur. These results
may be mainly caused by aquifer heterogeneity and/or measurement errors in the observed drawdown data. Finally,
different values of the specific yield and distance between
pumping well and observation well do not significantly affect the influence time of specific yield during the pumping.
These results may provide a useful reference for on-line
aquifer parameter estimation.
Acknowledgements
This study was partly supported by the Taiwan National Science Council under the grant NSC94-2211-E-009-015. The
authors thank two anonymous reviewers for constructive
comments and suggested revisions.
418
References
Batu, V., 1998. Aquifer Hydraulics. John Wiley & Sons, New York.
Boulton, N.S., 1954. Unsteady radial flow to a pumped well allowing
for delayed yield from storage. Int. Ass. Sci. Hydrol., Publ 37,
472477.
Boulton, N.S., 1963. Analysis of data from non-equilibrium pumping
tests allowing for delayed yield from storage. Proc. Inst. Civil
Eng. 26, 469482.
Chander, S., Kapoor, P.N., Goyal, S.K., 1981. Analysis of pumping
test data using Marquardt algorithm. Ground Water 19 (3), 275
278.
Charbeneau, R.J., 2000. Groundwater Hydraulics and Pollutant
Transport. Prentice Hall, Upper Saddle River, NJ.
Cooper Jr., H.H., Jacob, C.E., 1946. A generalized graphical
method for evaluating formation constants and summarizing
well field history. Transactions, American Geophysical Union 27
(IV), 526534.
Cooper, H.H., Jr., 1963. Type curves for nonsteady redial flow in an
infinite leaky artesian aquifer, in Bentall, Ray, complier, Shortcuts and special problems in aquifer tests, U.S. Geol. Survey
Water-supply Paper 1545C.
Cukier, R.I., Fortuin, C.M., Shuler, K.E., Petschek, A.G., Shaibly,
J.H., 1973. Study of the Sensitivity of Coupled Reaction System
to Uncertainties in Rate Coefficients: I. theory. J. Chem. Phys.
59 (8), 38733878.
Cukier, R.I., Fortuin, C.M., Shuler, K.E., 1975. Study of the
Sensitivity of Coupled Reaction System to Uncertainties in Rate
Coefficients: III Analysis of the approximations. J. Chem. Phys.
63 (3), 11401149.
Cukier, R.I., Levine, H.B., Shuler, K.E., 1978. Nonlinear sensitivity
analysis of multiparameter mode systems. J. Comp. Phys. 26 (1),
142.
Duffield, G.M., 2002. AQTESOLV for Windows. HydroSOLVE, Inc.,
Reston, VA.
Gooseff, M.N., Bencala, K.E., Scott, D.T., Runkel, R.L., Mcknight,
D.M., 2005. Sensitivity analysis of conservative and reactive
stream transient storage models applied to field data from
multiple-reach experiments. Adv. Water Resour. 28, 479492.
Hantush, M.S., Jacob, C.E., 1955. Non-steady radial flow in an
infinite leaky aquifer. Trans. Amer. Geophys. Union 36, 95100.
Hantush, M.S., 1964. Hydraulics of Wells. Adv. Hydrosci. 1, 281
442.
Jiao, J.J., Rushton, K.R., 1995. Sensitivity of drawdown to parameters and its influence on parameter estimation for pumping
tests in large-diameter wells. Ground Water 33 (5), 794800.
Kabala, Z.J., Milly, P.C.D., 1990. Sensitivity analysis of flow in
unsaturated heterogeneous porous-media - theory, numericalmodel, and its verification. Water Resour. Res 26 (4), 593610.
Kabala, Z.J., 2001. Sensitivity analysis of a pumping test on a well
with wellbore storage and skin. Adv. Water Res. 24, 483504.
Kabala, Z.J., El-Sayegh, H.K., Gavin, H.P., 2002. Sensitivity analysis
of a no-crossflow model for the transient flowmeter test.
Stochast. Environ. Res. Risk Assess. 16 (6), 399424.