Professional Documents
Culture Documents
C A P T D D: Onfined Quifer Arameters From Emporal Erivative OF Rawdowns by Sushil K. Singh
C A P T D D: Onfined Quifer Arameters From Emporal Erivative OF Rawdowns by Sushil K. Singh
OF DRAWDOWNS
By Sushil K. Singh1
ABSTRACT: A simple method that uses the time derivative of drawdowns is proposed for the evaluation of
confined aquifer parameters. Explicit expressions are proposed for evaluation of the aquifer parameters as well
as a graphical procedure. A reliable and accurate scheme to calculate the numerical derivative of drawdowns is
developed based upon an analytical approach. The method requires early drawdown data (u > 0.01, where u is
the argument of well function), and is shown to converge to the Cooper-Jacob method for late drawdowns (u
ⱕ 0.01). It does not require curve matching or an initial guess for the parameters. Calculations for the method
can be performed on a hand-held calculator. The method has been applied to published data sets and the results
Downloaded from ascelibrary.org by Texas Tech University on 05/13/15. Copyright ASCE. For personal use only; all rights reserved.
have been compared with those obtained using traditional methods. The method accurately estimates the aquifer
parameters using only early drawdown data, thereby indicating savings in time and money.
INTRODUCTION uifer parameters. The calculation for the method can be ac-
complished on a hand calculator.
Evaluation of aquifer parameters (i.e., transmissivity T and
storage coefficient S ), has been a continued field of research. PROPOSED METHOD
Continued efforts have been made for the development of sim-
ple methods for the evaluation of aquifer parameters since A well drilled in a homogeneous and isotropic confined aq-
Cooper and Jacob (1946) proposed their method. Khan (1982) uifer of infinite areal extent satisfying the assumptions in-
gave a procedure to evaluate aquifer parameters utilizing Ja- volved in the derivation of Theis’ (1935) equation, is pumped
cob’s approximation (u ⱕ 0.01, where u is the argument of at a constant rate. The drawdown at an observation well is
the well function) of the well function on a hand calculator. given by the Theis equation. The nondimensional form of the
Other methods use derivatives of drawdowns [e.g., Rai (1985), Theis equation is written as (Singh 2000)
冉 冊
El-Khatib (1987), Singh (1996), and Singh and Ram (1996)]
for the evaluation of aquifer parameters. These methods re- s 0.4348␣
= 1.5446W * (1)
quire the computation of derivatives of discrete drawdowns s ␣
*
with acceptable accuracy for reliable estimation of the param-
eters. The method by Rai (1985) requires construction of a where ␣ = t/r 2; and W(⭈) = well function described as
冕
graph between u and f (u), a function defined by him and read- ⬁
ing of the values of u for known f (u) from this graph for each e⫺x
W(u) = dx (2)
drawdown and its derivative. Yeh (1987) gave another ex- u x
pression of f (u) that incorporated logarithms of time and ad-
vocated that though both the expressions are analytically the The parameters s and ␣ correspond to the peak of the
* *
same, their finite-difference implementations may give differ- s/␣ versus ␣ curve and are respectively given by
ent results. This may be due to the fact that the expressions
Q
for the derivative are not based on an analytical approach. s = 5.152 ⫻ 10⫺2 (3)
* T
Therefore, it is observed that the methods using the derivative
of drawdowns suffer from errors in the computation of nu- S
merical derivatives. Spane and Wurstner (1993) gave a com- ␣ = (4)
* 1.7393T
puter program based on numerical scheme to calculate pres-
sure derivatives along with valuable literature review. Differentiating (1) with respect to ␣
冉 冊
Srivastava and Guzman-Guzman (1994) tested different avail-
able expressions for the derivatives, none of which had an 1 ⭸s 1.5446 ␣
= exp ⫺0.4348 * (5)
analytical basis. s ⭸␣ ␣ ␣
In this paper, a simple method for the explicit evaluation of *
aquifer parameters is proposed. The method uses the time de- Rearranging (5) and taking logarithm
rivative of drawdowns. A reliable scheme based upon an an-
alytical approach for calculating numerical derivatives with ac-
curacy is also proposed. The method uses early drawdowns (u
ⱖ 0.01). The method is shown to converge to the Cooper and
ln 冉
␣⭸s/⭸␣
1 冊 =⫺
2
␣
(6)
冉 冊
␣
⭸s
⭸␣ (␣i⫹␣i⫹2)/2
=
␣i ⫹ ␣i⫹2 si⫹2 ⫺ si
2 ␣i⫹2 ⫺ ␣i
(10)
Downloaded from ascelibrary.org by Texas Tech University on 05/13/15. Copyright ASCE. For personal use only; all rights reserved.
冉 冊
(1996) observed that the Cooper-Jacob method, when applied
to the drawdowns for which 2 /␣ > 0.01 in the neighborhood t⭸s/⭸t 2 r 2
log =⫺ (13)
of 0.01, may give erroneous values for the aquifer parameters. 1 2.303t
Also, when the observation well is at a great distance from
the pumped well, a large value for 2 /␣ is obtained. With and (10) is modified as
known values of 1 and 2, estimates of the aquifer parameters
can be obtained using (7) and (8). The drawdown data from
multiple observation wells are required to be arranged in as-
冉 冊
t
⭸s
⭸t (ti⫹ti⫹2)/2
=
ti ⫹ ti⫹2 si⫹2 ⫺ si
2 ti⫹2 ⫺ ti
(14)
cending order of ␣ before (10) is used to get the derivatives. The variable t⭸s/⭸t is plotted against 1/t on a semilog paper
The method yields estimates of aquifer parameters from early with 1/t on natural axis, to get the intercept as 1 and slope
drawdowns the collection of which requires less time and as 2 r 2/2.303. The use of (13) for a single observation well
money than standard methods. Another advantage of the reduces the computational efforts as compared to that of (12).
method is that the systematic errors in the drawdowns are au- El-Khatib (1987) proposed an equation similar to (13) which
tomatically removed while taking the derivatives. For exam- was used by Srivastava and Guzman-Guzman (1994). But both
ple, (10) uses the difference of two drawdowns observed at did not realize that the equation requires data for which 2 /␣
different times; hence, systematic errors are removed while > 0.01 (or preferably 0.1). For 2 /␣ ⱕ 0.01, a constant value
taking the difference of two drawdowns. of t⭸s/⭸t is obtained for all data (Fig. 1).
For the late drawdowns (2 /␣ ⱕ 0.01), (6) can be expressed
as Explicit Expressions
The least-square parameter estimation technique when ap-
⭸s
= 1 (11) plied to (6) yields the following normal equations:
⭸ ln(␣)
y = ln(1) ⫺ 2 x (15)
The above equation shows that s, when plotted against
ln(␣), a straight line is obtained—the slope of which is given yx = ln(1)x ⫺ 2 x 2
(16)
by 1. Thus, (11) is a mathematical representation of the Coo- Solution of (15) and (16) gives the explicit expressions for
per and Jacob (1946) method. Hence, the present method re- evaluating aquifer parameters. These are
duces to the Cooper-Jacob method for late drawdowns.
Q
T= exp(⫺y ⫺ 2 x) (17)
Graphical Method 4
冉 冊
where
␣⭸s/⭸␣ 2
log =⫺ (12) xy ⫺ x y
1 2.303␣ 2 = (19)
(x)2 ⫺ x 2
Eq. (12) shows that when ␣⭸s/⭸␣ is plotted against 1/␣ on
a semilog graph with 1/␣ on natural axis, a straight line with in which
constant (nonzero) slope is obtained. [This is true for 2 /␣
冘
n
1 1
much greater than 0.01 and preferably of the order of 0.1, x= (20)
n ␣i
otherwise nearly a constant value of ␣(⭸s/⭸␣) is obtained for i=1
冘 冉 冊
n
fitting the straight line, care must be exercised in giving less 1 1 ⭸s
xy = ln ␣ (22)
weight to those points for which 2 /␣ is <0.01. Eq. (12) is n i=1 ␣i ⭸␣
冘
i
used when data from more than one observation well are to n
lator. While using these equations, a check for 2 /␣ is nec- 1 1.0 0.20 — — —
2 1.5 0.27 1.5 0.10 0.15
essary in a similar way as discussed for the graphical method. 3 2.0 0.30 2.0 0.07 0.14
Thus, for the present method, the early drawdown data play a 4 2.5 0.34 2.5 0.07 0.175
vital role in obtaining the estimates of the aquifer parameters. 5 3.0 0.37 3.25 0.0467 0.152
The early drawdowns are important as they are less likely to 6 4.0 0.41 4.0 0.04 0.16
be corrupted by boundary effects, recharge, etc., and hence 7 5.0 0.45 5.0 0.035 0.175
should be given proper weight in determining the aquifer pa- 8 6.0 0.48 6.5 0.0267 0.174
9 8.0 0.53 8.0 0.0225 0.18
rameters. Very early-time drawdown data sometimes may be 10 10 0.57 10 0.0175 0.175
affected by gravel pack, well screen, and storage in the well 11 12 0.60 12 0.015 0.18
and may not follow the Theis solution. Often, these effects do 12 14 0.63 15 0.01167 0.175
Downloaded from ascelibrary.org by Texas Tech University on 05/13/15. Copyright ASCE. For personal use only; all rights reserved.
not influence the peak (considered in the proposed method) 13 18 0.67 19 0.009 0.171
especially when the observation well is at a larger distance 14 24 0.72 24 0.0075 0.18
from the pumped well. 15 30 0.76 32 0.005625 0.18
16 40 0.81 40 0.0045 0.18
17 50 0.85 50 0.0035 0.175
ERROR AND RELIABILITY CRITERIA 18 60 0.88 65 0.00267 0.174
19 80 0.93 80 0.002 0.16
There are different criteria for assessing the reliability of a 20 100 0.96 100 0.00175 0.175
model. In order to check the reliability of the estimated pa- 21 120 1.00 125 0.0016 0.20
rameters using the present method, the criteria of integral 22 150 1.04 150 0.001167 0.175
squared error (ISE) and standard error of estimate (SEE) have 23 180 1.07 180 0.001 0.18
been used. These are defined as 24 210 1.10 210 0.000833 0.175
25 240 1.12 — — —
冘
n
Note: Time unit is minute and length unit is meter.
ISE = (so ⫺ s)i2 (24)
i=1
冉 冊
0.5
ISE TABLE 2. Calculations for Derivatives for Set II
冏
SEE = (25)
n⫺2 1 si ⫺ si⫺2 ⭸s
t̄ = (ti ⫹ ti⫺2) t
i ti si 2 ti ⫺ ti⫺2 ⭸t t̄
where so = observed drawdown. Low values of ISE and SEE
show that the estimated values of the parameters are reliable. 1 2 0.002 — — —
2 4 0.015 — — —
3 6 0.023 4 0.0052 0.0210
APPLICATION AND DISCUSSION 4 10 0.061 7 0.0077 0.0537
5 15 0.103 10.5 0.0089 0.0933
Application of the method is demonstrated in three data sets. 6 20 0.146 15 0.0085 0.1275
Set I is from Todd (1980, p. 127), Set II is from Central 7 30 0.214 22 0.0074 0.1665
Ground Water Board (CGWB) (1982), Government of India, 8 40 0.266 30 0.0060 0.1800
and Set III is from Lohman (1972, p. 19). The detailed dis- 9 50 0.311 40 0.0048 0.1940
cussions are provided in the following subsections. 10 70 0.386 55 0.0040 0.2200
11 100 0.461 75 0.0030 0.2250
12 120 0.506 95 0.0024 0.2280
Set I 13 150 0.559 125 0.0020 0.2450
The pumping rate was 2,500 m3/day and the drawdowns Note: Time unit is minute and length unit is meter.
were observed at 60 m from the pumped well. While analyzing
Todd’s data, Khan (1982) estimated the aquifer parameters as
TABLE 3. Sample Calculations for Set II
T = 1,150 m2/day and S = 1.8 ⫻ 10⫺4 using the Cooper and
Jacob (1946) approximation of the well function and least- x1 y x1*y x1*x1
squares regression. This shows that for most of the data u is 0.2500 ⫺3.8632 ⫺0.9658 0.0625
<0.01. Therefore, the storage coefficient cannot be estimated 0.1429 ⫺2.9250 ⫺0.4179 0.0204
using the method developed above. It is restressed here that 0.0952 ⫺2.3716 ⫺0.2259 0.0091
the parameter t⭸s/⭸t tends to a constant limiting value for u ⱕ 0.0667 ⫺2.0596 ⫺0.1373 0.0044
0.01. This is illustrated in Table 1 along with the sample cal- 0.0444 ⫺1.7928 ⫺0.797 0.0020
0.0333 ⫺1.7148 ⫺0.572 0.0011
culations using the present approach. The average of this con- 0.0250 ⫺1.6399 ⫺0.0410 0.0006
stant limiting value from Table 1 (leaving the first 6 data) is 0.0182 ⫺1.5141 ⫺0.0275 0.0003
0.1769, and this gives T = 1,125 m2/day. Mention may be 0.0133 ⫺1.4917 ⫺0.0199 0.0002
made here that Srivastava and Guzman-Guzman (1994) fitted 0.0105 ⫺1.4784 ⫺0.0156 0.0001
a straight line with constant slope using these data and ob- 0.0080 ⫺1.4065 ⫺0.0113 0.0001
tained T = 1,145 m2/day and S = 2.09 ⫻ 10⫺4. This shows that 0.0643a ⫺2.0234a ⫺0.1817a 0.0092a
their estimate of S is not reliable because very few points exist Note: Time unit is minute and length unit is meter.
a
for which u is of the order of 0.1, and those too contain error. Denotes average.
Set II
Upper Yamuna River Basin, Haryana, India. The pumping test
In order to estimate the aquifer parameters from drawdown was conducted at a constant pumping rate of 1.8924 m3/min
data on a single observation well, the data obtained by CGWB for 7,000 min on a fully penetrating tube well in an alluvial
(1982) were considered. The CGWB conducted a pumping test aquifer under- and overlain by confining clays 24 and 50 m
in a confined aquifer at the ‘‘Mathana’’ site located in the thick, respectively. The drawdown data observed at an obser-
468 / JOURNAL OF HYDRAULIC ENGINEERING / JUNE 2001
Synthetic Data traditional methods show that the present method accurately
estimates the aquifer parameters even using only early draw-
In order to show the reliability of the method, drawdowns downs. While calculating ISEs for the present method, all data
were generated assuming Q = 1.5 m3/min, r = 100 m, T = including both early and late drawdowns were considered. The
0.50 m2/min, and S = 2.4 ⫻ 10⫺3 for different uniform time method is fully objective and yields reliable estimates of the
intervals. The drawdowns at different time intervals were se- aquifer parameters even when only early drawdowns are used.
lected for the analysis and the aquifer parameters were esti- Thus, the method avoids error involved in curve matching and
mated using the present method. The results with percentage does not require iterative tedious computation as needed by
error in the estimates of T and S are given in Table 7. For a software programs for aquifer parameter estimation, which
time interval of 15 min, error in the estimates of S is greater sometimes has to be used as a black box (users do not know
(5%). This may be because only eight drawdowns are available the calculation procedure). Also, it does not require initial es-
and from this six drawdowns are effectively used to arrive at timates for the parameter values.
the estimates of the parameters, leaving thereby fewer degrees
of freedom. Table 7 shows that the time interval for drawdown Identifying Nonideal Aquifer Condition
measurements does not have much effect on the parameter
estimates. With increase in the time interval, no systematic The ideal confined aquifer condition is denoted by the con-
increase in errors in the estimates of T and S was observed. stant limiting value of t⭸s/⭸t for greater time. At the most the
This is one of the advantages of the proposed method in which values may be randomly fluctuating over a mean value as ob-
the computation of the derivatives has an analytical basis. served in Table 1, where the values mostly range between
There exists an optimal time interval that may be approxi- 0.175 and 0.18. A decreasing trend in t⭸s/⭸t for greater time
mated as t /4, where t = r 2␣ *. shows the nonideal aquifer condition of the aquifer. Another
* *
The lower values of ISE and SEE than those obtained using test may be to estimate the aquifer parameters using early
JOURNAL OF HYDRAULIC ENGINEERING / JUNE 2001 / 469
冘
n r = distance of observation well from pumping well (L);
1 1 s = drawdown at observation well (L);
x1 = (27)
n i=1 ti s = parameter of Theis equation (L);
冘 冉 冊
*
n S = storage coefficient of aquifer (dimensionless);
1 ⭸s T = transmissivity of aquifer (L2 T ⫺1);
y1 = ln t (28)
n i=1 ⭸t i t = time measured since start of pumping (T);
冘 冉 冊
n u = 2 /␣ = argument of well function (dimensionless);
1 1 ⭸s W(⭈) = well function (dimensionless);
x1 y1 = ln t (29)
n i=1 ti ⭸t i ␣ = t/r 2 (L⫺2 T );
冘
n ␣ = parameter of Theis equation (L⫺2 T );
1 1 *
x 21 = (30) 1 = Q/(4T ) (L); and
n i=1 t 2i 2 = S/(4T ) (L⫺2 T ).