THE USEFULNESS OF MULTI-WELL AQUIFER

TESTS IN HETEROGENEOUS AQUIFERS*

Steven C. Young and Dudley J. Benton

Tennessee Valley Authority
Engineering Laboratory
Norris, TN

Joost C. Herweijer
GeoTrans, Inc.
Boulder, CO

Paul Sims
GeoTrans, Inc.
Louisville, KY

ABSTRACT

Three large-scale (100 meters) and seven small-scale (3-7 meters)

multi-well aquifer tests were conducted in a heterogeneous aquifer to
determine the transmissivity distribution across a one-hectare test
site. Two of the large-scale tests had constant but different rates of
discharge; the remaining large-scale test had a constant discharge that
was pulsed at regulated intervals. The small-scale tests were conducted
at two well clusters 20 meters apart. In order to efficiently and
objectively analyze the data, the program WELTEST was written. By using
the methods of non-linear least squares regression analysis and Broyden's
method to solve for non-linear extrema, WELTEST automatically determines
the best values of transmissivity and the storage coefficient. The test
results show that order of magnitude differences in the calculated
transmissivities at a well location can be realized by varying the
discharge rate at the pumping well, the duration of the aquifer test,
and/or the location of the pumping well. The calculated storage
coefficients for the tests cover a five-order magnitude range. The data
shows a definite trend for the storage coefficient to increase with the
distance between the pumping and the observation wells. This trend is
shown to be related to the orientation of high hydraulic conductivity
zones between the pumping and the observation wells. A comparison among
single-well aquifer tests, geological investigations and multi-well
aquifer tests indicate that the multi-well aquifer tests are poorly
suited for characterizing a transmissivity field.

INTRODUCTION

Contracted by the U.S. Air Force, the Tennessee Valley Authority (TVA) is
involved in research to improve the design of bioremediation and
pump-and-treat remediation. As part of this study, the TVA is
investigating the usefulness of multi-well aquifer tests for site

* Sponsored by the United States Air Force Engineering and Services

Center, Tyndall Air Force Base, FL
characterization in heterogeneous aquifers. In a heterogeneous aquifer,
a change in the test design alters the aquifer region included in the
cone-of-depres8 ion. This paper investigates the sensitivities of the
calculated hydraulic properties from multi-well aquifer tests to changes
in the design of the tests. In order to minimize any bias associated
with the data analysis, a computer program was developed to automatically
fit the best values of the storage coefficient and the transmissivity to
the observed drawdown in the observation wells. This paper presents the
results and discusses the trends for 10 multi-well aquifer tests.

DESCRIPTION OF THE TEST SITE

The test site occupies approximately one hectare of TVA's Columbus

Groundwater Research Test Site on Columbus Air Force Base (CAFB),
Mississippi. The site is located approximately 6 km east of the
Tombigbee River and 2.5 km south of the Buttahatchee River, and lies
above the 100-year flood plain of both rivers. The unconfined terrace
aquifer is composed of approximately 11 meters of Pleistocene and
Holocene fluvial deposits and primarily consists of irregular lenses of
poorly-sorted to well-sorted sandy-gravel and gravelly-sand. The
Quaternary deposits unconformably overlie the Cretaceous Age Eutaw
Formation that consists primarily of marine clay and silt.

Groundwater levels across the Columbus Groundwater Research Test Site

have been monitored since 1985 using single and multistage monitor
wells. The phreatic surface fluctuates seasonally from 2 to 3 meters.
The horizontal hydraulic gradient varies from approximately 0.02 (low
water table) to 0.05 (high water table). Upward and downward vertical
gradients, several orders of magnitude higher, have been observed over
most of the site. These vertical gradients are related to the spatial
variability in the hydraulic conductivity field and produce complex
groundwater flow patterns.

Using a geostatistical optimization technique, 37 fully-screened wells

were located within the one-hectare test site. Figure 1 shows the well
network. The optimization balanced several conflicting needs:
(1) availability of information on all scales (from 1 to 50 m) to
determine the spatial structure (variogram), (2 ) sufficient and even
areal coverage to ensure correct map interpolation, and (3) a radial
distribution of observation wells suitable for aquifer tests. This
technique is based on procedures given by Warrick and Myers (1987) and
Olea (1984).

THE HYDRAULIC CONDUCTIVITY FIELD

A depth profile of the hydraulic conductivity, representative for a local

area around a well bore, can be obtained during a single-well test by
using a borehole flowmeter to measure the contributions of the different
aquifer layers to the total rate withdrawn or injected. The concept of
the borehole flowmeter is illustrated in Figure 2. Basically, the method
 34 27 23 36*
•
10 0

90 -
3 6 9
•
31
60 26 •
■ 11 20
NORTH
co
70 37 ™ 16 - '
••1 5
cc i
2
• 18 19*. * .
1012 8
22 •
60 - 28 25 *
‘ 13
E
2
S 30
X . *24
17 21
50 • * * 32
•
1 4 7
« •
40

30

20 - 33
•
29
• 35 •
, l , , t 4___ I. ,1. 1, ,1 I , I , I , .J ___ ! ..
10
10 20 30 40 50 60 70 80 90 100 110
METERS
Figure 1. Well Network at Test Site.

Total dMwg* (torn

ttawM Q(totaQ

Z -
4

Z _
3

2-

Z -
t
1
r T -r
.001 .01 .1 1

CumuMJnFUw AvmgMHydnule CondueiMy

Figure 2. Schematic of Horizontal Flow to a Well and the Profiles

of the Cumulative Flow and the Calculated Hydraulic
Conductivities.
involves developing a constant flow to the well screen and then measuring
the vertical flow inside the well at designated depths. By successively
differencing the cumulative flow measurements, the incremental amount of
water leaving or entering a length of screen is calculated. Young
(1990a,b) explains the procedures used to install, develop, and test the
wells at CAFB.

At each of the 37 wells, the borehole flowmeter measurements were made at

0.3-meter intervals. The arithmetic mean, the geometric mean, and the
variance of the natural logarithm of 881 hydraulic conductivity
measurements is 0.26 cm/s, 0.032 cm/s, and 4.7, respectively. At a
typical well location, the hydraulic conductivity values range three
orders of magnitude. Figure 3 shows the hydraulic conductivity profiles
at Wells 1 through 6 . Orders of magnitude changes are common at
distances as short as 0.3 meters. Figure 4 shows areal cross sections of
depth-average hydraulic conductivities over two-meter intervals of the
aquifer.

The aquifer is composed of fluvial sediments from the Tombigbee and the
Buttahatchee Rivers. Aerial photographs of Columbus, Mississippi, and
vicinity show outlines of numerous river ox bows (Young, 1990a,b). One
of these ox bows lies within the well network. The location of the river
channel correlates very well with the band of high hydraulic
conductivities from 0.63 to 3.16 cm/s shown in Figure 4 at 60 to 62 m
MSL. Split spoon samples in this region show that these sediments are
predominantly gravels. These gravels probably represent the bed load of
the river channel. The region of lower hydraulic conductivity southwest
of f.he river channel represents pointbar materials formed under
catastrophic depositional events which included occasional floods.
(Herweijer and Young, 1990).

THE DELAYED YIELD PHENOMENA

During a pump test in an unconfined aquifer, a transition occurs between

water released from elastic storage and from water released, with some
delay, from the dewatering of the phreatic surface. The drainable water
essentially includes the drainable pore volume of the aquifer between the
original water table and the water table at steady state pumping. Steady
state vertical flow occurs only during late pumping and when the
"drainable" water has reached the falling water table. According to the
theory, when the aquifer is sufficiently stressed by pumping, the
drawdown curve characteristic of delayed yield will be sigmoid shaped,
with the flat segment representing the transitional phase of the flow
regime.

For homogeneous unconfined aquifers, Streltsova (1972) and Neuman (1972)

present a well defined physical concept for interpreting the
sigmoid-shaped curves. Streltsova's (1972) approximation implies that
the unconfined aquifer can be simulated by an aquitard overlying a
confined aquifer. The aquitard has a zero transmissivity and the storage
coefficient equal to the specific yield. The vertical resistance of the
 E-3 E-2 &1 E+0 E+1&4 E-3 &2 &1 E+0 E+1 E-4 E-3 E-2 E>1 E+0 E+1
K(cm/s) K (cm/s) K (cm/s)

Figure 3. Hydraulic Conductivity Profiles for Wells 1-6.

54 - 56 M (MSL) 56 - 58 M (MSL)
T— — 1-------- TS“ T-------- 1------
100 •

80
JÊ ** «
6 60 -
> OEPtH-fcVERAGEO
HYORAUUC CONDUCTIVITY
40 ip S r iT '
' -8.2 « LOG, g « 0.S
20 1 -0.3 < L0C,e < -0.2

•"■ii: -1.6 < L0G,e * -0-9

58 - 60 M (MSL) 60 - 62 M (MSL)
ii -2.3 < LOG,g < -1.6

-3 0 < LOC10 < 5

m -3

- J . T « LOG,0 < -3.0

m
* LOG,g * - 3 . 7
■ - 4 .4

mm
S.l < LOG,) « - 4 .4

60 80 100 60 80 100
X (m) X (m)
Figure 4. Cross'Sections of the Interpolated Three-Dimensional Hydraulic
Conductivity Field.
aquitard is three times smaller than the aquifer's vertical resistance.
Neuman's (1972) solution is refined: the vertical resistance of the
aquitard varies as a function of the distance to the pumping well. This
corrects for the differences in vertical trajectory at various distances
from the pumping well. Later, Neuman (1975) also described the
anisotropic case, giving type curves to determine the vertical hydraulic
conductivity Kz .

PUMPING OF A FICTITIOUS AQUIFER WITH A LENTICULAR ARCHITECTURE

The cross section sketched in Figure 5 is a possible representation of

the heterogeneous aquifer architecture at CAFB. Fully-screened wells are
assumed to be drilled in lenses of very high (Kl) and moderately high
(K2) hydraulic conductivity (all relative to K3). When pumping well B,
most of the water will be withdrawn from the high hydraulic conductivity
Kl lenses. Nearly simultaneous drawdown will be observed in well A,
which is connected to B by one of the lenses. Water will be released
from elastic storage in the lens. With some delay, a hydraulic gradient
will be established, resulting in vertical and lateral flow towards the
high conductivity lens. A storage coefficient based on early rising
portion of the curve will be in the order of 1 0 -s (the elastic
storage coefficient).

Well C, however, at the same distance from well B as well A, is not

connected by a lens to well B. Thus, the water level in well C does not
respond to pumping of well B as quickly as in well A. Analyzing the
early part of the curve would give storage coefficients higher than the
elastic storage. However, since a lens connected to B is rather close
by, the storage coefficient estimated from the early part of the curve
will be smaller than the specific yield.

Given the irregular pattern of lenses and hydraulic conductivities, the

transition between the release from elastic storage and the release from
the drainable pore volume will not be sharp and will depend primarily on
the orientation of the groundwater wells with respect to the geometry of
the aquifer lenses. In such heterogeneous aquifers, both the application
and the interpretation of results from analytical models similar to
Neuman's (1975) would be difficult. Therefore, a simple approach was
selected for our analysis. Drawdown curves were fitted based on the
Theis formula and Jacob's correction for unconfined aquifers was used.

THE WELTEST PROGRAM

WELTEST is a computer program that automatically determines the

transmissivity and storage coefficient values that produce a drawdown
curve which best matches the experimental data set. Program WELTEST uses
the method of nonlinear least-squares regression to obtain the best
match. The concept of applying a computerized analysis to determine the
"best" transmissivity and storage coefficient was introduced by
Vandenberg (1971). Vandenberg's (1971) program was specifically written
 LEGENO:
K » Hydraulle Conductivity
> >Kj>Kg> > K 4

WELL WELL
A B

3 r l

EZ

E l

CROSS-SECTIONAL VIEW
OF A FICTITIOUS AQUIFER
Figure 5. A Cross-Sectional View of a Fictitious Heterogeneous Aquifer.
for drawdown measurements in an observation well near a well pumping at a
constant rate from a semi-infinite, nonleaky aquifer. An examination of
the program reveals several potential problems, among which is the
inability to properly converge in some situations.

The WELTEST program was written to solve for best values of

transmissivity and storage coefficient for either a confined or an
unconfined aquifer, for either leaky or nonleaky conditions, and for
constant or variable pumping rates. WELTEST does not account for any
type of borehole storage or borehole skin effects. The WELTEST program
algorithms are more computationally efficient than the algorithms used in
the programs similar to Vandenberg's (1971). The modularity and speed of
WELTEST results by using Broyden's Method (Benton, 1990) to develop
derivative-free algorithms for locating extrema of nonlinear equations.

The input requirements for WELTEST include the experimental data, the
distance to the observation well, the pumping schedule, the thickness of
the aquifer, and whether the aquifer is confined or unconfined. The main
output from WELTEST is the "best" values for transmissivity and storage
coefficient. Optional output from WELTEST includes a sensitivity
analysis for each parameter, a plot comparing the predicted and the
observed pump-test curves, and/or a map of the residuals.

The map of the residuals provided by WELTEST illustrates the sensitivity

of the predicted time-drawdown response to transmissivity and the storage
coefficient values. To create a map of the residuals, program WELTEST
generates a series of hypothetical time-drawdown aquifer responses for
different sets of transmissivity and storage coefficient values. For
each of these time-drawdown responses, program WELTEST divides the
difference in the areas between the predicted and the observed time-
drawdown response by the total area beneath the observed time-drawdown
curve. For convenience, the residual is expressed as a percentage. A
residual of 0 . 1 0 means that the difference between area of the predicted
and the observed time-drawdown curves is 10 percent of the total area
beneath the observed time-drawdown curve. A residual of 0.0 means that
the predicted and the observed time-drawdown curves are exactly the same.

Figure 6 demonstrates an application of WELTEST and the importance of the

aquifer test design. The figure illustrates the residuals created for
observation wells at different distances from the pumping well for a
hypothetical aquifer with a transmissivity and a storage coefficient of
30 cm2/s and 0.03, respectively. Case 1 is for a constant pumping
rate of 40 L/min for 2k hours. Case 2 is for a 2-hour-on and 2-hour-off
pumping rate of 80 L/min for 24 hours. A comparison of the residuals for
the cases shows that the sensitivities of the solution to the well
equations to both storage coefficient and transmissivities are remarkably
improved by pulsing the pumping well.
 CASE 1: CONSTANT PUMPING AT 40 L/MIN FOR 24 HOURS
.08 T
LEGEND ZRESIDUAL LEGEND XRESIDUAL
.07

. 0G 2.5-7.5
C O E F FIC IE N T

.05 7.5-12.5

.04 12.5-17.5
17.5-22.5 17.5-22.5
STORAGE

.03

22.5-27.5 22.5-27.5
.02
27.5-30.0 27.5-30.0
.01 R=3a R=9m
MAXIMUM ORAHOOHN-0 MAXIMUM DRAWDOWN=
.00

CASE 2: CYCLICAL (2-HOUR-ON AND 2-HOUR-OFF) PUMPING AT 80 L/MIN FOR 24 HOURS

.0 8 — ------ 1----
LfGENDWESIOUAL LEGEHO XRES10UAL
.07 -
(" 2 2 3 G O
.06 - 2.5-7.5
C O E F F I C IE N T

.05 - 7.5-12.5:

.0 4 -
12.5-17.5 12.5-17.5
17.5-22.5 17.5-22.5
STORAGE

.03 -

22.5-27.5 22.5-27.5
.02 -
27.5-30.0 27.5-30.0
.01 -
R=3. R =9m
MAXIMUM DRAW D O W N S MAXIMUM DRAWDOWN^
00 i-------- 1 ------ 1-----
TR A N S M IS S IV ITY ICM2/ S E C ) T R A N S M IS S IV ITY ICM2/SEC1
ENG LAB 00/28/90:1602 FIG6.PLT

Figure 6. WELTEST Maps of the Cumulative Residual (as a Percentage' for a Fictitious Homogeneous Aquifer
with a Transmissivity and a Storage Coefficient of 30 cnr/s and 0.03, Respectively.
DESCRIPTION OF THE MULTI-WELL AQUIFER TESTS

Between June 1989 and August 1989, three large-scale multi-well aquifer
tests and seven small-scale multi-well aquifer tests were conducted. The
large-scale tests included pumping Well 5 (the center well) for about 6
days. The small-scale aquifer tests were conducted at well clusters with
closely spaced wells (3-6 meters) and had pumping durations of less than
3 hours.

Large-scale aquifer tests 1 and 3 involved constant pumping rates of

68 L/min and 112 L/min. Each test included manually monitoring the water
table at all 37 wells and automatically monitoring the water table at 9
wells. The second large-scale aquifer test included cyclic pumping at
Well 5 and measuring the drawdowns at the wells with pressure
transducers. The targeted average pumping rate was 68 L/min, which was
achieved by a series of pulses. A pulse included a period during which
the pumping rate was approximately 120 L/min and a period during which no
pumping occurred. The small-scale aquifer tests included pumping
Wells 12, 13, 16, 19, 24, 25, and 31 at a constant flowrate. During the
seven tests, a total of 47 observation wells were monitored for
drawdown. Of these 47 well records, 33 well records were suitable for
analysis.

DATA ANALYSIS

Effect of the Test Duration

Figures 7 and 8 show the transmissivities and storage coefficients

calculated for the large-scale aquifer tests 1 and 3 at times 2,000;
10,000; 50,000; and, 100,000 seconds. The figures show a trend for the
transmissivity to decrease and the storage coefficient to increase with
the duration of the test. Over the range duration of 2,000 to 100,000
seconds, the calculated transmissivities and storage coefficients vary by
factors up to two and ten, respectively.

Effect of the Design of the Aquifer Test

Figures 9 and 10 show the values of transmissivities and storage

coefficients calculated at the same well during different aquifer tests.
In order to minimize the effect of the different durations between the
small-scale and the large-scale tests, only the hydraulic values
calculated at10 ,0 0 0 seconds were used from the large-scale aquifer
tests. The small-scale tests had durations between 3,000 and 9,000
seconds. The figures show that the design of the aquifer tests has a
dramatic impact on the calculated hydraulic properties of the well. At
several well locations, the storage coefficient varies over five orders
of magnitude and/or the transmissivity varies over an order of magnitude.
 LARGE-SCALE AQUIFER TEST 1
[cm /si
TRANSMISSIVITY

LARGE-SCALE AQUIFER TEST 3

400
£3- WELL 2 WELL 20
350 - o - WELL 4 - A - WELL 25
WELL 6 WELL 30
WELL 14 - S ~ WELL 31
fcmVsl

300 - WELL 17
Q._
o ...
250 -
TRANSMISSIVITY

200 -

..s..
150
B -.
100 f- .. .. B ......... i.
... B...
50

0
10 - 10 < 10 s
TIME [ (seconds 1
Figure 7. Transmissivities as a Function of Time for Large-Scale Aquifer
Tests 1 and 3.
 L A R G E - S C A L E A Q U I F E R TEST 1
COEFFICIENT
STORAGE

L A R G E - S C A L E A Q U I FER TEST 3
COEFFICIENT
STORAGE

Figure 8. Storage Coefficients as a Function of Time for Large-Scale Aquifer

Tests 1 and 3.
 WELLS IN THE WESTERN REGION

PUMPING WELL
5 (ATI) 5 (AT2) 5 (AT3) 13 16 19 31
300 O * □ • ■ ▲ A

200
TRANSMISSIVITY (cm? /»)

100

■
50 S
O

30 * * *

▲
20 □

10
2 4 6 13 14 16 18 19 20 2«

300 WELLS IN THE EASTERN REGION

200 PUMPING WELL
5 (ATI) 5 (AT2) 5 (AT3) 12 24
O * □ • ■ i5

100
TRANSMISSIVITY (en?/s)

□
□

50 ▲ O
• □
A O Cl
«
►

* •
O

▲ • ▲
30 * *
— * $
A *
• * • ■
20 _ ■
*

*
10 1 .. I I ■ I 1 I I i i
10 ii 12 15 17 21 24 25 30
OBSERVATION WELL

Figure 9. Comparison of Transmissivities Calculated From Different Aquifer

Tests at the Same Well Location.
STORAGE COEFFICIENT WELLS IN THE WESTERN REGION

WELLS IN THE EASTERN REGION

1E+00
PUMPING W ELL
5 (AT2) 5 (AT3) 12 24
5 % T1> * □ • ■ A 2
IE-01 *
*
*
o *
*
o
1E-02 - a O □ *
*
STORAGE COEFFICIENT

▲
1
■ * O O
• ■
IE-03
• □
• •
A □ ▲
•
1E-04
▲ A •
•
IE-05

1E-06 ..l, . ___ J __ I . ... 1 1 l-

10 11 12 15 17 21 24 25 30

OBSERVATION WELL

Figure 10. Comparison of Storage Coefficients From Different Aquifer Tests

at the Same Well Location.
Effect of the Distance Between the Pumping and the Observation Wells

Figures 11 and 12 show the transmissivity and the storage coefficient

values as a function of the distance between the pumping well and the
observation well. No correlation is evident between transmissivity and
distance, but a correlation appears to exist between the storage
coefficient and distance. At distances of less than 10 meters, the
storage coefficients are within the range for confined aquifers (1 0 -6
to 10"2 ). Storage coefficients typical for unconfined aquifers
(1 0 ~ 2 to 1 0 “ 1) are consistently found only when at distances
greater than 20 meters. For distances between 10 and 20 meters, the
storage coefficients typically range from 1 0 - 4 to 1 0 -2.

DISCUSSION OF RESULTS

Given the very heterogeneous nature of the aquifer, the observed trends
depicted in Figures 7 and 8 can be explained in terms of crossflow; the
flow that may occur perpendicular to the radial flow to the well. When a
heterogeneous aquifer is pumped, hydraulic pressure changes occur first
in the zones of high diffusivities. If a fully-screened well intersects
one of the high diffusivity zones, then the potentiometric surface in a
well will more closely reflect the hydraulic pressure in the high
diffusivity zones rather than the average pressure in the aquifer at the
well location. Consequently, at early times, an analysis of the well
data will lead to estimates of transmissivities and storage coefficients
more representative of the zones of high diffusivity rather than of the
total aquifer thickness. As the duration of the pump test increases,
crossflow occurs from the zones of low diffusivity (high pressure) to
high diffusivity (low pressure). Over time, crossflow dissipates the
pressure differences between the different aquifer zones. Consequently,
at late times, an analysis of the well data leads to estimates of
transmissivity and storage coefficients less bias toward the zones of
high diffusivity than at early times.

The trends in Figures 9 and 10 show that the pumping rate, the
orientation of the pumping well to the observation well, and the distance
between the pumping well and the observation well affect the calculated
hydraulic properties. These trends can be attributed to very different
regional properties of the aquifer evident in Figure 4, created by the
different geological facies across the site. The wide range in storage
coefficient values and the trend shown in Figure 12 can be accounted for
in the conceptual model of an aquifer with lenticular architecture shown
in Figure 5.

SINGLE-WELL AND MULTI-WELL AQUIFER TRANSMISSIVITY FIELDS

During July 1989, single-well pump tests were conducted at each of the
37 wells. Each of these tests had pumping rates between 10 and 20 L/min
and lasted approximately 30 minutes. Traditionally, single-well tests
have been used primarily to determine only transmissivity values.
 DISTANCE (meter)
Figure 11. Transmissivity Values as a Function of Distance Between
the Observation and the Pumping Well.

DISTANCE (meter)
Figure 12. Storage Coefficients as a Function of Distance Between
the Observation and the Pumping Well.
Storage coefficients can be calculated but these coefficients are very
sensitive to several uncertain properties of the well. These properties
include, but are not limited to, the effective radius of the well, the
skin effect, and the storage capacity of the well.

Figures 13 and 14 show the depth-averaged hydraulic conductivity field

determined from the multi-well aquifer tests and the single-well tests.
For aquifer tests 1 and 3, and for the single-well pump tests, the
transmissivity values were determined with a Cooper-Jacob straight-line
equation (Cooper and Jacob, 1946).

For aquifer test 2, the transmissivity values were determined by

UELTEST. Figure 13 displays considerable less heterogeneity than
Figure 14. In comparing the two figures, order of magnitude differences
are not uncommon at different well locations. These differences indicate
that multi-well aquifer tests are not well-suited for determining the
architecture of the transmissivity fields in heterogeneous aquifers.

CONCLUSIONS

Three large-scale (100 meters) and seven small-scale (3-7 meters)

multi-well aquifer tests were conducted in a fluvial unconfined aquifer
across a one-hectare test site. The results show that order of magnitude
differences in the calculated transmissivities and storage coefficients
at a well location can occur by varying the pumping rate at the pumping
well, the duration of the test, and/or the location of the pumping well.
The sensitivity of calculated hydraulic properties to the design of the
aquifer tests is attributed to the aquifer's heterogeneities. One trend
is that as the test duration increases, the values of the calculated
storage coefficients increase and the calculated transmissivities
decrease. This trend is attributed to crossflow. Crossflow occurs
during an aquifer test if the heterogeneities in the aquifer architecture
produce pressure differences in the direction perpendicular to the radial
flow to the well.

The calculated storage coefficients ranged from 10- 6 to 10_1.

When the distance between the observation and the pumping wellsis less
than 10 m there is a high probability of the calculated storage
coefficient being less than 10-4. When the distance between the
observation and the pumping wells is greater than 20 m there is a high
probability of the calculated storage coefficient being greater than
10-2 . The trends observed between the calculated storage
coefficients and the distance is attributed to the lenticular
architecture of the aquifer. Low values are calculated for the storage
coefficients when a highly transmissive thin lens (or a series of lenses)
intersects both the pumping and the observation wells. High values are
calculated for the storage coefficients when the material between the
observation and the pumping wells are relatively homogeneous. The
likelihood of calculating low storage coefficients decreases with
Y (m)

D EPTH-AVERAGED
H YDR A U L I C CONDUCTIVITY

: 1 -8.2 « LOG,0 - 0.5

-8.9 < LOG,g < -0.2

-1.6 < LOG, g < -0.9

-2.3 < LOG,q < -1.6

m
ÜÜ

-3 0 < LOG| g < - 3 3

m -3.7 < LOC,e < -3.0

-4.4 * LOC,0 * -3.7

■
|j|||j -S.1 « LDGjg « -4.4

X (m)

Figure 13. Transmissivity Fields Based on the Large-Scale Aquifer Tests.

 100 D E P T H -A V E R A G E D
H Y D R A U L IC C O N D U C T IV IT Y

- -0 .2 < L O G ,g * 0 .5

80 -0 .9 < LOGl0 < - 0 . 2

-1 .6 < L O G ,g < - 0 . 9
E

füü -2 .3 < L O G ,g < - 1 . 6

60 - 3 0 < LOG, 0 < - 2 3

■
-3 .7 < LO G , g < - 3 . 0
■
-4 .4 * LOG, 0 « -3 .7
■
40 H tttT
rTTTTT
lllf IT - S . 1 « L O G j g < - 4 . 4
rTTTTT

20

20 40 60 80 100 120

X (m)
Figure 14, Transmissivity Field Based on the Low-Rate Pumping Test Results From the 37 Wells.

141
increasing distance between the pumping and the observation wells,
because at greater distances there is less likelihood that a thin lens of
high conductivity intersects both wells.

A comparison between the transmissivity fields calculated from the

multi-well and the single-well tests shows order of magnitude differences
at several well locations. Detailed geological investigations and seven
tracer tests at the test site confirm the transmissivity architecture
indicated by the single-well tests. This comparison indicates that
multi-well aquifer tests are not well-suited for determining the
architecture of the transmissivity field in heterogeneous aquifers.

REFERENCES

Benton, D. J., 1990, "An Improved Derivative-Free Algorithm for Locating

ExtremaBased on the Broyden Method," TVA Engineering Laboratory
Unpublished Report, Tennessee Valley Authority, Norris, TN.

Cooper, H. H., and C. E. Jacob, 1946, "A Generalized Graphical Method for
Evaluating Formation Constants and Summarizing Well-Field History,"
Transactions of the American Geophysical Union, 217, pp 626-634.

Herweijer,J. C., and S. C. Young, 1990, "Use of Detailed

Sedimentological Information for the Assessment of Aquifer Tests and
Tracer Tests in a Shallow Fluvial Aquifer," presented at the
Canadian/American Conference on Hydrogeology: Parameter Identification
and Estimation for Aquifer and Reservoir Characterization,"
September 18-20, Calgary, Alberta.

Neuman, S. P., 1972, "Theory of Flow in Unconfined Aquifers Considering

Delayed Response of the Water Table," Water Resources Research, Vol. 8 ,
No. 4, pp 1031-1045.

Neuman, S. P., 1975, "Analysis of Pumping Test Data From Anisotropic

Unconfined Aquifers Considering Delayed Gravity Response," Vol. 11,
No. 2, pp 329-342.

Streltsova, T. D., 1972, "Unsteady Radial Flow in an Unconfined Aquifer,"

Water Resources Research, Vol. 8 , No. 4, pp 1059-1079.

Olea, R. A., 1984, Systematic Sampling of Spatial Functions, Kansas

Geological Survey; Series on Spatial Analysis, No. 7.

Vandenberg, A., 1971, "Pump Testing in Heterogeneous Aquifers," Journal

of Hydrology, Vol. 34, pp 45-62.

Warrick, A. W., and D. E. Myers, 1987, "Optimization of Sampling

Locations for Variogram Calculations," Water Resources Research, Vol. 23,
No. 3.
Young, S. C., 1990a, "A Site Characterization Methodology for Aquifers in
Support of Bioreclamation Activities, Vol. I: Well Network Design, Well
Equation, and Aquifer Multiwell and Single-Well Tests," report in review,
United States Air Force Engineering and Services Center, Tyndall AFB, FL,
178 pp.

Young, S. C., 1990b, "A Site Characterization Methodology For Aquifers in

Support of Bioreclamation Activities, Vol. II: Borehole Flowmeter
Technique, Tracer Tests, Geostatistics, and Geology,” report in review,
United States Air Force Engineering and Services Center, Tyndall AFB, FL,
2 10 pp.