an integrated geophysical investigation of the hydrogeology … · although interpretation was...
TRANSCRIPT
An integrated geophysical investigation of the hydrogeology of an
anisotropic unconfined aquifer
S.K. Sandberga,*, L.D. Slaterb,1, R. Versteegc,2
aDepartment of Geosciences, University of Southern Maine, 37 College Avenue, Gorham, ME 04038, USAbDepartment of Geosciences, University of Missouri—Kansas City, 5100 Rockhill Road, Kansas City, MO 64110, USA
cLamont Doherty Earth Observatory, Route 9W, Palisades, NY 10964-8000, USA
Abstract
The predictive capability of groundwater flow models is frequently restricted by insufficient characterisation of a typically
heterogeneous and anisotropic subsurface. Trace levels of volatile organic compounds have been detected at municipal water
supply wells in Gray, Maine. Groundwater flow modelling based on available hydrogeologic data defines a dominant W–E
transport vector that is inconsistent with the apparent N40E transport of a plume emanating from a road salt storage facility, as
mapped with a terrain conductivity instrument.
A local-scale geophysical study at an undisturbed site in the glacial-marine delta deposit aquifer was conducted to investigate
the possible influence of anisotropy on flow through these unconsolidated sediments. Ground penetrating radar and terrain
conductivity measurements reveal evidence for structural features that are likely to promote preferential transport paths
orientated in a general NE–SW direction. Two conductive tracers, one deep and one surficial, were injected and monitored for
direct evaluation of the groundwater flow vector using resistivity and self potential methods. Although interpretation was
limited by an incomplete resistivity dataset, the results and modelling of both methods supports a general N30E–N40E flow
vector. Furthermore, consistent flow velocity estimates (,0.15 m/day) are obtained from the two methods. Analysis of this
integrated dataset suggests that anisotropy exerts a significant control on flow in this unconsolidated aquifer. Predictions of
sources of groundwater contamination at municipal wells based on flow models not accounting for this anisotropy will likely be
in error. This study illustrates the potential value of an integrated geophysical study, which will aid the development of accurate
flow models for unconsolidated aquifers. q 2002 Elsevier Science B.V. All rights reserved.
Keywords: Hydrogeology; Anisotropy; Electrical resistivity; Self potential methods
1. Introduction
A complex problem in hydrology is the determi-
nation of fluid transport behaviour such as flow
pathways, flow velocity and hydraulic conductivity in
a heterogeneous and anisotropic subsurface. Infor-
mation on fluid flow behaviour (both present and
predicted) is needed to assess the contamination impact
of spills and plumes and the planning and assessment of
remediation efforts. For example, in the state of Maine
residential groundwater supply wells are particularly
vulnerable to groundwater contamination. Geophysical
methods may provide a relatively low-cost approach to
hydrogeologic characterisation. Numerous papers
0022-1694/02/$ - see front matter q 2002 Elsevier Science B.V. All rights reserved.
PII: S0 02 2 -1 69 4 (0 2) 00 1 53 -1
Journal of Hydrology 267 (2002) 227–243
www.elsevier.com/locate/jhydrol
1 Fax: þ1-816-235-5535.2 Fax: þ1-845-365-8150.
* Corresponding author. Present address: Department of Geology,
University of South Florida, 4202 East Fowler Avenue, SCA 528,
Tampa, FL 33620, USA. Fax: þ1-813-974-2654.
E-mail addresses: [email protected]
(S.K. Sandberg), [email protected] (L.D. Slater),
[email protected] (R. Versteeg).
illustrate the utility of geophysical methods in defining
subsurface heterogeneity in aquifer properties (Kelly
and Mares, 1993). Other methods, notably azimuthal
resistivity (Ritzi and Andolsek, 1992), or electromag-
netic azimuthal resistivity (Slater et al., 1998) provide
information on aquifer anisotropy.
Geophysical tracer tests can provide additional
information on flow directions and permit estimations
of groundwater flow velocity at a scale defined by the
experimental configuration. Forced-gradient tests
involve pumping to induce an artificial flow field in
the system. A natural-gradient tracer test may yield
more significant results since the tracer should move
through the system in a similar manner to the
undisturbed groundwater. This test permits estimation
of the natural groundwater flow vector. Most previous
work involves monitoring of conductive tracer tests
using variations of the electrical resistivity method.
White (1988) described an effective study using a
surface resistivity array in a radial pattern away from the
injection well. White (1994) compared the effectiveness
of six surface resistivity arrays for tracer detection.
Other workers employ a modified mise-a-la-masse
method, incorporating an electrode placed within a
borehole (Bevc and Morrison, 1991; Osiensky, 1997).
This paper presents results of an integrated geophy-
sical investigation of the hydrogeology of an unconso-
lidated, unconfined aquifer system in the municipality
of Gray, Maine. Multiple geophysical methods were
used to investigate aquifer heterogeneity and aniso-
tropy. A localised geophysical tracer test was conducted
to estimate the groundwater flow vector. The paper first
introduces the regional hydrogeology of the aquifer and
further details the motivation for the study. The
geophysical characterisation of the hydrogeological
framework is then described and followed by the
geophysical tracer test. The likely significance of the
geophysical results to regional groundwater flow is
discussed, given recognised limitations of the geophy-
sical study. Finally, implications for groundwater
contamination of residential water supply wells in the
surrounding area are considered.
2. Study site and regional hydrogeology
The field site investigated in this study is a glacial-
marine delta deposit comprised of fine- to coarse-
grained sand in the municipality of Gray, Maine. Fig.
1 presents a map of the Gray project area. Key
elements in this study are a landfill, sand/salt storage
pile and municipal water supply wells. In recent years
trace levels of volatile organic contaminants have
been detected in monitoring wells between the landfill
and the municipal water supply wells. The landfill is
approximately 900 m from the wellfield. Hydrogeo-
logical studies have been undertaken over the course
of several years to determine whether the landfill is
the source of this contamination.
A 3D numerical flow model was constructed for
this area and designed to simulate head and flow
patterns in the vicinity of the Gray Water District
springs and the Gray landfill. Observations of head
and estimates of hydraulic conductivity from slug
tests (selected wells only) were obtained from
numerous wells in the study area and used to constrain
the flow model (C. Fitts, University of Southern
Maine, pers. comm., 1998). The model extended
horizontally to known surface water boundaries in
some areas and to mapped margins of the glacio-
fluvial sand aquifer in other areas. A five-layer model
was constructed consistent with observations of
sediment stratigraphy and bedrock surface topogra-
phy. This model indicates that the regional flow is
from west to east between the landfill and municipal
supply wells (Fig. 1). This result suggests that the
landfill is a likely source of groundwater contami-
nation observed in the supply wells. However, model
uncertainty arises close to the landfill due to boundary
condition definition, adding to uncertainty regarding
whether flow from the landfill is captured by the Gray
Water District wells.
Also within the study area, an EM34 terrain
conductivity survey was conducted in the vicinity of a
Health Clinic and an uncovered sand salt storage pile
nearby (Fig. 1). Electromagnetic data were collected in
the horizontal coplanar (vertical dipole) configuration
with a 10 m coil separation. Fig. 2 shows the terrain
conductivity contour map from that study. High terrain
conductivity values trend N40E, between the salt pile
and the health clinic, indicating the direction of
groundwater flow in that area. This geophysical estimate
of groundwater flow direction conflicts with the west–
east flow vector identified in the hydrogeological
modelling. This result casts doubt upon the validity of
the flow model and the interpretation that the landfill is
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243228
the site of groundwater contamination at the Gray
municipal supply wells.
An integrated geophysical study of this aquifer was
conducted to further investigate this discrepancy. A
sandpit between the landfill and municipal supply wells
was identified as a suitable undisturbed environment for
geophysical study (Fig. 1). The study had two primary
aims: first to identify possible evidence for anisotropy in
the aquifer material that could cause the flow vector to
deviate from that predicted by groundwater flow
modelling; secondly to directly determine the ground-
water flow vector using a geophysical tracer test.
3. Geophysical site characterisation
The geophysical grid established in the sandpit
consisted of 26 parallel lines placed 2 m apart, each
Fig. 1. Site map showing the locations of the water supply wells, the landfill, the sand/salt storage pile, and the sand pit where the tracer study
occurred. The regional hydraulic gradient as determined from hydrogeological modelling is approximately west to east.
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243 229
40 m in length, along a centreline oriented N40E
through existing monitoring well MW-201 (Fig. 3).
The long-axis of the grid is thus oriented consistent
with the trend of the terrain conductivity plume
migrating from the sand/salt storage pile (Fig. 2). Also
shown in Fig. 3 is a fan of resistivity radial array data-
collection lines (108 increments) used for tracer
monitoring, passing through MW-20l with a centre-
line orientation of N40E. Three shallow well points
were used to determine the water table gradient within
the sandpit, which was calculated as N60E. In the
absence of horizontal anisotropy, the horizontal
component of shallow groundwater flow would be
in this direction.
Resistivity imaging, terrain conductivity and
ground penetrating radar (GPR) were used to
investigate lithologic and stratigraphic variability at
the study site. Lithologic constraints were provided
from well cuttings recorded for MW-201. The 2D
electrical structure was investigated using dipole–
dipole resistivity measurements made with a 2 m
electrode spacing and a maximum dipole spacing (n )
of six. A resistivity model was obtained using an
algorithm combining finite element forward model-
Fig. 2. Terrain conductivity contour map of sand/salt storage pile area. Values are in mS/m, and station locations are shown as þ symbols. Data
were obtained with a Geonics EM34 system, using a 10 m coil separation in the horizontal coplanar (vertical dipole) orientation.
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243230
ling with regularised, weighted, least-squares iterative
inversion, as described in LaBrecque et al. (1996).
The model along Line 11 (Fig. 3) is shown in Fig. 4.
The lithologic interpretation from well cuttings at
MW-201 is shown for comparison. The resistivity
image depicts a two layer structure, with an upper
resistive layer thinning to the southeast. The boundary
correlates with the distinct interface between brown
coarse sand and underlying gray fine sand identified
from well cuttings. Note that the resistivity inversion
provides a smooth model of subsurface resistivity,
such that the boundary is not uniquely defined.
Terrain conductivity measurements were con-
ducted to investigate lateral heterogeneity within the
sandpit. EM31 data were collected in the horizontal
coplanar (vertical dipole) configuration, using a 4 m
line spacing and a 1 m station spacing. A contour plot
of terrain conductivity across the grid is shown in Fig.
5. These data show a general increase in terrain
conductivity toward the south. The data are consistent
with thinning of the high-resistivity surficial layer
(presumably coarser-grained sediments) toward the
southeast, as already discussed on Line 11 for the
resistivity image in Fig. 4. One explanation relates to
the regional model of glacial outwash known to
originate from the north and flow generally to the
south, with deposition of progressively finer grain
sizes. Such a model would presumably yield higher
conductivity values towards the distal (southern) end
of the deposit. Given this, Fig. 5 may indicate a plan
view of thinning toward the southeast. Preferential
groundwater flow (induced by such macro-aniso-
Fig. 3. Plan of the geophysical grid used in the tracer test. Also shown is well MW-201. The orientation of the axis of the grid is N40E. For
reference, the water table gradient based upon three wellpoints is oriented N60E. Key lines used in data interpretation are highlighted.
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243 231
Fig. 4. Two dimensional dipole–dipole resistivity model (Line 11) determined using smoothness-constrained inversion. Correlation with a
primary boundary determined from well cuttings obtained at MW-201 shown for comparison.
Fig. 5. Terrain conductivity map over geophysical grid obtained using a Geonics EM31 instrument in the vertical dipole orientation. Station
locations are shown as þ symbols.
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243232
tropy) may be generally toward the north–northeast,
where the coarser-grained sediment thickness
increases. However, other factors (e.g. variability in
fluid chemistry) could be responsible for the varia-
bility observed on this local-scale grid.
GPR has been used extensively to define the
hydrogeological framework of sites e.g. (Beres and
Haeni, 1991; Bridge et al., 1995). A pseudo 3D survey
was collected at the Gray site (consisting of tightly
spaced profiles of single offset 2D lines) using an
unshielded Ramac system. Data were collected using
both 100 and 200 MHz antennas. Velocity structure
was determined from common midpoint surveys
conducted at the site. Full results of this study are
reported in Versteeg and Birken (1998). Depositional
structures possibly promoting anisotropy within the
delta deposit were the primary target. Representative
cuts out of the 3D datasets are shown in Fig. 6(a) and
(b). Note that the main reflectors along Line 6 are
parallel (Fig. 6(a)), while there is a significant dip in
the reflectors in the line run perpendicular i.e. SW–
NE (Fig. 6(b)). These dipping reflectors are con-
sidered evidence of depositional foreset structures
(formed during delta deposition), probably dipping
towards the south as a result of general N–S glacial
transport (Fig. 6(b)). These imaged structures are
considered further evidence that macro-anisotropy
may be significantly influencing flow in the study
area. Given this, a tracer experiment was designed to
further investigate the relationship between the water
table vector and the groundwater flow vector at this
site.
4. Geophysical tracer test
Multiple geophysical methods are applicable for
detecting conductive tracers. In this study, the initial
intention was to concentrate on EM and GPR
methods. Consequently, pre-injection datasets were
obtained for these methods. Measurements obtained
immediately after tracer injection cast doubt on the
ability of the GPR and EM methods to resolve the
tracer in this environment. Consequently, resistivity
(radial array and modified pole–pole using a down-
hole electrode) and self potential (SP) surveys were
designed and implemented shortly after tracer injec-
tion. Interpretation is hence complicated by the
absence of pre-injection measurements. Although
this limits the dataset, the analysis described here
shows that information on tracer movement is still
resolvable. The study also serves to test the value of
geophysical methods in real world environments
when a pre-injection dataset is not available but
information on the flow vector is required. This is
more often than not the case when investigating
contaminant plumes migrating from old landfills and
industrial facilities.
Resistivity methods are appropriate for monitoring
tracer transport as the bulk resistivity of the earth is a
function of the fluid resistivity (the inverse of fluid
conductivity). In environments where matrix conduc-
tion is insignificant, fluid resistivity and bulk resis-
tivity are directly proportional and scaled by the
formation factor. Consequently, the replacement of in
situ fluid by conductive tracer is identifiable from
resistivity measurements. Numerous studies illustrate
the general applicability of the resistivity method for
detecting tracer transport (see for example, White,
1988, 1994). The SP method measures natural
electrical voltage potentials in the earth that arise
from small amounts of electrical current flow
associated with localised subsurface chemical,
hydraulic or heat gradients (Telford et al., 1990;
Sprunt et al., 1994). The concept of tracer transport
detection using SP is based upon the electrochemical
potential: if the concentration of electrolytes in the
ground varies locally, potential differences are set up
due to the difference in mobility of anions and cations
in solutions of different concentrations. Mapping of
this electrochemical potential is not extensively
recorded in the literature. One example is the
detection of a strong SP signal caused by the ionic
imbalance in concentrations at the boundary of a
landfill (Coleman, 1991). In the study reported here,
the electrochemical potential resulting from ionic
concentration gradients between tracer and in situ
water is used to identify tracer movement.
The choice of tracer concentration is an important
parameter in experiment design. A high salt concen-
tration optimises the success of geophysical detection.
However, this may also promote density driven flow
that is not typically representative of the flow regime
for in situ ground water. In this study, a relatively low
tracer concentration was used to prevent significant
unwanted density effects. In order to investigate
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243 233
possible deviation in the groundwater flow vector
with depth, two tracers were introduced at
different depths in the aquifer. First, 0.66 m3 of
water (conductivity 48 mS/cm) was extracted from
MW-201, mixed with 2.27 kg of ordinary table
salt, and re-injected (conductivity 7.1 mS/cm).
MW-201 is a 6.4 m deep 10 cm PVC-cased well,
with PVC slotted screen from 3.4 to 6.4 m depth
(Fig. 3). Second, a surficial tracer was introduced
on Line 5 of the geophysical grid, 10 m northwest
of MW-201. Table salt was placed and mixed in
standing water in the bottom of an excavated
0.5 m deep pit, and then buried. The water
conductivity before adding salt was 22 mS/cm,
and after adding salt, 35 mS/cm. The two tracers
were placed significantly far apart to permit
individual resolution in the geophysical data.
4.1. Radial array resistivity
In order to identify migration of the borehole
tracer, radial array resistivity data were collected on
radii 108 apart and centred on MW-201 (Fig. 3). This
electrode geometry conforms to a modified pole–
dipole array, with a current electrode at 4.6 m in the
well, a 5 m potential electrode a-spacing (dipole
length) and N (number of dipole distances away from
the well) ranging from 1 to 5. The lack of a pre-
Fig. 6. Representative radar images obtained along, and perpendicular to, the geophysical grid lines. (a) Line 6 radar result showing correlation
with lithologic log at MW-201. Note that the reflectors are primarily horizontal in this NW–SE direction. (b) Radar result run perpendicular to
geophysical grid lines (i.e. NE–SW). Note the dip in the reflectors. See Fig. 3 for line orientations and location.
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243234
injection dataset complicates interpretation of tracer
transport direction. However, the transport direction is
discernible from analysis of the post injection
datasets. Apparent resistivities were calculated based
on the measured resistance and the known geometric
conversion factor for the array. Consider first the plan
view of apparent resistivity for azimuthal angles
northeast of MW-201 on August 13 (Fig. 7(a)). In the
absence of significant subsurface heterogeneity, the
contours of apparent resistivity would be radially
symmetrical about the current electrode in the well. A
significant deviation from symmetry is observed and
highlighted by the boxed area in Fig. 7(a). The result
indicates a heterogeneity that is most developed along
the N30E radius. Apparent resistivity at N ¼ 3 as a
function of azimuth is expanded in Fig. 7(b). A wide
zone of high apparent resistivity with a slight decrease
in the middle of the zone, centred at N30E is flanked
on either side by lower apparent resistivity, which is
the expected pattern for a conductive subsurface body.
Current flow lines are drawn into the conductor,
causing perpendicular isopotential lines to converge
on either side of the conductive region. This causes an
increased number of isopotentials per radial distance,
resulting in a measured higher voltage, and a higher
apparent resistivity calculation on either side of the
conductor.
Analysis of temporal changes in the radial array
dataset suggests that this conductor is indeed the
tracer, rather than some other subsurface heterogen-
eity at the site. To illustrate, apparent resistivity as a
function of N separation along N30E is shown for data
collected on August 8, August 13, September 20, and
November 8. The general trend from high values at
small N separation to lower values at larger N
separation reflects the increasing penetration depth
obtained with increasing N separation. Data at low N
separation are more influenced by the upper resistive
layer (Fig. 7(c)). Temporal changes in the apparent
resistivity profile are evident in Fig. 7(c). Apparent
resistivity increases with time are indicative of a
conductive slug moving away from the well and away
from potential electrodes. This effect is most obvious
in the N ¼ 1 data. Apparent resistivity at N ¼ 2 for
August 8 and 13 are near identical, indicating that the
tracer had not migrated far enough to influence the
data at this time. However, recovery to higher
apparent resistivity at N ¼ 2 for September 20 and
November 8 is indicative of tracer moving away from
these electrodes between these times. Temporal shifts
in apparent resistivity are also observed along other
azimuths for N ¼ 1 data. However, the apparent
resistivity recovery is slower over time for the N30E
orientation than that for all other azimuths (i.e.
measured apparent resistivity is perturbed for a longer
time on the electrodes orientated N30E). This further
indicates that the elongation and transport of the
saltwater slug centred on this azimuth.
4.2. Self potential
Self potential (SP) data were collected on the
geophysical grid, at approximately 5–9 day intervals,
commencing 3 days after tracer introduction.
Measurements were obtained using Cu/CuSO4 non-
polarizable electrodes, and a high-impedance digital
voltmeter. Data were collected using a fixed reference
electrode, and a roving electrode at stations on the
geophysical grid. The reference electrode was located
east–northeast of MW-201, as shown in Fig. 3. Base
station readings near this reference electrode were
reoccupied in order to account for thermal drift in the
signal.
Clearly detectable SP anomalies are associated
with tracer addition and subsequent transport. SP data
obtained on Line 6 (Fig. 3) are shown in Fig. 8(a).
These data illustrate the development of a low SP
anomaly at 210 m, consistent with the surface
injection location. The sharp anomaly in the August
7 dataset increases in amplitude and expands laterally
in the August 16 dataset, then narrows with decreased
amplitude in the August 21 dataset. An SP anomaly
near the well, at Station 0, appears offset toward the
northeast. The process of borehole injection may have
forced tracer up-gradient and in preferential direc-
tions, in part controlled by local heterogeneity near
the well screen. Consequently, the interpretation of
borehole tracer movement is based on Line 8 away
from the well and these localised complicating effects.
SP data obtained on Line 8 are shown in Fig. 8(b).
Since Line 8 is 6 m down gradient from the tracer
injection points, the August 2 dataset was used as a
background reference. The assumption is that the
tracers have not influenced data on Line 8 by this time,
and the August 2 data are indeed background values.
Natural voltage variability exists throughout the
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243 235
survey area, as evidenced by noisy raw voltage
readings as shown in Fig. 8(a) for Line 6. A direct
subtraction of background voltages (in this case, data
from August 2) accounts for this variability. In
addition, a three-point running average filter was
applied to these data, further reducing noisy
responses. Lastly, base station readings obtained
near the reference electrode, were used to level
datasets from different days. The data in Fig. 8(b)
appears smoother than that in Fig. 8(a) due to this
processing.
For data from Line 8, the temporal variation,
indicating movement, is evident for both surficial and
well tracers, shown at Stations 29.5, and 0,
respectively. These data, collected 6 m down grid
from tracer introduction (Line 8 in Fig. 3), clearly
show that tracer migration is almost directly down the
axis of the geophysical grid, at N40E, as any other
flow direction would have significantly shifted the
anomaly centres from Stations 210 and 0.
The SP signal at 210 m on Line 6 is plotted as a
function of sample day in Fig. 9. By assuming that the
peak anomaly correlates with the condition of
maximum tracer concentration at this point, it is
possible to derive a simple estimate for a 1D seepage
velocity. A smooth curve was fit to the sparse SP
readings in Fig. 9, allowing an estimate of the time of
peak concentration (tpeak) on Line 6, 2 m down grid
from the surficial injection point (Fig. 9). Using
tpeak < 13 days, the estimated flow velocity is 2 m/13
days < 0.15 m/day. A similar calculation for the SP
data on Line 8 (Fig. 8(b)) for the well tracer (Station
0) yields a flow velocity of 0.17 m/day.
4.3. Resistivity modelling
Modified pole–pole data were obtained 3 and 10
days after tracer injection. One current electrode was
placed at 4.6 m in MW-201, with the other at 35.1 m
on Line 30. Voltages on the grid were measured
relative to a reference electrode at 12 m on Line 15.
Three-dimensional finite-difference modelling of this
dataset was attempted using MODFLOW (McDonald
and Harbaugh, 1988). As the governing equations of
electrical flow and groundwater flow are directly
analogous, it is possible to substitute electrical
current, electrical potential and electrical conduc-
tivity, for discharge, hydraulic head and hydraulic
conductivity (Jansen and Taylor, 1995). Examples of
the use of MODFLOW for 3D modelling of resistivity
data include Osiensky and Williams (1996) and
Osiensky (1997).
The 2D resistivity model obtained from inversion
of the dipole–dipole dataset (Fig. 4) was used to
define a pre-injection MODFLOW model. This model
implies a constant resistivity structure in the N40E–
S40W direction. Terrain conductivity data (Fig. 5)
indicate variation across the grid, although variability
is minimal within the region bounded by the tracer
injection and Lines 6 and 8. It was necessary to define
sharp model boundaries from the smooth inverse
model. The lithologic log from MW-201 was used to
constrain these boundaries. The major boundary is at
3.1 m, corresponding to the interface between med-
ium coarse sand and underlying silty sand. These
layers were assigned resistivities of 1000 and
525 V m, respectively, based on 1D and 2D resistivity
modelling results. To optimise the fit of data away
from the injection site, it was necessary to include an
upper 0.2 m resistive layer (4000 V m likely to
correspond with unsaturated sand). Although this
layer is not evident in Fig. 4, it was resolved with a
Schlumberger sounding (utilising smaller electrode
spacings) at the site.
The primary objective was to obtain a second
geophysical estimate of transport velocity for com-
parison with that obtained using the SP method. This
was achieved by modelling resistivity perturbations
caused by tracer addition in terms of tracer location at
3 and 10 days after tracer injection. This modelling
focuses on the surface tracer, as the SP data provide
evidence for complexity in the geometry of the
borehole-applied tracer at early times after injection
(discussed in the previous section). The fit of the
Fig. 7. Resistivity radial array results showing: (a) apparent resistivity contour map where values are plotted at the midpoint of the receiver
dipole. The dashed box indicates anomalous apparent resistivity, signified by non-radial symmetry. (b) An expansion of N ¼ 3 apparent
resistivity plotted versus azimuthal angle. (c) Apparent resistivity as a function of N separation showing variation over time for an azimuthal
angle of N40E.
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243 237
dataset to the estimated model 3 days after injection is
shown in Fig. 10. Data between 220 and 25 m were
excluded as the model overestimates the measured
apparent resistivity, probably due to the thinning of
the upper resistive layer in this direction (Fig. 4).
Otherwise, the model provides a close fit to the
measured data. Perturbations of measured apparent
resistivity from the model curve around 210 m and
0 m are considered the response of the injected
tracers. The modelling exercise focuses on the larger
perturbation associated with the near-surface
injection.
In order to model the tracer at 3 days after
injection, a conductive body was added to the
model. Thirty model perturbations (varying the
location, dimensions and resistivity of the body)
were analysed. The range of models was constrained
by consideration of the known initial tracer conduc-
tivity, tracer volume and estimated formation factor
(0.35). As a consequence of the low tracer concen-
tration, density effects were avoided and therefore not
considered in the modelling. The best fit to the
measured data, as shown in Fig. 11(a), was obtained
for Model B in Fig. 11(b). Although the fit is not
perfect, the shape of the model curve matches the
shape of the field data, supporting the presence of a
small volume of tracer. To enhance the perturbations,
Fig. 11(b) also shows the difference between (1)
measured data and the initial model, and (2) model B
and the initial model. Data on Line 8 (6 m north of the
injection point) show no significant deviation from the
initial model, indicating that, at this time, tracer had
not migrated far enough for detection.
Data collected 10 days after injection show a
suppressed perturbation (relative to 3 days after
injection) on Line 6 at 210 m and tentative evidence
of a response on Line 8 (Fig. 11(b)). This response is
consistent with tracer displacement along N40E.
Thirty model perturbations were again analysed,
again constrained by the N30E–N40E flow direction.
The best fit to the measured data, as shown in Fig.
11(a), was obtained for Model C in Fig. 11(b). The
differences between (1) measured data and the initial
model, and (2) model B and the initial model, are also
shown. Again, the fit to Line 6 is imperfect but
consistent with further tracer movement N40E. The
Fig. 8. Self potential voltage versus distance along Lines 6 and 8. (a) Line 6 voltage data points are plotted without smoothing. (b) SP data from
Line 8 are plotted after smoothing (three-point running average filter), levelling (using base station readings), and the 2 August data has been
subtracted (assuming that August 2 is background for this Line).
Fig. 9. Self potential voltage versus days after injection for Line 6 Station 210 m.
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243 239
subtle response on Line 8 is not resolved in the model.
A simple estimation of tracer seepage velocity was
obtained by calculating the modelled displacement in
the centre of mass of the tracer between the two
datasets (Fig. 11(b)). The result is a seepage velocity
of 0.14 m/day in a direction N40E, consistent with the
calculation from the SP data. There is considerable
uncertainty in this calculation as the simple model
does not completely reproduce the field data. In
addition, the tracer is likely to have a diffuse interface,
rather than the sharp interface modelled here.
However, the agreement with the SP calculation
supports the geophysical estimate of seepage velocity.
At the very least, the SP and resistivity calculations
are likely to provide a credible order of magnitude
estimate of seepage velocity at this site.
5. Discussion
The geophysical results described here indicate
that groundwater flow within the study area is not well
described by existing models based on available
hydrogeological data. The geophysical data and
interpretation support the argument that anisotropy
within the delta deposit exerts a significant control on
flow, such that the resultant flow vector does not
parallel the resultant hydraulic gradient vector. The
work has significant implications, as municipal supply
wells within the Gray project area are threatened by
groundwater contamination. It is necessary to define
the likely sources of contamination if these water
resources are to be adequately protected.
Geophysical studies at different scales generally
support a dominant flow vector orientated N30E–
N40E, which contrasts with the dominant W–E flow
predicted from hydrogeologic modelling of the
aquifer system. At a local scale, geophysical moni-
toring of a tracer injected within a sandpit resolved a
N30E–N40E flow vector, despite a N60E local
resultant hydraulic gradient vector in the vicinity of
the injection. Site characterisation using GPR and EM
methods identified structures within the delta deposit
that are consistent with the interpretation of a
principal axis of anisotropy orientated approximately
NE–SW. The argument is strengthened by the terrain
Fig. 10. Fit of pole–pole resistivity data to background finite difference (MODFLOW) model. (a) Background model superimposed on the
measured data for August 2, 4 days after tracer injection. The influence of the tracer mass injected at 210 m is apparent in Line 2 m north but
absent in Line 6 m north. (b) Model parameters (not to scale) based on the results of 1D inversion of dipole–dipole data in Fig. 4 used for the fit
shown in (a).
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243240
Fig. 11. Results from finite difference (MODFLOW) resistivity modelling. (a) Lines 6 and 8 modelling results for the August 2 dataset. The
tracer dimensions are shown as B in (b). (b) Lines 6 and 8 modelling results for the August 9 dataset. The tracer dimensions are shown as C in
(b). (c) Schematic showing a plan view of the modelled tracer migration for the surficial tracer injected at Station 210 m on Line 0.
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243 241
conductivity mapping of a larger scale conductive
anomaly apparently emanating from a road salt
storage pile. Plume transport again follows a pre-
dominantly N40E orientation. This suggests that the
results of the local geophysical study in the sandpit are
applicable at some sub-regional scale within the
project area.
Interpretation of the geophysical tracer monitoring
was limited by the relatively sparse temporal
sampling applied. The absence of a complete pre-
tracer injection dataset was an unfortunate compli-
cation. In addition, the low tracer concentration
employed in this study resulted in a subtle resistivity
response. However, modelling of two different
geophysical datasets, electrical resistivity and SP,
provided estimates of tracer velocity in close agree-
ment. Although there is significant uncertainty in the
modelling of both datasets, it is likely that the result at
least represents an order of magnitude estimate of flow
velocity at the study site. The degree to which this
estimate is applicable at a regional scale remains
unclear.
One important aspect of this study is the successful
detection of a clear electrochemical SP anomaly
associated with the tracer. The temporal change in
shape of this anomaly resolved a groundwater flow
vector in close agreement with that obtained from
resistivity modelling. Whereas the response of the
resistivity methods to the small tracer volume and
concentration added is subtle, the SP signal is quite
clear. As far as the authors are aware, such use of the
SP method is not reported in the literature. The results
obtained here support further studies on this promis-
ing hydrogeological application of the SP method.
6. Conclusions
This study demonstrates the value of geophysics in
hydrogeological investigations of unconsolidated
aquifers. Hydrogeological models frequently offer
limited prediction performance due to insufficient
characterisation of subsurface hydraulic properties. In
this case, a geophysical study provided evidence that
anisotropy within a delta system is exerting a strong
control on the direction of groundwater flow. This
statement is supported by multiple methods at
different scales. Geophysical characterisation of
deltaic structures, as well as monitoring of an injected
tracer, support the argument. Current models of
groundwater flow developed for this study area do
not account for this anisotropy. It is likely that
predictions of the sources of contamination at affected
wells within the study area will be in error.
Geophysical methods have the potential to constrain
hydraulic parameters and improve groundwater flow
model prediction. In this case, for example, multiple
geophysical tests within the region could assess
spatial variability in anisotropy, flow direction and
flow rate across the groundwater model domain.
Acknowledgements
S.K. Sandberg and L. Slater were supported by the
Maine Science and Technology Foundation under
cooperative agreement 95-06N. R. Versteeg was
supported by the Schlumberger Foundation as the
Schlumberger Junior Scholar. A USM student intern
(Evan Sanborn) was funded by the Gray Water
District and assisted with groundwater modelling.
We thank Jim Foster of the Gray Water District for
access to the water supply wells and springs, as well
as other logistical and moral support. Dr C Fitts
(University of Southern Maine) gave valuable advice
on the hydrogeology of the Gray area. Dr A. Binley
(Lancaster University, UK) provided the 2D resis-
tivity inversion algorithm used to process dipole–
dipole data. We thank the following students who
assisted with geophysical and hydrogeologic field-
work: Jon Drasdis, Stacy Towne, Matthew Doughty,
Kelly Rust, and Evan Sanborn.
References
Beres, M., Haeni, F.P., 1991. Application of ground penetrating
radar methods in hydrogeologic studies. Ground Water 29,
375–386.
Bevc, D., Morrison, H., 1991. Borehole-to-surface electrical
resistivity monitoring of a salt water injection experiment.
Geophysics 56, 769–777.
Bridge, J.S., Alexander, J., Collier, R.E.L., Gawthorpe, R.L., Jarvis,
J., 1995. Ground penetrating radar and coring used to study the
large scale structure of point bar deposits in three dimensions.
Sedimentology 42, 839–852.
Coleman, A.R., 1991. The use of the self-potential method in the
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243242
delineation of a reclaimed landfill site. Proceedings of the
Conference on Planning and Engineering of Landfills, 10–11
July 1991, University of Birmingham, The Midlands Geotech-
nical Society, U.K.
Jansen, J., Taylor, R., 1995. Using MODFLOW as an interpretive
tool for several geophysical methods. Proceedings of the
Symposium on the Application of Geophysics to Environmental
and Engineering Problems (SAGEEP), 699–708.
Kelly, W.E., Mares, S., 1993. Developments in Water Science 44:
Applied Geophysics in Hydrogeological and Engineering
Practice, Elsevier, London, p. 289.
LaBrecque, D.J., Miletto, M., Daily, W., Ramirez, A., Owen, E.,
1996. The effects of noise on Occam’s inversion of resistivity
tomography data. Geophysics 61, 538–548.
McDonald, M.G., Harbaugh, A.W., 1988. A modular three
dimensional finite-difference ground-water model. US Geologi-
cal Survey Techniques of Water Resources Investigations
Report, Book 6, Chapter A1, p. 576.
Osiensky, J.L., 1997. Ground water modelling of mise-a-la-masse
delineation of contaminated ground water plumes. Journal of
Hydrology 197, 146–165.
Osiensky, J.L., Williams, R.E., 1996. A two-dimensional MOD-
FLOW numerical approximation of mise-a-la-masse electrical
flow through porous media. Ground Water 34, 727–733.
Ritzi, R.W. Jr, Andolsek, R.H., 1992. Relation between anisotropic
transmissivity and azimuthal resistivity surveys in shallow,
fractured, carbonate flow systems. Ground Water 30, 774–780.
Slater, L.D., Sandberg, S.K., Jankowski, M., 1998. Survey design
procedures and data processing techniques applied to the EM
azimuthal resistivity method. Journal of Environmental and
Engineering Geophysics 3, 167–177.
Sprunt, E.S., Mercer, T.B., Djabbarah, N.F., 1994. Streaming
potential from multiphase flow. Geophysics 59, 707–711.
Telford, W.M., Geldart, L.P., Sheriff, R.E., 1990. Applied
Geophysics, second ed, Cambridge University Press, Cam-
bridge, p. 770.
Versteeg, R., Birken, R., 1998. Multifrequency three dimensional
radar facies acquisition and analysis. Proceedings on the
Seventh International conference on Ground Penetrating
Radar (GPR98), Lawrence, Kansas, USA.
White, P.A., 1988. Measurement of ground-water parameters using
salt-water injection and surface resistivity. Ground Water 26,
179–186.
White, P.A., 1994. Electrode arrays for measuring groundwater flow
direction and velocity. Geophysics 59, 192–201.
S.K. Sandberg et al. / Journal of Hydrology 267 (2002) 227–243 243