Sie sind auf Seite 1von 10

GEOPHYSICS, VOL. 71, NO. 3 共MAY-JUNE 2006兲; P. B75–B84, 7 FIGS., 2 TABLES.

10.1190/1.2196873

Induced-polarization detection and mapping of contaminant plumes

John A. Sogade1, Francesca Scira-Scappuzzo1, Yervant Vichabian1, Weiqun Shi1,


William Rodi1, David P. Lesmes2, and Frank Dale Morgan1

ABSTRACT nants of interest are benzene and EDB, partly because of their
health risk and partly because they present the highest concen-
Several laboratory and scaled model investigations suggest trations 共2400 and 1000 ␮g/L, respectively兲 among the plume
that organic contaminants affect the surface electrical proper- constituents and are therefore more likely to be related to the
ties of exposed soils/rocks and therefore produce measurable polarization source. IP data were acquired along a survey line
induced polarization 共IP兲 signatures. However, there is little that partially transects the plume extending over contaminated
field evidence of an IP methodology for contaminant mapping. and uncontaminated zones and were inverted to give 2D resis-
A 2D time-domain IP method is developed for mapping the FS- tivity and chargeability plots to 100 m depth and a horizontal
12 contaminant plume at the Massachusetts Military Reserva- extent of 400 m. By separately inverting IP data derived from
tion 共MMR兲 located in Cape Cod, Massachusetts. The FS-12 time windows located at short and long decay times, a time-
plume consists of approximately 265 m3 of fuel that erupted domain gross 共spectral兲 chargeability difference is produced.
from a broken underground pipeline in the early 1970s. Ben- Both the chargeability and gross spectral chargeability differ-
zene and ethylene dibromide 共EDB兲 are the primary contami- ence show good agreement with the known location of the
nants at FS-12, with concentrations exceeding the allowed plume from monitoring wells, with the IP chargeability section
maximum concentration levels 共MCL兲, while other constituents suggesting contaminant distribution detail that cannot other-
of the plume did not exceed their MCL. Therefore, the contami- wise be inferred from the sparse borehole distribution.

INTRODUCTION 共1961兲, Madden and Cantwell 共1967兲, Ward and Frazer 共1967兲,
Angoran 共1975兲, Angoran and Madden 共1977兲, Morgan 共1981兲,
Induced polarization is a phenomenon that takes place at the in- Klein et al. 共1984兲, Olhoeft 共1985兲, Lesmes and Morgan 共2001兲,
terface between an electrolyte and a mineral grain. It is a current- and Lesmes and Frye 共2001兲.
induced effect observed as a delay of the voltage response of the The surfaces of most solids usually possess a net charge by ad-
ground. Conrad Schlumberger proposed the induced polarization sorption of essentially fixed ions. The net charge is a few molecular
共IP兲 effect in 1913 共Allaud and Martin, 1977兲. The systematic sci- layers thick and constitutes what is called the fixed layer. The net
entific study of the IP phenomenon mostly began in the 1950s by surface charge on the solid is a property measured by the ion ex-
Bleil 共1953兲, Hallof 共1957兲, Madden and Neves 共1957兲, Vacquier change capacity 共IEC兲, and it attracts charges of opposite polarity
et al. 共1957兲, Madden and Marshall 共1958, 1959a, b兲, Seigel to form a diffuse layer next to the fixed layer. This double layer
共1959兲, and Wait 共1959兲, who observed that IP could monitor the presents capacitive impedance to current passage across the solid
interface phenomena between pore-fluid electrolytes and conduc- interface. It is now very clear that ionic conduction paths via the
tive materials in the ground, such as metal bodies or metallic min- pore fluid electrolyte primarily control current conduction in resis-
erals. Laboratory investigations aimed at understanding the tivity. The IP effect, however, is linked to the buildup of excess
physico-chemical mechanisms responsible for the impedance of charge that results when the normal pore conduction path is
the mineral electrolyte interface have been conducted by many blocked or the mobility of the normal ionic charge carriers is
workers, among whom are Marshall and Madden 共1959兲, Madden slowed. Electrode polarization and membrane polarization are two

Manuscript received by the Editor April 12, 2003; revised manuscript received August 8, 2005; published online May 24, 2006.
1
Massachusetts Institute of Technology, Earth Resources Laboratory, Department of Earth, Atmospheric and Planetary Sciences, 42 Carleton Street, Room
E34-412, Cambridge, Massachusetts 02142. E-mail: sogade@erl.mit.edu; yerv@erl.mit.edu; rodi@erl.mit.edu; morgan@erl.mit.edu.
2
Boston College, Department of Geology and Geophysics, Chestnut Hill, Massachusetts 02167.
© 2006 Society of Exploration Geophysicists. All rights reserved.

B75
B76 Sogade et al.

generally acceptable mechanisms that explain the observed IP ef- and Vanhala, 1994兲. In the time domain, this means at times greater
fects for natural rock materials at low frequencies 共below 1 kHz兲. than 1 ms. For membrane polarization, the reduction in ion mobil-
Electrode polarization is the result of the interface impedance ity and therefore the IP effect are most easily observed for potential
that results when conducting mineral grains block pore paths. The variations 共⬍1 Hz兲 much slower than the time of diffusion of ions
mode of current conduction changes from ionic to metallic 共elec- between adjacent membrane zones. To avoid induction coupling
tronic兲 at the mineral electrolyte interface, and current is carried problems and based on the results of field-tested S/N ratios, we
across the interface either capacitively via the double layer or by have found it preferable to choose times greater than 300 ms, with
charge-transfer reactions. If the rate-limiting step of these surface the implication that the significant contributions to the observed IP
reactions is diffusion, then the reaction impedance is usually de-
effect derives only from the very-low-frequency 共⬍4 Hz兲 part of
scribed by the Warburg type, with an inverse square root depen-
the spectrum.
dence on frequency. If the reaction is kinetics limited, the reaction
Field data arising from IP targets in a heterogeneous earth repre-
impedance may be different from a Warburg type 共Olhoeft, 1986兲.
sent a nonlinear coupling of responses from pure or intrinsic IP,
Membrane polarization is caused by the interface impedance
that results when, in certain regions of the normal conducting pore electrical resistivity, and the geometry of the host medium. The re-
path, the diffuse layer is thick enough to block the path, thereby se- covery of the pure IP properties of targets from field measurements
lectively passing ions of certain size and polarity, reducing the mo- requires the solution of forward and inverse problems. The forward
bility of charge carriers and causing charge buildup. It rests on the and inverse problems should therefore accommodate the complex
assumption that these regions of sufficiently thick diffuse layers of interactions between IP properties, electrical resistivity, and geom-
ions, called membrane zones, exist, and that the pore width is not etry. Along these lines, a robust 2D time-domain IP inversion code
too wide. One mechanism that can produce membrane zones is has been developed based on a 3D dc resistivity forward and in-
clay/clay-sized materials with high IEC that partially blocks the verse algorithm developed at the Earth Resources Laboratory
ionic solution path. Therefore, membrane polarization is of impor- 共ERL兲 by Zhang et al. 共1995兲 and Shi 共1998兲.
tance in rocks with a few percent clays distributed throughout the The measured IP response depends on the distribution, concen-
rock matrix and dispersed among other larger mineral grains — for tration, and types of contaminants present in the subsurface. Pre-
example, as a thin surface coating 共Vacquier et al., 1957兲. It is very liminary results are shown here, based on the analysis of 2D IP
common for natural rock minerals, including clay, to be coated by data using inversion that reveals very close correlation between the
organic matter, which often also has high IEC 共Jenne, 1977兲. Mem- locations of polarization and contaminant concentration 共benzene
brane polarization is a more likely mechanism for the IP effect ob- and ethylene dibromide 关EDB兴兲 anomalies. Therefore, the current
served in contaminant mapping.
case study demonstrates an ability of the IP method to map a con-
Therefore, the surface properties of solid mineral grains and ad-
taminant plume 共benzene and EDB兲 in the subsurface. Gross spec-
sorption are very important to the IP effect. When contaminants en-
tral chargeability difference, determined by subtracting inversion
ter the rock matrix, surface reactions between the rock/soil grain
surface and surface reactive contaminant solute affect the grain- results of early and late time IP decay data, is indicative of spectral
surface electrical characteristics of the rock/soils. Such surface re- information and may provide diagnostic information to discrimi-
actions or sorptive processes include adsorption 共solute sticks to a nate polarization sources when fully developed.
solid surface by physical or chemical means兲, ion exchange 共attrac- A limitation of conventional IP is that disseminated clays and
tion between ions in the pore fluid and clay/clay-sized minerals in metal bodies or metallic minerals have significant IP signatures
the soil兲, chemisorption 共solute is incorporated in the structure of a and, in the absence of other information, cannot be distinguished
sediment/rock兲, and absorption 共solute diffuses into the sediment/ from contaminants. Spectral IP, when fully developed, may dis-
rock/organic matter兲. These processes tend to remove the contami- criminate the IP signatures of contaminants and geologic and/or
nants from the free phase in the soil and are affected by the satura- cultural noise sources such as clays and metal bodies. However,
tion of the soil. Groundwater adsorption is a very important and nonlinear complex resistivity, in which the relation between mea-
common sorptive process 共Drever, 1997兲. For example, the dis- sured potentials and applied current is nonlinear, have been re-
solved components of aromatic and chlorinated hydrocarbons, ported to discriminate IP effects attributed to redox reactions from
which are usually hydrophobic, are strongly adsorbed by the solid those of ion exchange and/or clay-organic reactions 共Olhoeft,
phases in the subsurface. The important substrate here is the or- 1985兲. A more serious limitation arises from electromagnetic in-
ganic coating on rock/soil minerals. The consequent adsorption, ductive coupling 共EMC兲, which is not accounted for by present dc-
usually hydrophobic driven, has been recognized as probably the based IP models but is treated as data distortion to be removed or
most important mechanism for groundwater contaminant transport
corrected 共Wynn and Zonge, 1975; Pelton et al., 1978; Wait and
of aromatic and chlorinated hydrocarbons 共Drever, 1997兲. The link
Gruszka, 1986兲. Routh and Oldenburg 共2001兲 report recent efforts
between the IP response and adsorbed species has long been
aimed at the removal of inductive coupling from frequency-do-
known 共Morgan, 1981; Olhoeft, 1986兲; therefore, because the ad-
sorbed species include the contaminant solutes, the induced polar- main IP data.
ization method could, in principle, detect and map the contaminant Until recently, IP was used almost exclusively in the mineral ex-
plume. ploration field. However, over the past two decades IP measure-
The IP literature suggests that the most important contributions ments have improved, and new applications of IP have emerged in
of surface impedance to the IP effect exist at frequencies below the environmental field. Recent reports of applications of IP to en-
1 kHz 共Madden and Cantwell, 1967; Klein et al., 1984; Börner et vironmental problems include, but are not limited to, Olhoeft
al., 1993; Vanhala, 1997a, b兲, although there is evidence that useful 共1986兲, Barker 共1990兲, Vanhala 共1997a, b兲, Shi 共1998兲, and Morgan
information exists beyond 1 kHz 共Börner et al., 1993; Soininen et al. 共1999兲.
Induced-polarization mapping of contaminant plumes B77

REVIEW OF PERTINENT THEORY oped 2D time-domain codes that perform inversions for charge-
abilities based on the scheme outlined in steps 1–4 and represented
The following sections discuss pertinent theories, including in- by the operator ⌽−1.
trinsic IP, macroscopic IP, and inversion.
DATA ACQUISITION AND ANALYSIS
Response of a homogeneous media — intrinsic IP
Fuel spill 12 (FS-12)
The time- or frequency-domain response of a homogeneous po-
larizable material to an applied electric field is well known. Various The IP data used in this study were acquired at the MMR FS-12
relaxation models fit measured spectral complex resistivity data plume located in western Cape Cod, Massachusetts, about 60
␳a共 ␻兲 = 1/␴a共 ␻兲 over spatially uniform materials, where miles south of Boston. The MMR has been in existence since 1912
␳a共 ␻兲/␴a共 ␻兲 are the apparent complex resistivity/conductivity re- and was a major facility for the United States Air Force from
sponse functions. Among the most popular are those of Seigel 1948–1973. Since 1973, the primary users of the MMR have been
共1959兲, Madden and Cantwell 共1967兲, Zonge 共1972兲, and Pelton et the Massachusetts National Guard and the United States Coast
al. 共1978兲. More recent models include those of Lesmes and Mor- Guard. In 1986, the National Guard Bureau’s Installation Restora-
gan 共2001兲, but the most widely accepted is the Cole-Cole model tion Program was initiated to investigate suspected contaminant
introduced to the IP literature by Pelton et al. 共1978兲. plumes at the MMR. Following the investigation, it became listed
as a Superfund site on the National Priority List on November 21,
1989. The FS-12 is one of ten major plumes located on the MMR
Response of heterogeneous media — macroscopic IP
and was caused by a leak from an underground fuel pipeline carry-
We limit discussion to time-domain IP response because the ing JP-4 jet fuel and aviation gasoline 共AVGAS兲. The leak is esti-
field data are collected in this mode. Figure 1 shows a schematic of mated to be 265 m3 and occurred in 1972 and 1973. Seventeen
the potentials for time-domain IP analysis. Immediately after the years later, in 1990, the Sandwich water district drilled some mu-
current is turned on, ⌽␴ is the instantaneous voltage developed nicipal water supply wells in the area. The groundwater analysis
across the potential electrodes, and ⌽␩ is the primary voltage mea- indicated high concentrations of volatile organic compounds
sured just before the current is turned off. After the current is 共VOC兲 — in particular, benzene derived from JP-4 and EDB de-
turned off, the voltage drops to a secondary level ⌽s, then decays rived from AVGAS. The FS-12 plume extends downgradient for
with time t as ␸s共t兲 关i.e., ⌽s = ␸s共0兲兴. We adopt here Siegel’s about 1524 m outside a portion of the eastern boundary of MMR,
共1959兲 model of the time-domain IP response, which assumes that reaching Camp Good News, a privately owned summer camp in
the primary effect of polarization is to lower the effective conduc- the village of Forestdale within the township of Sandwich, Massa-
tivity of a homogeneous medium, i.e., conductivity ␴ is trans- chusetts. Figure 2a shows a plan view of the plume site, indicating
formed to an effective conductivity 共 ␴ − ␴␩兲, where ␩ = ⌽s /⌽␩, the existing wells, the IP survey line, and the extent of EDB con-
called the chargeability of the medium, solely characterizes its po- tamination as of November 1995 共AFCEE, 1997兲. Similarly, Fig-
larization properties. Siegel 共1959兲 extends his time-domain re- ure 2b shows the extent of benzene contamination as of November
sponse over a homogeneous material to include heterogeneous me- 1995 共AFCEE, 1997兲.
dia, which can be described as a juxtaposition of n different
materials 共i.e., those generated by spatial discretizations to achieve
numerical solutions兲 of different conductivities ␴i,i = 1, . . . ,n, Geologic and hydrological setting
and chargeabilities ␩i, i = 1, . . . ,n. In the absence of polarization, The FS-12 plume lies entirely in the Mashpee Pitted Plain. Geo-
the primary voltage is expressible as ⌽␴ = ⌽共 ␴兲, and in the pres- logic data, mostly derived from borings, indicate that the substrata
ence of polarization it is ⌽␩ = ⌽共 ␴␩兲, where ␴ = ␴i,i = 1, . . . ,n is
the intrinsic conductivity distribution and ␴␩ = ␴␩i = ␴i共1 − ␩i兲,
i = 1, . . . ,n is the transformed conductivity distribution after po-
larization. Note that ⌽ is the nonlinear dc forward-mapping opera-
tor that converts models ␴ to data ⌽␴, and the formula ⌽s /⌽␩ now
gives the apparent chargeability ␩a.

Inversion
Similar to the algorithm suggested by Oldenburg and Li 共1994兲
共steps 2–4 below兲, a scheme is developed to calculate chargeability
␩ for time-domain IP data as follows:

1兲 Using an exponential function fit, calculate ⌽s from field


measurements 共see Appendix A兲 and form ⌽␴ = ⌽␩ − ⌽s.
2兲 Calculate ␴ = ⌽−1共 ⌽␴兲 using the nonlinear inverse mapping
operator ⌽−1 that converts data ⌽␴ to models ␴. Figure 1. Schematic of the IP decay curve showing ⌽␴ 共the instan-
taneous potential at current on兲, ⌽␩ 共the potential affected by
3兲 Calculate ␴␩ = ␴共1 − ␩兲 = ⌽−1共 ⌽␩兲 from the field data ⌽␩. current-induced polarization兲, and ⌽s 共the polarization potential at
4兲 Calculate ␩ = 共 ␴ − ␴␩兲/␴. the start of the IP decay兲. Also, t0 denotes the time delay between
stoppage of current injection and sampling of the IP decay, and in
We have adapted the 3D dc resistivity forward and inverse codes conjunction with t1, t2, t3, and t4 defines the four time windows
developed at ERL 共Zhang et al., 1995; Shi, 1998兲 and have devel- used to determine chargeability.
B78 Sogade et al.

consist of outwash sands and gravels with discontinuous lenses of zone of 61 m with an average depth to the water table of 25 m. The
fine sand, silt, and clay down to an average depth of 67 m 共Op area has an estimated porosity of 30%. The horizontal groundwater
Tech, 1996兲. The FS-12 plume site is underlain by the Cape Cod flow velocity is 102 m/year for the plume to the south-southeast.
aquifer, which has an estimated total thickness of the saturated The survey line was chosen to be in the proximity of and parallel to
a line of seven monitoring wells 共GMW-12/GMW-16, GMW-18/
GMW-30, GMW-32/GMW-44, and GMW-33兲 to facilitate com-
parison with well data 共see Figure 2a and b兲. Figure 3a and b, re-
show the EDB and benzene isoconcentration maps overlapping the
vertical geologic cross section below the survey line.

FS-12 plume characteristics


Under the direction of the Hazardous Waste Remedial Actions
Program 共HAZWRAP兲 of the United States Department of Energy
共DOE兲, Advanced Sciences Inc. 共ASI兲 implemented a remedial in-
vestigation work plan in 1993 共ASI, 1995兲. The investigation result
concluded that the highest concentration of contaminants detected
in the FS-12 plume as of April 1993 were benzene 共1600 ␮g/L兲
and EDB 共597 ␮g/L兲. However, the contaminated groundwater
also showed the presence of toluene, xylenes, naphthalene, ethyl
benzene, 2-methylnaphtalene, and others 共summarized in Table 1兲,
all at much lower concentrations than either benzene or EDB. The
FS-12 plume is migrating through glacial outwash sands and grav-
els, with its top descending from the water table to as much as
40 m below and extending 1524 m from the source. Plume con-
taminants have been detected in the upper 3–7 m of the glacial
lacustrine sediments underlying the outwash sands and gravels.
In November 1995, data supplemental to those of ASI 共Op Tech,
1996兲 found that the highest concentration of benzene detected in
the FS-12 plume has changed from 1600 to 2500 ␮g/L 共MCL of
5 ␮g/L and from 597 to 1000 ␮g/L for EDB 共MCL of 0.2 ␮g/L兲.
For example, benzene concentration increased in GMW-30/18
from 180 ␮g/L 共April 1993兲 to 2400 ␮g/L 共November 1995兲, and
EDB increased in the same well from an estimated 211 ␮g/L
共April 1993兲 to 1000 ␮g/L 共November 1995兲. Additionally, the
data indicate that both the benzene and EDB plumes migrated
south-southeast.

Acquisition
The data presented in this study were collected in Camp Good
News in the late fall of 1996. The high concentration of contami-
nants and the type of pollution 共organic substances兲 at FS-12 led to
the choice of IP for the data acquisition. Additionally, continuous
observations of the subsurface at the site by the groundwater moni-
toring wells 共GMW兲 and the construction of new wells by the In-
stallation Restoration Program at MMR are thought to be very use-
ful for validation purposes. Finally, the convenient location for the
fieldwork, far from power lines, and the relatively uniform geology
made FS-12 a very suitable plume on which to focus.
The data were acquired in the time domain using an IRIS
ELREC-T transmitter-receiver system with a 250-W dc/dc con-
verter and a 12-V car battery. The IP data were taken at four time
windows during the relaxation time lasting 4 s, where t0 = 300,
t1 = 800, t2 = 1600, t3 = 2600, t4 = 3800, and tm = 2000 ms 共see
Figure 1兲. The time delay before the start of the current on sam-
pling was tm = 2000 ms, while the delay before start of recording
Figure 2. Plan view of the FS-12 plume site, indicating the existing the decay was t0 = 300 ms. Data were gathered using a dipole-
wells, geologic section line marked CC⬘ 共shown as a 2D vertical
cross section in Figure 3兲, IP survey line, as well as the 共a兲 EDB dipole array with elemental copper electrodes for current transmis-
concentration plot 共AFCEE, 1997兲 and 共b兲 benzene concentration sion and silver/silver Ccloride 共Ag/AgCl兲 porous pot electrodes
plot 共AFCEE, 1997兲. for potential. The survey line was 560 m in extent with 45 elec-
Induced-polarization mapping of contaminant plumes B79

trode locations. The separation of the potential and current elec- seen in Table 2. These results suggest that the IP contributions from
trodes was 24 m 共80 ft兲 each. The potential electrodes were moved the vadose zone are relatively unaffected by saturation, a conclu-
12 m 共40 ft兲 down the line per measurement for each current di- sion which further implies that the source of the significant IP ef-
pole, and 13 current dipole sources were used, each consisting of fects lies in the saturated region of the water table where the con-
16 potential measurements. Figure 4 shows a schematic of the taminant is.
pseudosection array used with an indication of the expected loca-
tion of the plume, based on the groundwater monitoring wells.
Analysis
Care was taken to separate all wires connecting electrodes to the
instrument to minimize crosstalk and capacitive coupling errors. In The inverse dc operation that converts measured potential data
addition, the equipment monitors contact resistance for each elec- into the resistivity distribution needed in the inversion scheme is
trode pair to ensure it was kept below based on the 2D/3D forward modeling and inversion code devel-
20 k⍀, thereby minimizing leakage cur-
rents. Because the transmitter current was
off while sampling the IP decay, inductive
coupling distortion could be avoided if
the sampling started after the rapidly de-
caying inductive coupling signals were
extinguished. Several field tests and
simple synthetic model studies suggest
that a sampling onset delay of about
300 ms is more than adequate for this
site.

Preliminary analysis of data


The apparent resistivity decreases with
pseudodepth, from 10,000–30,000 ohm-
m at the surface to 1000–2000 ohm-m at
depth. 共Note that the pseudodepth is equal
to half the spread between the current and
potential electrode pairs.兲 This associa-
tion is possible because the penetration
depth of the current is a function of the
spreading of the electrode pairs. The per-
cent rms of the measurements is on the
order of 1% for measurements repeated
several times a few minutes apart. The IP
data indicate that the chargeability in-
creases with depth by a factor of three,
from 3–4 mV/V at the surface to
14–15 mV/V at depth. The rms for shal-
low measurements is very small 共⬍1%兲
but increases with depth, where it reaches
values of up to 15%. Table 2 presents the
rms of the apparent resistivity and IP
measurements, along with the minimum
and maximum values for each quantity.
The resistivity measurements are, in gen-
eral, more stable and repeatable than IP,
especially at depth. However, resistivity
measurements are more sensitive to
ground conditions such as water content.
For example, measurements repeated a
few days after an intensive rain produced
a large variation of the resistivity com-
pared to those taken on a relatively dry
day. The IP measurements, on the other
hand, were barely affected by the wet-dry Figure 3. 共a兲 Geologic cross section around and under the survey line as well as EDB isocon-
centration maps adapted from HAZWRAP, DOE 共Op Tech, 1996兲. 共b兲 Geologic cross section
conditions, as shown by the much smaller around and under the survey line as well as benzene isoconcentration maps adapted from
rms error of the wet-dry measurements HAZWRAP, DOE 共Op Tech, 1996兲.
B80 Sogade et al.

oped at the Earth Resources Laboratory by Zhang et al. 共1995兲 and Raw data processing is necessary to prepare the data for inver-
Shi 共1998兲. The forward model is based on the transmission net- sion, and it entails fitting the decay from any contiguous combina-
work analogy applied to discretize the earth into a network system tion of the four time windows to an exponential function, whereby
comprising network nodes and impedance branches. The resulting ⌽s and ⌽␴ are recovered 共see Appendix A兲. Two resistivity inver-
system of equations is solved using the biconjugate gradient sions are performed using steps 2 and 3 of the inverse scheme to
method. The nonlinear inversion is posed as a Tikhonov regular- recover ␴ and ␴␩ = ␴共1 − ␩兲, while step 4 calculates the charge-
ized least-squares problem, and the resulting inverse steps solved ability ␩ for each numeric block. A measure of the fit between data
using an iterative nonlinear conjugate gradient algorithm. Typical
and model for each of ⌽␴, ⌽␩ and ␩a is achieved by the normalized
inversions last about 12 iterations and take 1–2 minutes on a mod-
rms, defined as
ern PC 共2.8-GHz Pentium processor兲.

Table 1. Summary of the chemical constituents of the FS-12 plume derived from ground monitoring wells GMW12/16, 18/30,
32/44, and 33, (ASI, 1995; Op Tech, 1996).

MCLs GMW-12/16 GMW-18/30 GMW-32/44 GMW-33

VOCs
Benzene 5 ␮g/L 1600 ␮g/L 180 ␮g/L — —
Toluene 1000 ␮g/L — 1J — —
EthylBenzene 700 ␮g/L — 2J — —
Xylenes 共total兲 10,000 ␮g/L 200 J 15 ␮g/L — —
Chloroform 5 ␮g/L 1J — 2J —
2-Butanone 350 ␮g/L 3J — —
TICs — 2250 ␮g/L 216 ␮g/L 38 ␮g/L —
SVOCs
Phenol 4000 ␮g/L — 2 J — —
Napthalene 20 ␮g/L 2 J/18 ␮g/L 2 J — —
Di-n-butyl-phtalate 800 ␮g/L 0.7 J/0.9 J 0.7 J — —
Bis共2-Ethylhexyl兲pthalate 6 ␮g/L — 2 J/3 J 1J —
2-methylnapthalane 10 ␮g/L 1J — — 2J
TICs — 8/178 ␮g/L 79/2 ␮g/L 10 ␮g/L —
EDB 0.2 ␮g/L 597.76 J 211.63 J/0.41 ␮g/L 24.34 J —
Metals 共unfiltered and
above background兲
Al 50–200 ␮g/L 878 ␮g/L 145/3310 ␮g/L 933/488 ␮g/L 2190 ␮g/L
Fe 300 ␮g/L 694/79.3 ␮g/L 163/3910 ␮g/L 1310/782 ␮g/L 2450 ␮g/L
Hg 2 ␮g/L — 0.26 ␮g/L/N — 0.39 J
Pb 15 ␮g/L — N/5.8 ␮g/L — —
Zn 5000 ␮g/L 28.6/25.6 ␮g/L 40.4 ␮g/L/N — —
As 10 ␮g/L — N/2.9 ␮g/L N/5.1 J —
Metals 共filtered and
above background兲
Fe 300 ␮g/L N/77.0 ␮g/L — — —
Zn 5000 ␮g/L 57.7 ␮g/L/43.6 J 23.0 ␮g/L/49.2 J —
As 10 ␮g/L — — N/5.7 ␮g/L —
Physical properties
Temperature 共°C兲 12.9/11.9 10.0/12 7.7/11.7 —
Conductivity 共 ␮S兲 55/92.2 46/45 39.9/61
Chemical properties
pH 7.7/NA 8.1/7.7 6.2/7.9
Dissolved oxygen 10.2/7.2 0.6/0.5 7.2/9.4 —
Acronyms: VOC 共volatile organic compounds兲, SVOC 共semivolatile organic compounds兲, MCL 共maximum contaminant level兲.
Legend: N 共none兲, J 共estimated value兲.
Note: MCLs used default to Massachusetts standard if different from EPA’s.
Induced-polarization mapping of contaminant plumes B81

nrms共%兲 = 100 ⫻ 冑 兺j
冉冊
ej
dj
N
2

,
very well with the IP anomalies seen in the tomogram of Figure 5b.
The fact that negligible concentration of any contaminant 共zero for
benzene and ⬍5 ␮g/L for EDB兲 was found in GMW-33 is also
borne out by the 2D IP plot of Figure 5b. However, the anomaly
depicted in the IP tomogram does not extend all the way to the left
where the data points span j = 1, ¯ , N, and e j is the data residual edge of the image as expected, based on the concentration data
or the disparity between measurement and model at data point d j. from GMW12/16 in Figure 3. To check if this shortcoming was
The value of nrms 共%兲 will approach 0% for perfect fit between caused by the deficiency of data in the vicinity of the left edge of
data and model and 100% if the prediction error at each data point the image 共border effect兲 for the pseudosection array used, an in-
equals the size of the measurement at that point. It exceeds 100% version was performed on synthetic resistivity and IP data. The
for cases in which the prediction error is greater than the measure- chosen rectangular target approximates the location, shape, and
ment at many data points. The values of nrms共%兲 for the inversions size of the plume as depicted by the groundwater monitoring wells
are 3.42% 共from a start value of 406%兲 for ␴, 3.36% 共from a start along profile C–C⬘. The results 共not shown here兲 clearly confirm
value of 393%兲 for ␴␩, and 9.55% 共from a start value in the range that the edge effect seen for the IP result is from distortion in the
of 100%兲 for ␩. The inversion results are displayed as a 2D plot of inverted image caused by a lack of data in the edges for the pseu-
the IP properties 共measured by the chargeability兲 of all blocks, dosection array. We have since improved on the pseudosection ac-
which we call a 2D IP plot. quisition array geometry to reduce this edge distortion.
There is another reason why there might be slight discrepancies
in the inverted IP image and the contaminant concentration data.
RESULTS
During the one year between when the groundwater samples were
Figure 5a is a 2D plot of the resistivity taken in November 1995 共Op Tech, 1996兲 and our measurements in
under the survey line, showing two rela-
tively conductive zones 共a 300-ohm-m
zone located 20–100 m laterally, 40–80
m vertically, and another 300-ohm-m
zone 130–230 m laterally, 60–100 m ver-
tically兲, which could be contaminated
zones or clay-rich soils. These 300-ohm-
m zones are likely the result of the pres-
ence of clay or sandy clay, as attested by
the geology map which expects a clay
layer or zone below 60-m depth 共see Fig-
ure 3a and b兲. The high background resis-
tivity is typical of the Cape Cod area of
Massachusetts. Figure 5b is a 2D charge-
ability plot derived from the time-domain
IP data, showing three distinct anomalous
zones 共130 mV/V located 20–80 m later- Figure 4. Schematic of the pseudosection array used for the survey. In terms of conventional
ally, 15–60 m vertically; 50 mV/V lo- dipole-dipole configuration, the potential/current electrode separation a = 24.38 m 共80 ft兲
and n = 0.5, . . . ,8 in steps of 0.5. The 400-m section shown in the inversion results of Fig-
cated 200–240 m laterally, 40–80 m ver- ures 4 and 7 is indicated by the arrows. The insert depicts the 100-␮g/L EDB isoconcentra-
tically; and 80 mV/V located 300–340 m tion, adapted from the GMW data.
laterally, 30–60 m vertically兲 and some
other smaller features.
Figure 6 is a summary of groundwater Table 2. Error Analysis for the resistivity and induced-polarization field
measurements.
samples from the three monitoring wells
displaying concentrations of benzene and
EDB. It is shown as the egg-shaped Field data Potential 共mV兲 Apparent resistivity 共 ⍀ . m兲 Chargeability 共mV/V兲
plume adapted from the concentration Range of min max min max min max
data provided by HAZWRAP DOE 共Op Measurement 2.2 3496.0 1000.0 30000.0 3.0 15.0
Tech, 1996兲. The chargeable regions of
the ground from the IP tomogram corre- Single day measurements Resistivity 共ohm-m兲 Chargeability 共mV/V兲
lates quite well with the matching con- Mean 2831.4 10.3
centration data from the groundwater rms 26.62 1.49
monitoring wells, both in magnitude and
% change 0.94 14.3
spatial extent. Also, the depth and extent
of the benzene and EDB concentrations Wet/dry 共different days兲 Resistivity 共ohm-m兲 Chargeability 共mV/V兲
from the two wells at GMW-30 共extend- Mean 2419.9 11.2
ing from 20–65 m deep兲 and GMW-32 rms 823.15 1.92
共extending from 30–65 m deep兲 correlate % change 34.02 17.44
B82 Sogade et al.

November/December 1996, the plume may have moved downgra- benzene-contaminated shaly sandstone is stronger 共28 mrad兲 than
dient with the groundwater flow and therefore more toward the east those of the clean sample 共16 mrad兲. The investigated band of fre-
end of the profile. In addition, the bottom of the plume may have quencies spans four decades, starting from 0.001 Hz. Results of
descended deeper below the water table, following the expected spectral IP properties of toluene-contaminated clay 共Sadowski,
plume migration pattern detailed in Op Tech 共1996兲. 1998兲 and glacial till 共Vanhala et al., 1992兲 samples have been re-
Laboratory measurements were not performed on any of the ported, with significant 共50 mrad兲 IP phase response at higher fre-
samples at the investigation site to constrain the results. However, quencies 共above 10 Hz兲. Similarly, the spectral IP response of phe-
laboratory investigations of the spectral IP properties of rocks/soils nol has been reported 共Jones, 1997兲. Most of the results mentioned
contaminated with organic compounds have been carried out by a above involve significantly higher concentrations than the ppb and
number of authors 共Sadowski, 1988; Vanhala et al., 1992; Börner ppm ranges that occur in this investigation. However, McKinley
et al., 1993兲. The published laboratory measurements were 共2003兲 shows evidence that measurement on in-situ tetrachloro-
ethene 共PCE兲/trichloroethene 共TCE兲 contaminated clayey sand
searched for the contaminants of FS-12 共Table 1兲 to determine their
samples at concentrations of a few parts per million resulted in sig-
possible IP response. With data adapted from the results of Börner
nificantly high polarization response 共160 mrad兲. This observation
et al. 共1993兲, it can be shown that the peak IP 共phase兲 response of
is corroborated by our experience at many other investigation sites
and underlies the important point that polarization anomalies of
field-contaminated samples results not only from concentration of
the contaminants but also from the resulting complex biophysical
and geochemical interactions at the rock/soil grain interface occa-
sioned by the long exposure times.
The literature search found no laboratory results for EDB. In the
absence of any other information, it is very difficult to ascertain
which plume constituent共s兲 contributed to the IP response. How-
ever, it is reasonable to argue that benzene 共with a proved polariza-
tion response兲 and EDB, which have concentrations much stronger
than any of the other plume constituents 共see Table 1兲, are the
likely source of the polarization anomalies. To rule out the possi-
bility that geologic variations could have caused the polarization
anomalies, it is argued that, because the same materials are seen in
the three borehole locations along the survey line as revealed in the
geologic map of Figure 3, the polarization response should have
been the same around the three boreholes, which is not the case.
The GMW wells have metallic content and are a likely cause of
concern; however, most of the wells were at least 5 m away from
the survey line, and no evidence from the IP inversion suggest that
the wells affected the inversion results. In a different investigation
Figure 5. 共a兲 2D resistivity section of the plume, showing 300- carried out at another site, IP inversion clearly imaged the metal
ohm-m conductive zones located below 50 m depth. The 300- well, which is not seen in the current investigation.
ohm-m zone is likely the result of the presence of clay or sandy
clay, as attested to by the geology map which expects a clay layer It is known that the frequency content of the polarization decay
or zone at about 60 m depth 共see Figure 3兲. Shallower conductive response is roughly inversely proportional to the decay time 共Sum-
zones of 1000–3000 ohm-m are likely related to groundwater. 共b兲 ner, 1976兲. Therefore, it is expected that a polarization distribution
A 2D chargeability section cutting through both the plume as well derived by inversion using the time windows located near the onset
as a contaminant-free zone. The plot shows a distribution of
chargeability anomalies that correlate closely with the contaminant of the decay will be rich in the high-frequency content of the polar-
concentration anomalies from GMW data as shown in Figure 6.

Figure 7. A 2D section showing chargeability difference 共inverted


chargeability using the first two time windows; inverted charge-
ability using the last two time windows兲. The section cuts through
the plume as well as a contaminant-free zone. The plot reveals evi-
dence of spectra where the difference values are high, which inci-
Figure 6. Extrapolated plume concentration data for both benzene dentally correlates with the polarizability and contaminant concen-
and EDB based on the HAZWRAP DOE ground-monitoring wells. tration anomalies.
Induced-polarization mapping of contaminant plumes B83

ization response. Similarly, the distribution derived from the time pable of accommodating data at short times 共time domain兲 and
windows near the end of the decay will be rich in the low-fre- high frequencies 共frequency domain兲.
quency content of the polarization response. Based on this idea,
Figure 7 shows a gross 共spectral兲 chargeability difference derived ACKNOWLEDGMENTS
by subtracting a low-frequency 共long time, comprising the last
three of the four time windows兲 2D chargeability distribution from We thank EPA Northeast Hazardous Substance Research Center
a high-frequency 共short time, comprising the first three of the four for the financial assistance provided to facilitate this research work
time windows兲 2D chargeability distribution. There is evidence of through grant NJ1T 9-92601. We also acknowledge Dennis
spectra, indicated by the zones of positive and negative differences McLaughlin of the Department of Civil and Environmental Engi-
in Figure 7, which correlates with the polarized regions. This extra neering at the Massachusetts Institute of Technology for his sup-
step to deduce spectral information from time-domain data is inno- portive role at the initial stage of this project.
vative.
APPENDIX A

CONCLUSION ESTIMATION OF THE


In general, the images from the 2D IP section agree quite well SECONDARY POTENTIAL ⌽S
with the distribution of the plume from the well data. This is re- Any function chosen to fit ⌽s共t兲 in order to extrapolate for
markable, considering that the plume might have migrated further ⌽s共t = 0兲 is at best an approximation. We have chosen to fit the de-
in the space of the one year separating the IP measurements and the cay ⌽s共t兲 to an exponential function, which is then used to extrapo-
recording of the well data. In the chosen cross section, the presence late for ⌽s共t = 0兲. The basis for this approach is that the portions of
or absence of contaminants correlates well with the chargeable or the decay for times t ⱖ 160 ms have in our experience been well
nonchargeable regions of the ground, obviating the need for a con- fitted to exponential functions. Such a treatment biases the deriv-
trol measurement in a contaminant-free zone. The monitoring well able information and limits it to certain types of IP targets. It also
samples are accurate but only provide information at the well loca- imposes a restricting model on the IP response, acting like a filter.
tions, and the plumes generated from well data are usually extrapo- Consequently, extending the fit backward to extrapolate for ⌽s共t
lated based on too few wells. IP may overcome these shortcomings = 0兲 will introduce errors that will not, however, mar the useful in-
and provide better resolution covering an entire investigation site. formation about relative magnitudes needed for mapping purposes.
The chief benefit of an IP technology is cost reductions from the Rather than measure the decay at discrete instants tk, resulting in
optimization of well sampling achieved by integrating IP into site instantaneous apparent chargeability, most instruments measure
characterization and monitoring programs. IP has the potential to time averaged apparent chargeability in a time window ⌬tk. For the
reduce greatly both the number and frequency of sampling and to IRIS ELREC-T instrument used to collect the field data discussed
better locate wells for monitoring and characterization. Traditional here, the formula used to calculate the time averaged apparent
site characterization and monitoring techniques use a phased ap- chargeability M for each window is
proach of data collection on a statistical basis or grid to develop

冫冕
and modify conceptual models. Each phase focuses on locating the ti+1 ti+1
contaminant plume and follows a preapproved work plan. Random
sampling to obtain data that are adequate, in terms of their spatial M i = 103 · 冕 ⌽s共t兲dt ⌽␩dt in mV/V.
and temporal frequencies, to define variability in contaminant con- ti ti
centration at polluted sites is too expensive and labor intensive. For
共A-1兲
example, DOE 共expedited site characterization, DOE/EM-0420,
1998兲, estimated that a planned traditional site characterization
The IRIS instrument returns, at most, four values of chargeabilities
technique involving the use of drilling and monitoring wells ap-
M, calculated via the four time windows ⌬ti+1 = 共ti+1 − ti兲, i
plied to DOE’s Pantex site will involve 54 penetrations at a cost of
= 0,1,2,3.
$11 million. IP characterization of the same site will involve an es-
Using the exponential function ⌽s共t兲 = ⌽se共−pt兲, M i can be writ-
timated six initial penetrations for calibration and up to an addi-
ten as
tional six to verify any substantial changes in the IP anomalies dur-


ing monitoring at a total cost of $3.0 million, or more than 60% ti+1
savings.
However, results need to be verified for accuracy and reliability
in order to be used in decision making. In the FS-12 field example,
M i = 103 · 冕 ⌽se共−pt兲dt ⌽␩ · 共ti+1 − ti兲. 共A-2兲
ti
IP chargeability and gross IP chargeability difference appear to
show more subsurface complexity than the well sampling data, but Using two, three, or four M values of M 1, M 2, M 3, M 4, equation
currently we do not know their accuracy. A determination of the ac- A-2 can be inverted to find p; thus, ⌽s can be deduced. The result-
curacy will require a dense distribution of validation wells at a ing equation to solve for p is given by
known contaminated site, with IP performed before and after well
placement. M 1⌬ t 1
Also, to overcome the barrier imposed by electromagnetic cou- B= 4 ,
pling problems, there is a need to further develop forward and in- ⌺ M i⌬ t i
verse 2D/3D time- and frequency-domain codes/algorithms ca- i=1
B84 Sogade et al.

共1 − B兲 + Be−p共t4−t0兲 − e−p共t1−t0兲 = 0. 共A-3兲 Marshall, D. J., and T. R. Madden, 1959, Induced polarization — A study
of its causes, Geophysics, 24, 790–816.
McKinley, K. S., 2003, Nonlinear complex resistivity to investigate clay-
A root-bracketing routine is used to isolate possible roots; then a organic reactions at the Savannah River site, South Carolina: M.S. thesis,
root-solving algorithm is applied to find the roots p in equation Colorado School of Mines.
Morgan, F. D., 1981, Electrodics of sulfide minerals: Implications for in-
A-3. In all cases investigated thus far, only one root is returned; duced polarization: Ph.D. thesis, Massachusetts Institute of Technology.
thus, the problem of choosing the appropriate root from multiple Morgan, F. D., F. Scira-Scappuzzo, W. Shi, W. Rodi, J. Sogade, Y. Vicha-
choices has not been confronted. bian, and D. Lesmes, 1999, Induced polarization imaging of a jet fuel
plume: Symposium on the Application of Geophysics to Engineering and
Environmental Problems 共SAGEEP兲, Environmental and Engineering
Geophysical Society, Proceedings, 541–548.
REFERENCES Oldenburg, D. W., and Y. Li, 1994, Inversion of induced polarization data:
Geophysics, 59, 1327–1341.
Advanced Sciences, Inc., 共ASI兲, 1995, Final remedial investigation report Olhoeft, G. R., 1985, Low-frequency electrical properties: Geophysics, 50,
of FS-I 2 study area, Massachusetts Military Reservation: Advanced Sci- 2492–2503.
ences, Inc. Olhoeft, G. R., 1986, Direct detection of hydrocarbon and organic chemi-
Air Force Center for Environmental Excellence 共AFCEE兲, 1997, Plume re- cals with ground-penetrating radar and complex resistivity: Petroleum,
sponse program final S-12 containment system design report, vol. 1. hydrocarbons and organic chemicals in ground water — Prevention, De-
Allaud, L. A., and M. H. Martin, 1977, Schlumberger, The history of a tection, and Restoration, NWWA/API, Proceedings, 284–305.
technique: John Wiley & Sons, Inc. Operational Technologies Corporation 共Op Tech兲, 1996, Installation resto-
Angoran, Y. E., 1975, Induced polarization of metallic minerals: A study of ration program, plume containment design data gap field work: Techni-
its chemical basis: Ph.D. thesis, Massachusetts Institute of Technology. cal memorandum vol. I. 共draft兲, Massachusetts Military Reservation,
Angoran, Y., and T. R. Madden, 1977, Induced polarization: A preliminary Cape Cod, Massachusetts. Prepared for HQ ANGKEVR, Andrews Air
study of its chemical basis: Geophysics, 42, 788–803. Force Base.
Barker, R. D., 1990, Investigation of groundwater salinity by geophysical Pelton, W. H., S. H. Ward, P. G. Hallof, W. R. Sill, and P. H. Nelson, 1978,
methods, in S. H. Ward, ed., Geotechnical and environmental geophys- Mineral discrimination and removal of inductive coupling with multifre-
ics, vol. II: SEG Investigations in Geophysics 5, 201–211. quency IP: Geophysics, 43, 588–609.
Bleil, D. F., 1953, Induced polarization: A method of geophysical prospect- Routh, P. S., and D. W. Oldenburg, 2001, Electromagnetic coupling in
ing: Geophysics, 18, 636–661. frequency-domain induced polarization data: A method for removal:
Börner, F., M. Gruhne, and J. Schön, 1993, Contamination indications de- Geophysical Journal International, 145, 59–76.
rived from electrical properties in the low frequency range: Geophysical Sadowski, R. M., 1988, Clay-organic reactions: M.S. thesis, Colorado
Prospecting, 41, 83–98. School of Mines.
Drever, J. I., 1997, The geochemistry of natural waters: Surface and Seigel, H. O., 1959, Mathematical formulation and type curves for induced
groundwater environments, 3rd ed.: Prentice-Hall, Inc. polarization: Geophysics, 24, 547–565.
DOE 共U. S. Department of Energy兲, 1998, Expedited site characterization: Shi, W., 1998, Advanced modeling and inversion techniques for three-
Characterization, monitoring, and sensors technology crosscut program dimensional geoelectrical surveys: Ph.D. thesis, Massachusetts Institute
and subsurface contaminants focus area: DOE/EM-0420. of Technology.
Hallof, P. G., 1957, On the interpretation of resistivity and induced polar- Soininen, H., and H. Vanhala, 1994, Spectral induced polarization studies
ization results: Ph.D. thesis, Massachusetts Institute of Technology. of a Ni–Cu deposit: 56th Meeting, EAEG, Extended Abstracts.
Jenne, E. A., 1977, Trace element sorption by sediments and soils — Sites Sumner, J. S., 1976, Principles of induced polarization for geophysical ex-
and processes, in W. Chappel and K. Petersen, eds., Symposium on mo- ploration: Elsevier Science Publishing Company, Inc.
lybdenum in the environment, vol. 2: Marcel Deckker, 425–553. Vaquier, V., C. R. Holmes, P. R. Kintzinger, and M. Lavergne, 1957, Pros-
Jones, D. P., 1997, Investigation of clay-organic reactions using complex pecting for ground water by induced polarization: Geophysics, 22, 660–
resistivity: M.S. thesis, Colorado School of Mines. 687.
Klein, J. D., T. Biegler, and M. D. Horne, 1984, Mineral interfacial pro- Vanhala, H., 1997a, Mapping oil-contaminated sand and till with the spec-
cesses in the method of induced polarization: Geophysics, 49, 1105– tral induced polarization 共SIP兲 method: Geophysical Prospecting, 45,
1114.
Lesmes, D. P., and K. F. Frye, 2001, The influence of pore fluid chemistry 303–326.
on the complex conductivity and induced polarization responses of Be- ——–, 1997b, Laboratory and field studies of environmental and explora-
rea sandstone: Journal of Geophysical Research, 106, 4079–4090. tion applications of the spectral induced polarization 共SIP兲 method:
Lesmes, D. P., and F. D. Morgan, 2001, Dielectric spectroscopy of sedi- Ph.D. thesis, Helsinki University of Technology.
mentary rocks: Journal of Geophysical Research, 106, 13392–13346. Vanhala, H., H. Soininen, and I. Kukkonen, 1992, Detecting organic chemi-
Madden, T. R., 1961, Electrode polarization and its influence on the electri- cal contaminants by spectral-induced polarization method in glacial till
cal properties of mineralized rocks: Ph.D. thesis, Massachusetts Institute environment: Geophysics, 57, 1014–1017.
of Technology. Ward, S. H., and D. C. Fraser, 1967, Conduction of electricity in rocks, in
Madden, T. R., and T. Cantwell, 1967, Induced polarization, a review, in J. S. H. Ward, ed., Mining geophysics: SEG, 197–223.
S. Sumner, ed., Mining geophysics, vol. II: SEG, 373–400. Wait, J. R., 1959, Overvoltage research and its geophysical applications:
Madden, T. R., and D. J. Marshall, 1958, A laboratory investigation of in- Pergamon Press, Inc.
duced polarization: U. S. Atomic Energy report RME-3156. Wait, J. R., and T. P. Gruszka, 1986, On electromagnetic coupling removal
——–, 1959a, Electrode and membrane polarization: U. S. Atomic Energy from induced polarization surveys: Geoexploration, 24, 21–27.
report RME-3157. Wynn, J. C., and K. L. Zonge, 1975, EM coupling, its intrinsic value, its re-
——–, 1959b, Induced polarization — A study of its causes and magnitudes moval and the cultural coupling problem: Geophysics, 30, 831–840.
in geologic materials — Final report: U. S. Atomic Energy Report RME- Zhang, J., R. L. Mackie, and T. R. Madden, 1995, 3-D resistivity forward
3160. modeling and inversion using conjugate gradients, Geophysics, 60,
Madden, T. R., and A. S. Neves, 1957, Background effects in the induced 1313–1325.
polarization method of geophysical exploration: Annual progress report Zonge, K. L., 1972, Electrical properties of rocks as applied to geophysical
for 1956–57, U. S. Atomic Energy Commission. prospecting: Ph.D. thesis, University of Arizona.

Das könnte Ihnen auch gefallen