FIRST BREAK VOL 10, NO 2, FEBRUARY 1992/53

A simple algorithm for electrical

imaging of the subsurface
R. Barker '

Summary changes in resistivity and the effects of these can be car-

A simple algorithm for the inversion of electrical tomo- ried down into the back projection to cause erroneous
graphy data in the form of pseudosections has been de- results. The technique discussed by Smith (1986) and
veloped. This uses the measured apparent resistivity Lowry and Shive (1990) for locating subsurface cavities
data as the starting model for a two-dimensional finite can suffer from this problem. Other techniques which in-
difference iterative imaging scheme. Differences be- volve interactive optimization of simple starting models
tween the computed and measured pseudosections are (Tripp et al. 1984) suffer currently from the problem of
used to apply corrections to the model so that over about requiring large amounts of computer time.
ten iterations the model provides an accept abIe inver- Many of the disadvantages of published imaging
sion of the observed data. techniques using surface arrays can be overcome using
Example surveys across a landfill site, an igneous an extension of a simple and novel technique'published
intrusion covered by sediment and a variety of models by Zohdy (1989) for the automatic inversion of resistiv-
show th at useful and valid electrical images of the ity sounding curves. This paper briefly reviews Zohdy's
subsurface can be produced. technique and shows, with examples, how it can be
applied to two-dimensional (2D) invers ion of Wenner
Introduction pseudosections.
There is considerable interest, particularly in archaeol-
ogy and mineral exploration, in imaging the distribution The Zobdy technique
of electrical properties in the subsurface. In the medical Zohdy's (1989) technique is essentially one of least
field electrical imaging techniques have also been de- squares optimization in which a starting model is succes-
veloped for investigation of the human torso. In all these sively adjusted until the difference between the ob-
applications, most success has been obtained with the served and model pseudosections is reduced to a
use of electrodes which surround or partially surround minimum. The novel feature of the technique is the
the area to be imaged (Barber and Seagar 1987). Barber starting model and how it is derived. The individu al
and Brown (1984) reviewed the use in medicine of arrays stages of the technique are illustrated in Fig. 1.
of electrodes surrounding the body and simple back Zohdystarts by assuming that th ere are as many sub-
projection techniques to produce approximate cross- surface layers as points on the sounding curve. In the
sectional images of the body. They refer to these initial model the resistivity ofthe first (shallowest) layer
techniques as 'applied potential tomography'. In geol- is taken as the apparent resistivity of the first point on the
ogy, tomographic surveys normally employ arrays of curve, the second layer takes the apparent resistivity of
electrodes in boreholes for the data collection . Although the second point, and so on along the curve. The mean
back projection techniques have been applied, methods dep th of each layer is taken as the electrode spacing at
which optimize an initial starting model have also been which the resistivity was measured multiplied by a con-
investigated (Dailey and Yorkey 1988; La Brecque and stant. The value of this constant is that which reduces the
Ward 1988; Beasley and Ward 1988;Shima and Saito difference between the observed and model apparent re-
1988). sistivity curves to a minimum. Zohdy determines it by
Deployment of electrodes in boreholes is not always trial and error.
convenient. However, the use of linear surf ace arrays or The starting model is used to genera te a theoretical
grids of electrodes has not yet resulted in a simple practi- sounding curve which then is compared with the field
cal imaging system, mainly because these techniques do curve. If the correct multiplying constant has been used,
not lend themselves to simple back projection. The the two curves will be 'in ph ase' but will have different
pole-dipole or dipole-dipole arrays which are normally amplitudes. An iterative process is then carried out to
employed are very sensitive to near-surface lateral adjust the resistivities of the modellayers while keeping
the boundaries fixed. Each layer resistivity is adjusted by
'School of Earth Sciences, University of Birmingham, Edgbaston, a factor equal to the ratio of the observed apparent re-
Birmingham B15 2IT, UK. sistivity to the calculated apparent resistivity (Zohdy
I1

54/FIRST BREAK VOL 10, NO 2, FEBRUARY 1992

1989) as follows: vantage-it can. be modified to operate in two dimen-

{j) Po{j) sions for inversion of electrical imaging data sets
(j') _
Pi+ 1 - Pi Pc,.{j)' (1)
collected in the form of pseudosections.
where i is the number of iterations already carried out, j
is the number of the layer (or electrode spacing), Pi{j) is Pseudosection inversion
the resistivity of layer j after i iterations, Po{j) is the A pseudosectien is a display technique devised by Hallof
observed apparent resistivity at the jth spacing on the (1957) which involves plotting resistivity traverse data as
sounding curve, and Pei{j) is the calculated apparent a depth section, with each apparent resistivity being
resistivity at the jth spacing for the ith iteration. plotted as if it were the true resistivity of a point im-
After each iteration the apparent resistivity curve is mediately below the centre of the electrode 'array at a
recalculated. The process is repeated until therms dif- depth proportional to the electrode spacing. The
ference between the observed and model curves reaches contoured data provide an approximate picture of the
a minimum. Alternative methods for adjusting the layer resistivity distribution in the plane of the section. The
resistivities may be used (Koefoed 1979), although for principles of the technique are shown in Fig. 2.
typical Wenner or Schlumberger data (1) provides well- A technique for transforming a pseudosection from
behaved, if slow, convergence. apparent resistivity to true resistivity may be considered
Although the technique is simpie, it probably holds to have the same basic stages as described by Zohdy
little attraction over other sounding inversion (1989), although a 2D pseudosection modelling program
techniques which yield simpler three- or four-layer must be used instead of the resistivity curve modelling
modeis. However, the method has one important ad- algorithm. Firstly a finite difference network (Dey and

50

E
I
E
s:
o
10~--==---

50
>-
t:
>
I- b
en shifted
(i.j
w layering
cc

10 100 1000
ELECTRODE SPACING, a, or DEPTH

Fig. 1. Automatic sounding inversion technique: (a) Observed data and initiallayering. (b) Shifted layering and resulting model sounding curve.
Difference e between model curve and observed curve is used to apply a correction c to the layering. (c) The finallayering and resulting model
curve, which is closely similar to the observed data.

lil.
 FIRST BREAK VOL 10, NO 2, FEBRUARY 1992/55

Station 3
I
I I

C1 P1 P2 C2
3a 3a 3a
I

Station 2
I

Cl P11 P2
I
2a 2a I za ~2
1\

Station 1
I
I I
Cl Pl P2 C2
I a I a I a I
A
n = 1 •1 • • •
n = 2 • •
2
n=3 •
3
• • • •
n= 4 • •
4
n=5 •
5

Fig. 2. An example measurement sequence for building up a pseudosection.

Morrison 1979) is designed so th at a single element or situations this has important practical advantages.
block of elements coincides with each point on the Wenner data can be easily collected using computer-
pseudosection. The apparent resistivity of this point is controlled systems (Griffiths et al. 1990) and as th is elec-
used as the initial resistivity value of the associated trode array is less sensitive to spurious lateral effects
element. This initial model is then used to generate a than dipole-dipole arrays (Barker 1979), it pro duces
theoretical pseudosection data set. Differences between pseudosections which are already very smoothed
the model and observed pseudosection apparent resis- approximate images of the subsurface (Olayinka and
tivities are used to modify the true resistivity of each Barker 1990).
network element according to (1). The process is re- The first stage in the inversion procedure is the design
peated until the observed data are modelled to within a of an initial model and to this end a semi-infinite half
pre-defined error limit or until the rms difference space is discretized to a rectangular non-uniform grid of
reaches a minimum. 185 horizontal and 18 vertical elements. The central por-
One of the problems is how best to re late electrode tion of the grid is used to model the pseudosection with
spacing to depth in order to bring the computed appa- one datum point of the latter corresponding to four hori-
rent resistivities 'in-phase' with the measured data. For zontally adjacent elements of the grid as shown in Fig. 3.
resistivity sounding Zohdy resolved the problem by ap- The depth of the central point of the block of four ele-
plying a range of shift factors and selecting th at which ments is equal to the electrode spacing a multiplied by
gave the smallest error between the measured sounding the constant 0.5. Outside the range ofthe pseudosection
curve and the initial model curve. Barker (1989) recog- the grid elements are given the resistivity value of
nized that the shift applied by Zohdy was directly related the nearest vertical or horizontal element on the
to the depth of investigation, and th at for most resistivity pseudosection.
distributions fast convergence could be obtainèd by ap- The potential distribution at the surface from current
plying a fixed shift factor equal to the median depth of in- ftowing into the ground at selected points is computed
vestigation as defined by Edwards (1977) for the general using the finite difference procedure described by Dey
case of a homogeneous earth. For the Wenner array this and Morrison (1979). By modelling the field survey a
factor is 0.5 (i.e. layer depth = 0.5 a, where a is the theoretical pseudosection can be built up.
spacing between adjacent electrodes). Although this The model pseudosection is then compared with the
value is correct only for a homogeneous earth, the main field data and adjustment of the true resistivities of each
effect of applying it to a hetergeneous earth is merely a of the model grid elements is carried out according to
reduction in the convergence rate. (1).
In the examples of the procedure presented below we The whole process is repeated iteratively until the rms
consider only the Wenner electrode array as in many difference drops to a predetermined level, or until the
56/FIRST BREAK VOL 10. NO 2. FEBRUARY 1992

0.5
1 .5 .
25
Electrode positions

•.-
;:~:
" '.
'!i(
20m

n=

?!"he [,
15

25

I~
~~ 3
. ~ ~
\'\
3

45

55

Fig. J. Sectien of finite difference model showing how individual values of resistivity in the pseudosection are al\ocated to blocks of four adjacent
elements. Surrounding edge elements (.) take on the value of the nearest pseudosection resistivity.

error reaches a minimum. With good quality data the 100 ohm mand a lowerlayer oflOOOohm m. Alsoshown
process is surprisingly stabIe , in Fig. 4, for comparison, is the layering obtained with
A simple test with a theoretical pseudosection corn- the Zohdy technique applied to a sounding over the
puted for a single horizontal interface gives results which same structure. There is a smalI, but hardly significant,
are closely similar to those obtained with the same te eh- improvement in performance of the sounding inter-
niques applied to a resistivity sounding. Figure 4 shows pretation, this probably being the result of the greater
the layering obtained after ten iterations in modelling a accuracy with which the sounding curves are computed.
Wenner pseudosection measured over a single interface A Wenner pseudosection measured over a horizont-
at a depth of 45 m between an upper layer of resistivity ally layered structure provides a simple smooth function

100,-,---,,--"-,----,--,--,-1' -- -- - - - - - - - - - - - -- - - ,

initia! mOdel-i

I1 sounding curve............. »:
trom tinal >- /'
I
I interpreta tio';}:/'
E I
I

/
I
E
s:
o
)If
~
/ '-........
values of
1 /4f"'" apparent resistivity
(" trom initial model

_/
/", I
101==-~-~=~-~.====t--""'''' ------'
Interpretation trom:
_.- pseudosection
--- sounding

10 100
DepthIelectrode spacing, m

Fig. 4. Verticalline of pseudosection apparent resistivity values interpreted as a conventional sounding and as part af a 2D pseudasection.
 FIRST BREAK VOL 10, NO 2, FEBRUARY 1992/57

(a) Uertical fault ohserued data

i,;, ,i,li"

••• U,Qfi "i,:,!,,!I!1 lli! !ij

',,:;,;j,i ,i,j,'" ,i,i,i! i:i
n = & ij !lH ',' i" ',' i" ~" i, I" lil! !II
......
fault
(b) fault trialS

§I ~<';'/
(c) fault trial 18 electrode positions

>\" ;/:
I &810
':: . :;:',: :

!H,!H!i!!HH ::H!::H!!H!! .... 1__ l1li ... _

8 12 2S 58 'IS 125
RESISTIVITY In oho-.
Fig. S. (a) Pseudosection of apparent resistivity computed for the Wenner array across vertical fault separating material of 10ohm m on the left
from material of 100 ohm m. Electrode positions and fault shown. Electrode spacing = 20 m. (b) Resistivity depth section obtained after ten
iterations.

(a) BLOeI(ohserued data

.... :::::: : ... ' .: -: .; .: . ;. : .; .. : : : : : .; : : : : : : : : : ~: : ... : : :

';; .... ;:;:':':.!' ••!.!••••di!!!I!II!llil!i!!,; •••• :::::·::,:···:::··;···

n = 4
(b) BLOeI(trial &

J:
l-
n,
w , ,
o :: :
:: :
BIlII
<, block
(c) BLOeI(trial 18 electrode positions

:. :

: :dl(lli'
~ii!:i:i~l:;:;:mmmmm •••• 1111111111 ••••
8 1& 32 128
RESISTIVITY in oho-.
Fig. 6. (a) Pseudosection of apparent resistivity computed across 20 bleek for the Wenner array. Block has a resistivity of IOOohmm in a medium
of 10 ohm m. Block and electrode positions shown. Electrode spacing = 20 m. (b) Resistivity depth section after six iterations. (c) Resistivity
depth section after 18 iterations.
58/FIRST BREAK VOL 10, NO 2, FEBRUARY 1992

120

100

.....----original bloek

80 model

60

40

electrode positions

Fig. 7. Calculated resistivities at 50 m depth across the block structurc of Fig. 5 compared after every four iterations. The element structure is
shown for tbe optimum inversion after 18 iterations.

and good results can be expected. Measurements over a rms difference is reached af ter six iterations. This is pos-
vertical fault provide a much more complex apparent re- sibly because different parts of the sectian reach accept-
sistivity pseudosection with lobes to each side ofthe fault able solutions and minimum errors of fit at different
(Fig. Sa), However, the proposed modelling technique rates. With poorly defined or noisy anomalies the pro-
appears quite stable and convergence continues for cess appears not to converge on a solution but to pass
more than 20 iterations. Computed images after five and through a solution, then la overshoot and eventually to
ten iterations (Fig. Sb and c) demonstrate the pro- become unstable.
gressive sharpening of the fault image, As with real data there is generally na known solution:
In fact, the influence of the side lobes of the anomaly it is acceptable to stop iterating when the smallest rms
on the inversion cannot be wholly rernoved. Although difference between the measured and ca1culated pseudo-
the technique heads towards the truc solution, it is not sections is reached.
reached because of inherent ambiguity in the systern, I have tested the technique on real data and have ob-
This is less of a problem where structures do not con- tained pleasing results. The following examples illus-
tinue to the surface and so a faulted structure covered trate successful applications and highlight same of the
with a surface layer may be more successfully imaged. Iimitations of the method.
Another interesting structure to consider is a horst
block model such as th at shown in Fig. 6, The shaded Example I: Quarry extenslon
pseudosection measured across the block (Fig. 6a) Tarmac Raadstone's Old Cliffe Hili Quarry in Leicester-
shows a gentie structure of small amplitude, However, shire is a major souree of aggregate. In recent years a
af ter eighteen iterations the interpreted resistivities new quarry has been opened in the same rock type
(Fig. 6c) are similar ta thase of the model, although the (microdiorite) 2 km to the west (Croxall et al. 1989; Bell
dimensions, while having some resemblance to the and Hopkins 1985), As part of the geological site investi-
model, do not exhibit the same sharpness. Horizontal gations, a resistivity pseudosectien was measured using
profiles of resistivity camputed across the block at a a Campus MRT (microprocessor-controlled resistivity
depth of 50 m at each iteration are compared in Fig. 7, It traversing) System along part of a line joining the two
is c1ear th at the very small amplitude anomaly used as quarries, An electrode spacing of 25 m was employed,
the initial model has been adjusted near to the true the aim being the determination of the thickness of the
resistivity at this level af ter 18 iterations. c1ay overburden along the line. The observed apparent
Although the model at 18 iterations appears close st in resistivity pseudosection of Fig. 8a was adjusted over 15
appearance to the known structure, the smallest tata I iterations to achieve a minimum error of fit. Af ter 15
 FIRST BREAK VOL 10. NO 2. FEBRUARY 1992/59

(a) studtaMl obserued data electrodes

. : . : : .. iiiiij!!lii!!!1 iHiHiiiiiiii! ••••••• l1li1•••

16 32 64 128 2S6 512
RESISTIUITY in oho-.
Fig. 8. (a) Wcnner apparent resistivity pseudosection measured across buried microdiorite at Stud Farm, Cliffe Hill, Leicestershire. Electrode
spacing = 25 m. (b) Resistivity depth section obtained after 15 iterations. Positions of boreholes and observed depths to microdiorite are
indicated.

iterations the error started to rise indicating that parts of resistivity zone suggests that there is an area of low re sis-
the section we re unstable (Fig. 9). Such instability oe- tivity material (possibly sandstone saturated with
curs where the data contain errors of observation and leaehate) bel ow the known base of the fill.
strong near-surface lateral effects.
The irregular granite topography indicated by the
final section of Fig. 8b agrees well with wh at is known of
this area, including depths to weathered microdiorite
proved by four boreholes at the positions shown. The
image of Fig. 8b probably provides a better image of the
subsurface than a model involving sharp interfaces (e.g.
Griffiths and Barker 1989), as the granite is known to be
deeply weathered at pi aces along its surface with the
resistivity of the top of the microdiorite graduaUy
increasing downwards. In this example, differences
between the model and the boreholes are also likely to

\\
be due to the faet th at the measured apparent resis- g•
tivities are affeeted by unknown depth variations outside ~ STUD FARM
w
the line of the section (i.e. the structure is not 20). ~
"
Example 2: Landfill site
An electrical imaging survey was carried out across a
landfill site in Nottinghamshire, England. Here there
was thought to be downward migration of leaehate from
the landfiU into the underlying porous sandstone. The
-.<,
"".----._----.
approximate dimensions of the landfill are shown in Fig. s -·---e_e_._._._. ---·
lOc and are based on past surveyors' records.
The survey employed a unit electrode spaeing of 10 m
and data were eoUeeted with multiples of n = 1 to n = 6
(cf Fig. 2). The observed pseudosection is shown in Fig.
IOa. The data were inverted using the proposed
s 20
teehnique and the minimum rms error (6.1 %) was at- TRIAL
tained af ter eight iterations. The resulting eleetrieal
image (Fig. lOb) outlines the landfill as a low resistivity ERROR CURVE FOR ITERATtVE INVERSION
OF STUD FARM SURVEY
zone surrounded by higher resistivity sandstone. The
lateral boundaries agree weil with the known dimen- Fig. 9. Error curve for the iterative inversion of the Cliffe Hili data of
sions, although the greater vertieal extent of the low Fig.8a.
 60/FIRST BREAK VOL 10, NO 2, FEBRUARY 1992

(a) landr i11 cbser-eed data

(b) I.Ddrill trial 8

16 24 32 48 96
RESISTIUITY in ohn-a

(c) landr ill

LANDFILL

" /

',------~ LEACHATE

CaNT AMINA TlON

Fig. 10. (a) Wenner apparent resistivity pseudosection measured across a landfill. Electrode spacing = 10 m. (b) Resistivity depth section
obtained after eight iterations. (c) Approximate section across the landfill based on existing information.

Noisy data
The problems of noisy data are very much those de- noise cao be removed using a simpte 3-point triangular
scribed by Zohdy (1989), A datum point which is spuri- filter (0,25, 0,5, 0,25) applied along the rows. I prefer to
DUS, either through containing significant error or apply the filter horizontally as this preserves a generally
through being strongly affected by near-surface changes horizontally layered geological structure.
in resistivity, wilt result in oDe element of the image Alternatively, the procedure adopted by Zohdy
decreasing or increasing to anomalous resistivity values. (1989) can be applied. That is, when the point is reaehed
The result is that the imaging process becomes unstable where the error starts to increase, the inversion is recorn-
af ter a limited number of iterations and the total rms meneed by using the last model pseudosection as this is
error starts to increase. always a smooth section with little of the irregularity of
Near-surface changes in resistivity pose a problem be- the original data,
eause, although their origin lies outside the area imaged, An example of the application of this technique is
their effects can of ten be quite strongly seen on the illustrated in Fig. 11 where the results of processing
observed pseudosection. Unless the near-surface is in- obtained over an alluvium-filled valley at Oicks Creek,
cluded in the image by making many measurements of New South Wales, Australia are illustrated. The ob-
apparent resistivity at very sm all electrode spacings, the served data of Fig. 11a have been processed to provide
computer processing attempts to accommodate these the smallest error of fit af ter ten iterations. The pro-
effects in the resistivity change at greater depth, This cessed image (Fig. 11b) appears noisy, however, and the
again leads to instability in the processing technique. fitting error is relatively high (Fig. 12), At this point the
Errors of observation are always present to a small de- processing is reeommeneed using the pseudosection cal-
gree and will always impose a limit on how weil the data culated from the model af ter ten iterations as the new
can be modelled. However, with noisy data mueh of the starting model. Now the inversion is very sta bie and the

•
 FIRST BREAK VOL 10, NO 2, FEBRUARY 1992/61

(a) Dicks Creek, observed data

n = &
(b) trial 10 electrodes

;:'i~il
. '''mIl
.: :

1500
(c)~t~ria~l~l~o~s~m~o~o~th~e~d~~~~ ~~ ~~~ _

1500
(d) trial 20 smoothed

hllrw~.··
,::li\Wj

1500 ..
'
. . ' : ![iiiiillli!l!! mmmmm ••• 1 .... ' ••••

32 48 &4 128 192

RESISTIUITY in ohIrM
Fig. 11. (a) Wenner pseudosection measured across fractured bedrock overtain by thin alluvium at Dicks Creek, Australia. Electrode spacing
= 5 m. (b) Resistivity depth section obtained after ten iterations. (c) Resistivity depth section obtained after ten iterations using the pseudo-
section calculated from the depth section in (b) as a starting point. (d) As for (c) but with 20 iterations.

error continues to reduce, albeit slowly, even after 20 this may be resolved by restarting the process with the
iterations (Fig. 12). model pseudosection calculated at the minimum error.
The result after ten iterations since adopting the new The technique is most successful where the initial
starting model (Fig. 11c) is clearly much smoother than pseudosection has some resemblance to the final inter-
the image in Fig. lIb. Af ter 20 iterations the image (Fig. pretation. Electrode arrays such as the pole-dipole,
lld) has not changed greatly. Although some interesting which provides relatively complex pseudosections over
structures are becoming apparent, their significance can simple structures, may not be the most suitable for the
only be deterrnined by drilling. application of the technique. Although I have obtained
good results with the Wenner array, it is possible that
Conclusions alternative two-electrode or other arrays might be
A simple and powerful method is presented for the fully superior.
automatic modelling of resistivity pseudosections. The
technique produces elect ri cal images which are geo- Acknowledgements
logically significant, especially when good quality data I should like to thank Dr J. Turnbull for permission to
can be recorded. Pseudosection data which include the use the Cliffe Hili data and Tarmac Roadstone Holdings
effects of measurement and near-surface resistivity vari- Lid for permission to publish them, Prof D.H. Griffiths
ations can cause the technique to become unstable, but for the use of the landfill data, and Dr I. Acworth of
 62/FIRST BREAK VOL 10, NO 2, FEBRUARY 1992

15 BEASLEY,C.W. and WARD,S.H. 1988. Cross-borehole resistivity in-

version. 58th Meeting, Society of Exploration Geophysicists,
Expanded Abstracts, 198-200.
BELL, A.R. and HOPKINSD.A. 1985. From farmland to Tarmac.
Extractive Industry Geology 1985, P.W. Scott (ed.), Geological
Society Miscellaneous Papers 18, 19-26.
CROXALL,M.A., HARRIS, D.1. and HILLS, R.L. 1989. Cliffe Hili
;f'. Quarry: a geotechnical study. Extractive 1ndustry Geology 1989,
IX
o J.W. Gaskarth and A.C. Lumsden (eds), 177-190. Institution of
IX
IX Geologists, London.
W
al
DAILEY,W. and YORKEY,T.J. 1988. Evaluation of cross-borehole re-
::. sistivity tomography. 58th Meeting, Society of Exploration
IX

Geophysicists, Expanded Abstracts, 201-203.

DEY, A. and MORRIsON,H.F. 1979. Resistivity modelling for arbitrary
shaped two-dimensional structures. Geophysical Prospecting 27,
106-136.
EDWARDS,L.S. 1977. A modified pseudosection for resistivity and in-
duced-polarization. Geophysics 42, 1020-1036.
HALLOF,P.G. 1957. On the lnterpretation of Resistivity and lnduced
Polarization Measurements. PhD thesis, Massachusetts Institute
2 3 4 5 6 8 9 10 11 12 13 14 15 16 17 18 of Technology.
TRIAL
GRIFFITHS,D.H. and BARKER,R.D. 1989. Electrical imaging and
Fig. 12. Error curves for iterative inversion of Dicks Creek Survey: examples of its application. Extractive Industry Geology 1989,
upper using original data; lower using the pseudosection calculated J.W. Gaskarth and A.C. Lumsden (eds), 251-254. Institution of
from the depth section obtained after eight iterations of the original Geologists, London.
data. GRIFFITHS,D.H., TURNBULL,J. and OLAYINKA,A.1. 1990. Two-
dimensional resistivity mapping with a computer-controlled
array. First Break 8,121-129.
KOEFOED,O.O. 1979. Geosounding Principles, 1, Resistivity Sounding
Measurements. Elsevier, Amsterdam.
the Centre for Groundwater Management and Hydro- LA BRECQUE,D.J. and WARD,S.H. 1988. Two-dimensional inversion
of cross-borehole resistivity data using movable boundaries. 58th
geology, University of New South Wales, Australia, for
Meeting, Society of Exploration Geophysicists, Expanded
permission to publish the Dicks Creek data. Abstracts, 194-197.
LOWRY,T. and SHIVE,P.N. 1990. An evaluation of Bristow's method
Received 22 May 1991; accepted 28 October 1991 for the detection of subsurface cavities. Geophysics 55,514-520.
OLAYINKA, A.1. and BARKER,R.D. 1990. Borehole siting in crystalline
basement areas of Nigeria with a micro-processor controlled
resistivity traversing system. Ground Water 28, 178-183.
References SHIMA,H. and SAlTO,H. 1988. Application of resistivity tomography
BARBER,D.e. and SEAGAR,A.D. 1987. Fast reconstruction ofresis- for detection of faults and evaluation of this hydraulic continuity:
tance images. Clinical Physics and Physiology Measurements 8, some numerical experiments. 58th Meeting, Society of Explora-
47-54. tion Geophysicists, Expanded Abstracts, 204-207.
BARBER,D.C. and BROWN,B.H. 1984. Applied potential tomo- SMITH,D.L. 1986. Application of the pole-dipole resistivity technique
graphy. Joumal of Physics, E: Scientific lnstruments 17, 723-733. to the detection of solution cavities beneath highways.
BARKER,R.D. 1979. Signal contribution sections and their use in resis- Geophysics 51,833-837.
tivity studies. Geophysical lournal of the Royal Astronomical TRIPP, A.C., HOLMANN,G.W. and SWIFT,e.M. 1984. Two-dimen-
Society 59, 123-129. sional resistivity inversion. Geophysics 49, 1708-1717.
BARKER,R.D. 1989. Depth of investigation of collinear symmetrical ZOHDY,A. A. R. 1989. A new method for the interpretation of Schlum-
four-electrode arrays. Geophysics 54, 1031-1037. berger and Wenner sounding curves. Geophysics 54,245-253.