Sunteți pe pagina 1din 8

Archaeological Prospection

Archaeol. Prospect. 6, 179±186 (1999)

A Discussion on the Resolution of


Two-Dimensional Resistivity Modelling
N. COSKUN1* AND J. E. SZYMANSKI2
1
Geophysical Engineering Department, Karadeniz Technical University,
61080, Trabzon, Turkey
2Department of Electronics, University of York, Heslington, York YO1 5DD, UK

ABSTRACT Theoretical resistivity profiles corresponding to a subsurface resistivity structure probed via an
assortment of arrays have been calculated using the finite difference method. Some of the
assumptions implicit within the calculation are illustrated by examining the effects of the precise
relative positions of an electrode array and a buried feature. The responses for the Wenner,
pole–pole and dipole–dipole arrays were carried out by using the horizontal profiling technique.
Four different possible relative array–feature positions were modelled with the finite difference
method. Copyright * c 1999 John Wiley & Sons, Ltd.

Key words: geoelectric prospecting; two-dimensional modelling; numerical experiment.

Introduction purposes, corresponding to the transition in


materials at the boundary between human-made
Horizontal resistivity profiling is an important foundations and the surrounding soil, or at the
technique for civil engineering and environ- sides of in-filled ditches or drainage systems
mental applications and mineral prospecting. (Scollar et al, 1991). As with large-scale geo-
Investigated features have vertical or near-vertical exploration, archaeological resistivity techniques
boundaries Ð faults, dykes, fissures, dipping are rather sensitive to near-surface conductivity
veins, or the sides of large three-dimensional variations (Telford et al, 1990), leading to con-
bodies. Hence, vertical models are well- siderable data variations with changes in local
established as being important for the interpret- subsurface moisture content associated with
ation of field resistivity data (Telford et al, 1990; drainage, the water table and recent weather
Chunduru et al, 1991). The resistivity method has conditions (Scollar et al, 1991).
also been applied widely in non-invasive field There would be no prior knowledge of the
archaeology for the location, and potential ident- precise position of any feature relative to the
ification, of buried remains and structures. probe array on the ground. In the numerical
Although the scale of the structures involved on modelling the discretization of the target region
archaeological sites is much smaller than those has to be determined; otherwise, the iterative
being examined in large-scale geophysics, the inversion algorithm shifts the place and changes
spacing of the electrodes is reduced correspond- the size of feature.
ingly and there are considerable similarities
between the approaches. In particular, vertical
boundaries are important for archaeological
Data interpretation and modelling
*Correspondence to: N. Coskun, Geophysical Engineering
Department, Karadeniz Technical University, 61080, Trabzon, The quantitative interpretation of resistivity data
Turkey. can be a difficult problem. A full mathematical

CCC 1075±2196/99/040179±08$17.50 Received 3 July 1998


Copyright # 1999 John Wiley & Sons, Ltd. Accepted 17 June 1999
180 N. Coskun and J. E. Szymanski

Figure 1. The finite difference mesh.

Figure 2. The enlargement of ABCD area shown in Figure 1.

inversion is computationally expensive, ill- (1966). The effects of a range of structures are
conditioned and noise-sensitive (Menke, 1984). shown in Kunetz (1966). More recently, a FOR-
Basic interpretation is often carried out by com- TRAN program has been provided for computing
paring field results with anomalies calculated for apparent resistivities over a perfect vertical dyke
simple theoretical models Ð but most formal of arbitrary thickness and with different media on
theoretical approaches are applicable only to a either side (Chunduru et al, 1991). Analogue tank
very limited range of structures and situations, modelling studies have been carried out by
such as simple plane-layered models (Sharma, McPhar Geophysics Limited (1967 to 1970),
1986; Telford et al, 1990), a perfect infinite dyke amongst others, and simple three-dimensional
(Chunduru et al, 1991; Telford et al, 1990; Van bodies have been investigated using Wenner,
Nostrand and Cook, 1966) or hemispherical sinks pole±pole and unipole arrays by Apparao and
and buried spheres (Van Nostrand and Cook, Roy (1970). Most of these works have been
1966). Analytical solutions have been given for concerned with mineral exploration, but studies
some vertical and inclined contacts for the more appropriate for archaeological situations
dipole±dipole array, in particular by Al'pin have been carried out by Bernabini et al (1987),

Copyright # 1999 John Wiley & Sons, Ltd. Archaeological Prospection, Vol. 6, 179±186 (1999)
Two-Dimensional Resistivity Modelling 181

large regions. This lack of flexibility is deliberate,


and allows simple interpretations to be con-
structed from even highly noisy data, as the full
inverse problem is reduced to the estimation of a
small number of parameters (region resistivity
values and dimensions). In effect, the assumption
of a specific structural form is incorporating a
priori knowledge and hence applying consider-
able restrictions to the subspace of acceptable
inversions that are compatible with the experi-
mental data within the noise threshold. However,
even minor extensions can make a model result
invalid. For example, the recent treatment of
Chunduru et al, (1991) is inapplicable to any case
except an infinitely deep, outcropping, vertical
dyke. In that, the effect of an overburden layer is
not incorporated into the model.
A restricted modelling approach ensures that a
solution can be found, but it is far from
guaranteed that it is the `correct' interpretation:
the main problem is that expectations are being
encoded in a somewhat arbitrary fashion via the
original assumptions about the model structures.
Some assumptions are certainly necessary,
because the incorporation of prior knowledge is
the only consistent way to overcome the null-
subspace problem, as has already been recog-
nized in image reconstruction problems (Hanson,
1987), but the form of the assumptions can be far
less restrictive and the information can be
incorporated as prior local or global knowledge
Figure 3. Four different possible relative array–feature within a regularised inversion algorithm. This
positions (a–d), modelled with the FDM. In each case the approach is used very commonly in image
configuration unit is displaced by a further 1/4 of the dipole restoration and tomographic applications, with
length.
Bayesian methods simultaneously providing
both the least arbitrary and the most powerful
tools (Skilling and Gull, 1985, 1987; Aykroyd and
who calculated responses for resistive parallele- Green, 1991). A Bayesian approach using
piped bodies by using an integral approach for covariance matrices to carry prior knowledge
the `classic' Wenner and Schlumberger arrays, about parameters and their constraints has been
their focused versions, and the focused tri- applied to resistivity inversion by Pous et al,
electrode, tetraelectrode, pentaelectrode arrays; (1987) and, although still not a general approach,
and by Spiegel et al, (1980), who have studied it clearly allows an increased flexibility in the
resistivity anomalies corresponding to tunnels/ initial model.
voids and have considered the effects of varia- Discrete numerical techniques, such as the
tions in surface terrain variations. finite element method (FEM) and the finite
The limitations intrinsic to analytic models difference method (FDM), offer greater flexibility
include the facts that only simple, or highly for the study of the interaction between applied
specific, structures can be considered, and that currents at the boundary of a complex inhomo-
resistivity values are assumed to be constant over geneous medium. An adequately fine mesh can

Copyright # 1999 John Wiley & Sons, Ltd. Archaeological Prospection, Vol. 6, 179±186 (1999)
182 N. Coskun and J. E. Szymanski

Figure 4. Theoretical FDM Wenner anomalies over a resistive outcropping target. Illustrating the anomaly variations as the
relative array–feature positions are changed as indicated in Figure 3. x-axes are in arbitrary units.

overcome the model-dependent limitations vertical elements was used and boundary con-
described above, at the cost of increased computa- ditions were handled by the use of four further
tional storage and time: the methods can cope non-uniform elements added at each side and
with mixed horizontal and vertical structures, five elements to the bottom of the main area to
overburden effects, non-vertical structures and provide a termination strip. This sort of segmen-
isolated bodies, as well as gradual parameter tation copes very well with boundary conditions.
variations. Detailed theoretical information is given in the
The finite difference method has been dis- above works and is not repeated here.
cussed in detail by Mufti (1976) and Dey and In summary, the aims of the work reported
Morrison (1979). For the purposes of this work here are: to use the FDM to investigate the effect
the algorithm of Dey and Morrison (1979) has on an anomaly of the precise relative array±target
been used because of its flexibility and the sim- crossing position; and to provide information
plicity of its approximate formulation. A square towards using the FDM within an iterative
uniform mesh of 100 horizontal elements and 10 inversion algorithm.

Copyright # 1999 John Wiley & Sons, Ltd. Archaeological Prospection, Vol. 6, 179±186 (1999)
Two-Dimensional Resistivity Modelling 183

Figure 5. Theoretical FDM pole–pole anomalies over a resistive outcropping target. Illustrating the anomaly variations as
the relative array–feature positions are changed as indicated in Figure 3. x-axes are in arbitrary units.

Electrode arrays studied applied current, and G is the so-called geometric


constant of the electrode array (Sharma, 1986).
An infinite number of four-probe (or more than For a perfectly homogeneous conducting med-
four probes) array configurations are possible for ium the calculated value of the resistivity is
the purpose of resistivity profiling: the most independent of the probe positions and spacing
commonly used arrays have been reviewed by but, in general, this is not the case and
Ward (1990). Here, the dipole±dipole, pole±pole measurements recorded will vary in a compli-
and Wenner arrays have been studied over cated fashion with the relative positions of the
various complex structures, but only an example electrodes. Under these circumstances, the use of
is given to show the effect of the precise relative the geometric constant becomes a position-
array±target crossing position on the calculated dependent approximation: the values of resistiv-
apparent resistivity profiles. ities calculated under these circumstances
For each set of probe positions, the resistivity are referred to as apparent resistivities and
is usually calculated by r ˆ GDV/I, where DV designated by ra : the SI unit of ra is the
is the measured potential difference, I is the ohm-metre.

Copyright # 1999 John Wiley & Sons, Ltd. Archaeological Prospection, Vol. 6, 179±186 (1999)
184 N. Coskun and J. E. Szymanski

Figure 6. Theoretical FDM dipole–dipole anomalies over a resistive outcropping target. Illustrating the anomaly variations
as the relative array–feature positions are changed as indicated in Figure 3. x-axes are in arbitrary units.

Results and discussion grid segment values are 1 for all 100 (Figures 1
and 2).
Theoretical apparent resistivity profile curves Two potential restrictions implicit in the FDM
resulting from a model have been computed for arise from possible modelling inadequacies
three conventional collinear arrays with the FDM associated either with the use of too coarse a
approach. The FDM mesh used was of a mesh or with the restrictions of only being able to
reasonably high resolution relative to the probe place the probes at boundary nodes of the FDM
spacing (108 divisions horizontally by 15 verti- mesh. The former problem has been addressed
cally). above, and the importance of being able to
The grid intervals were at the following values represent accurately the relative array±feature
of z (z direction was vertical, increasing down- positions are illustrated in Figure 3. Four different
wards): 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 8, 16, 64, 512. relative probe-array±feature crossing positions
The y direction was horizontal, and the grid are possible with the resolution of the FDM
intervals were at the following values of y: 1000, mesh used here (Figure 3): a rather coarser mesh
100, 20, 4, A . . . B, 4, 20, 100, 1000, where A . . . B might force the assumption that all vertical

Copyright # 1999 John Wiley & Sons, Ltd. Archaeological Prospection, Vol. 6, 179±186 (1999)
Two-Dimensional Resistivity Modelling 185

boundaries appear directly below a probe, for target depth and the resistivity characteristics of
example. Here, initially, anomaly responses were any overburden layer. In other words, the block
calculated with the model placed in the middle of size of the finite difference mesh should be equal
the mesh, and then the probe array was moved to 1/4 (0.25 by 0.25 units) of the minimum
1/4, 1/2, 3/4 probe units to the right, allowing electrode spacing.
four different array±feature crossing positions. This information forms a sound basis for the
Figures 4±6 illustrate the responses of Wenner, use of the approach within an iterative inversion
pole±pole, and dipole±dipole arrays, respect- algorithm, although the application to field data
ively, to a single, near-surface, isolated resistive is likely to require, in addition, the use of a priori
outcropping feature (resistivity contrast 1:10) of information to assist in the regularisation of the
unit dimensions: the importance of the precise ill-conditioned iterative inverse problem.
relative positioning is immediately apparent. In
practice, there would be no prior knowledge of
the precise position of any feature relative to the
probe array. The variability of the calculated References
anomalies shows that any modelling approach
will need to be fine and flexible if an accurate Al'pin, L. M. (1966). The Theory of Dipole-dipole
Sounding. New York: Consultant Bureau Enterprises
iterative inversion of field data is desired. Incorporated.
The sensitivity of anomaly magnitude, shape Apparao, A. and Roy, A. (1970). Resistivity model
and breadth to the precise target depth and near- experiments Ð 2. Geoexploration 9: 195±205.
surface/overburden resistivity characteristics, as Aykroyd, R. G. and Green, P. J. (1991). Global and local
well as target size, shape and resistivity contrast, priors, and the locations of lesions using gamma-ray
imagery. Philosophical Transactions of the Royal Society
will place a major restriction on iterative inver- of London, Series A 337: 323±342.
sion approaches, given the constraints imposed Bernabini, M., Brizzolari, E. and Mazzola, C. (1987).
by noise, measurement limitations and model- Anomalies due to resistive parallelepiped .bodies in
dependent features introduced by the forward resistivity profiles. Bolletino di Geofisica Teorica ed
modelling stage. The inherent ill-conditioning Applicata XXIX 114: 133±146.
Chunduru, R. K., Nagendra, R. and Patangay, N. S.
will require the use of a priori information to (1991). RESDYK Ð a FORTRAN program for com-
compensate for some of these difficulties and to puting apparent resistivity over an infinitely deep
suppress the creation of large-scale artefacts by outcropping vertical dike. Computers and Geosciences
the inversion process. 17: 1395±1408.
Dey, A. and Morrison, H. F. (1979). Resistivity
modelling for arbitrarily shaped two-dimensional
structures. Geophysical Prospecting 27: 106±136.
Conclusion Hanson, K. M. (1987). Bayesian and related methods in
image reconstruction from incomplete data. In Stark,
With a view towards the use of the FDM as a H. (editor), Image Recovery: Theory and Applications
forward model within an iterative inversion (New York: Academic Press). 79±125.
Kunetz, G. (1966). Principles of Direct Current Resistivity
algorithm, the sensitivity of the calculated Prospecting. Berlin: Gebruder Borntraeger.
anomaly curves for a range of probe arrays has Menke, W. (1984). Geophysical Data Analysis: Discrete
been examined, with particular emphasis on the Inverse Theory. London: Academic Press.
model-dependent effects that could arise from Mufti, I. R. (1976). Finite-difference resistivity model-
constraining either the horizontal or vertical ling for arbitrarily shaped two-dimensional struc-
tures. Geophysics 41: 62±78.
target boundaries to too coarse a mesh or to Pous, J., Marcuello, A. and Queralt, P. (1987). Resis-
specific positions. The results show that high tivity inversion with a priori information. Geophysical
horizontal resolution is required because array Prospecting 35: 590±603.
responses are highly dependent on the relative Scollar, I., Tabbagh, A., Hesse, A. and Herzog, I.
array±target positions, and that a high vertical (1991). Archaeological Prospecting and Remote Sen-
sing Ð Topics in Remote Sensing, Vol. 2. Cambridge:
resolution is also required (at least in the near- Cambridge University Press.
surface region) because the anomaly magnitudes Sharma, P. V. (1986). Geophysical Methods in Geology,
and breadths are also highly dependent on the 2nd edn. New York: Elsevier.

Copyright # 1999 John Wiley & Sons, Ltd. Archaeological Prospection, Vol. 6, 179±186 (1999)
186 N. Coskun and J. E. Szymanski

Skilling, J. and Gull, S. F. (1985). Algorithms and Spiegel, R. J., Sturdivant, V. R. and Owen, T. E. (1980).
applications: maximum entropy and Bayesian Modelling resistivity anomalies from localized voids
methods in inverse problems. In Smith, C. R. and under irregular terrain. Geophysics 45: 1164±1183.
Grandy, W. T. (editors), Proceedings of the First and Telford, W. M., Geldart, L. P. and Sheriff, R. E. (1990).
Second International Workshops on Maximum Entropy Applied Geophysics, 2nd edn. Cambridge: Cambridge
and Bayesian Methods in Applied Statistics, University University Press.
of Wyoming, 8±10 June 1985 (Reidel). 83±132. Van Nostrand, R. G. and Cook, K. L. (1966). Inter-
Skilling, J. and Gull, S. F. (1987). Prior knowledge must pretation of resistivity data. United States Geological
be used: maximum entropy and Bayesian methods Survey Professional Paper 499, 310 p.
in inverse problems. In Smith, C. R. and Grandy, Ward, S. H. (1990). Resistivity and induced polaris-
W. T. (editors), Proceedings of the Third International ation methods. Geotechnical and environmental
Workshops on Maximum Entropy and Bayesian geophysics. SEG (Society of Exploration Geophysics),
Methods in Applied Statistics, University of Wyoming, Investigations in Geophysics 5: 147±189.
1±4 August 1987 (Reidel). 162±172.

Copyright # 1999 John Wiley & Sons, Ltd. Archaeological Prospection, Vol. 6, 179±186 (1999)

S-ar putea să vă placă și