1Department of Engineering Geology, Lund University, Sweden ... DC resistivity imaging is applied broadly in ... geophysical textbook (e.g. Sharma, 1997). The.
IAGA WG 1.2 on Electromagnetic Induction in the Earth Extended Abstract 19th Workshop Beijing, China, October 23-29, 2008
Inversion comparison study in the cases of different array configuration and inversion method Dahai Zhang1,2, Torleif Dahlin1, Roger Wisén1 1 2
Department of Engineering Geology, Lund University, Sweden
Shanghai Offshore Oil&Gas Company, SINOPEC, Shanghai, China
SUMMARY By inversion comparison study in the cases of different arrays (WN, SH, PP, DD, PD, GN) and inversion methods, 3-D effects on 2-D resistivity imaging were showed via numerical simulation of synthetic models and subsequent inversion of synthetic datasets and field datasets. Inversion was executed using 2-D and 3-D inversion software. 3-D finite difference forward modeling was carried out, and datasets were extracted as 2-D pseudosection along parallel profiles in one direction of the models. These datasets were inverted individually using a 2-D inversion algorithm and combined using a 3-D algorithm. In both cases inversions were carried out with L1-norm (robust) inversion as well as L2-norm (least-square) inversion. Keywords: DC, 3-D inversion, quasi-3D, array configuration, inversion parameter. effects in 2-D surveying are much smaller than
INTRODUCTION
2-D effects in 1-D surveying (Dahlin and Loke, Today 2-D resistivity surveying plays an
1997). The results presented here show that,
important
area
nevertheless, there is an advantage in performing
investigations. DC resistivity imaging is applied
3-D inversion compared to 2-D inversion. Of
broadly in the fields of engineering and
course, this also takes into consideration the
environmental surveying. We show that where a
extraordinary development of computers since
3-D environment is evident, the 3-D resistivity
the former studies.
role
in
many
large-scale
survey, as presented here, with subsequent 3-D inversion can give a much increased detail and
For the results presented here more tests, using
accuracy of the resulting resistivity model than
more experimental factors than for the earlier
the 2D inversion. In many cases this will most
studies (model parameters, array configurations,
likely be motivated by the aim and against the
inversion parameters, etc.), have been made.
additional costs.
Numerical simulation has been made for several resistivity models. Results from one of these
One of the most important short-comings for
models, that are representative for the entire
3-D inversion is that it is time-consuming and
study, are presented here. Also, a field example
computer exhaustive. The existence of 2-D
that clearly show the advantage of 3-D inversion
effects on 1-D resistivity modeling is a well
are presented.
known problem(e.g. Dahlin and Loke,1998). Former studies of 3-D effects on 2-D inversion,
METHOD
utilizing numerical simulation, show that 3-D
19th IAGA WG 1.2 Workshop on Electromagnetic Induction in the Earth Beijing, China, October 23-29, 2008
1/6
ZHANG, et al, 2008, Inversion Comparison Study
S.1_2
Techniques for acquisition and methods for
order to better understand the differences
interpretation of resistivity measurements have
between 2-D and 3-D inversion. Synthetic data
been
used
for different array configurations have been
extensively for several decades. Continuous
inverted using 2-D inversion with L1 and L2
profiling has been used to obtain a 2D image of
norm and 3-D inversion with L1 norm.
the
continuously
subsurface
developed
resistivity
and
and
is
a
well-documented method (Griffiths and Turnbull,
In the field examples, resistivity data was
1985; Overmeeren and Ritsema, 1988; Dahlin
collected as CVES data with different versions
1993; Griffiths and Barker 1993; Dahlin 2001).
of the ABEM Lund Imaging System (Dahlin, 1996). This CVES system allows for very
The
pole-pole
pole-dipole
(PP),
(PD),
dipole-dipole Wenner
(DD),
(WN)
and
flexible data collection. Based on the target and the
actual
geophysical
configuration
known
programmed and, if desirable, altered during the
described
in
any
standard
geophysical textbook (e.g. Sharma, 1997). The
data
density
can
any
Schlumberger (SH) configurations are well and
and
problem,
be
measurement sequence.
depth of investigation and sensitivity problems of different arrays are discussed by many authors (e.g.
Edwards,
1977;
Loke,
2002).
SYNTHETIC RESULTS
The
application of the multiple gradient (GD) array
The synthetic models are constructed in 6 layer
(Figure 1) for multi-channel measurement
slices, with a 41 by 21 electrode grid as shown in
systems was introduced by Dahlin and Zhou
2. The layer depths are 0.3, 0.7, 1.5, 2.4, 3.5 and
(2004). They show that this array has a high
4.7 m respectively. This model is a little bit
signal to noise ratio and at the same time a good
complex and can be regarded as a horst structure
distribution of the sensitivity. In this study we
with a contact between two rock units with
tested these six different array configurations.
different resistivity, 1000 Ωm and 300 Ωm respectively. The horst structure is overlain by a layer of 70 Ωm resistivity, representing a soil
layer. This layer also contain two smaller bodies
a A
M na
of low (30 Ωm) and high (1000 Ωm) resistivity.
B
N ma
Within the high resistive rock unit there is a low
sa
resistive structure (10 Ωm) representing a weathered fracture zone in the hard rock. Figure 1. Sketch
of
gradient
array
layout
showing the position of the electrodes for a
The 2D robust inversion result with WN, SH,
measurement
PD, and PP arrays are showed in Figure 3.
with separation factor
s = 7,
n-factor 2 and midpoint factor m = -2. The n-factor is here defined as the relative spacing
The location and geometry of the anomaly body
between the potential dipole and the closest
were showed from the slice map in different
current electrode.
array configurations, but the shape of anomaly body was distorted severely in the lower part for
The numerical simulations were carried out in
PD arrays, and for PP arrays it shows blur
19th IAGA WG 1.2 Workshop on Electromagnetic Induction in the Earth Beijing, China, October 23-29, 2008
2/6
ZHANG, et al, 2008, Inversion Comparison Study
S.1_2
anomaly boundary in the shallow part. Generally
depth of 4.3 meter, but the resistivity of
speaking, inversion result with WN arrays shows
surrounding media is a bit lower than true one,
the higher resolution compared to synthetic
even in 3-D inversion this problem still can be
model in the lower or shallow part.
found. Anyway, Wenner and gradient arrays inversion show the best representation ability in the
description
of
underground
electrical
property distribution among 6 different kinds of array configurations in the synthetic model experiment. FIELD DATASET Recently we made another field electrical resistivity measurement in Mörrum, South Sweden. The measurement net parameters are: seven 160m-long profiles, 5m electrode spacing, and 10m line spacing. The whole profile is west-eastward, and a weathering outcrop of blocky rock with slight topographic uphill is Figure 2. The depth slice map of conductive horst model with noisy surface model.
included in our work area, and gradient array was employed in this survey. Figure 24 shows the depth slice map of 2-D robust inversion and 3-D inversion results. From the 2D result, it is
Simultaneously we showed the inversion results
very obviously that a high resistivity (~ 10000
in general gradient array (Fig. 4). First, let’s
Ohm m) anomaly blocky body is located in the
have a quick look for 2-D inversion result slice
south part of the work area, extends to the deep
maps. In the shallow part, a conductive and
and disappeared at ~20m; about the same depth
resistive rectangle anomaly body is showed
place another high resistivity region emerges at
clearly in 2-D LS inversion, and horst structure
the northeast corner of work area; based on
in lower part is also represented as a stable and
above-discussed
continuous conductive anomaly extended to the
resistive anomaly body emerged on the surface
deeper; and better effect could obtain in 2-D
with conductive surrounding layer stems from
robust inversion, more strictly constrained
lower high resistivity basement located at the
condition made the border of anomaly body
northeastward depth of 20~30m, that is to say,
more clear and improve the inversion quality,
the whole high-resistive structure body obtruded
but which has the influence on the inversion
and exposed to the surface from lower northeast
effect on near-surface. Gradient arrays result
direction. At the same time we processed the
show better ability of survey depth than Wenner
datasets by 3-D robust inversion (see right map
arrays, even in the case of 2-D robust inversion,
in Fig. 24). The 3D result shows the similar
we still can distinguish the geometry shape,
electrical property distribution with 2D inversion
position and resistivity of conductive body in the
but higher resolution in lower part, it is the
19th IAGA WG 1.2 Workshop on Electromagnetic Induction in the Earth Beijing, China, October 23-29, 2008
we can deduce that the high
3/6
ZHANG, et al, 2008, Inversion Comparison Study
S.1_2
reason that 2-D inversion got the fewer data
electrode arrays, Geophysical Prospecting, 52,
points than 3-D. But the 3-D result supplies the
379-398.
further clearer testimony to abovementioned deduction.
Dahlin, T., Loke, M. H., 1997, Quasi-3D resistivity
CONCLUSION AND DISCUSSION
imaging
–
mapping
of
three
dimensional structures using two dimensional DC resistivity techniques, Procs. 3rd meeting of
By lots of synthetic geological modeling and
Environmental and Engineering Geophysics
field
Society, Denmark, 143-146.
datasets
inversion,
the
research
consequences show 3-D effects exists in the 2-D surveying, but it is much less than 2-D effects in
Edwards, L. S., 1977, A modified pseudosection
the 1-D surveying. To decrease 3-D effect as
for
possible as we can, the choice of the array
Geophysics, 42, 1020-1036.
resistivity
and
induced-polarization,
configurations and inversion parameter have a important role in the inversion results, Wenner
Griffiths,
D.H.
and
Barker,
and Gradient arrays inversion show the best
Two-dimensional
resolution to the resistivity anomaly structure,
modeling in areas of complex geology, Journal of
and dipole-dipole array gets the worst. 3-D
Applied Geophysics, 29, 211-226.
resistivity
R.D., imaging
1993, and
inversion is sensitive to large-scale resistivity structure than 2-D inversion, and 2-D L1-norm
Griffiths, D.H. and Turnbull, J., 1985, A
inversion also presents the robust advantages in
multi-electrode array for resistivity surveying,
depth and anomaly body border. Combining 2-D
First Break, 3(7), 16-20.
and 3-D advantages, prudent selection of array type and inversion are helpful for improving the
Loke, M. H., 2002, Tutorial: 2D and 3D electrical
final geological model.
imaging surveys. Overmeeren, R.A. van and Ritsema, I.L., 1988,
REFERENCE
“Continuous vertical electrical sounding,.” First Dahlin, T., 1993, On the Automation of 2D
Break, 6(10), 313-324.
Resistivity Surveying for Engineering and Environmental Applications, Dr.Thesis, ISRN
Sharma,
LUTVDG/TVDG--1007--SE,
engineering geophysics, ISBN 0-521-57632-6,
ISBN
91-628-1032-4, Lund University, 187p.
P.V.,
1997,
Environmental
and
Cambridge University Press, UK.
Dahlin, T., 2001, The development of electrical imaging techniques, Computers and Geosciences, 27(9), 1019-1029. Dahlin, T and Zhou, B., 2004, A numerical comparison of 2D resistivity imaging with ten
19th IAGA WG 1.2 Workshop on Electromagnetic Induction in the Earth Beijing, China, October 23-29, 2008
4/6
ZHANG, et al, 2008, Inversion Comparison Study
S.1_2
Figure 3. Horst model inversion slice map in WN, SH, PP, PD array configuration, respectively.
2-D inversion (L2-norm)
2-D inversion (L1-norm)
3-D inversion (L1-norm)
Figure 4. Horst model inversion slice map in gradient array configuration.
19th IAGA WG 1.2 Workshop on Electromagnetic Induction in the Earth Beijing, China, October 23-29, 2008
5/6
ZHANG, et al, 2008, Inversion Comparison Study
S.1_2
Figure 5. 2-D (Left) and 3-D (Right) inversion result slice map of gradient field dataset in Mörrum, Sweden.
19th IAGA WG 1.2 Workshop on Electromagnetic Induction in the Earth Beijing, China, October 23-29, 2008
6/6