Inversion of Magnetic Data from Remanent and Induced Sources
Robert G. Ellis Barry de Wet Ian N. Macleod
Geosoft Inc. Ivanhoe Australia Ltd. Geosoft Inc.
Suite 810, 207 Queens Quay West, Level 13, 484 St Kilda Road Suite 810, 207 Queens Quay West,
Toronto, ON, Canada Melbourne, VIC, 3004, Australia Toronto, ON, Canada
[email protected] [email protected] [email protected] Several authors have reported progress toward magnetic data
inversions including remanent effects (for example, Shearer
SUMMARY and Li 2004, Kubota and Uchiyama 2005, Lelièvre and
Oldenburg 2009). In this work we report further progress in
Magnetic field data are of fundamental importance in this direction with a technique we call Magnetization Vector
many areas of geophysical exploration with 3D voxel Inversion (MVI), which incorporates both remanent and
inversion being a common aid to their interpretation. In induced magnetization without prior knowledge of the
the majority of voxel based inversions it is assumed that direction or strength of remanent magnetization. In the
the magnetic response arises entirely from magnetic following sections, we extend conventional scalar
induction. However, in the last decade, several studies susceptibility inversion to a magnetization vector inversion,
have found that remanent magnetization is far more that is, we allow the inversion to solve for the source
prevalent than previously thought. Our experience with magnetization amplitude and direction. While this increases
numerous minerals exploration projects confirms that the the number of variables in the inversion we will show by
presence of non-induced magnetization is the rule rather example that the same regularization principles that allow
than the exception in base metals exploration. compact targets to be resolved in highly unconstrained scalar
susceptibility inversion also apply in vector inversion.
In this work we show that failure to accommodate for
remanent magnetization in 3D voxel-based inversion can Perhaps our most significant finding is that MVI, or more
lead to misleading interpretations. We present a technique generally, inversion including all forms of magnetization,
we call Magnetization Vector Inversion (MVI), which significantly improves the interpretation of the majority of
incorporates both remanent and induced magnetization minerals based magnetic field inversions. Unfortunately, the
without prior knowledge of the direction or strength of surprising degree of improvement in interpretability cannot be
remanent magnetization. We demonstrate our inversion adequately presented in a paper and can only be verified by
using model studies and field data. Successful application direct experience. Consequently, while we have applied MVI
to numerous minerals exploration surveys confirms that to a large number of magnetic field surveys and find the
incorporating remanent magnetization is essential for the results to be significantly superior to conventional scalar based
correct interpretation of magnetic field data. inversion, in this paper we are forced to limit our attention to a
synthetic case and field data from the Cu-Au Osborne deposit
Key words: inversion, 3D, remanent magnetization, located approximately 195km SE of Mount Isa, in Western
magnetization vector inversion. Queensland, Australia.
INTRODUCTION
METHOD AND RESULTS
The utility of magnetic field data in many areas of geophysical
exploration is well-known as is the application of 3D voxel Let us begin with the very general assumption that the
inversion to aid in magnetic data interpretation (for example, magnetic properties of the earth can be represented by a
Li and Oldenburg 1996, Pilkington, M., 1997, Silva et al. volume magnetization, (Telford et al. 1990). We make
2000, Zhdanov and Portniaguine 2002, to cite just a few). In no assumptions about whether source of the magnetization is
the majority of voxel based inversions it is assumed that the induced, remanent, or otherwise.
magnetic response arises entirely from magnetic induction.
However, in the last decade, studies have found that remanent From magnetostatics, the magnetic field at point resulting
magnetization is far more prevalent than previously thought from a volume containing magnetization, , is given by
(McEnroe et al. 2009) and affects crustal rocks as well as
zones of mineralization. Unfortunately, remanent
magnetization can seriously distort inversion based on the
assumption that the source is only induced magnetization. The
severity of the distortion is due to the highly non-unique
This expression shows directly that the magnetization vector
nature of potential field inversion making it extraordinarily
is the natural parameter for inversion. This is a crucial
easy for a potential field inversion to produce a seemingly
observation.
plausible model which agrees satisfactorily with the observed
data, even when a fundamental assumption in the inversion is
If the volume consists of a collection of sub-volumes
flawed.
each of constant magnetization then
nd
22 International Geophysical Conference and Exhibition, 26-29 February 2012 - Brisbane, Australia 1
Magnetization Vector Inversion Ellis, De Wet, Macleod
simplicity, the data were simulated at 20m constant clearance
and on a regular 8m grid.
Inverting the TMI data in Figure 2 yields the model shown in
Figure 3 which should be compared to the true model shown
This defines the forward problem: given a set of sources in Figure 1. There is some variability in the magnetization
then is the predicted magnetic field direction but the predominant direction is clearly EW, in
anomaly at points, . Note that the coordinate agreement with the true model.
index is summed over indicating that we are free to choose
the most computationally convenient internal coordinate
system. It also suggests that a coordinate invariant quantity,
such as the amplitude, will be most robustly
determined from the data.
For conciseness, we will represent Eq (2) simply as
The vector magnetization inverse problem is defined as
solving for given subject to an appropriate regularization
condition. Although there are many choices for the
regularization (see for example, Zhdanov 2002), we choose
without loss of generality, the familiar Tikohonov minimum
gradient regularizer. The inverse problem becomes solving Figure 1: The buried prism model with magnetization vector
for in, orientation (Easterly) shown by the green cones. Side=100m
where in the first line, the total objective function is the sum
of a data term and a model term with a Tikohonov
regularization parameter, . The second line defines the data
objective function in terms of the data equation (3) and the
error associated with each data point, . The third line gives
the model objective function in terms of the gradient of the Figure 2: The TMI data simulated over the magnetization vector
model shown in Figure 1. The axes are in metres.
model and the amplitude of the model, with weighting
terms as required, . The fourth line indicates that the
Tikohonv regularization parameter is chosen based on a
satisfactory fit to the data in a chi-squared sense, . In
addition, other constraints, such as upper and lower bounds,
can be placed on as appropriate to the specific exploration
problem.
Example - Buried Prism
Although the buried prism model is far too simplistic to have
exploration significance, it does make an excellent
pedagogical example, so we follow tradition and begin by
considering the inversion of simulated TMI data over a buried
prism with magnetization vector perpendicular to the earth
field. The model consists a cube with side length 40m buried
Figure 3: The MVI recovered model for comparison with Figure
with a depth to top of 20m and a magnetization vector in the 1. The magnetization vector orientation shown by the green cones.
EW direction, ( as shown in Figure 1.
Vector magnetization models in 3D are difficult to interpret
Simulated TMI data are shown in Figure 2 for Earth field with directly in all the but the simplest cases. In real-world
inclination 90 and amplitude 24000 nT. Cardinal directions exploration we need some simpler derived scalars which
have been chosen only for simplicity of explanation; any highlight the important information in the vector model. As
directions could be chosen with equivalent results. Also for suggested by Eq(1), the most robust and meaningful scalar is
the amplitude of the vector magnetization and this should be
nd
22 International Geophysical Conference and Exhibition, 26-29 February 2012 - Brisbane, Australia 2
Magnetization Vector Inversion Ellis, De Wet, Macleod
the primary quantity used in interpretation. However, since the exploration environments based on experience from many
magnetization vector direction is the earth field direction for magnetic surveys, however this cannot be shown here.
induced sources, it is tempting to attempt to use the directional
information recovered in MVI to generate scalars related to
the earth field direction.
There are many possibilities but we have found that three
useful derived scalars for exploration are: the amplitude of the
magnetization, the earth field projection of the magnetization,
and the amplitude of the perpendicular-to-earth-field
components of the magnetization. In exploration problems,
the amplitude is robust by being independent on of any Figure 5: (a) A cross section through the true model, (b) the
assumptions regarding the earth field, while the amplitude recovered scalar susceptibilty. The color bar shows the
perpendicular is an approximate indicator of non-induced susceptibility magnitude in SI.
magnetization. To support our findings, these three derived
scalars are shown in Figure 4b, c, d for an East-West slice Example - Osborne
through the model volume bisecting the target in the true The preceding pedagogical study of MVI on simulated data
model. over a prism provides a solid basis for the much more
important application of MVI to field data. As mentioned in
In exploration situations it is convenient to present MVI the Introduction, it is hard to appreciate fully the impact on
output normalized by the amplitude of the earth's magnetic magnetic data interpretation by including non-induced
intensity in the area of interest. That is, our results are magnetic sources. However, to motivate our assertion, we
displayed as where is the amplitude of the earth's present typical results taken from TMI data collected over the
magnetic intensity in the area of interest. By using this Osborne deposit.
normalization in an area of purely induced magnetization, the
numerical values returned by MVI inversion will be directly The history of the Osborne mine is well described elsewhere,
comparable to those of scalar susceptibility inversion, in our see for example, Rutherford et al. 2005. Briefly, significant
case in SI. Cu-Au mineralization beneath 30-50m of deeply weathered
cover was confirmed in 1989. Intense drilling between 1990
and 1993 defined a total measured and indicated resource of
11.2 Mt at 3.51% Cu and 1.49 g/t Au. Exploration since 1995
has delineated high-grade primary mineralization dipping
steeply East to some 1100 m vertical depth. As of 2001, total
mined, un-mined and indicated resources are reported to be
about 36 Mt and 1.1%Cu and 1 g/t Au (Tullemans et al.
2001). Current exploration is focussed on mapping the high-
grade mineralization to greater depths and mapping similar
structures in the surrounding area. The geophysics includes
total magnetic intensity (TMI) data over the property, which is
shown in Figure 6. The TMI data were acquired in 1997 flown
at 40m clearance on 40m line spacing.
Figure 4: (a) A cross section through the true model, (b) the
recovered amplitude of the magnetization vector, (c) the
amplitude of the perpendicular-to-earth-field components of the
magnetization, (d) the projection of the magnetization on to the
earth field direction. The colour scales indicate the MVI
magnetization in normalized to SI (see text).
For completeness, and to show the contrast between MVI and
conventional scalar inversion, Figure 5b shows the equivalent
section through a model produced by an inversion which
assumes only induced magnetization. As should be expected,
the recovered model using scalar inversion is a very poor Figure 6: The observed TMI data acquired over the Osborne
representation of the true model, which in real-world property. The axes are in metres. The color scale shows the TMI
exploration ultimately adds significant confusion to the amplitude in nT.
interpretation process.
Magnetization Vector Inversion of the Osborne TMI data
This simple prism example demonstrates the power of yields the magnetization vector amplitude earth model shown
magnetization vector inversion and its advantage over scalar in Figure 7. Superimposed (in black) is the subsequently
susceptibility inversion in cases where the magnetization discovered mineralization from extensive drilling and
vector direction deviates from the earth field direction. We underground mining. For comparison, Figure 8 shows the
argue that this situation predominates in real-world corresponding scalar susceptibility inversion. Comparing
nd
22 International Geophysical Conference and Exhibition, 26-29 February 2012 - Brisbane, Australia 3
Magnetization Vector Inversion Ellis, De Wet, Macleod
Figure 7 and Figure 8 shows that inverting for the ACKNOWLEDGMENTS
magnetization vector provides a much better model for
interpretation. The scalar inversion fails to represent reality in The authors would wish to thank Geosoft Inc. and Ivanhoe
this case suggesting, most likely, that the scalar assumption is Australia Ltd. for permission to publish this work.
violated: a common occurrence in mineral exploration in our
experience. In contrast the MVI model is consistent with the REFERENCES
drilling results, and furthermore, indicates a steeply dipping
volume on the Eastern flank. The strong near surface anomaly Butler, R. F., 1992, Paleomagnetism: magnetic domains to
to the west of the dipping zone is known banded ironstone. geologic terranes, Blackwell Scientific Publications.
Kubota, R., and Uchiyama A., 2005, Three-dimensional
magnetization vector inversion of a seamount, Earth Planets
Space, 57, 691–699
Li, Y., and D. W. Oldenburg, 3-D inversion of magnetic data,
Geophysics, 61, 1996, 394-408.
Lelièvre, P. G., and Oldenburg, D. W., 2009, A 3D total
magnetization inversion applicable when significant,
complicated remanence is present, Geophysics, 74, L21-L30
McEnroe, S. A., Fabian, K., Robinson, P., Gaina, C., Brown,
L., 2009, Crustal Magnetism, Lamellar Magnetism and Rocks
that Remember, Elements, 5, 241-246.
Figure 7: An EW section through the recovered MVI model
amplitude at the Osborne property with the now known Pilkington, M., 1997, 3-D magnetic imaging using conjugate
mineralization shown in black. The color bar gives the normalized gradients, Geophysics, 62, 1132-1142.
amplitude in SI. The axes are in metres.
Rutherford, N. F., Lawrance, L. M., and Sparks, G., 2005,
Osborne Cu-Au Deposit, Clonclurry, North West Queensland,
CRC LEME Report.
Shearer, S., and Y. Li, 2004, 3D Inversion of magnetic total
gradient data in the presence of remanent magnetization: 74th
Annual Meeting, SEG, Technical Program, Expanded
Abstracts, 23, 774-777.
Silva, J. B. C., Medeiros, W. E., and Barbosa, V. C. F.,
2001, Potential-field inversion: Choosing the appropriate
technique to solve a geologic problem, Geophysics, 66, 511 -
520.
Telford, W. M., Geldart, L. P., Sherriff, R. E., and Keys, D.
Figure 8: The same section as in Figure 7 for the scalar model A., 1990, Applied Geophysics, Cambridge University Press.
with drilling and mineralization in black. The color bar gives the
susceptibility in SI. The axes are in metres.
Tullemans, F. J., Agnew P., and Voulgaris, P., 2001, The Role
of Geology and Exploration Within the Mining Cycle at the
CONCLUSIONS Osborne Mine, NW Queensland, in Monograph 23 - Mineral
Resource and Ore Reserve Estimation - The AusIMM Guide
We have argued that remanent magnetization must be to Good Practice, Australian Institute of Mining and
included in magnetic field data inversion in order to avoid
Metallurgy, Melbourne, 157-168.
seriously misleading interpretations. To support this argument
we demonstrated the value of Magnetization Vector Inversion Zdhanov, M. S., 2002, Geophysical Inverse Theory and
using model studies, and field data from the Osborne property. Regularization Problems, Method in Geochemistry and
The degree of improvement afforded by using MVI in all
Geophysics 36, Elsevier Science B.V., Amsterdam, The
areas of magnetic field data inversion may seem surprising,
Netherlands.
however recent advances in understanding remanent
magnetism suggest that non-induced magnetization plays a far Zhdanov, M. S., and Portniaguine, O., 2002, 3-D magnetic
more important role than previously thought in the origin of
inversion with data compression and image focusing,
magnetic anomalies. Successful application to numerous
Geophysics, 67, 1532-1541
minerals exploration surveys confirms that incorporating
remanent magnetization is recommended for the correct
interpretation of the majority of magnetic field data.
nd
22 International Geophysical Conference and Exhibition, 26-29 February 2012 - Brisbane, Australia 4
