GEOPHYSICS, VOL. 62, NO. 3 (MAY-JUNE 1997); P. 807–813, 4 FIGS.

Automatic conversion of magnetic data to

depth, dip, and susceptibility contrast
using the SPI (TM) method

Jeffrey B. Thurston∗ and Richard S. Smith∗

3) asymmetric anomalies caused by a nonvertical magneti-

ABSTRACT zation direction, and
The Source Parameter Imaging (SPITM ) method com- 4) asymmetric anomalies caused by a nonvertical dip.
putes source parameters from gridded magnetic data.
Thus, magnetic-data interpretation generally requires a spe-
The method assumes either a 2-D sloping contact or a
cialist in potential-field methods who also has an excellent un-
2-D dipping thin-sheet model and is based on the com-
derstanding of the geology.
plex analytic signal. Solution grids show the edge lo-
There are a large number of methods designed to simplify
cations, depths, dips, and susceptibility contrasts. The
magnetic data by removing one or more of the above complex-
estimate of the depth is independent of the magnetic
ities. Reduction to the pole (Baranov, 1957) removes the asym-
inclination, declination, dip, strike and any remanent
metry caused by nonvertical magnetization direction, while the
magnetization; however, the dip and the susceptibility
pseudogravimetric transform (Baranov, 1957) reduces a dipo-
estimates do assume that there is no remanent magneti-
lar field to a monopolar field. Subsequently, the source edges
zation. Image processing of the source-parameter grids
are typically identified by a zero crossing on a second-vertical
enhances detail and provides maps that facilitate inter-
derivative profile or by maxima on a horizontal-gradient or
pretation by nonspecialists.
analytic-signal amplitude profile. Sample profile plots of these
The SPI method tests successfully on synthetic pro-
quantities can be seen in Figure 1b. This example is a good
file and gridded data. SPI maps derived from aeromag-
illustration of how the zero crossing or maxima of the deriva-
netic data acquired over the Peace River Arch area of
tives can be offset if the dip is nonvertical. The maximum of the
northwestern Canada correlate well with known base-
analytic-signal amplitude, however, is independent of inclina-
ment structure and furthermore show that the Ksituan
tion, declination, remanent magnetization and dip—provided
Magmatic Arc can be divided into several susceptibility
the sources are 2-D (i.e., have effectively infinite strike length).
subdomains.
The disadvantage of the analytic-signal amplitude is that the
mapped features are broad compared to the features on a
INTRODUCTION second-vertical derivative map.
Another way of avoiding the above complications in the
Magnetic data are used commonly to map thin magnetic magnetic field and its subsequent transformations and deriva-
sheets and/or contacts such as faults. An example of this is tives is to deconvolve the total-field response into quantities
in petroleum exploration where magnetics is used to identify that describe the causative body (the source parameters). This
faults in the basement that may control the depositional history can be done manually for each anomaly using characteristic
of the sedimentary basin. In this case, it is important that the curves (e.g., Peters, 1949; Grant and West, 1965). Characteristic
interpreter has an understanding of the structural and litho- curves are still used widely (Spector and Lawler, 1995), as they
logical framework and the importance of regional deposition do not require a computer, can be performed in the field, allow
patterns. However, interpretation of the magnetics is compli- the user to differentiate between noise and signal, and can be
cated by: applied to old analog data sets. Forward and inverse modeling
1) the dipolar nature of the induced field; methods are more sophisticated than characteristic curve tech-
2) possible offsets between the edges of the causative niques, but are usually more time consuming. Generally, these
body and characteristic points on the response (peaks, methods are not used on aeromagnetic surveys because of the
crossovers); large quantity of digital data now being collected.
Manuscript received by the Editor October 24, 1995; revised manuscript received August 28, 1996.
∗
Geoterrex, a Division of CGG Canada Ltd., 2060 Walkley Road, Ottawa, Ontario K1G 3P5, Canada.
°c 1997 Society of Exploration Geophysicists. All rights reserved.

807
808 Thurston and Smith

Automatic source-parameter estimation can be accom- is that they do not require profiles perpendicular to 2-D
plished either on a profile basis or on a grid basis. Profile structures.
techniques include the Naudy method (Naudy, 1971), Werner We have developed a technique that works on gridded data,
deconvolution (Hartman et al., 1971; Jain, 1976) and 2-D Eu- the results from which can be processed easily using off-the-
ler deconvolution (Thompson, 1982). Such profile techniques shelf imaging processing software. One of our images can be
assume a 2-D structure and that the structure is perpendicular used to determine the source location and the source depth.
to the flight line. In some cases, it may be possible to correct Unlike the second-vertical-derivative method, there is no off-
the result when these assumptions are not satisfied, but this is set of the estimated location (the cross-over) and the actual
normally done manually, making it a time consuming task. Pro- location. The image is not dipolar in nature and is independent
file methods such as these are generally capable of estimating of the magnetic inclination, the declination, the dip, the strike,
the depth, and in some cases, the magnetization, and the dip and any remanent magnetization. Our only assumption is that
(assuming no remanence). the causative body is a 2-D contact or a thin sheet. If we also
There are a number of methods that work on a grid; assume there is no remanent magnetization, it is possible to
for example 3-D Euler deconvolution (Reid, 1990), the generate two additional images that show estimates of the dip
3-D analytic-signal technique (Roest et al., 1992), and the and the susceptibility contrast.
enhanced analytic-signal technique (Hsu et al., 1996). The Our goal is that these images can be interpreted by someone
results from these methods are usually displayed by plotting who is not familiar with magnetic data interpretation, but is an
a symbol at the interpreted source location; the size of the expert in the local geology.
plotted symbol generally increases as the interpreted depth
to the source increases. The advantage of grid methods METHODOLOGY

Our method extends the theory of the complex analytic

signal (Nabighian, 1972, 1984; Atchuta Rao et al., 1981; Roest
et al., 1992), by computing three complex attributes from
which source parameters can be computed. In the analysis
of temporal series these are: the instantaneous amplitude,
analogous to the analytic-signal amplitude computed by Roest
et al. (1992); the instantaneous phase; and, the instantaneous
frequency (see e.g., Taner et al., 1979). Prefacing the attribute
name with instantaneous makes the distinction clear between
these quantities and their spectral namesakes. As magnetic
data are spatial, rather than temporal, we use the term local
instead of instantaneous.
The use of the complex analytic signal, which assumes
2-D sources, has two additional appealing features. First, this
complex function can be computed for either total-field or
horizontal-gradient data. The former case is used for source-
parameter estimation assuming a thin-sheet model, while the
latter case is for sources that are assumed to be sloping con-
tacts with effectively infinite depth extent. Thus, the assumed
model can be altered by changing only the quantity that is in-
put to the algorithm. Further, use of the complex analytic signal
means that depth can be computed without a priori knowledge
of the magnetization direction. As with other automatic meth-
ods, there is no easy way of determining which model is most
appropriate.

Complex attributes
FIG. 1. (a) The total-field anomaly for the dipping contact
model shown at the bottom of the figure. The dip is 135◦ , the Nabighian (1972) defined the complex analytic signal in two
depth to the top is 100 m, the ambient field strength is 60 000 nT, ways: (1) in terms of the horizontal and vertical derivatives
the susceptibility contrast is 0.01 SI, and the declination is zero and (2) in terms of the total field and its Hilbert transform. The
degrees (for this particular model 2K Fc sin(d) = 848 nT and
2I − d − 90 = −75◦ ). (b) Profiles of the analytic-signal am- former definition is
plitude, horizontal derivative, vertical derivative (all left axis) ∂ M(x, z) ∂ M(x, z)
and the second-vertical derivative (right axis). Note that the A(x, z) = −j , (1)
horizontal derivative maximum and the second-vertical deriva- ∂x ∂z
tive zero crossing are both displaced from the top of the con- where M(x, z) is the magnitude of the total magnetic field, j is
tact (zero). The analytic-signal amplitude peaks at zero, with a
value of 8.48 nT/m. (c) The wavenumber and phase response. the imaginary number, z and x are Cartesian coordinates for
At x = 0, the wavenumber is 0.01 m−1 , and the phase is 75◦ . the vertical direction and the direction perpendicular to strike.
Using equations (11), (13), and (14), the local depth, the local Equation (1) is equivalent to
dip, and the local susceptibility contrast are 100 m, 135◦ , and
0.01 SI, respectively. A(x, z) = |A| exp( jθ ), (2)
 Magnetic SPI Maps 809

where arguments are in degrees. Substituting equations (8) and (9)

s
µ ¶2 µ ¶2 into the expression for the local wavenumber (7) yields
∂M ∂M h
|A| = + , (3) κ= , (10)
∂x ∂z h2 + x 2
and where the coordinate system has been defined such that x = 0
· Á ¸ directly over the edge. Equation (10) makes it evident that
∂M ∂M
θ = tan−1 . (4) maxima of the local wavenumber are independent of the mag-
∂z ∂x netization direction. Thus, the peaks outline source edges, and
at these locations x = 0. At x = 0, we can calculate “local
By definition, the analytic-signal amplitude and the local phase depth” defined by
are given in equations (3) and (4), respectively (see e.g.,
1
Bracewell, 1965). Atchuta Rao et al. (1981) have described h= . (11)
a manual source-parameter estimation process that works on κ
a profile basis and uses the analytic-signal amplitude and phase. To compute the “local dip,” denoted d, expressions for the
Our method also uses the local frequency, denoted f , which is gradients of the sloping contact are substituted into the expres-
defined as the rate of change of the local phase with respect to sion for the local phase [equation (4)], giving
x (Bracewell, 1965). This quantity is given by µ ¶
· Á ¸ −1 x cos(2I − d − 90) − h sin(2I − d − 90)
1 ∂ ∂M ∂M θ = tan .
f = tan−1 . (5) h cos(2I − d − 90) + x sin(2I − d − 90)
2π ∂ x ∂z ∂x (12)
In the analysis of potential fields, it is often more convenient The local dip can be estimated by setting x = 0 and rear-
to use local wavenumber, denoted by κ, rather than f where ranging equation (12) to give

κ = 2π f. (6) d = θ + 2I − 90. (13)

Making this substitution, and using the differentiation rule In deriving equation (13), and in the following, we assume no
d(tan−1 φ)/d x = 1/(1 + φ 2 ) gives remanent magnetization. Finally, the “local susceptibility” is
µ 2 ¶ given by setting x = 0 and substituting equations (8) and (9)
1 ∂ M ∂M ∂2 M ∂ M into equation (3). If the local depth and local dip are used in
κ= − . (7)
|A|2 ∂ x∂z ∂ x ∂ x 2 ∂z place of depth and dip, respectively, then
|A|
Complex attributes can also be computed for the analytic K = . (14)
signal of the thin-sheet model. This is done by replacing the 2κ Fc sin d
horizontal gradient in the real part of equation (1) with the Equations (13) and (14) give the local dip and local suscepti-
total-magnetic field, and the vertical gradient in the imaginary bility contrast only at the source boundaries (x = 0). Because
part with the Hilbert transform of the total-magnetic field. these functions are not necessarily extrema at these locations,
it is necessary first to determine the edge positions from the
Source-parameter estimation maxima of the local wavenumber.
As an example of the usefulness of these quantities, we have
We now show how the complex attributes can be used di- plotted the wavenumber, and the phase on Figure 1c. Clearly,
rectly to determine the edge location, depth, dip, and suscep- the peak in the local wavenumber represents the contact posi-
tibility contrast. We begin with the expressions for the vertical tion, where x = 0. Using the wavenumber, the phase and the
and horizontal gradients of a sloping contact analytic-signal amplitude at this position, the local depth, the
local dip and the local susceptibility contrast can be derived
∂M as 100 m, 135◦ , and 0.01 SI, respectively. These quantities all
= 2K Fc sin d
∂z agree exactly with the source parameters used in generating
x cos(2I − d − 90) − h sin(2I − d − 90) the synthetic data.
× , (8) In the derivation of the local depth, dip, and susceptibility,
h2 + x 2
it has been assumed that there is no interference from adja-
∂M cent anomalies. This type of interference corrupts the results
= 2K Fc sin d
∂x from all automatic algorithms and generally requires that de-
h cos(2I − d − 90) + x sin(2I − d − 90) tailed modeling be used for an accurate quantitative analysis.
× , (9) Hartman et al. (1971) discuss a method for dealing with inter-
h2 + x 2
ference on a profile basis. However, since we are calculating
(Nabighian, 1972), where K is the susceptibility contrast at second-order derivatives to generate the images, the interfer-
the contact, F is the magnitude of the Earth’s magnetic field, ence will be much less than if we were using total-field or first-
c = 1 − cos2 i sin2 α, α is the angle between the positive x-axis derivative data.
and magnetic north, i is the ambient-field inclination, tan I = For the thin sheet model, the total field and its Hilbert trans-
tan i/ cos α, d is the dip (measured from the positive x-axis), form are given by the right-hand sides of equations (9) and
h is the depth to the top of the contact and all trigonometric (8), respectively (Reford, 1964; Nabighian, 1972). The only
810 Thurston and Smith

difference is that the susceptibility contrast (K ) is replaced FIELD DATA AND DISCUSSION
by the susceptibility-thickness (kw). With this substitution, all
We demonstrate this technique on nonexclusive data that
subsequent equations are identical.
are part of a larger survey flown by Geoterrex in 1989. These
Source-parameter imaging data were acquired with the intention of mapping Precambrian
basement underlying the sedimentary section in northwestern
Estimation of source parameters can be performed on grid- Alberta and northeastern British Columbia. Tests conducted
ded magnetic data. This has two advantages. First, this elim- prior to the survey revealed that a constant barometric ele-
inates errors caused by survey lines that are not oriented vation of 1200 m above sea level suppressed the majority of
perpendicular to strike. Second, there is no dependence on a the cultural effects, and still provided good resolution of base-
user-selected window or operator size, which other techniques ment anomalies. The lines are oriented north-south, at a spac-
like the Naudy and Euler methods require. In addition, grids ing of 1.6 km, and the tie-lines are spaced at 4.8 km. Diurnal
of the output quantities can be generated, and subsequently variations have been removed with a standard line/tie-line lev-
image processed to enhance detail and provide structural in- eling procedure, the IGRF has been subtracted, and the to-
formation that otherwise may not be evident. tal field has been gridded at an interval of 250 m. These data
In practice, we have accomplished this on gridded data by are shown in Figure 3a. For reference, the Ksituan Magmatic
estimating the strike direction at each grid point. We calculate Arc, as interpreted in Ross et al. (1991), is labeled on this
the vertical gradient in the frequency domain, and the horizon- map.
tal derivatives are computed in the direction perpendicular to Source-parameter images have been generated for this data
strike using the least-squares method of Thurston and Brown set assuming the sources to be sloping contacts. Compared with
(1994). The problem therefore reduces to that described in the analytic-signal amplitude (3b), there is greater resolution
equations (1) through (14) above. in the local-wavenumber image (3c). Thus, we present the local
The effectiveness of the method can be demonstrated on wavenumber as an alternative to the analytic signal amplitude
synthetic gridded data. The model used is a parallelepiped as both have the appealing feature of being independent of the
2500 × 2500 m, buried 500 m, and with a depth extent of 20 km. ambient field direction and the dip, but the local wavenumber
The susceptibility contrast of the model with the surrounding provides greater resolution and has maxima that are inversely
medium is 0.13 SI units. The total field is shown in Figure 2a proportional to depth.
and the analytic-signal amplitude, calculated using the method Once the maxima of the local wavenumber are determined,
in Roest et al. (1992), is shown on Figure 2b. The maxima of we can image the local depth, local dip, and local susceptibil-
the analytic-signal amplitude outlines the vertical projection ity contrast along the contacts (Figures 4a, 4b and 4c). The
of the edges of the body (these edges are at an angle of 20◦ to interpretation of these images is aided by superimposing the
the cardinal points of the compass). Along the southern edge, local depth, dip, and susceptibility contrast (as color) over a
the peak is diffuse, making it difficult to identify edge loca- grey-scale image of the sun-illuminated total field.
tions and to estimate the source depths using the half width From these images, an interpreter can recognize variations
of the analytic-signal-amplitude anomaly (Nabighian, 1972; that may occur along strike. Note, for example, the depth varia-
Roest et al., 1992). By contrast, the peaks of the local wavenum- tion along the Dunvegan Fault (the linear feature that forms the
ber (Figure 1c), which are defined by the hotter colors, are re- southeastern boundary of the Ksituan Magmatic Arc). Crude
solved more sharply. The color bar of Figure 1c has been labeled contouring of the well-log information indicates that the top of
with the inverse of the local wavenumber, so the depth can be the basement occurs at about 2 km below sea level along the
read off the image at the peak position. Using image processing southeastern edge of the Ksituan Arc (Cant, 1988). The depths
algorithms, it is also possible to display the inverse of the local from the source-parameter images indicate the sources that
wavenumber only at the position where the local wavenumber are 2-D are deeper than 3.3 km below the sensor, which corre-
is a maximum, thereby generating a map on which all points sponds to more than 2.1 km below sea level (which is within the
that are not null represent contact positions and the colors give basement). Further, the depth image correlates with variations
an indication of the depth of the contacts. This is what is dis- between known blocks in the basement. To demonstrate this
played on the local depth map (Figure 2d). The depths derived we have overlain (on Figure 4a) the trace of a horst and graben
in this fashion are about 500 m (the true value), except along the system (the Axial Graben; Cant, 1988) that underlies the axis
north edge, where the estimated depths are about 700 m. These of the Peace River Arch. The uplifted portion of the basement
poorer solutions occur where the total-field image (Figure 2a) corresponds to relatively shallow estimated depths (redder),
shows significant curvature and our assumption of two dimen- while the adjacent basement lows correspond to deeper depths
sionality is somewhat violated. For this example, the images of (greener). Additional structural information is provided by the
local dip and local susceptibility contrast (also plotted where local-dip image: Figure 4b indicates that most of the coherent
the wavenumber is a maximum) give answers close to the true 2-D features that comprise the Ksituan Domain are near ver-
values of 90◦ and 0.13 SI units, respectively. For this example, tical. This near-vertical structure is consistent with statements
the maximum variation from the true value is about 20◦ for the made in Zelt and Ellis (1989), indicating that our assumption
dip and a factor or two for the susceptibility contrast. Again, of no remanent magnetization in this area is reasonably well
the poorer estimates of local dip and susceptibility contrast satisfied.
are where the data are not 2-D. At present, the algorithm is Unfortunately, a paucity of basement lithological data pre-
not able to determine whether or not the sources are adher- cludes detailed ground truthing of the local susceptibility con-
ing to the assumption of two-dimensionality. This decision is trast estimates (Figure 4c). However, we can use this image
currently left to the interpreter. to draw new conclusions about the Ksituan Magmatic Arc.
 Magnetic SPI Maps 811

FIG. 2. (a) The total magnetic anomaly of a synthetic parallelepiped model 2500 m by 2500 m, with a depth to top of 500 m, and a
depth to bottom of 20 km. The sides of the model are vertical, but are oriented 20◦ to the east of grid (and magnetic) north. In SI
units, the susceptibility contrast at the edge is 0.13. The Earth’s magnetic field has an inclination of −61◦ (southern hemisphere),
a declination of zero, and a field strength of 56 000 nT. The map area shown is 12.8 km square and geographic north is to the
top of the figure. (b) The image of the analytic-signal amplitude (assuming a contact model). The peaks of the analytic-signal
define the contacts (or edges) of the block. Note that the peak is diffuse on the southern edge of the body. (c) The inverse of the
local wavenumber, the peak position of which also defines the edges of the block. On this image, hot colors represent large local
wavenumbers or small inverse local wavenumbers. The color bar is labeled with the inverse local wavenumber. (d) The local depth,
which is the inverse of the local wavenumber. The only positions imaged are those where the local wavenumber peaks, where we
assume that x = 0. (e) The local dip of the contact. Contacts with a component of dip to the east have dips in the range zero to 90◦
(vertical) and dips to the west are in the range 90◦ to 180◦ . (f) The apparent susceptibility contrast at the contact. The estimates of
local depth, dip, and susceptibility contrast assume the data are 2-D (no variation along strike). The extent to which the violation
of this assumption affects the results can be seen on the northern edge, where the magnetic-field image shows significant curvature
and is thus not 2-D.
812 Thurston and Smith

FIG. 3. (a) The total magnetic field, (IGRF) removed, for an FIG. 4. (a) The estimated local depth for the same area as
area in the Western Canadian Sedimentary Basin that is ap- Figure 3. This depth (below the sensor) is shown in color at
proximately 160 × 160 km. The sensor is 1.2 km above sea-level the contact position and has been superimposed on a greyscale
and geographic north is to the top of the figure. (b) The ana- image of the magnetic shaded relief. The white lines on the im-
lytic-signal amplitude of Roest et al. (1992), which implicitly age represent faults, interpreted from existing well-log data
assumes a contact model. (c) The local wavenumber: warm col- in Cant (1988). They show a horst and graben system fol-
ors indicate high wavenumber, cool colors low wavenumber. lowing the Peace River Arch. The down-thrown side is in-
The color bar has been labeled with the inverse of the local dicated by the small tick marks. (b) The estimate of the lo-
wavenumber so it has units of km. This image is particularly cal dip of the contacts. Contacts with a component of dip to
useful for extracting structural information as it has been en- the east are cool colors (blue is 0◦ dip), vertical dips are yel-
hanced with a sun-angle illumination. At a peak in the local low, and westerly dips are hot colors (red is 180◦ dip). (c)
wavenumber (x = 0), the inverse of the local wavenumber will The estimate of the local susceptibility contrast at the contact.
correspond to the depth of the contact below the sensor. The The range of the local susceptibility contrast is compressed
peaks of the analytic-signal amplitude and the local wavenum- using the common logarithm. The local depth, dip, and sus-
ber can both be used to map the contacts or edges of mag- ceptibility contrast are only plotted at the peak of the local
netic bodies. However, the local wavenumber corresponds to wavenumber, where we assume that x = 0. The local suscepti-
a higher derivative and thus shows more features and greater bility contrast and local dip maps assume there is no remanent
resolution. magnetization.
 Magnetic SPI Maps 813

The image shows that the Arc is not necessarily homogeneous REFERENCES
(Ross et al., 1991), it can be subdivided further into several Atchuta Rao, D., Ram Babu, H. V., and Sanker Narayan, P. V., 1981,
subdomains on the basis of the variation in the susceptibility Interpretation of magnetic anomalies due to dikes: The complex
contrast. gradient method: Geophysics, 46, 1572–1578.
Baranov, V., 1957, A new method for interpretation of aeromagnetic
maps: pseudo-gravimetric anomalies: Geophysics, 22, 359–383.
CONCLUSION Bracewell, R., 1965, The Fourier transform and its applications:
McGraw-Hill Book Co.
Automatic estimation of the source parameters can be done Cant, D. J., 1988, Regional structure and development of the Peace
River Arch, Alberta: A Paleozoic failed-rift system? Bull. Can. Petr.
rapidly and easily. The estimates are independent of the mag- Geol., 36, 284–295.
netic inclination and free from effects associated with the di- Grant, F. S., and West, G. F., 1965, Interpretation theory in applied
polar nature of the magnetic response. The local wavenum- geophysics: McGraw Hill Book Co.
Hartman, R. R., Teskey, D. J., and Friedberg, J. L., 1971, A system for
ber map more closely resembles the geology than the total- rapid digital aeromagnetic interpretation: Geophysics, 36, 891–918.
field magnetic map and its derivatives. Correct estimates of Hsu, S.-K., Sibuet, J.-C., and Shyu C.-T., 1996, High-resolution detec-
tion of geologic boundaries from potential-field anomalies: An en-
the source parameters are only obtained when the structure is hanced analytic signal technique: Geophysics, 61, 373–386.
2-D and corresponds to the chosen model (sloping contact or Jain, S., 1976, An automatic method of direct interpretation of magnetic
dipping thin sheet). The estimates of dip and susceptibility are profiles: Geophysics, 41, 531–541.
Nabighian, M. N., 1972, The analytic signal of two-dimensional mag-
only reliable if the remanent magnetization is small. netic bodies with polygonal cross-section: Its properties and use for
The estimates of source parameters we obtain from the im- automated anomaly interpretation: Geophysics, 37, 507–517.
ages are consistent with the known source information (for ——— 1984, Toward a three-dimensional automatic interpretation of
potential-field data via generalized Hilbert transforms: Fundamental
the synthetic data) and the known geological information (for relations: Geophysics, 49, 780–786.
the field data). Furthermore, in the field example we obtained Naudy, H., 1971, Automatic determination of depth on aeromagnetic
additional information that extends our understanding of this profiles: Geophysics, 36, 712–722.
Peters, L. J., 1949, The direct approach to magnetic interpretation and
geologically complex area considerably. its practical application: Geophysics, 14, 290–320.
The images presented can be interpreted easily by a non- Reid, A. B., Allsop, J. M., Granser, H., Millet, A. J., and Somerton, I.
W., 1990, Magnetic interpretation in three dimensions using Euler
specialist familiar with the local geology and the current ex- deconvolution: Geophysics, 55, 80–91.
ploration concepts. There is no requirement for a detailed un- Reford, M. S., 1964, Magnetic anomalies over thin sheets: Geophysics,
derstanding of the principles of magnetic data interpretation 29, 532–536.
Roest, W. R., Verhoef, J., and Pilkington, M., 1992, Magnetic interpre-
or a knowledge of the technicalities of the source-parameter tation using the 3-D analytic signal: Geophysics, 57, 116–125.
estimation algorithm. Ross, G. M., Parrish, R. R., Villeneuve, M. E., and Bowring, S. A.,
1991, Geophysics and geochronology of the crystalline basement of
the Alberta Basin, western Canada: Can. J. Earth Sci., 28, 512–522.
ACKNOWLEDGMENTS Spector, A., and Lawler, T. L., 1995, Application of aeromagnetic data
to mineral potential evaluation in Minnesota: Geophysics, 60, 1704–
The authors would like to thank GEOTERREX, a Division 1714.
of CGG Canada Ltd., for permission to publish images of a por- Taner, M. T., Koehler, F., and Sheriff, R. E., 1979, Complex seismic
trace analysis: Geophysics, 44, 1041–1063.
tion of the nonexclusive data set from western Canada. Dave Thompson, D. T., 1982, EULDPH: A new technique for making
Daggar kindly generated the synthetic data shown in Figure 2 computer-assisted depth estimates from magnetic data: Geophysics,
47, 31–37.
and assisted us with the imaging processing algorithms. We Thurston, J. B., and Brown, R. J., 1994, Automated source-edge lo-
would like to thank David A. Chapin, Afif H. Saad and Harold cation with a new variable pass-band horizontal-gradient operator:
L. Yarger for constructive reviews that greatly improved the Geophysics, 59, 546–554.
Zelt, C. A., and Ellis, R. M., 1989, Seismic structure of the crust and
quality of the manuscript. Source Parameter Imaging and SPI upper mantle in the Peace River Arch region, Canada: J. Geophys.
are trademarks of Geoterrex. Res., 94, No. B5, 5729–5744.