Anda di halaman 1dari 21

Physics of the Earth and Planetary Interiors 150 (2005) 6383

Series and parallel transformations of the magnetotelluric


impedance tensor: theory and applications
Jose M. Romo , Enrique Gomez-Trevino, Francisco J. Esparza
Division de Ciencias de la Tierra, Centro de Investigacion Cientca y de Educacion Superior de Ensenada, B.C. (CICESE),
Km 107 Carretera Tijuana-Ensenada, Ensenada 22860, Baja California, Mexico
Received 17 July 2003; received in revised form 10 March 2004; accepted 16 August 2004

Abstract
The basic magnetotelluric (MT) impedance tensor transforms into a set of physical and geometrical parameters that maintain
their validity regardless of dimensionality. In two dimensions (2D), the traditional TM and TE impedances rearrange into an
equivalent pair, series and parallel, which complement each other and together represent the original tensor. The series equivalent
relates to TM and the parallel counterpart to TE. We show how the series- and parallel-impedance concepts can be applied in
three dimensions (3D), overcoming some of the current limitations of TE and TM 2D concepts. The series response function is
mainly affected by galvanic effects related with current flow across interfaces, while the parallel impedance is more sensitive
to inductive effects associated with current flow along interfaces. An intrinsic and most convenient property of the series and
parallel impedances is that they do not depend on the measuring axes, as do the individual tensor elements, as well as the TE
and TM impedances in the 2D case. The directional sensitivity of the new representation is provided by two angular parameters
that complete the equivalency. Formally, a forward transformation operates over the original tensor elements in the traditional
xy domain, and produces parameters in what can be called the SP domain, where S stands for series and P for parallel. The
existence of the inverse transformation for going from the SP to the original xy domain guaranties that there is no loss of
information when going from one representation to the other. We illustrate the performance of SP quantities using forward
computations on multi-dimensional models and 2D inversions of synthetic and field data.
2004 Elsevier B.V. All rights reserved.
Keywords: Magnetotellurics; Impedance tensor; Series and parallel transformation; Magnetotelluric inversion

Correspondence to: Earth Science Division, CICESE, P.O. Box 434843, San Diego, CA 92143 4843, USA; Earth Science Division, CICESE,
Casa International Brokerage Inc., 9355 Airway RD, San Diego, CA 92154, USA. Tel.: +52 646 175 0500; fax: +52 646 175 0567.
E-mail address: jromo@cicese.mx (J.M. Romo).

0031-9201/$ see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.pepi.2004.08.021

64

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

1. Introduction
The basic data of a magnetotelluric (MT) survey is
usually expressed by a tensor. The recognition that the
appropriate representation for the earths impedance
required a tensor approach came early in the history
of magnetotellurics. The original scalar approach for
layered media (Tikhonov, 1950; Cagniard, 1953) was
of limited practical use. Field experiments showed the
scalar impedance depended not only on the spatial variation of resistivity, but also on the polarization of the
source fields. This limitation was removed by properly considering that the horizontal components of
the electromagnetic field are linearly related through
a second-order impedance tensor (Cantwell, 1960;
Rokityansky, 1961; Bostick and Smith, 1962; Swift,
1967), i.e.,
  


Ex
Zxx Zxy
Hx
=
(1)
Ey
Zyx Zyy
Hy
This model made possible the practical use of magnetotellurics, but it also brought new challenges with
it.
The impedance tensor comprises the interactions of
horizontal electric currents in all possible directions
regardless of the polarization of the inducing fields. It
is a simple but general representation of the earths response function for arbitrary polarization of the natural
electromagnetic field.
Over horizontally layered media the information included in the second-order tensor is consistent with
the original scalar version of MT. The diagonal elements are zero because the induced current is not deflected into the media, and the off-diagonal elements
have identical magnitude and opposite sign, since the
impedance is the same in all directions. Thus, onedimensional (1D) environments are properly handled
by the impedance tensor directly.
In two-dimensional (2D) structures, two orthogonal directions are needed to describe the current
flow. These directions define the familiar TE (transverse electric) and TM (transverse magnetic) modes
for 2D models, where current flows along and across
strike, respectively. If the coordinate system is properly aligned with the geo-electric strike, the two diagonal elements that were zero in the 1D case remain so.
The off-diagonal elements of the tensor correspond to

the impedance associated with the TE and TM modes.


Thus the impedance tensor leads to quantities that can
be easily recognized and properly assigned to conduction modes in a 2D physical model. If the coordinate
axes are not aligned with the strike of the model, all
four elements of the tensor are different from zero, but
a rotation of axes recovers the standard TE and TM
modes. In practice, as first done by Swift (1967), the
rotation angle is chosen to minimize a combined norm
of the diagonal elements.
The simplicity and sound physical basis of Swifts
(1967) approach led to its wide application even when
not fully applicable. In general, in the presence of threedimensional (3D) current flow, it is generally impossible to find an angle of rotation that effectively reduces
the diagonal elements to a reasonable near-zero level.
The resulting pseudo-TE and pseudo-TM modes are no
longer a complete representation of the physical system
impedance. This problem has motivated a vast number
of investigations in the magnetotelluric literature.
General 3D situations necessitate the analysis of the
full tensor. The efforts to simplify the analysis of the
four complex frequency functions, which also change
with rotation of the reference frame, have evolved
trough different routes. Some insight on the relationships among tensor elements and their behavior under
axes-rotation has been gained trough graphical representations, from the early work by Sims (1969), to
the Mohr-diagrams representation proposed by Lilley
(1993a,b, 1998). Another approach includes the efforts
to exploit the relative advantages for interpretation of
a number of rotation-invariant quantities formulated
with the tensor elements (e.g. Fischer and Masero,
1994; Szarka and Menvielle, 1997; Weaver et al., 2000;
Szarka et al., 2000). A very successful line of research
has been to assume the measured impedance tensor
as composed of a true, regional 2D tensor distorted in
some way by local effects, (e.g. Larsen, 1977; Zhang et
al., 1987; Bahr, 1988, 1991; Groom and Bailey, 1989,
1991; Chave and Smith, 1994; Smith, 1995, 1997).
These ad hoc decomposition methods have proved useful for a large number of practical cases whenever the
natural situations are close to the assumed model, i.e.,
that any 3D heterogeneity can be considered as local
geological noise.
Another approach makes no assumption about the
regional background. It recognizes that the full tensor describes the 3D earth, and uses mathematical

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

transformations in an effort to find more useful representations that are equivalent to the original tensor. Examples of these are the eigenstate formulations developed by Eggers (1982), Cevallos (1986), LaTorraca et
al. (1986) and Counil et al. (1986), the tensor factorization worked out by Spitz (1985) and the canonical decomposition proposed by Yee and Paulson (1987). The
main advantage of these methods is that they are applicable to general 3D environments. Unfortunately, the
connection between the resulting earth-responses and
the physical model is not as obvious as it is for Swifts
analysis of the 2D problem. These difficulties, added
to the complexities of the analysis itself, are the probable causes that these methods have not been widely
used.
It is correct to say that finding equivalent representations of the impedance tensor is redundant, since we
know the system response of a multi-dimensional earth
is entirely contained in the four original elements. However, while this is formally true, it is not so much in
practice. Consider for instance Swifts (1967) simple
transformation {Zxx , Zxy , Zyx , Zyy }{ZTE , ZTM , },
where stands for the rotation angle that minimizes
the tensor diagonal. Assuming a perfect 2D situation,
it provides an exact equivalent of the original tensor,
but certainly one a lot easier to relate to the physical
system. This is an example where an equivalent transformation proved extremely useful. Is there something
similar in 3D?
We began by considering that the flow of electric
current in the earth can be separated to be either along
or across boundaries, and started a search for parameters that could relate to this pattern. Eventually this lead
us to two new impedance functions: series and parallel;
the first intended to characterize currents across boundaries, and the second for currents parallel to them.
These functions could play in 3D the roles that TE and
TM have in 2D.
The mathematical development is presented in four
steps. The first consists of a complex transformation
similar to the one we used for the analysis of the magnetic transfer function (Romo et al., 1999). In the second we obtain a representation for the tensor in terms
of the series impedance. In the third, a similar representation in terms of the parallel impedance is obtained.
Then, in the fourth step, series and parallel representations are compressed into a single one, which includes
two angular response functions to complete the trans-

65

formation. The basic properties of the new parameters


are explored by inspecting anomalies produced by simple 2D and 3D models. The paper ends with the application of a 2D inverse code to SP synthetic and field
data.

2. Impedance transformations
2.1. Standard transformation
Standard impedance rotation assumes that electromagnetic fields observed in the xy coordinate frame
are transformed to a new coordinate system x y by
applying a rotation matrix R to both the electric and
magnetic fields. Using Eq. (1), the above transformation can be written as
RE = RZRT RH,

(2)

where Z is the impedance tensor, E and H the horizontal electromagnetic fields, and R is the rotation matrix
given by


cos
sin
R=
.
(3)
sin cos
RT stands for the transposed rotation matrix and is
the clockwise angle between both coordinate systems.
Thus, the impedance tensor in the new coordinate
system can be obtained from the original by
Z = RZRT .

(4)

Swifts (1967) procedure finds the optimal angle capable to reduce Z to an off-diagonal form.
2.2. Complex transformation
Now consider applying a transformation Re to the
horizontal electric field E, while Rh , possibly different
from Re , is applied to the horizontal magnetic field H,
i.e.,
Re E = Re ZRTh Rh H,

cos f
where Rf =
sin f

sin f
cos f

(5)


,

(6)

and subscript f stands for either e or h, as R is applied to E or H, respectively. Two reference frames,

66

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

one for the electric field and a different one for the magnetic field, have been used in the past by Counil et al.
(1986) and more recently by Lilley (1998). However,
their approaches derive in parameterizations distinct
from those proposed in this work. We follow our earlier work about the magnetic transfer function (Romo
et al., 1999), allowing f to be complex. Hence, Rf becomes a complex transformation capable to introduce
not only axes rotations, but also ellipticity transformations (phase shifts). It is not difficult to prove that assuming = + i, the complex matrix Rf can be written as a true rotation matrix multiplied by a phase shift
matrix formed by hyperbolic functions and imaginary
off-diagonal elements,



cos f
cosh f
sin f
i sinh f
Rf =
sin f cos f
i sinh f cosh f
(7)

Z1 + Z2 = (Zxy + Zyx ) cos(e + h )


+ (Zyy Zxx ) sin(e + h ),
Z1 Z2 = (Zxy Zyx ) cos(e h )
+ (Zyy + Zxx ) sin(e h ),
0 = (Zxx Zyy ) cos(e + h )
+ (Zxy + Zyx ) sin(e + h ),
0 = (Zxx + Zyy ) cos(e h )
(Zxy Zyx ) sin(e h ).

(11)

Using the last two equations in (11) it is easy to show


that
Zyy Zxx
,
(12)
tan(e + h ) =
Zxy + Zyx
and
Zxx + Zyy
.
Zxy Zyx

The complex arguments provide Rf with new capabilities still preserving the form and mathematical properties of a simple unitary matrix.
After the transformation proposed in (5), the transformed impedance can be written in terms of the original as

tan(e h ) =

Z1 + Z2 =

(Zxy + Zyx )
(Zyy Zxx )
=
,
cos(e + h )
sin(e + h )

(14)

Z = Re ZRTh .

Z1 Z2 =

(Zxy Zyx )
(Zyy + Zxx )
=
.
cos(e h )
sin(e h )

(15)

(8)

(13)

The first two equations in (11) can now be written as

Using Eq. (8) we can find the values of e and h required to anti-diagonalize Z, that is to make


0
Z1
T
Re ZRh =
(9)
Z2 0

The explicit equation for Z1 is obtained by adding Eqs.


(14) and (15), whereas Z2 is given by a subtraction of
the same equations:


(Zxy Zyx )
1 (Zxy + Zyx )
Z1 =
+
(16)
2 cos(e + h ) cos(e h )

The four equations represented in (9) can be written


explicitly as

and

Z1 = Zxx Ce Sh Zyx Se Sh + Zxy Ce Ch + Zyy Se Ch ,


Z2 = Zxx Se Ch + Zyx Ce Ch + Zxy Se Sh + Zyy Ce Sh ,
0 = Zxx Ce Ch + Zyx Se Ch + Zxy Ce Sh + Zyy Se Sh ,
0 = Zxx Se Sh Zyx Ce Sh Zxy Se Ch + Zyy Ce Ch ,
(10)
where Cf = cos f and Sf = sin f are both complex valued functions. It is convenient to combine these equations in order to obtain



(Zxy Zyx )
1 (Zxy + Zyx )

.
Z2 =
2 cos(e + h ) cos(e h )

(17)

Applying similar algebraic steps to Eqs. (12) and (13)


we can find explicit expressions, not given here, for e
and h . The main result of the above transformation is
that the original impedance can be equivalently represented by a new set of four complex quantities, i.e.,
{Zxx , Zxy , Zyx , Zyy } {Z1 , Z2 , e , h }.

(18)

The quantities Z1 and Z2 are closely related to the


characteristic values found by LaTorraca et al. (1986),
who used singular value decomposition (SVD) in their

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

67

analysis. Furthermore, Z1 and Z2 are equivalent to the


principal values obtained with the canonical decomposition proposed by Yee and Paulson (1987), whose
analysis show that Z1 and Z2 yield extreme estimates
of the impedance magnitude, which means that the associated apparent resistivity values are the maximum
and minimum possible. The remaining two quantities
of the above transformation, e and h are complex
parameters that, as shown later, are associated with geometrical properties of the media. Using Eqs. (5) and
(6) we can write
  


E1
0 Z1
H1
=
.
(19)
E2
Z2 0
H2

2.3. Series impedance

The first row in Eq. (19) means that the electric field E1
depends only on magnetic fields in the direction of H2 ,
and, similarly, in the second row the electric field E2
is sensitive to magnetic fields only in the direction of
H1 . Just like SVD, or the canonical decomposition, the
above transformation is an effective way to decouple
the 3D magnetotelluric equations into two polarization states. It can be thought of as a natural extension
of the 2D analysis of Swift (1967). Eq. (19) generalizes Swifts analysis in two ways: first, it considers
elliptically polarized fields instead of linearly polarized
fields; and second, electric and magnetic fields are not
necessarily perpendicular to each other.
In what follows we use Eq. (19) as well as e and
h to obtain two other transformations similar to (18).
The motivation now is to obtain parameters that can
characterize currents that flow across and along interfaces. Again, it is convenient to consider the experience gained in 2D interpretation, as recently reviewed
by Berdichevsky et al. (1998). They emphasize the
complementary nature of the information provided by
the TE and TM polarization modes. In particular, they
show that TM-anomalies produced by resistive features
are relatively stronger than those caused by conductive
bodies. Conversely, TE-anomalies produced by conductive features are stronger than those caused by resistive bodies. The physics behind these observations is
very similar to what happens in a network of series and
parallel resistors. The highest resistance dominates the
voltage drop in a series array, whereas the lowest resistance in a network of parallel resistors dominates the
output voltage. In what follows we present the transformations suggested by these analogies.

where E = E1 + E2 , and R is, once again, a complex


transformation. This time we seek a magnetic field capable of producing the required E, by insisting that the
transformed impedances be 180 out of phase. We call
this equivalent impedance ZS , where the subscript S
stands for series. We look for a matrix R such that
  

Z2
ZS
=
.
(22)
R
Z1
ZS

In an array of series resistors voltages are added


and the result divided by the common current to obtain an equivalent resistance. The analogy suggests
the addition of electric fields to obtain the equivalent
impedance. The sum of E1 and E2 from Eq. (19) is
E1 + E2 = Z1 H2 + Z2 H1 .
This equation can be rewritten as


H
1
E = ( Z2 Z1 )RT R
,
H2

(20)

(21)

This condition leads to


2ZS = (Z2 Z1 ) cos + (Z2 + Z1 ) sin ,
0 = (Z2 + Z1 ) cos (Z2 Z1 ) sin ,
with given by


Z2 + Z1
tan =
.
Z2 Z 1

(23)

(24)

Replacing (24) in (23), the series impedance ZS can be


written as
ZS =

Z2 Z 1
Z 2 + Z1
=
.
2 cos
2 sin

(25)

In a situation close to 1D, Z1 = Z2 = ZS . On the other


hand, as the difference Z2 Z1 increases, the weighted
average given by (25) magnifies ZS so that it is always closer to the maximum. Consequently, the series estimator mainly depends on the higher principal
impedance rather than on the lower one.
Using Eqs. (24) and (25) we can compute Z1 and Z2
from ZS and , and vice versa. Thus, the transformation
represented in (18) can be modified to a transformation

68

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

of the form
{Z1 , Z2 , e , h } {ZS , , e , h }

(26)

ZP =

2.4. Parallel impedance


A transformation involving parallel impedance can
be obtained by following the same basic steps as in the
series case. In an array of parallel resistors we add currents, multiply by the common voltage, and proceed to
obtain the equivalent resistance. The analogy suggests
that the parallel impedance can be obtained by adding
magnetic fields. For this it is convenient to invert Eq.
(19) such that

 
 
H1
0 Y1
E1
=
,
(27)
H2
Y2 0
E2
where Y1 and Y2 , admittance functions in this case, are
simply the reciprocals of Z2 and Z1 , respectively. Now
we add magnetic fields as
H1 + H2 = Y1 E2 + Y2 E1 ,
which is equivalent to

H = ( Y2

Y1 )RT R

E1
E2

(28)

,

(29)

where H = H1 + H2 and R is, as usual, a complex transformation. Now we define an electric field capable of
producing the required H by balancing the admittance
values Y1 and Y2 . We call the equivalent impedance
ZP , where the subscript P stands for parallel. For convenience we find ZP by first finding the equivalent admittance YP = ZP1 . We look for a matrix R such that
  

Y2
YP
R
=
.
(30)
Y1
YP
This implies that


Y2 + Y1
tan =
,
Y2 Y 1

(31)

and
YP =

Y2 Y1
Y 2 + Y1
=
.
2 cos
2 sin

Since Y1 = Z21 and Y2 = Z11 , it follows that = ,


hence we write Eq. (32) as

(32)

2Z1 Z2 cos
2Z1 Z2 sin
=
.
Z2 Z 1
Z2 + Z 1

(33)

The behavior of ZP according to Eq. (33) is easier to inspect through the properties of YP as given in Eq. (32).
For instance, it can be seen that YP is an arithmetic average dominated by the maximum admittance (minimum
impedance). In other words, the parallel estimator is
more sensitive to the lower principal impedance rather
that to the higher one.
Using Eqs. (31) and (33) we can compute Z1 and Z2
from ZP and , and vice versa. Thus, the transformation
represented in (18) can be modified to a transformation
of the form
{Z1 , Z2 , e , h } {ZP , , e , h }

(34)

2.5. Seriesparallel transformation


The transformations represented by relations (26)
and (34) can be compressed into a single one by replacing in (26) by ZP . This is possible because we
can compute ZP from Eq. (33) when is known. The inverse transformation is also true, as will be shown in the
next section. We can thus write {Z1 , Z2 , e , h } {ZS ,
ZP , e , h }, or equivalently, we have the transformation
{Zxx , Zxy , Zyx , Zyy } {ZS , ZP , e , h }, connecting the
original impedance tensor elements with the new parameters.
The reason for constructing a representation that
combines the series and parallel concepts is that they
complement each other, just as the concepts of TE and
TM do in 2D. Furthermore, ZS can be identified with
ZTM and ZP with ZTE . However, it must be noticed that
neither ZS reduce to ZTM nor ZP to ZTE in 2D. As shown
later, both ZS and ZP are particular invariant averages of
ZTM and ZTE . On the other hand, e and h are required
for the transformation to be a complete representation,
in the sense that there should be no loss of information
with respect to the original tensor. However, instead of
using e and h as such it is more convenient to use
their average and their difference.
The tangent of the angular difference given in Eq.
(13) is not completely unknown in the magnetotelluric
literature; its absolute value is the impedance skew
(Swift, 1967; Vozoff, 1972, 1991). It is clear that in

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

2D there must be no angular difference between the


E and H rotations, since a rotation by the same real
angle brings both coordinate frames to the principal
system. On the other hand, in a general 3D situation it
is inevitable that the E and H coordinate frames rotate
separately, producing an angular difference. We thus
have the angular difference  = e h as a natural
3D indicator. From Eq. (13) we can write


Zxx + Zyy
.
(35)
 = arctan
Zxy Zyx
Notice that the argument of this complex quantity is
the quotient of the trace and the off-diagonal difference,
two quantities known to be invariant under rotation.
Hence, the angular difference  is also invariant under
rotation.
Let us now consider Eq. (12) for the tangent of the
angular sum. This can be identified as the impedance ellipticity (Word et al., 1971; Vozoff, 1991), a parameter
sometimes used as a 3D indicator, for it is complex and
distinct from zero for 3D, it is real for the general 2D
case, and it is equal to zero when the reference frame is
aligned with the 2D strike. Instead of the angular sum it
is convenient to examine the arithmetic average given
by


h = 1 arctan Zyy Zxx
= e +
2
2
Zxy +Zyx
(36)



B
= 21 arctan A
= 21 arctan BA2 ,
|A|

where A = Zxy + Zyx , B = Zyy Zxx , and A* denotes complex conjugate. After some algebra it can be found that
the real part of is


= 1 arctan 2 e(BA ) k ,
e()
(37)
4
2
|A|2 |B|2
which is the rotation angle defined by Swift (1967) and
by Sims (1969).
Finally, we notice that the imaginary part can be
written as a function of the same parameters A and B
as
 2

1
|A| + |B|2 + 2 m(BA )
m() = ln
(38)
8
|A|2 + |B|2 2 m(BA )
which is distinct from zero only in the 3D case, when
m(BA* ) = 0.

69

In summary, using definitions (35) and (36), the


seriesparallel transformation can be represented as
}
{Zxx , Zxy , Zyx , Zyy } {ZS , ZP , ,

(39)

Finally, we state that ZP , ZS ,  and m() must all


be invariant under rotation. We mentioned that  is an
invariant because it is the ratio of two invariant quantities. By using geometrical insight we can see that this
is how it should be:  must be an invariant because
it is the difference of two angles measured in the same
reference frame, hence it does not matter if this reference frame is rotated, the angular difference remains
the same. It can be proved that a rotation by a real angle do not alter the imaginary part of neither e nor
h , and consequently m() is also rotation-invariant.
Obviously this is not true for e(). In the case of ZP
and ZS their invariant character derives from the particular way these functions are constructed. The simple
process of adding two orthogonal components of the
same field erases the directional character of the resulting impedance. Then ZP and ZS must represent nondirectional properties of the earth. This result can be
formally established by reworking the formulas for ZP
and ZS into more familiar terms. Consider for instance
Eq. (25) for the case of ZS . It is not difficult to show that
ZS2 =

Z12 + Z22
,
2

(40)

which, using (14) and (15), can be written as the sum of


squared elements of the original tensor, ssq(Z). That is,
ZS2 =

1 2
2
2
2
+ Zyx
+ Zyy
)
(Z + Zxy
2 xx

(41)

In a similar way, the squared parallel admittance can


be written as
YP2

Y12 + Y22
2

Z2 2 .
= Z12Z +
2Z 2
2

(42)

Using the fact that Z1 Z2 = det(Z), it is straightforward


to show that
ZP2 =

2det2 (Z)
ssq(Z)

(43)

Both, det(Z) and ssq(Z) are well-known invariant quantities (Yee and Paulson, 1987; Szarka and Menvielle,
1997), hence ZS and ZP are also invariant.

70

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

2.6. The inverse transformation

C = (2ZS (ZS ZP ))1/2 cos()

The
transformation
{Zxx , Zxy , Zyx , Zyy }
} can be formalized as follows. First,
{ZS , ZP , ,
using Eqs. (35), (36), (41) and (43), the transformed
parameters are obtained from the original tensor
elements by

D = (2ZS (ZS ZP ))1/2 sin().


ZS =
ZP =

2 + Z2 + Z2 + Z2
Zxx
xy
yy
yx

1/2
,

Zyx Zxy Zxx Zyy


1/2

2 + Z2 + Z2 + Z2 )
(Zxx
xy
yy
yx


Zyy Zxx
1
= arctan
and
2
Zxy + Zyx


Zxx + Zyy
 = arctan
.
Zxy Zyx

(44)

To obtain the inverse transformation it is convenient to


define
A = Zxy + Zyx ,

B = Zyy Zxx ,

C = Zxy Zyx

and D = Zxx + Zyy .

(45)

Using Eqs. (14) and (15) it can be shown that

A = (Z1 + Z2 ) cos(2),
C = (Z1 Z2 ) cos()

and
(48)

Using the set of Eq. (45), the original tensor elements


can be recovered with the inverse transformation


ZS (ZS ZP ) 1/2
sin()
Zxx =
2


ZS (ZS + ZP ) 1/2
sin(2),

2


ZS (ZS ZP ) 1/2
cos()
Zxy =
2


ZS (ZS + ZP ) 1/2
+
cos(2),
2


ZS (ZS ZP ) 1/2
Zyx =
cos()
2


ZS (ZS + ZP ) 1/2
+
cos(2) and
2


ZS (ZS ZP ) 1/2
Zyy =
sin()
2


ZS (ZS + ZP ) 1/2
sin(2).
(49)
+
2

B = (Z1 +Z2 ) sin(2),


and

D = (Z1 Z2 ) sin().

(46)

In addition, it is not difficult to prove that


(Z1 + Z2 )2 = ssq(Z) 2 det(Z)
= 2ZS2 + 2ZS ZP

and

(Z1 Z2 )2 = ssq(Z) + 2 det(Z)


= 2ZS2 2ZS ZP .

(47)

Eq. (47) show that knowing ZS and ZP , the principal


impedances Z1 and Z2 can be calculated. Substituting
these relations into the set of Eq. (46), the variables A,
B, C and D can be expressed as
A = (2ZS (ZS + ZP ))1/2 cos(2),
B = (2ZS (ZS + ZP ))1/2 sin(2),

The existence of an inverse transformation confirms that there is no loss of information if one uses
} instead of the original set {Zxx , Zxy ,
{ZS , ZP , ,
Zyy , Zyy }.
Let us consider the 2D case as an example. When
the axes are aligned with the strike, Zxx = Zyy = 0, thus
Eq. (44) reduce to

1/2
2 + Z2
Zxy
yx
,
ZS =
2

2Zyx Zxy
ZP =
,
= 0,
 = 0.
2 + Z2 )1/2
(Zxy
yx
(50)
In the transformed domain, the 2D case with the axes
aligned to strike is equivalent to assuming =  =
0. In this case Eq. (49) for the inverse transformation

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

reduce to


Zyx

resulting equations are less compact and a lot less


intuitive.

1/2

ZS (ZS ZP )
2

1/2
ZS (ZS + ZP )
+
,
2


ZS (ZS ZP ) 1/2
=
2


ZS (ZS + ZP ) 1/2
+
and Zyy = 0.
2

Zxx = 0,

Zxy =

3. Series and parallel anomalies

(51)

It is clear that the concepts of ZS and ZP are


quite different from those that lead to the traditional
2D impedance functions Zyx (TE) and Zxy (TM). It
comes as no surprise that in 2D situations the series and parallel impedances do no reduce to the TM
and TE. In fact, ZS and ZP remain, as in 3D, independent of the coordinate system, while Zyx (TE)
and Zxy (TM), by definition, depend on direction.
However, the two sets of responses share a common feature: they are both physically sound representations of the impedance tensor. Even in nearly
2D cases, where TE and TM are adequate concepts, SP responses are no less valid alternatives for
interpretation.
Eqs. (44) and (49) represent, in compact form, the
forward and inverse transformations of the impedance
tensor, respectively. Even though ZS and ZP can be computed using the compact quadratic forms shown in (44),
they can yield unnecessary phase wrapping. For this
reason we prefer the calculation via the parameter ,
that is
tan() =
ZS =

C cos(2)
,
A cos()
A

2 cos() cos(2)

(52)
,

(53)

and
ZP =

71

det(Z)
.
ZS

(54)

In the same way, we can avoid using the quadratic


forms (47) by going back to Eq. (25), which uses the
same parameter . Again, this may avoid unnecessary
phase wrapping in the inverse computations, but the

In this section we show how the transformation performs with synthetic data in 2D and 3D situations. In
the 2D case we inspect the four transformed parameters
considering variations in frequency and space, as presented in the form of pseudo-sections. In the 3D case
we inspect only part of the transformation by comparing two apparent resistivities, xy and yx , with their
counterparts in the SP domain S and P , using plan
views at a single frequency for the comparison.
3.1. Two-dimensions
It is not difficult to predict the qualitative behavior of the series and parallel responses on the basis of
our experience with the TE and TM impedances. It is
clear that the higher impedance will dominate the series mode and the smaller the parallel. On the other
hand, we know that TM anomalies are due to both inductive and galvanic effects, while TE anomalies are
only of inductive nature. Thus, the parallel impedance,
as TE, is most sensitive to conductors, whereas conductors and resistors affect about equally the series
impedance, so as the TM. This is illustrated in Fig. 2
for the 2D model presented in Fig. 1. The responses are
compared using standard pseudo-sections of apparent
resistivity and phase. It can be observed that the TE and
parallel pseudo-sections look very much alike; in both
of them the effect of the conductive body dominates
the response. It can also be observed that the series response looks very much like the TM mode, as expected.
Contrasting series with parallel, we can see that they
complement each other in very much the same way as
TE and TM. These similarities are most welcome for
they provide a bridge between the two representations.
We can use, at least qualitatively, the same arguments
of induction to understand parallel anomalies, and of
galvanic effects to explain the series responses.
The example also shows that the intensity of series and parallel anomalies is somewhat weaker than
the conventional TM and TE. This is simply because
the former are averages of the latter, as expressed
by Eq. (50). With the averaging process we gain the

72

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

the 3D modeling algorithm developed by Mackie et al.


(1994).
3.3. Perpendicular conductive plates

Fig. 1. Two-dimensional model consisting of resistive (5000  m)


and conductive (2  m) vertical blocks embedded in a 100  m homogeneous media.

non-directional character of the series and parallel


impedances at the expense of smoothing the responses.
Nevertheless, considering that defining the geo-electric
strike is the most difficult task prior to 2D interpretation, a slight dimming of the anomalies is a small price
to pay for having pseudo-sections that remain identical
regardless of the axes directions. It is reassuring that in
spite of the smoothing effect, in a true 2D situation the
series and parallel pseudo-sections contain exactly the
same information as the traditional TE and TM modes.
3.2. Three-dimensions
In 3D there are no true TE and TM modes, since
there is no geometrically invariant direction. In general,
all four elements of the impedance tensor are different
from zero, and so are the two angular parameters in the
SP equivalency. We seek here to learn how to look at
the series and parallel impedances in terms of familiar
physical effects. To do this we use the model shown in
Fig. 3 that consists of two vertical plates placed perpendicularly to each other and immersed in an otherwise homogeneous medium of resistivity equal to
100  m. We experimented by changing the resistivity of the plates to illustrate different properties of the
responses. In all the cases the electromagnetic fields
produced at the top of the model were computed using

Fig. 4 shows a plan view of the apparent resistivity


anomaly for a period of 10 s, for the case when both
plates are conductive and have a resistivity of 2  m.
Conventional apparent resistivities in (x, y) coordinates
are labeled xy and yx , where the first subscript corresponds, as usual, to the direction of the measured electric field. The first thing to notice is that the strongest
anomaly in both xy and yx is produced by the plate situated along the measured electric field direction, which
corresponds with the main current flow. The anomaly
is due both to local induction in the plate and to current
channeling (e.g. Park, 1985). When the plates are situated across the main current flow, we notice that they
produce much weaker anomalies. In this case there is
a poor coupling for both inductive and current channeling effects. The complementary nature of xy and
yx is fairly evident in this example. The plate situated
along the x direction produces a strong local TE-like
effect in xy , while the plate situated along y causes a
strong local TE-like effect in yx . Corresponding TMlike effects are evident by lesser anomalies. It must be
mentioned that some significant effects are also present
in the diagonal elements xx and yy (not shown), particularly at the corners of the plate that is situated perpendicular to the inducing magnetic field. In summary,
the apparent resistivities xy and yx are complementary because each is most sensitive to the plate which
is situated along the electric field direction.
The accompanying series and parallel anomalies in
the lower part of the same figure can not be complementary in the same way, for any information dependent on direction has been removed by the transformation. They must complement in some other way,
for we know that both are required for the inverse
transformation. Notice that the parallel anomalies are
stronger and less elongated than the series. This observation and a comparison with xy and yx , as discussed
above, reveals that the parallel anomaly is more sensitive to the local TE-like modes of xy and yx . Similarly, the series anomalies are more sensitive to the local
TM-like modes of xy and yx . Thus, parallel apparent
resistivities are more sensitive than the series counterparts to currents that flow along boundaries, while

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383
Fig. 2. Apparent resistivity and phase pseudo-sections calculated from the conventional TETM and new seriesparallel impedances. (a) TE mode; (b) parallel impedance; (c) TM
mode; (d) series impedance.

73

74

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

Fig. 3. Three-dimensional model consisting of two vertical plates placed perpendicularly to each other at a depth of 1.0 km within a 100  m
homogeneous media. The dimensions of the plates are indicated in the images projected at the coordinate planes.

the latter responds better to currents that flow across


them. This suggests that we can see the series
impedance as a TM-like mode for 3D, and the parallel
impedance as its TE-like counterpart.
Corresponding plan views for the phase are shown
in Fig. 5. Phase anomalies reflect the same features
described above for apparent resistivities. In the conventional responses the plate situated along the main
current flow (electric field direction) produces a large
effect while the plate situated across the electric field
direction produces weak anomalies. The corresponding series and parallel phase anomalies in Fig. 5, again,
show the averaging effects discussed above for apparent resistivities. The parallel phase anomaly is stronger
than the series for both conductive plates.
The example above indicate that the series
impedance tends to reflect local TM-like effects, while
local TE-like anomalies are better reflected by the parallel impedance, all this independently of the orientation of the conductive body.
3.4. Resistive and conductive plates
We now consider another case for testing the hypothesis that we can think of the series impedance as
a TM-like mode and the parallel impedance as its TElike counterpart. The model consists of the same two
plates shown in Fig. 3, but now the plate that lies along
the x direction is more resistive (5000  m) than the
background. Fig. 6 illustrates that this model produces

a fairly small effect in xy for both conductive and resistive plates, and rather intense anomalies in yx . In
the first case, where the anomalies are small, the electric current in the x direction crosses the conductive
plate and flows parallel to the resistive one. This situation produces minor induction and current gathering in
the conductive plate, and also minor deflection of currents around the resistive one. Conversely, in the yx
case electric current flows in the y direction, moving
parallel to the conductive plate and across the resistive one. In this case, both plates are optimally oriented
for larger induction and current gathering in the conductor, as well as for larger current deflection around
the resistor, consequently both plates produce strong
anomalies.
The series and parallel anomalies are shown in the
bottom images of Fig. 6. Here we would like to examine
how xy , poorly coupled, and yx , optimally coupled,
distribute themselves into S and P . A simple inspection reveals that the minimum produced by the conductive plate is stronger in the parallel response as compared with the anomaly of the same plate in the series
response. This reinforces the view that P responds primarily to local TE-like effects of the conductive plate.
In the same way, the maximum produced above the
resistive plate is larger for the series response as compared with the parallel anomaly of the same plate. This
in turn confirms that S responds more to the local
TM-like effects of the resistive plate. In this case the
difference is small and somewhat difficult to observe

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383
Fig. 4. Plan views of apparent resistivity on top of the 3D model of Fig. 3, for the case when both plates are conductive (2  m). The four images represent anomalies for a period
of 10 s, as derived from the xy and yx conventional impedances and from the series and parallel impedances. (a) XY; (b) YX; (c) series; (d) parallel.

75

76
J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

Fig. 5. Phase response calculated from the same set of impedances used in Fig. 4. (a) XY; (b) YX; (c) series; (d) parallel.

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383
Fig. 6. Plan views of apparent resistivity on top of the 3D model of Fig. 3, for the case when the plate oriented along x direction is resistive (5000  m), and that oriented along y
direction is conductive (2  m). The four images represent anomalies for a period of 10 s, as derived from the xy and yx conventional impedances and from the series and parallel
impedances.
77

78

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

in our example, nevertheless it is there, as it should be,


since the series impedance is dominated by the higher
resistivity.
The above numerical simulations confirm the expectation that we can think of the series impedance as a
TM-like mode and of the parallel impedance as its TElike counterpart, and that series and parallel anomalies
do no depend on the degree of directional coupling, but
only on geometry and electrical conductivity.
Berdichevsky et al. (1998) have demonstrated that
the effects produced by conductors are better sensed
by the apparent resistivity related to the longitudinal
mode (local TE). They have also shown that the effects produced by resistive bodies can only be sensed
by the apparent resistivity resulting from the transversal mode (local TM). Our calculations regarding xy
and yx confirm these findings. Accordingly, our identification of S with the local TM and of P with the
local TE extend their findings to the series and parallel
responses. This means that SP qualitative analyses of
curves and maps could benefit from the accumulated
experience of years of TE and TM analyses. However,
whereas this is a desired feature for any new prospect
for 3D interpretations, the new parameters need to be
tested quantitatively in a variety of situations before
they can be considered as suitable alternatives. Among
other issues, what is at stake is the usefulness of the
formulation for the inversion of data. We explore this
subject in the following section.

4. 2D inversion of SP impedances
As shown before, TE and TM traditional responses
from 2D models are easily converted to SP quantities, and so are their partial derivatives with respect
to model parameters. This makes it possible for existing inverse codes to be readily transformed into corresponding inverse codes for SP data. We adapted
the GaussNewton inversion algorithm by Rodi and
Mackie (2001) and applied it to both, synthetic and
field data. In the first case we explore how the new parameters perform in a 3D situation, by applying the 2D
inversion algorithm to the SP responses of a hypothetical 3D model. Because SP impedances are invariants
under rotation, they do no change with the choice of
strike. We define profiles across a 3D model and simply assume that they are normal to strike. For the field
data experiment we apply the 2D inversion to the COPROD2 data set in two modalities of SP data. In one
we use the processed TE and TM impedances as used
by other workers, and in the other we use the original
full tensor prior to corrections.
4.1. 3D synthetic data
In this section we examine the ability of the SP
impedances to produce reliable models when interpreted with 2D inversion algorithms. Synthetic data
produced by a 3D model were selected along three

Fig. 7. The 3D model consisting of two parallel blocks 1.2 km deep in a 100  m homogeneous media. The resistive block is 5000  m and the
conductive one is 2  m. The dimensions of the blocks are indicated in the images projected at the coordinate planes; 2D inversion of SP data
was performed along profiles A, B and D.

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

profiles crossing the anomalous bodies as shown


in Fig. 7. The model consists of a resistive block
(5000  m) and a conductive block (2  m) embedded
in a 100  m homogeneous half-space. Line A, coinciding with the axes of symmetry of the model, has the
best conditions for 2D inversion. In contrast, line B runs
along the edges of the bodies and line D draws a diagonal across, both awkwardly suited for conventional
2D interpretations. The 3D forward modeling code of
Mackie et al. (1994) was used in the computations. All
four elements of the impedance tensor were included
to compute the series and parallel impedances.
Fig. 8 shows the results of the inversion and the
misfit attained for each site. Input data consisted of apparent resistivity and phase with 5% random noise. We
entered data at nine periods logarithmically spaced between 0.1 and 1000 s. Both, series and parallel modes
were inverted simultaneously using a regularization
factor = 3.
The solution obtained for line A recovers quite well
the geometry and resistivity of both anomalies, although the bottoms of the blocks are not as well resolved as their tops. The bottom of the resistor is the
worst resolved, probably because of the screening effect of the relatively more conductive surrounding media. A spurious conductive stripe appears beneath sites
4 and 6, possibly caused by numerical inconsistencies
during the inversion process. The misfit remains close
to 2 standard deviations (S.D.) in most of the sites,
except in sites 16 and 47, where the misfit increases
to 3 S.D. As 5% uncertainties were assumed for both,
resistivity and phase data, one standard deviation correspond to a 5% misfit.
Results for line B show the consequence of applying
2D inversion to data with strong 3D effects. The closeness of this line to the edge of both blocks produces a
2D image of poorer quality. However, both anomalies
are still discernible from the surrounding media. The
lower boundary of the conductive block is not very well
resolved, with the conductive anomaly spuriously extending deep into the half-space. Surprisingly, the geometry of the resistive block is better estimated. Misfit
remains around 4 S.D. (20%) for most of the sites except for sites 34 and 50 where it increases to 7 S.D.
The solution for line D produces a very good image of both anomalies, in spite of the systematic 3D
effect caused by the angle made between the lateral
edges of the blocks and the 2D strike assumed for the

79

inversion. We consider that this is a consequence of the


non-directional character of the SP impedances. The
resistive anomaly is the worst resolved, both in geometry and resistivity, probably because it is more sensitive
to the noise added to the data. An experiment with noise
free data produced an image with quality comparable
to that obtained for line A. The misfit remained within
the range of 24 S.D. for all sites.
The above experiments with 3D synthetic data reinforce the view that SP impedances may play in
3D the role that TE and TM impedances play in 2D.
They show that series and parallel impedances are
amenable to quantitative interpretations without the
need to determine strike directions or make tensor modifications. While this is also true for other rotational
invariant responses, SP impedances also complement
each other qualitatively, in a similar way as TE and TM
impedances do in 2D geometries, an interesting property that allows the extension of the familiar TE and
TM-like behavior to 3D interpretation.
4.2. COPROD2 eld data
We present here the application of the 2D inverse
code to SP impedances of the COPROD2 data set described by Jones (1993). We show results from two
experiments: first we inverted series and parallel quantities computed with Eq. (50), using only TE and TM
impedances. Then, in a second example, we inverted
SP data computed using the full impedance tensor as
required by Eq. (44).
The COPROD2 data set consists of 35 sounding along a 400 km east-west profile in southern
Saskatchewan and Manitoba, Canada. We first considered the TE and TM modes as identified by Jones
(1993) and computed the corresponding SP apparent
resistivities. Many data points were removed as recommended by the warnings in the data file, following
the practice of other workers that have used the set. A
total of 2716 data points were used in the inversion.
The model shown in Fig. 9a reproduces the observed
data with a misfit of about 5%. The strongest anomaly
near the center of the profile is a feature known as the
North American Central Plains conductivity anomaly
(NACP) and to the eastern extreme is a second basement anomaly, associated with the Thomson Nickel
Belt (TOBE). Both anomalies are well recovered from
the SP data as computed from the provided TE and

80

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

Fig. 8. Resistivity models recovered by the 2D inversion of: (a) line A, (b) line B and (c) line D. Misfit at every site is indicated at the top of
each model. Outlines of the target blocks are indicated with dashed line.

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

81

Fig. 9. Resistivity models resulting from the 2D inversion of the SP impedances computed from the COPROD2 data set. (a) SP data computed
from TE and TM provided impedances. (b) SP data computed from the full impedance tensor. NACP and TOBE anomalies are well recovered
in both SP data sets. Both resulting models are also comparable with those discussed by Jones (1993).

TM impedances. The resulting model compares well


with those discussed by Jones (1993).
In the second experiment we computed SP apparent resistivities using all the elements of the tensor. In
this case the data were not edited to remove bad points.
We simply selected every other data point, without any
care about their quality. For many practical purposes it
could very well be considered a different data set. In
this case the number of data was 2920. The misfit was
of about 10%, twice as large as before, partly because
of the increased scattering of the unedited data. The
resulting model is shown in Fig. 9b. It can be observed
that both anomalies are again well recovered.

The first of the above two applications to the roughly


2D COPROD2 data set, illustrates that inverting SP
impedances, as computed from the TE and TM modes,
produces models that compare well with those obtained
by a straight inversion of the traditional TE and TM
modes. The second experiment, on the other hand, illustrates that SP impedances, straightforwardly computed from the full tensor without elaborated corrections, produce again results that compare well with
those obtained from the more elaborated TETM pair.
The application of the new approach to cases with more
severe 3D-distorted data, will be a future task, as will
be the delimitations of using only the SP impedances

82

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383

as opposed to the complete set that includes the angular


parameters.

5. Conclusions
The magnetotelluric impedance tensor can be rigorously transformed into an equivalent representation
with clear-cut physical features. The bases of the representation are the traditional series and parallel equivalencies of network theory, which we apply to the
impedance tensor through appropriate adaptations. Series and parallel equivalencies are sound concepts that
lead to response functions sensitive to global properties
independently of the orientation of the coordinate axes.
The two angular parameters and  that complete the
mathematical analysis also have physical significance
useful in 3D environments. Investigation of these angular functions and their properties is a future task.
From a mathematical point of view, the transformation operates on the original tensor elements in the traditional xy domain, and produces four complex quantities in what can be called the SP domain, where S
stands for series and P for parallel. The existence of an
inverse transformation for going from the SP to the
original x-y domain assures that there is no loss of information when going from one representation to the
other. This means that the same information is available in either representation. However, the response
functions in the SP domain provide physical insights
beyond those offered by the raw tensor elements in the
original x-y domain. Furthermore, because of the complementary nature of the two SP impedances, they
may play in 3D the role that TE and TM impedances
play in 2D.
The claim that SP impedances could play, regardless of dimensionality, the roles that TE and TM play
in 2D, was explored experimenting with 2D inversion of both synthetic and field data. Our experiment
with 3D synthetic data, confirm that series and parallel impedances are amenable to quantitative interpretations. It shows that SP impedances are a viable alternative that can be used in routine inversions of magnetotelluric data, whether in two or three dimensions.
The applications of the 2D inversion code to field
data are also encouraging. The two models obtained
under different circumstances are close enough to each
other, and also to others reported in the literature.

While the main purpose of the series and parallel


transformation is not the 2D inversion of 3D data, it
represents a viable alternative that competes with other
rotation invariants, with the benefit of a close relation
to the physics of inductive and galvanic effects. We
believe that the potential of our approach resides in that
the SP impedances, next to complementing each other,
by themselves summarize the impedance information
of the tensor, and that along with the angular functions
and , represent a complete set that fully represent
a general 3D situation. This may provide some guide
for the selective choice of the response functions to use
in 3D inversions.
Acknowledgments
The authors are grateful to Alan G. Jones, Laszlo
Szarka and one anonymous reviewer for their valuable
suggestions to improve this paper, as well as to Randy
Mackie and two anonymous reviewers who revised a
previous version of the manuscript. We are also grateful
to Keeva Vozoff, whos challenging questions, a couple
of years ago, motivated our work on the tipper and
as a logical consequence, the present work. Financial
support for this work was provided by Conacyt grant
No. 25792-T.
References
Bahr, K., 1988. Interpretation of the magnetotelluric impedance tensor: regional induction and local telluric distortion. J. Geophys.
62, 119127.
Bahr, K., 1991. Geological noise in magnetotelluric data: a classification of distortion types. Phys. Earth Planet. Inter. 66, 24
38.
Berdichevsky, M.N., Dimitriev, V.I., Pozdnjakova, E.E., 1998. On
two-dimensional interpretation of magnetotelluric soundings.
Geophys. J. Int. 133, 585606.
Bostick, F.X., Smith, H.W., 1962. Investigation of large-scale inhomogeneities in the Earth by the magnetotelluric method. In:
Vozoff, K. (Eds.). Magnetotelluric methods. Geophysics Reprint
Series, No. 5, Society of Exploration Geophysicists, 1986, Tulsa,
Oklahoma, pp. 148155.
Cagniard, L., 1953. Basic theory of the magneto-telluric method of
geophysical prospecting. Geophysics 18, 605635.
Cantwell, T., 1960. Detection and analysis of low frequency magnetotelluric signals. Ph.D. Thesis, Massachusetts Institute of Technology.
Cevallos, C., 1986. Magnetotelluric interpretationanother approach. Ph.D. Thesis, Macquarie University.

J.M. Romo et al. / Physics of the Earth and Planetary Interiors 150 (2005) 6383
Chave, A.D., Smith, J.T., 1994. On electric and magnetic galvanic distortion tensor decompositions. J. Geophys. Res. 99,
46694682.
Counil, J.L., Le Mouel, J.L., Menevielle, M., 1986. Associate and
conjugated direction concepts in magnetotellurics. Ann. Geophysicae. 4 (B2), 115130.
Eggers, D.E., 1982. An eigenstate formulation of the magnetotelluric
impedance tensor. Geophysics 47, 12041214.
Fischer, G., Masero, W., 1994. Rotational properties of the magnetotelluric impedance tensor: the example of the Araguainha
impact crater. Brazil Geophys. J. Int. 119, 548560.
Groom, R.W., Bailey, R.C., 1989. Decomposition of magnetotelluric
impedance tensors in the presence of local three-dimensional
galvanic distortion. J. Geophys. Res. 94, 19131925.
Groom, R.W., Bailey, R.C., 1991. Analytic investigations of the effects of near-surface three-dimensional galvanic scatterers on MT
tensor decomposition. Geophysics 56, 496518.
Jones, A.G., 1993. The COPROD2 dataset: tectonic setting, recorded
MT data, and comparison of models. J. Geomagn. Geoelectr. 45,
933955.
Larsen, J.C., 1977. Removal of local surface conductivity effects
from low frequency mantle response curves. Acta Geod. Geophys. Acad. Sci. Hung. 12, 183186.
LaTorraca, G.A., Madden, T.R., Horringa, J., 1986. An analysis of
the magnetotelluric impedance tensor for three-dimensional conductivity structures. Geophysics 51, 18191829.
Lilley, F.E.M., 1993a. Magnetotelluric analysis using Mohr circles.
Geophysics 45, 833839.
Lilley, F.E.M., 1993b. Tree-dimensionality of the BC87 magnetotelluric data set studied using Mohr circles. J. Geomagn. Geoelectr.
45, 11071113.
Lilley, F.E.M., 1998. Magnetotelluric tensor decomposition. Part I.
Theory for a basic procedure. Geophysics 63, 18851897.
Mackie, R.L., Smith, J.T., Madden, T.R., 1994. Three-dimensional
electromagnetic modeling using finite difference equations: the
magnetotelluric example. Radio Sci. 29, 923935.
Park, S.K., 1985. Distortion of magnetotelluric sounding curves by
three-dimensional structures. Geophysics 50, 785797.
Rodi, W., Mackie, R.L., 2001. Nonlinear conjugate gradients algorithm for 2D magnetotelluric inversion. Geophysics 66, 174187.
Rokityansky, I.I., 1961. On the application of the magnetotelluric
method to anisotropic and inhomogeneous masses. In: Vozoff,
K. (Ed.), Magnetotelluric methods. Geophys. Reprint Ser. 5, Soc.
Expl. Geophys., pp. 143147.

83

Romo, J.M., Gomez-Trevino, E., Esparza, F., 1999. An invariant


representation for the magnetic transfer function in magnetotellurics. Geophysics 64, 14181428.
Sims, W.E., 1969. Methods of magnetotelluric analysis. Ph.D. Thesis, University of Texas at Austin.
Smith, J.T., 1995. Understanding telluric distortion matrices. Geophys. J. Int. 122, 219226.
Smith, J.T., 1997. Estimating galvanic-distortion magnetic fields in
magnetotellurics. Geophys. J. Int. 130, 6572.
Spitz, S., 1985. The magnetotelluric impedance tensor properties
with respect to rotations. Geophysics 50, 16101617.
Swift Jr., C.M., 1967. A magnetotelluric investigation of an electrical
conductivity anomaly in the southwestern United States. Ph.D.
Thesis, Massachusetts Institute of Technology.
Szarka, L., Menvielle, M., 1997. Analysis of rotational invariants
of the magnetotelluric impedance tensor. Geophys. J. Int. 129,
133142.
Szarka, L., Menvielle, M., Spichak, V.V., 2000. Imaging properties
of apparent resistivities bases on rotational invariants of the magnetotelluric impedance tensor. Acta Geod. Geophys. Hung. 32,
149175.
Tikhonov, A.N., 1950. On determining electrical characteristics of
the deep layers of the Earths crust. In: Vozoff, K. (Ed.), Magnetotelluric Methods. Geophys. Reprint Ser. 5, Soc. Expl. Geophys.,
pp. 23.
Vozoff, K., 1972. The magnetotelluric method in the exploration of
sedimentary basins. Geophysics 37, 98141.
Vozoff, K., 1991. The magnetotelluric method. In: Nabighian, M.N.
(Ed.), Electromagnetic Methods in Applied Geophysics 2, Application Soc. Expl. Geophys., pp. 641711.
Weaver, J.T., Agarwal, A.K., Lilley, F.E.M., 2000. Characterization
of the magnetotelluric tensor in terms of its invariants. Geophys.
J. Int. 141, 321336.
Word, D.R., Smith, H.W., Bostick Jr., F.X., 1971. Crustal investigation of the magnetotelluric tensor impedance method. In: Heacock, J.G. (Ed.). The structure and physical properties of the
earths crust. Am. Geophys. Union, Geophys. Monogr. Ser. 14,
pp. 145167.
Yee, E., Paulson, K.V., 1987. The canonical decomposition and its
relationship to other forms of magnetotelluric impedance tensor
analysis. J. Geophys. 61, 173189.
Zhang, P., Roberts, R.G., Pedersen, L.B., 1987. Magnetotelluric
strike rules. Geophysics 52, 267278.

Anda mungkin juga menyukai