Professional Documents
Culture Documents
A New Formula To Compute Apparent Resistivities FR
A New Formula To Compute Apparent Resistivities FR
net/publication/228077453
CITATIONS READS
6 132
2 authors:
Some of the authors of this publication are also working on these related projects:
Ph.D: Towards Geologic inversion using machine learning: Petrophysically and Geologically Guided Geophysical Inversion View project
A numerical upscaling framework for the quasi-static Maxwell's equations View project
All content following this page was uploaded by Douglas W. Oldenburg on 23 May 2014.
INTRODUCTION 0IH0
a = , 共1兲
Among the marine electromagnetic 共EM兲 methods used to in- 4 R 2B
vestigate resistivity structures below the seafloor, the magnetomet-
ric resistivity 共MMR兲 method has unique characteristics 共Edwards where B is the measured azimuthal magnetic field at the receiver,
and Nabighian, 1991兲. The method essentially involves measuring R is the radial distance between the Rx and Tx wire, 0 is the per-
the magnetic field associated with manmade, noninductive 共low- meability of free space and nonmagnetic seafloor, I is the current
frequency or pseudo-dc兲 current flow energized into the seawater strength in the transmitter wire, H denotes the thickness of the sea-
and seafloor through two vertically separated electrodes 共bipole兲. water, and 0 is the resistivity of the seawater, which is presumably
The magnetic field measured at the ocean-bottom magnetometer known. Two assumptions are required in the derivation of this
depends upon the total current flow at the seafloor and in the sea- equation: First, the range R must be large compared to the sea
Manuscript received by the Editor August 27, 2004; revised manuscript received October 6, 2005; published online May 19, 2006.
1
Formerly University of British Columbia, Vancouver, Canada: presently Schlumberger-EMI Technology Center, 1301 South 46 Street, Building 300, Rich-
mond, California 94804. E-mail: jchen16@slb.com.
2
University of British Columbia, Department of Earth and Ocean Sciences, 6339 Stores Road, Vancouver, British Columbia, Canada. E-mail:
doug@eos.ubc.ca.
© 2006 Society of a Exploration Geophysicists. All rights reserved.
G73
G74 Chen and Oldenburg
depth H; second, the integrated conductivity of the sea layer 0H trode C is placed at the interface z = zs between layers s and s + 1
must be large compared with the parameter 1R, where 1 is sea- to simplify the mathematics. Each layer has a constant conductiv-
floor conductivity 共i.e., H ⬍ R ⬍ 0 /1H兲. These assumptions are ity j with thickness h j and a magnetic permeability equal to free
necessary so that the bipole current is channeled out to relatively space. There are a total of N − 1 interfaces, with N as the termi-
large distances by the sea. Typical values of conductivities are 0 nating half-space. In the source-free region, the magnetic field B
= 3.3 S/m and 1 = 0.01 to 2.0 S/m. Therefore, the Chave et al. obeys
共1991兲 method for calculating apparent resistivity works best in
a shallow ocean, as in the MOSES experiment 共Edwards et al., 1
1985兲, where H = 200 m. However, in a deep-sea survey, R/H ⵜ⫻ ⫻ B = 0. 共3兲
ranges from 0.1–3.0 共Evans et al., 1998; Evans et al., 2002兲. As a
result, the first assumption fails, and the formula will not provide a
The problem is axisymmetric, and B has only an azimuthal com-
good approximation 共see Figure 1兲. Clearly, the apparent-resist-
ponent in cylindrical coordinates r, ,z. For simplicity, we use B to
ivity curve provides no indication of the layered-resistivity struc-
represent the azimuthal component in the following derivations.
ture of the seafloor, especially the deep, low-resistivity zone.
Expanding equation 3 and neglecting the conductivity, because
The second formula is given in Wolfgram et al. 共1986兲, where a
is a constant in each layer, yields
is obtained by
0I 0 H 2B 1 B 1 2B
a = − 0 . 共2兲
冑
2RB H + R22
r2
+
r r
−
r2
B +
z2
= 0. 共4兲
The formula ignores the effect from the electrode on the sea sur- Following the Hankel transform method 共Edwards and Nabig-
face by assuming that the top electrode is located at infinity. To de- hian, 1991兲, we define a Hankel transform pair as
rive this equation, R must be smaller than H 共i.e., R ⬍ H兲. This as-
冕 ⬁
sumption limits the formula’s use in deep-sea MMR because the
Tx-Rx separation R can be greater than H. As Figure 1b illustrates, B̃共,z兲 = rB共r,z兲J1共r兲dr 共5兲
the Wolfgram et al. 共1986兲 method offers a poor indication of 0
three-layer structure. When R/H is small 共0.1 in this example兲, the
apparent resistivity approaches the true value 共7 ⍀.m兲; otherwise, and
the formula provides an inadequate approximation.
冉 冊 冉 冊
B̃ j兩共,z兲兩z=z j = B̃ j+1兩共,z兲兩z=z j 共9兲 B̃+共,H兲 = . 共14兲
2 0 0 −2H
1+ − 1− e
1 1
and
The total field resulting from the bipole is B̃共 ,H兲 = B̃+ + B̃−. The
Ẽrj 兩共,z兲兩z=z j = Ẽrj+1兩共,z兲兩z=z j . 共10兲
denominator term in both equations can be expanded in a binomial
approximation as
In addition, we have the constraints
1
ŨN = 0, 共11兲 冉 冊 冉 冊
1+
0
1
− 1−
0 −2H
1
e
冉 冊
冤 冥
i.e., there are no upcoming fields in the last layer, and 0
1−
兩Ez兩z=0 = 0. 共12兲 1 1 −2H
冉 冊 冉 冊
= 1+ e + HOT , 共15兲
0 0
1+ 1+
Constraint 12 requires that there is no current crossing the air- 1 1
seawater interface. Therefore, the 2N unknown coefficients can be
determined from the 2共N − 1兲 + 2 equations. Further details can where HOT stands for higher-order terms of e−2H. Substituting
be found in Chen and Oldenburg 共2004兲. equation 15 into B̃+ and B̃− yields
Apparent-resistivity formulas 0
⬁
0I 1
Suppose the Tx bipole extends from the sea surface to the sea
bottom 共length L = H兲 and the seafloor is a uniform half-space
B̃共,H兲 =
2
兺 ne−nH ,
0 n=0
共16兲
with resistivity 1 共see Figure 3兲. Using the wavenumber method 1+
1
above, we can compute the magnetic field at the seafloor 共depth H兲
resulting from one semi-infinite wire terminating at the sea surface
where
共the top electrode is assumed negative兲 in the wavenumber do-
main,
冦 冧
1 n=0
0 n = − 2␥ m−1
n = 2m − 1 共17兲
4 e −H
0I 1 ␥ 共1 + ␥兲 n = 2m, m = 1,2, . . .
m−1
冉 冊 冉 冊
B̃−共,H兲 = − , 共13兲
4 0 0 −2H
1+ − 1− e
1 1 and
Figure 2. Schematic of a semi-infinite wire source buried in a lay- Figure 3. A two-layer 共including seawater兲 model for defining ap-
ered earth. parent resistivity.
G76 Chen and Oldenburg
Transforming back to the spatial domain and making use of the in-
冉 冊
R 0 2 1+␥
冑 冉冊 冑 冉 冊
tegral identity F2 , = − + ␥,
H 1 R 2
R 2
1+ 1+
冕 冋 册
⬁
H 2H
1 nH
e−nHJ1共R兲d = 1− 共19兲 共23兲
0 R 冑R 2
+ 共nH兲2
and so on. Unfortunately, this is not a good approach because of
yields the oscillating nature of the coefficients n, illustrated in Figure 5,
where ratio 0 /1 = 0.01 is used. Interestingly, if we look at F1 in
equation 22, we find that the Wolfgram et al. 共1986兲 formula ig-
冦 冤 冑 冉 冊 冥冧
⬁
0I 0 1 nores 共1/冑1 + 共R/H兲2兲 − 1; in other words,
B共R,H兲 =
2R 0 + 1
1+ 兺 n 1− 2
.
冉 冊
n=1 R
1+ R 0 1
冑 冉冊
nH Fw , = , 共24兲
H 1 R 2
共20兲 1+
To obtain a simple approximate relation between B and 0 /1, we
H
define F共R/H, 0 /1兲, which is a function of R/H and 0 /1, to rep-
resent the content within the brackets: which does a better job than F1 to approximate to the infinite se-
ries.
Following that insight, we begin to develop a formula using two
冉 冊
冤 冑 冉 冊冥
⬁ terms in the series expansion 共n = 2兲. As we note from Figure 5, F2
R 0 1
F ,
H 1
=1+ 兺 n
n=1
1−
R 2
. 共21兲 is not a good approximation, but when we delete terms ␥ − 共 ␥ /
冑1 + 共R/共2H兲兲2兲 from equation 23, the remainder performs better
1+
nH in terms of getting closer to the true F when R/H increases. As a
Function F共R/H, 0 /1兲 is displayed in Figure 4, where R/H ranges further modification, we replace unity by ␣ in the second term, so
from 0.01 to 10 while the ratio 0 /1 is 0.001, 0.01, 0.1, and 0.5, re- our expression has the form
冉 冊
spectively. When R/H ⱕ 0.2, the function F is independent of
0 /1. In addition, when 0 /1 ⱕ 0.01, F depends only on R/H. We R 0 2 ␣
冑 冉冊 冑 冉 冊
F␣ , = − . 共25兲
will take advantage of this feature to develop a simple relationship H 1 R 2
R 2
in the following derivation. 1+ 1+
One approach to obtain an approximate form of equation 21 is to H 2H
truncate the infinite series in that equation at some value of n and
represent the result as Fn共R/H, 0 /1兲. For example, The unknown ␣, a function of R/H and 0 /1, can be obtained by
fitting to the curves shown in Figure 4. To conveniently pick an ␣
冉 冊
value, Figure 6a shows the lookup curve of ␣ versus R/H and
R 0 2 0 /1, and Figure 6b is a contour map for ␣. Alternatively, ␣ can be
冑 冉冊
F1 , = − 1, 共22兲
H 1 R 2 computed through an explicit expression
1+
H
冦 冧
R value of ␣ is not extremely sensitive to 0 /1. From Figure 6a, even
1.0 if ⬍ 0.2 where 0 /1 varies almost three decades 共0.001 to 0.5兲, ␣ only
H
冉冊 冉冊
␣= , changes in the range of 0.75–1.0. This means that even a poor esti-
R 0
˜␣共x,y兲, x = log10 , y = log10 otherwise mation of 0 /1 will not make a significant impact on selecting an
H 1 ␣ from the lookup curve. In this sense, determination of ␣ in equa-
共26兲 tion 26 is robust and stable. Finding a good truncated function F␣,
we can define the corresponding apparent resistivity by
where
冦冑 冉 冊 冑 冋 册冧
0I 0 2 ␣
˜␣共x,y兲 = 共− 0.023 − 0.055y + 0.101y 2兲 + 共− 0.012 a = − − 0 .
2R B R 2
R 2
− 0.029y + 0.005y + 0.087y 兲x + 共0.023
2 3 1+ 1+
H 共2H兲
+ 0.080y − 0.124y 2兲x2 + 共0.021 + 0.064y 共28兲
− 0.136y 2兲x3 + 共0.005 + 0.016y + 0.964y 2兲x4 .
Equation 28 is actually the simplest situation encountered in a
共27兲 practical survey. Because of the bathymetry of the ocean bottom,
the lower electrode of the transmitter might be located at a depth
All of the coefficients are obtained by fitting a polynomial of order different from the depth of the receiver. A general model can be
four in x and order three in y in a least-squares sense. presented by locating both the lower electrode and the receiver at
In general, we do not know exactly what 0 /1 is; fortunately, the different depths in the seawater. Depending upon the relative
depth, we consider the problem in two cases, as shown in Figure 7.
Case A
In this case, the magnetometer is located at depth Zr共Zr ⱕ L
ⱕ H兲, simulating the situation in which the transmitter is near the
sea bottom while the receiver is at a hill because of bathymetry of
the seafloor. Generally, we follow the same procedure as above to
derive the magnetic field. However, this derivation is more compli-
cated because we have two additional depths, Zr and L, and have
more combinations among Zr, L, and H. More importantly, truncat-
ing the infinite series in the spatial domain has proven unsatis-
factory because of its oscillating behavior. We have resorted to a
slightly different method to find an optimum ␣共R/H, 0 /1兲 in this
case. First, we truncate the infinite series directly in the wavenum-
ber domain and retain exponential terms up to n = 2 for the mag-
netic field 共e.g., we only have exponential terms such as e−H, e−2H,
e−Zr, e−共2H−Zr兲, etc.兲. Second, we transform these related terms into
the spatial domain using the integral identity 共equation 19兲. Finally,
assembling them yields the total magnetic field
B共R,Zr兲 =
0I 1 0
4R 0 + 1 1
冋
A1 + A2 , 册 共29兲
−
冑R 2
4H
+ 共2H兲 2
−
冑R 2
Zr + L
+ 共Zr + L兲2 B共R,Zr兲 =
0I 1 0
4R 0 + 1 1
冋
A4 + A5 , 册 共35兲
2H − Zr − L
− 共30兲
冑R2 + 共2H − Zr − L兲2 where the coefficients are
冑 冉冊 冑 冋 册
A1 = 2
+ 2
R R 0IA5
1+ 1+ A6 = 1 − . 共39兲
Zr 共2H − Zr兲 4RB
␣ L − Zr
冑 冋 册
+ +
R 2 冑R 2
+ 共L − Zr兲2
After correction, coefficient A4 reads
1+
共2H + Zr − L兲 2 2
冑 冉冊 冑 冋 册
A4 = +
2␣ ␣ R 2
R 2
冑 冋 册 冑 冋 册
− − 1+ 1+
R 2
R 2
Zr 共2H − Zr兲
1+ 1+
共2H兲 共Zr + L兲 ␣ ␣
冑 冋 册 冑 冋 册
− 2
− 2
2H − Zr − L R R
− , 共34兲
冑R 2
+ 共2H − Zr − L兲 2 1+
共L + Zr兲
1+
共2H + L − Zr兲
Zr − L 2H − Zr − L
where ␣ can be approximately determined by equation 26. The ef- − − 共40兲
fect of ␣ on A2 and A3 is very small and can be neglected. 冑R2 + 共Zr − L兲2 冑R2 + 共2H − Zr − L兲2 .
Apparent resistivity for marine MMR G79
As a consistency check, when Zr = L ⫽ H, cases A and B should The transmitters are located 2700 m below the sea surface, while
be identical. In other words, we will have A1 ⬅ A4 and A2 ⬅ A5. the magnetometer is at 2500 m depth. This simulates the case
This is true for our derivations. where the magnetometer is situated on the ridge axis without tak-
ing the bathymetry into account. Figure 8a shows the azimuthal B
versus the normalized distance R/H for the on-axis magnetometer.
Verification For comparison, the off-axis magnetic field is plotted also. Surpris-
As a verification, we use the new formula derived in equation 28 ingly, the amplitudes for the on-axis receiver are much larger
to compute the apparent resistivity for the three-layer model shown 共about one order of magnitude兲 than those for the off-axis receiver.
in Figure 1a. We assume 0 /1 = 0.01, and corresponding values of This significant difference results purely from the vertical shift of
␣ are computed from equation 26. The new sounding curve cor- receiver location. Without taking the transmitter-receiver geomet-
rectly reveals the three-layer structure and gives a good approxi- ric difference into account, the derived resistivities of the on-axis
mation to both the first layer and basement resistivity. More impor- response varies from 0.2–15 ⍀.m, while the off-axis response
tantly, there is no restriction on the normalized distance R/H. The yields values from 10–18 ⍀.m 共solid dots and circles in Figure 8b,
new formula works over a wider range 共0.01 ⱕ R/H ⱕ 10兲. obtained using Wolfgram et al., 1986兲. Obviously, these results are
unsatisfactory. When we use the new formula to obtain the appar-
ent resistivity, both curves 共solid and dashed lines in Figure 8b兲 of-
APPLICATIONS fer a good approximation to the model value 共10 ⍀.m兲.
The new apparent-resistivity formula provides a useful tool to
address some practical issues that arise in a marine MMR survey.
First, we look at the effect of the relative vertical offset between Effect of normalized range
the transmitter and the receiver. We then show that it is necessary Analysis of the apparent-resistivity curve reveals the importance
to acquire data over a large range of R/H. Finally, we apply the de- of the normalized distance on the data interpretation and survey de-
rived formula to field data from the East Pacific Rise.
sign. As shown in Figure 9a, the apparent-resistivity curve ob- range兲 clearly shows a three-layer model of the seafloor. The inset
tained with a normalized range 0.04 ⱕ R/H ⱕ 4 共labeled full is the true 1D model. When we carry out a 1D inversion based
upon a generalized cross-validation tech-
nique with full-range data, the recovered
structure reveals the lower-resistivity
layer in a three-layer model 共see Figure
9b兲. Conversely, if we only use the data in
the normalized range 0.3 ⱕ R/H ⱕ 4 共la-
beled small range兲, the recovered model
from a 1D inversion indicates a two-layer
structure. While this can be explained
easily from the apparent-resistivity curve,
it is not as obvious if we look at the
magnetic-field curve. This simple exam-
ple suggests that if the normalized dis-
tance is not covered widely enough, we
will likely miss the shallow-resistivity in-
formation, resulting in a poor 1D model.
In this regard, choice of the normalized
range has a definitive impact on the sur-
Figure 9. Effect of the normalized range on data inversion and survey design. 共a兲 apparent re- vey design.
sistivities for the three-layer seafloor model. 共b兲 Recovered 1D resistivities with a small range
共R/H = 0.3–4兲 and a full data range 共0.04–4兲.