Sei sulla pagina 1di 8

Geophysical Prospecting, 2012, 60, 153160 doi: 10.1111/j.1365-2478.2011.00948.

On the computation of the electrical potential inside a


horizontally-layered half-space
Alexis Maineult1 and Jean-Jacques Schott2
1 Institut
de Physique du Globe Sorbonne Paris Cite, CNRS et Universite Paris Diderot, 1 rue Jussieu, 75005 Paris, France, and 2 Institut de
Physique du Globe, Universite de Strasbourg/EOST, 5 rue Descartes, 67000 Strasbourg, France

Received September 2010, revision accepted December 2010

ABSTRACT
A new formulation is proposed for the electrical potential developed inside a
horizontally-layered half-space for a direct current point-source at the surface. The
recursion formula for the kernel coefficient in the potential integral is simpler than the
generally used two-coefficient recursion. The numerical difficulties that may occur
during the computation of the integrals and near the source axis are examined and so-
lutions are proposed. The set of equations permits a stable and accurate computation
of the tabular potential everywhere in the medium.

Key words: Electrical potential, Geoelectrical methods, Layered earth, Resistivity


modelling.

2005, and references therein), the boundary element method


INTRODUCTION
(e.g., Okabe 1981; Xu, Zhao and Yi 1998; Ma 2002; Blome,
Geoelectrical prospecting is a well-known and frequently Maurer and Schmidt 2009) and the popular finite-element
used method for quantitative and non-destructive subsurface (e.g., Coggon 1971; Loke and Barker 1996; Li and Spitzer
exploration up to depths of a few hundred metres. How- 2002; Erdogan, Demirci and Candansayar 2008; Marescot
ever, the use of the geoelectrical prospecting method has et al. 2008; Blome et al., 2009) and finite-difference (Mufti
long been restricted due to the difficulty to model three- 1976, 1978; Dey and Morrison 1979; Spitzer 1995; Zhang,
dimensional arbitrarily-shaped structures correctly. One- Mackie and Madden 1995; Loke and Barker 1996; Zhao
dimensional modelling and inversion have been historically and Yedlin 1996b; Wang and Mezzatesta 2001; Moucha and
used but this approach is too restricted for the majority of field Bailey 2004; Cardarelli and Fischanger 2006; Erdogan et al.
data, since the natural distribution of resistivity is rarely ho- 2008) methods. Nevertheless, two difficulties remain more or
mogeneous or tabular. Even though laterally constrained 1D less unsolved. The first is the problem of the singularity of
inversion (e.g., Auken et al . 2005) can be satisfactorily used the potential near the injection electrode and the second is the
for 2D structures with smooth lateral variations, data acquired correct implementation of the boundary conditions. Some au-
on objects with more complex geometry, such as cavities for thors (see for example Lowry, Allen and Shive 1989; Dabas,
instance, cannot be correctly inverted under the 1D or pseudo- Tabbagh and Tabbagh 1994; Zhao and Yedlin 1996a; Li and
2D assumption. In the past decades, significant advances have Spitzer 2002) suggested splitting the conductivity distribution
been made by implementing numerical methods to solve the and thus the electrical potential into two terms, with only one
two and three-dimensional electrical problem, in particular being singular but in such a way that it is analytically solvable.
the integral-equations method (e.g., Boulanger and Chouteau Zhao and Yedlin (1996a) and Lowry et al. (1989) proposed
to use a homogeneous half-space as a reference medium. Li
and Spitzer (2002) and Dabas et al. (1995) suggested using
E-mail: maineult@ipgp.fr a layered half-space instead, as it would be in many cases


C 2011 European Association of Geoscientists & Engineers 153
154 A. Maineult and J.-J. Schott

geologically more relevant than using a homogeneous one. In the conservation of the electrical charge:
this case, the potential of the reference medium can be solved
numerically only. The electrical potential of a horizontally r ( ) = 0. (1)
layered earth (or tabular potential) for a point source inject-
At the interface between the medium and the air (i.e., for
ing a direct current (DC) at the surface has been known for
z = 0), the normal current density must be zero, except at the
a long time (e.g., Stefanesco, Schlumberger and Schlumberger
source location:
1930; Slichter 1933; Pekeris 1940; Mooney et al., 1966) but
 
generally the use of the solution was restricted to the potential
= I (x xS , y yS ) , (2)
distribution at the surface only since the interpretation of z z=0
the data acquired at the surface does not require any knowl-
edge of the potential values below the surface. The numerical where is the two-dimensional Dirac distribution. It is to be
computation of the potential for the whole half-medium is noted that the Dirac distribution modelling the point source is
not simple (it involves a double recursion, for example) and two-dimensional here but, in general, it would depend on the
can easily diverge numerically, as we experienced when intro- parametric representation of the surface. In the case described
ducing the splitting method with a layered reference medium here, an obvious parametrization is (x, y, 0). On the remain-
into a finite-difference modelling code. Thus, we developed a ing boundaries, theoretically localized at infinite distance, the
new formulation of the solution, which is easier to use for the potential tends toward zero:
computation of the tabular potential everywhere in the lay-
lim = 0, (3)
ered half-space. We also examine the ways to circumvent the r S +

numerical problems that may arise. We argue that the whole


where rS (x, y, z) stands for the Euclidian distance between the
set of equations that is reported here could be of interest for
current point (x, y, z) and the source (xS , yS , 0).
any numerical code that includes the decomposition of the
As stated in the introduction, two majors problems encoun-
resistivity distribution into a layered reference medium and a
tered in solving the set of equations (1)(3) are:
perturbation.
i the source singularity in equation (2), which makes the prob-
lem not solvable in the classical domain of the C2 -functions.
GENERAL EQUATIONS FOR THE ii the difficulty to establish satisfying boundary conditions,
ELECTRICAL POTENTIAL since the nullity of the potential is valid only at infinite
distance.
We consider a conductivity distribution (x, y, z) in an open
half-space  = {z 0}, where (x, y, z) are the Cartesian coor- To reduce the effects of the source singularity and to improve
dinates, the z-axis being positive downwards. A DC-current I the boundary conditions, Lowry et al. (1989) suggested to
is injected at the surface point source (xS , yS , 0) (Fig. 1a). The split the potential into two terms, i.e., = s + r . The
resulting electrical potential field (x, y, z), for z > 0, obeys singular potential s (x, y, z) is the response of a homogeneous
half-medium of constant conductivity h . The regular poten-
tial r (x, y, z) is the response of the conductivity perturbation
(x, y, z) = (x, y, z) h . In this case, s obeys equa-
d0=0
source location 1
tions (1)(3) with h instead of . The problem is still singular
t1
d1 but in the case of a homogeneous half-space, the solution is
yS surface t2 2
d2 well-known and given by:
xS x
I 1
y s (x, y, z) = , (4)
2 h r S
cartesian dL-2 where rS (x, y, z) = [(x xS )2 + (y yS )2 + z2 ]1/2 is the distance
half-space z tL-1 L-1
dL-1 to the source. Combining the splitting of the potential and of
L the conductivity into equation (1) provides the equations for
(a) (b) z
the regular potential:
Figure 1 a) Cartesian half-space and coordinates. b) Definition of the
thickness (t), resistivity () and depth (d) of the L layers. r ( r ) = r ( s ) , (5)


C 2011 European Association of Geoscientists & Engineers, Geophysical Prospecting, 60, 153160
Full computation of electrical layered-earth potentials 155

 
with the boundary condition (3) and for the interface between 1 l 1 l+1
l {1, . . . , L 1} = . (11)
air and the medium: l z z=dl l+1 z z=dl
 r

= 0, (6)
z z=0
CLASSICAL SOLUTION
instead of equation (2). This problem is no longer singular
A demonstration of the general solution of equations (7)(11)
and can be easily solved numerically once s is known.
at any point (x, y, z) such as dl1 z dl , l {1, . . . , L} is
Lowry et al. (1989) proposed for h the average conductiv-
given by Koefoed (1979). The potential writes:
ity over the entire domain . Zhao and Yedlin (1996a) argued

that it would be more accurate to take the conductivity in the   1 I +    
l r Sh , z = (1 + al ()) e z + bl ( ) e z J0 r Sh d ,
close vicinity of the source. This last method was recently ap- 2 0
plied by Oldenborger and Ruth (2006) and Wu, Xiao and (12)
Wang (2003) for instance. Li and Spitzer (2002) and Dabas where rS h (x, y) = [(x xS )2 + (y yS )2 ]1/2 is the Euclidian
et al. (1995) proposed to use as a reference medium a lay- radial distance (Fig. 1a) and J0 the 0-order Bessel function of
ered earth and the associated potential for s . This provides the first kind. The coefficients a( ) and b() are given by the
boundary conditions more accurate than the average or lo- following recursion, which results from continuity conditions
cal conductivity would do, for many field applications such (10) and (11):
as the localization of finite buried conductive bodies or void S1 1
cavities embedded in a layered medium (situations often en- a1 = b1 = , (13a)
2
countered in environmental, archaeological or karst studies). and
For instance, a complementary vertical sounding carried out
i {1, . . . , L 1}
far from the searched objects could provide the input infor-    
mation required for the computation of s . The equations for
1 i+1 1 i+1 2di
a
i+1 +1 = 1+ (ai +1) + 1 e bi
the tabular potential are developed hereafter. 2 i 2 i
    .

1 i+1 2di 1 i+1

bi+1 = 1 e (ai +1) + 1+ bi
2 i 2 i
EQUATIONS FOR A HORIZONTALLY
(13b)
LAYERED HALF-MEDIUM
Slichters kernel Si ( ) is computed using the Sunde-Pekeris
The stratified medium of reference contains L homogeneous recursion:
layers, the Lth layer being the substratum. Hereafter l , dl and

S =1
tl denote the resistivity, the bottom-depth and the thickness L
of the lth layer, l {1, . . . , L1}, respectively and L the Si+1 + i tanh ( ti ) . (14)

i {L 1, . . . , 1} S = i+1
i i
+ S tanh ( t )
resistivity of the substratum (Fig. 1b). By convention, we set i+1 i+1 i

d0 = 0, dL = + and tL = +. Equations (1)(3) can be


written as (e.g., Koefoed 1979):
NEW RECURSION
l {1, . . . , L} l = 0, (7)
Although equations (12), (13a) and (13b) associated with re-
 cursion (14) are theoretically sufficient to compute the po-
1
= 1 I (x xS , y yS ) , (8) tential everywhere (see for instance Li and Spitzer (2002) for
z z=0
an application but with a slightly different formulation of the
recursion), numerical problems may arise when implementing
l {1, . . . , L} lim l = 0, (9)
r S + equation (12). Since factor ez tends towards infinity when z
increases, it is better to compute product bi ( )ez instead of
where l is the restriction of to the lth layer (i.e., dl1
computing factors ez and bi ( ) separately. Note however that
z dl , l {1, . . . , L}). Moreover, the potential and the nor-
this numerical problem is not encountered when computing
mal component of the current density at each interface are
the potential at the surface.
continuous:
To circumvent the possible numerical problems, we intro-
l {1, . . . , L 1} l )z=dl = l+1 )z=dl , (10) duce the modified coefficients u( , z) and v(, z) (actually


C 2011 European Association of Geoscientists & Engineers, Geophysical Prospecting, 60, 153160
156 A. Maineult and J.-J. Schott

functions of and z, valid, with respect to z > 0, in the whole The set of equations (14) and (19)(21) is sufficient to com-
half-space but useful only in the ith layer): pute the tabular potential everywhere in the half-space, is
simpler than the classical solution and avoids the numerical
ui = (1 + ai ( )) e z
. (15) problems that may arise when computing equation (13a,b).
vi = bi ( ) e z

The recursion (13a,b) then becomes:


PREVENTING NUMERICAL DIVERGENCE
u1 = S1 +1 e z OF THE INTEGRALS
2
, (16a)
v1 = S1 1 e z
2 Another numerical difficulty can occur in the calculation of
integrals (20) and (21) (and also 12), when the term tanh(ti )
and
in the Sunde-Pekeris relation (equation 14) becomes too large,
i {1, . . . , L 1} i.e., when kernel Si is close to 1. In this case, asymptotic ap-
   
1 i+1 1 i+1 2(di z) proximations for ul and l , denoted ul and l have to be

u = 1 + u + 1 e vi
i+1 2 i
i
2 i used. The first-order asymptotic approximation of Si ( ) for
    . ti tending towards infinity, denoted Si ( ), writes:

1 i+1 2(di z) 1 i+1

vi+1 = 1 e ui + 1+ vi
2 i 2 i SL = 1
(16b) , (22)
i {L 1, . . . , 1} S = 1 2R e2ti
i i
The coefficient v(, z) can be expressed as a function of
u( , z) as: where
i i+1
Si 1 2 (zdi1 ) i {L 1, . . . , 1} Ri = . (23)
i {1, . . . , L} vi = e ui . (17) i + i+1
Si + 1
The proof of equation (22) is given in Appendix A. Replacing
by introducing equation (15) into the following relation be-
equation (22) in recursion (19) yields:
tween a and b (Koefoed 1979):
 
2t1 e z
1 + ai Si + 1 2di1 u1 = 1 R1 e


i {1, . . . , L} = e . (18)
bi Si 1 i {1, . . . , L1} . (24)




Introducing equation (17) into equation (16) provides a sim- ui+1 = 2 1+ i+1
1
i
+ 1 i+1
i
Ri
R e2 ti 1
ui
i
pler recursion:
The potential in the l layer, l {1, . . . , L1}, then becomes:
th

S1 +1 z  
u1 = e 

2   1 I + Rl e2(zdl )  

l r Sh , z = 1 ul (, z) J0 r Sh d ,
i {1, . . . , L1} . (19) 2 0 1Rl e2tl

    (25)

1 i+1 i+1 Si 1 2ti

ui+1 = 1+ + 1 e ui
2 i i Si +1 and for the substratum:

  1 I +  
The potential for z such as dl1 z dl , l {1, . . . , L1} L r Sh , z = u L (, z) J0 r Sh d. (26)
is obtained by introducing equations (15) and (17) into equa- 2 0
tion (12), which gives: So when ti (respectively t1 ) is too large, u1 (respectively ui+1 )
   has to be calculated using relation (24) instead of relation (19)
 h  1 I + Sl ( ) 1 2 (zdl1 )
l r S , z = 1+ e and the potential using relations (25) and (26) instead of re-
2 0 Sl ( ) +1
  lations (20) and (21). In order to bind the absolute difference
ul (, z) J0 r Sh d (20)
between the hyperbolic tangent and its approximate expres-
Finally, we similarly obtain the potential in the substratum sion by an arbitrary value , quantity ti must satisfy the
(dL1 z) by taking into account that vL = 0: condition:
 
  
  1 I +   1 8
L r Sh , z = u L (, z) J0 r Sh d . (21) ti > ln  1+ 1 . (27)
2 0 2


C 2011 European Association of Geoscientists & Engineers, Geophysical Prospecting, 60, 153160
Full computation of electrical layered-earth potentials 157

This provides a suitable criterion to decide when ti is too Equations (28a,b,c) have to be used for small distances (gen-
large. erally, once the solutions (25) or (26) starts to decrease or
oscillate).

PREVENTING NUMERICAL DIVERGENCE


NEAR THE SOURCE COMPUTATION OF THE KERNEL

A last numerical difficulty can arise when computing the po- The integral in the expression of the tabular potential (equa-
tential near the vertical axis defined by the source, i.e., for tions (12), (20), (21), (25) and (26)) is a Hankel Transform of
rS h 0. Note also that a natural divergence occurs for order 0 of a particular kernel function. In general, the Hankel
rS 0, i.e., at the electrode location, where the potential is Transform cannot be calculated analytically. Linear filtering
infinite. In this case, the first-order asymptotic approximation methods have long been introduced in 1D electrical mod-
of J0 (equivalent to 1) yields: elling to compute the Hankel Transforms (Ghosh 1971a,b;
1 I Anderson 1979; Koefoed 1979). Many efficient algorithms are
1 (rs , z = 0) = 1 (rs , z = 0) , (28a) now available (Anderson 1979; Christensen 1990; Mohsen
r Sh 0 2r S
and Hashish 1994; Guptasarma and Singh 1997). For the nu-
merical computation, we used the 120-point filter designed by
l {1, . . . , L 1} and z = 0 Guptasarma and Singh (1997), as it gives better results than
 +
the other cited algorithms when applied to analytically known

l (r S , z) 2 1I
1 + SSll ()1 e2 (zdl1 ) ul (, z) d
r Sh 0
0 ()+1
Hankel Transforms.
   + ,

1 I Rl e2 (zdl )
l r Sh , z 2 0
1 1R e2 tl
u
l (, z) d
r Sh 0 l

EXAMPLE
(28b)
To illustrate the interest of our approach, we applied the clas-
1 I
 + sical formula (equation (12)), the modified integrals (equa-

(r , z) u L (, z) d
L S h r S 0
2 0 tions (20) and (21)) and the modified integrals with asymptotic
 + . (28c) developments (equations (28a,b,c)) to a simple three-layer

1 I
uL (, z) d
L (r S , z) h 2 0
r S 0 medium. The thickness and resistivity of the first layer are

layer 1 layer 2 layer 3


a surface c depth 0.5 m e depth 5.5 m

100

1
electrical potential (V)

equ. (12), (20), (25)


equ. (20) and (25) equ. (21) and (26)
equ. (12), (20), (25) equ. (25) equ. (26)
0.01

b depth 0.25 m d depth 3 m f depth 10.5 m

100

1
equ. (12), (20), (25)
equ. (20) and (25) equ. (21) and (26)
equ. (12), (20), (25) equ. (25) equ. (26)
0.01
0.01 1 100 0.01 1 100 0.01 1 100
radial distance to the electrode axis (m)

Figure 2 Results of the computation (electrical potential versus radial distance to the vertical axis defined by the electrode position) for the
three-layer model (t1 , t2 ) = (0.5,5) m and ( 1 , 2 , 3 ) = (50, 250, 25) m (see Fig. 1b). In the first layer (a,b), all formulations (equations (12),
(20) and (25)) produce the same results in the considered distance range (open circles). In the other layers (c,f), the classical formula (12)
diverges and the solution is given by both equations (20) or (21) and equations (25) or (26) only (grey squares). Equations (20) or (21) can also
diverge and the solution is then given by equations (25) or (26) only (black diamonds).


C 2011 European Association of Geoscientists & Engineers, Geophysical Prospecting, 60, 153160
158 A. Maineult and J.-J. Schott

0.5 m and 50 m respectively (e.g., wet surface layer), 5 m 8


x 10
and 250 m for the second layer (e.g., dry sandy layer) and the a surface
3
resistivity of the substratum is equal to 25 m (e.g., saturated

electrical potential (V)


sandy aquifer). We applied a DC-current of 1 A. Note that
we implemented the equations in Fortran 90 and executed the 2

R
code on a simple PC (processor Intel CoreTM duo 2.93 GHz
with 2 Go RAM). 1
equ. (25)
Figure 2 displays some potential profiles. In the distance
equ. (28a)
range we considered (i.e., 0.005150 m), the classical and
0
modified formulae provide the same results (at least until the
seventh decimal) in the first layer (Fig. 2a,b). In the second
33.8762
layer (Fig. 2c,d), the numerical results obtained with equa- b depth 0.5 m
tion (12) diverge (i.e., produce oscillation, unphysically large
number or even not-a-number, i.e., numerical overflow) be-
33.8761
low 0.3 m at 0.5 m depth and 15 m at 3 m. The modified
formula (20) enables computation for a smaller distance but
also diverges (at 0.02 m and 0.1 m, respectively). But the use
33.8760
of asymptotic development (equation (28b)) circumvents the
equ. (25)
problem efficiently as will be discussed below (note that we
equ. (28b)
used an value of 1015 in equation (27)). In the substratum,
33.8759
the classical formula is not usable anymore (Fig. 2e,f) but the 10
-7
10
-5
10
-3

modified formulae work well. radial distance to the electrode axis (m)
For a very small distance to the vertical axis defined by the
Figure 3 Example of results for a small radial distance (same model
electrode, the numerical results obtained with equation (25) as in Fig. 2). a) At the surface, the numerical results obtained with
can also diverge, as shown in Fig. 3(b), except at the surface equation (25) (dots) are in close agreement with the equivalent devel-
(Fig. 3a), where the computations are in close agreement with opment of the potential (equation (28a)) (black line). b) In the other
the equivalent development of the potential (equation (28a)). layers, the results of equation (25) (dots) or equation (26) converge
towards the asymptotic value (black line) given by equations (28b)
When the divergence occurs, the values of the potential for all
and (28c) at around 104 m but then numerically diverge for a smaller
the layers but the surface have to be replaced by the asymp-
distance (here, below 105 m) and thus the asymptotic values have to
totic values given in equations (28b) and (28c) (solid line in be used instead.
Fig. 3b). Note that such small distances are never required for
the interpretation of field data but asymptotic values can be
required for numerical modelling (for instance if the (x, y) co-
ACKNOWLEDGEMENTS
ordinates of the electrode coincides with the (x, y) coordinates
of some nodes of the computation grid). The authors sincerely thank Associate Editor Oliver Ritter
and the anonymous reviewers who helped us to improve this
paper. This is IPGP contribution n 3109.
CONCLUSIONS

We presented here simpler and more stable formulae for the


computation of the electrical potential developed in a lay- REFERENCES
ered half-space in response to a surface DC current injection. Anderson W.L. 1979. Computer program. Numerical integration of
We also examined the numerical difficulties than can occur i) related Hankel transforms of order 0 and 1 by adaptive digital
during the computation of the integral for the potential and filtering. Geophysics 44, 12871305.
ii) near the injection electrode. We proposed solutions that Auken E., Christiansen A.V., Jacoben B.H., Foged N. and Srensen
K.I. 2005. Piecewise 1D laterally constrained inversion of resistivity
prevent from numerical divergence. A simple but realistic ex-
data. Geophysical Prospecting 53, 497506.
ample showed the interest of this approach. Thus we suggest Blome M., Maurer H.R. and Schmidt K. 2009. Advances in three-
using it when applying the potential splitting method with a dimensional geoelectric forward slover techniques. Geophysical
layered-earth medium as a reference to resistivity modelling. Journal International 176, 740752.


C 2011 European Association of Geoscientists & Engineers, Geophysical Prospecting, 60, 153160
Full computation of electrical layered-earth potentials 159

Boulanger O. and Chouteau M. 2005. 3D modelling and sensitivity Okabe M. 1981. Boundary element method for the arbitrary inhomo-
in DC resistivity using charge density. Geophysical Prospecting 53, geneities problem in electrical prospecting. Geophysical Prospect-
579617. ing 29, 3959.
Cardarelli E. and Fischanger F. 2006. 2D data modelling by electrical Oldenborger G.A. and Routh P.S. 2006. Theoretical development
resistivity tomography for complex subsurface geology. Geophysi- of the differential scattering decomposition for the 3D resistivity
cal Prospecting 54, 121133. experiment. Geophysical Prospecting 54, 463473.
Christensen N.B. 1990. Optimized fast Hankel transform filters. Geo- Pekeris C.L. 1940. Direct method of interpretation in resistivity
physical Prospecting 38, 545568. prospecting. Geophysics 5, 3142.
Coggon J.H. 1971. Electromagnetic and electrical modeling by the Slichter L.B. 1933. The interpretation of the resistivity prospecting
finite element method. Geophysics 36, 132155. method for horizontal structures. Physics 4, 307322.
Dabas M., Tabbagh A. and Tabbagh J. 1994. 3-D inversion in sub- Spitzer K. 1995. A 3-D finite-difference algorithm for DC resistivity
surface electrical surveying I. Theory. Geophysical Journal Inter- modelling using conjugate gradient methods. Geophysical Journal
national 119, 975990. International 123, 903914.
Dey A. and Morrison H.F. 1979. Resistivity modeling for arbitrarily Stefanesco S., Schlumberger C. and Schlumberger M. 1930. Sur
shaped three-dimensional structures. Geophysics 44, 753780. la distribution electrique potentielle autour dune prise de terre
Erdogan E., Demirci I. and Candansayar M.E. 2008. Incorporating ponctuelle dans un terrain a couches horizontales, homogenes et
topography into 2D resistivity modeling using finite-element and isotropes. Journal de Physique et le Radium serie VII(1), 132
finite-difference approaches. Geophysics 73, F135F142. 140.
Ghosh D.P. 1971a. The application of linear filter theory to the direct Wang T. and Mezzatesta A. 2001. 3-D dc resistivity modeling us-
interpretation of geoelectrical resistivity sounding measurements. ing the sweeping-seed, conjugate-gradient method. Geophysics 66,
Geophysical Prospecting 19, 192217. 441447.
Ghosh D.P. 1971b. Inverse filter coefficients for the computation Wu X., Xiao Y., Qi C. and Wang T. 2003. Computations of secondary
of apparent resistivity standard curves for a horizontally stratified potential for 3D DC resistivity modelling using an incomplete
earth. Geophysical Prospecting 19, 769775. Choleski conjugate-gradient method. Geophysical Prospecting
Guptasarma D. and Singh B. 1997. New digital linear filters 51, 567577.
for Hankel J0 and J1 transforms. Geophysical Prospecting 45, Xu S.-Z., Zhao S. and Yi N. 1998. A boundary element method for
745762. 2-D dc resistivity modeling with a point current source. Geophysics
Koefoed O. 1979. Geosounding principles 1. Resistivity sounding 63, 399404.
measurements. In: Methods in Geochemistry and Geophysics. El- Zhang J., Mackie R.L. and Madden T.R. 1995. 3-D resistivity forward
sevier. modeling and inversion using conjugate gradients. Geophysics 60,
Li Y. and Spitzer K. 2002. Three-dimensional DC resistivity forward 13131325.
modelling using finite elements in comparison with finite-difference Zhao S. and Yedlin M.J. 1996a. Some refinements on the finite-
solutions. Geophysical Journal International 151, 924934. difference method for 3-D resistivity modelling. Geophysics 61,
Loke M.H. and Barker R.D. 1996. Rapid least-squares inversion 13011307.
of apparent resistivity pseudosections by a quasi-Newton method. Zhao S. and Yedlin M.J. 1996b. Multidomain Chebyshev spec-
Geophysical Prospecting 44, 131152. tral method for 3-D DC resistivity modelling. Geophysics 61,
Lowry T., Allen M.B. and Shive P.N. 1989. Singularity removal: 16161623.
A refinement of resistivity modeling techniques. Geophysics 54,
766774.
Ma Q. 2002. The boundary element method for 3-D dc resistivity APPENDIX A
modeling in layered earth. Geophysics 67, 610617.
Marescot L., Palma Lopes S., Rigobert S. and Green A.G. 2008. Non- The first-order approximation of Si for tanh(ti ) tending to-
linear inversion of geoelectric data acquired across 3D objects using wards 1 can be written:
finite-element approach. Geophysics 73, F121F133.
Mohsen A.A. and Hashish E.A. 1994. The fast Hankel transform. Si+1 + i tanh (ti )
Geophysical Prospecting 42, 131139. Si = i
i+1

i+1
+Si+1 tanh (ti ) tanh(ti )1
Mooney H.M., Orellana E., Pickett H. and Tornheim L. 1966. A re-
sistivity computation method for layered earth models. Geophysics Si+1 i  
1 i+1
tanh ( ti ) 1 = Si . (A1)
31, 192203. Si+1 + i
i+1
Moucha R. and Bailey R.C. 2004. An accurate and robust multigrid
algorithm for 2D forward modelling. Geophysical Prospecting 52,
Introducing the first approximation for the hyperbolic tangent
197212.
Mufti I.R. 1976. Finite-difference resistivity modeling for arbitrarily for ti tending towards infinity:
shaped two-dimensional structures. Geophysics 41, 6278.
Mufti I.R. 1978. A practical approach to finite-difference resistivity exp (2ti ) 1
tanh ( ti ) = 1 2 exp (2 ti ) , (A2)
modelling. Geophysics 43, 930942. exp (2ti ) + 1 ti +


C 2011 European Association of Geoscientists & Engineers, Geophysical Prospecting, 60, 153160
160 A. Maineult and J.-J. Schott

 
we obtain: i i+1 i i+1
Si =12 +2 Ri+1 e 2ti+1
e2ti .
i + i+1 (i + i+1 )2
i i+1 Si+1 2ti
Si = 1 2 e . (A3) (A6)
i + i+1 Si+1
As the second term in the brackets is negligible when ti+1
Now, let us assume that for any i such as L1 > i > 1 the tends towards infinity, equation (A6) becomes (to the first
following relation is valid: order):
i i+1 2ti

Si+1 = 1 2Ri+1 e2ti+1 , (A4) Si = 1 2 e = 1 2Ri e2ti . (A7)
i + i+1
Finally, SL1 can be straightforwardly derived from SL , the
with Ri defined by equation (23). Then equation (A3) can be
asymptotic expression of the kernel for the substratum, which
rewritten using the approximated form (equation (A4)) for
is equal to SL and thus to 1. We have:
Si+1 :
  L1 L SL 2tL1
i i+1 1 2Ri+1 e2ti+1 2ti SL1 = 12 e

Si = 1 2  e . (A5) L1 + L SL
i + i+1 1 2Ri+1 e2ti+1 L1 L 2tL1
= 12 e = 1 2RL1 e2tL1 , (A8)
Since the term 2Ri+1 exp(2ti+1 ) tends towards 0, equa- L1 + L
tion (A5) can be expanded to the second order as: so relation (22) is demonstrated.


C 2011 European Association of Geoscientists & Engineers, Geophysical Prospecting, 60, 153160

Potrebbero piacerti anche