Available online at www.pelagiaresearchlibrary.

com

Pelagia Research Library

Advances in Applied Science Research, 2012, 3 (1): 508-519

ISSN: 0976-8610
CODEN (USA): AASRFC

Analysis of kernel function by the transformation of field resistivity data

J. C. EGBAI

Department of Physics, Delta State University, Abraka, Nigeria

_____________________________________________________________________________________________

ABSTRACT

The interpretation process has been a major problem in geophysical investigation of the subsurface of the earth in
terms of lithological variations with depth. The resistivity kernel function performs as an intermediary in the process
of interpreting apparent measurements in terms of lithological variations with depth. The kernel function cannot be
measured in the field but must be obtained from a transformation of measured electric potentials or apparent
resistivities. The function is more traceable mathematically than the apparent resistivity function, it has become the
basis of a method of quantitative interpretation. In this work, some literature based on kernel function with their
peculiar problems were highlighted. The theoretical derivation where the necessary information about the
configuration of the earth were fully explained. The Schlumberger electrode configuration was adopted for data
collection in the field. The field data obtained from Ughelli in Delta State, Nigeria were matched with theoretical
curves from kernel function. The results obtained from kernel function agreed very well with well logs interpretation
in Ughelli.
_____________________________________________________________________________________________

INTRODUCTION

The resistivity kernel function performs as an intermediary in the process of interpreting apparent measurements in
terms of lithological variations with depth. It depends solely on the resistivities and thickness of an earth which is
considered to be locally stratified in homogeneous and isotropic layers. It is independent of the electrode
configuration unlike the apparent resistivity function. The kernel function cannot be measured in the field but must
be obtained from a transformation of measured electric potentials or apparent resistivities. The interpretation process
involves matching practical field curves with theoretically generated curves from resistivity kernel functions (Egbai,
1997).

The resistivity kernel function has undergone a series of modifications. It has developed into the resistivity
transform function which is similar in shape and magnitude to the apparent resistivity function from which it is
derived. This transform function has been used as the basis for an analogy with electric filter theory which permits
data to be exchanged readily between the apparent resistivity and transform domains.

The natural kernel function method has been the linear filter which is able to exchange data efficiently between the
apparent resistivity and resistivity transform domains.

Linear filter theory provides a rapid means of calculating resistivity transforms and apparent resistivities, thereby
facilitating resistivity sounding interpretation. The coefficients associaited with the method previously was
inadequate for reflection coefficient approaching minus one. The new coefficient of Schlumberger electrode
configuration significantly reduce the inadequacy thereby rendering the linear method an accurate and rapid means
of calculating apparent resistivity computations (Neil,1975).

508
Pelagia Research Library
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________
It has been found that the wiener-Hoft least squares method is a very successful tool for the determination of
resistivity sounding filters. The values of the individual filter coefficients differs quite appreciably from those
obtained by the Ghosh procedure. These differences in the filter coefficients, however , have only a negligible effect
on the output of the filter. It seems that these differences in the coefficient correspond to a filter function of a rather
narrow frequency band around the Nyquist frequency, which is only very weakly present in the input functions
(Koefoed and Dirks, 1979).

The ‘raised’ kernel function in an intermediate function in the interpretation of resistivity sounding data and the
methods have been described both for the determination of the raised kernel function from the apparent resistivity
function, and the determination of the layer distribution from the raised kernel function (Koefoed, 2006).

The procedure is described by which the second step in the interpretation method result in the determination of layer
distribution of layer distribution from the raised kernel function is considerably accelerated.

A power series expansion can be used to obtained the kernel from apparent resistivities for an arbitrary electrode
configuration. Three types of function are most appropriate for this purpose. The expansion coefficient can be by a
least-square method. In this case, orthonomalization of the functions is of great advantage. ( Kohlbeck,2006).

The linear filter method has, in fact, revolutionized the interpretation of resistivity soundings. It provides a simple,
rapid and accurate means of calculating theoretical resistivity sounding curves for any assumed earth model
comprising homogeneous and electrically isotropic horizontal layers (Egbai, 1997).

Electrical prospecting is an active component of geophysical exploration, and its scope is constantly widening and
now encompasses surface measurements, single-borehole measurements, crosshole measurements, sea-button,
tunnel, or gallery measurements (Straub, 1995a). The kernel function, as a function of the depth coordinate, is the
solution of a 1-d differential equation. The conventional procedure for the calculation of the kernel function consists
in applying a recursive scheme. This procedure according to Straub is effective from a computational point of view
but becomes cumbersome from an analytical point of view, especially in the case of an arbitrary number of layers
for arbitrary positions of the source and measurement points.

The kernel function plays an important role in the 1-D problem because of the spectral representation of the electric
potential for a stratified model with a point source (Sraub, 1995b).

The kernel function deals with theoretical thickness and theoretical resistivities of earth layers for the aim of
drawing theoretical apparent resistivity curves. For purpose of interpretation, these curves are matched with actual
field data (Egbai, 1997).

Slichter (1993) and Langer (1993) arrived at a solution to the problem of using surface potential data, V(r ) , to give
()
directly the conductivity variation with depth, σ z . Assuming conductivity to be continuous and analytic function
of depth only, they derive the solution by applying Ohms law and Laplace’s equation as:

∞
Iρ 1
k (λ )J 0 (λ r )dλ
2π ∫0
V( r ) = (1)

where k (λ ) is the Slichiter kernel, J 0 is the zero-order Bessel function of the first kind, r is distance and
1 1

λ =  ρ t ρ  σ
=  l  = anisotropy, ρ is resistivity while σ = conductivity. The subscript l means
2 2

 l   σt 
parallel to the layers while t means tangential to the layers. The integral was inverted by a Hankel transform to
define the kernel, k (λ ) , in terms of the measured potentials.

509
Pelagia Research Library
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________
∞
2πλ ∞
k (λ ) = V(r ) J 0 (λ r )rdr
Iρ 1 ∫0
(2)

Koefoed (1965a) worked on the resistivity kernel function by removing two particular deficiencies in previous
kernel investigations.

(i) The kernel had always been defined in terms of observed potentials, and
(ii) The kernel had been computed by numerical integration. He applied Hankel inversion to the apparent resistivity
formula of Stefanesco et al, (1930) to define the kernel function in terms of observed apparent resistivities.

∞
θ (λ ) = ∫ r {ρα (r ) − ρ }J (λr )dr
1 1
(3)
2ρ1
1 1
0

where J 1 = first order Bessel function of the second kind.

Koefoed observed that numerical integration could be made more efficient by redefining the kernel as:

∞
ρn − ρ1
θ (λ ) = ∫ r {ρα (r ) − ρ }J (λr )dr
1 1
+ (4)
2ρ1 2ρ1
n 1
0

As r increases, the intgrand approaches zero.

Koefoed (1965b) subsequently abandoned curve-matching in the kernel domain in favour of the graphical method of
Perkeris which he restated in terms of a “modified kernel function”.

θ (λ )
G (λ ) =
1 + θ (λ )
(5)

G (λ ) is asymptotic to k1 exp(− 2λh1 ) at large λ . A plot on log-linear graph paper gives the layer thickness h
and the reflection coefficient k1 . The top layer is removed and a reduced G (λ ) is calculated by recursion to give
the parameter of the next layer.

Koefoed, on the significance of kernel function in resistivity interpretation introduced a “raised kernel function”
shown as

H * (λ ) = θ (λ ) + 1 = ∑ ∆ θ (λ ) (6)
2
This he defined by:
∞
H * (λ ) = ∫ r ρα (r )J (λr )dr
1 1
(7)
2ρ1
1
0

Koefoed (1970) adopted the “resistivity transform function” as a supplement for the “raised kernel function.” The
function is shown

T (λ ) = 2 ρ 1 H * (λ ) (8)

This he defined recursively for an assumed earth model. It also depends only on the thickness and resistivities of
subsurface layers and is independent of the electrode array used in the field.

510
Pelagia Research Library
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________
It should be stated clearly that no universally accepted kernel function has been fully identified to date.
Interpretation are made by matching field curves with one type of kernel function curve or the other.

In this work, the field curves were compared with well log data and a very high positive correlation were achieved.

THEORY
The theory on kernel function is based on the work of Egbai (1997). The potential at the surface of an n-layer earth
having arbitrary resistivities and thickness could be obtained by applying separation of variables to Laplace’s
equation in cylindrical coordinates written as:

∞
Iρ 1 1 
V (r ) =  + 2 ∫ θ n (λ )J 0 (λ r )dλ  (9)
2π r 0 

The apparent resistivity for a horizontally stratified homogeneous and isotropic layer earth could be written in terms
of Bessel function as

∞
  1  
ρ a (r ) = ρ 1 1 + 4 r ∫ θ n (λ )− 2 θ n  λ   J 0 (λ rr )dλ 
1
(10)
 0  2  

where J 0 (λ r ) = Bessel function of zero order and first kind

θ n (λ ) = kernel function.
n = number of layers.
By differentiating equation (9), the Schlumberger apparent resistivity over an n-layer earth becomes:

∞
 
ρ a (r ) = ρ 1 1 + 2 r 2 ∫ λθ n (λ )J 1 (λ r )dλ  (11)
 0 

where J 1 = first order Bessel function of the first kind

Equation (10) could be analyzed or evaluated by writing the kernel function as a ratio of polynomials

ρ n (λ )
θ n (λ ) =
H n (λ ) − Pn (λ )
(12)

Recursively, with U = e −2 λ , we have

( )
P1 (U ) = P1 U −1 = 0
H1 (U ) = H (U ) = 1
1
−1

511
Pelagia Research Library
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________

H m (U ) = H m − 1 (U ) − k m − 1U λm − 1 H m − 1 U −1 ( )
Pm (U ) = Pm − 1 (U ) + k m − 1U λm −1 Pm − 1 { (U ) + H (U )}, m = 2 ,...n
−1
m −1
−1

( )
Pj + 1 (U ) = Pj (U ) + k j U Zj H j U − 1 , j = 1 ,2 ,..., n − 1
H j + 1 (U ) = H j (U ) + k jU Zj P (U )
j
−1

For n = 2 , we have explicitly

k1e −2 h1λ
θ 2 (λ ) = (13)
1 − k1 e − 2 h1λ

and n = 3 , we have

k1 e −2 h1λ + k 2 e −2 h2λ
θ 3 (λ ) =
1 − k1 e 2 h1λ + k1 k 2 e − 2 (h2 − h1 )λ (14)

where k = reflection factor, h = depth

If ρ a (r ) is known, the kernel function θ n (λ ) could be obtained from equation (10). Introducing the functions

1 
f (λ ) = θ n (λ ) − 1 θ n  λ  (15)
2 2 

and

C (λ ) = ρ 1 {4 f (λ ) + 1} (16)

From equation (10), we have

∞
ρ a (r ) = ∫ C (λ )J 0 (λ r )dλ (17)
0

Applying Fourier –Bessel integral (Bowman, 1958)

∞ ∞
 
F (s ) = ∫  ∫ F ( x )J 0 ( xt ) ( xt ) xdxJ 0 (st )tdt (18)
0 0 
If we multiply equation (17) by J 0 (λ r ) and integrating over r , we have

C (λ )
∞ ∞
∞ 
∫ ρ a (r )J 0 (λr )dr = ∫ r  ∫ C (t )J 0 (tr )dt  J 0 (λr )dr =
0 0  0  λ
(19)

Thus,

512
Pelagia Research Library
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________
∞ ∞ ∞ ∞
t
C (λ ) = λ ∫ ρ a (r )J 0 (λ r )dr = ∫ ρ a   J 0 (t )dt (20)
0 0 λ

From equation (16), we have

C (λ ) − ρ 1
f (λ ) = (21)
4ρ1

Hence f (λ ) can be calculated from equations (20) and (21). From equation (15), θ n (λ ) can be derived as

∞
θ n (λ ) = ∑
1 1
k
f k (22)
k =0 2 2
1 λ
∴θ n (λ ) = f (λ ) + θ n  
2 2
λ λ 1 λ 
θn   = f   + θn   (23)
2 2 2 4
Thus,

θ n (λ ) = f (λ ) +
1 λ
2
( )
1  λ 
2  4 
1
2
1
4
( )
 f 2 + θ n    = f (λ ) + f λ 2 + θ n λ 4 ( ) (24)

and so on. The series in equation (22) is convergent since f (λ ) → constant when λ → 0 . The value for C (λ )
is given by equation (20). For a two layer earth, we have

∞ ∞
 t t  t
C (λ ) = ∫  ρ a   − ρ end   J 0 (t )dt + ∫ ρ end   J 0 (t )dt (25)
0  λ λ   0 λ 

where ρ end (r ) is the apparent resistivity curve at the end for a two-layer earth
∞
t  λ 
C end (λ ) = ∫ ρ end   J 0 (t )dt = ρ end  4θ end (λ ) − θ end   + 1
1
(26)
0 λ  2 2 
where
kend e −2 hend λ
θ end (λ ) = (27)
1 − k end e − 2 hend λ

From equation (25), we have

∞
 t t 
C (λ ) = ∫  ρ a   − ρ end   J 0 (t ) dt + C end (λ ) (28)
0  λ λ 

ρ a (t ) − ρ end (t ) = 0 for t ≥ rend

513
Pelagia Research Library
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________
For equation (28), we can put

∞ ∞
 t t 
T (λ ) = ∫  ρ a   − ρ end   J 0 (t )dt = ∑ T j (λ ) (29)
0  λ λ  j =0

where
z j+1
 t t 
T j (λ ) = ∫  ρ a   − ρ end   J 0 (t )dt (30)
zj  λ λ 
( Z j = the j th zero order of J 0 (t ), Z 0 = 0 )

The series in equation (29) will form an alternating series with decreasing terms. Applying Euler transformation to
speed up the convergence. For equation (28), C (λ ) = C end (λ ) for small values of λ . If we assumed

C (λ ) = C end (λ ) for λ ≤ λ 1 (31)

Calculation of kernel function θ n (λ ) for λ ≤ λ1 .

Applying equations (21, 22 and 31), we have

θ n (λ ) = ∑
∞
1  λ  ∞ 1 C 2m − ρ 1
λ
( )
∞
λ C end 2λm − ρ 1 ( )
f m =∑ m =∑ m for λ ≤ λ 1
m =0 2
m
2  m =0 2 4 ρ1 m=0 2 4 ρ1

Applying equation (26) and knowing that

λ λ
− hend − 2 hend
1 ∞
1 kend e 2m ∞
kend ρ 2m

∑
2 m =0 2 m − hend
λ
=∑
m =0 − 2 hend
λ
1 − kend e 2m 1 − kend ρ 2m

we have
λ
− 2 hend
ρ end − ρ 1  ρ end  kend ρ 2m
θ n (λ ) = +   λ (32)
2 ρ1  ρ1  −2h
1 − kend ρ
end
2m

Calculation of kernel function θ n (λ ) for λ→∞

When λ → ∞ , f (λ ) → 0 . We find a value λ = λ 0 such that f (λ ) ≤∈1 for λ ≥ λ 0

We then calculate
 λ 
f  k0−1  for k = 1 ,2 ,...
2 

514
Pelagia Research Library
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________
λ0
until
2 k0 −1 ≤ λ0
 λ0 
Let k = k 0 , then applying equation (32), we obtain the value of θn k −1 
2 0 
From equation (22) we obtain
 λ0   λ  1 λ 
θn  k −1 
= f  k0− 1  + θ n  0k  , k = k0 − 1 ,...1 (33)
2  2  2 2 

The theoretical apparent resistivity curves could be computed with equation (32) for λ ≤ λ1 and equation (33) for
λ → ∞.
METHODOLOGY

Ughelli, an oil town of Delta State, Nigeria lies between latitude 5 o 30' N and 5 o 48' N and longitudes 5 o 48' E
o
and 6 05' E and covers an area of about 150 square kilometers. The vegetation is of the Mangrove and Rain forest.
The town is drained by a river which flows from North to South and crosses the centre of the town. The area is made
of top white sand, clay followed by alluvial sand, sandstone and clay.

The Schlumberger electrode configuration was utilized for the purpose of data acquisition. The geophysical
instrument used was the Abem Terrameter SAS 300B. The current electrode spacing reach a maximum of 400m
while the potential electrode spacing were varied as the need arose. A total of 84 vertical electrical soundings were
carried out in the six locations. The six locations are as follows:

Location
A – Market square
B – Along Ughelli Patani Road
C – Government College Site
D – Post Office Road
E – Catholic Church area
F – Olori Road, Ughelli.

The apparent resistivity equation for field data is given by

πV (l 2 − a 2 )
ρa = (34)
4I a
If we incorporate equation (32) and (33) into a computer program (Egbai and Ekpekpo, 2003) in line with the basic
language for generating automatic field curves and matching it with equation (34), theoretical kernel curves will be
obtained.

RESULTS AND DISCUSSION

Quantitative interpretation was done firstly by curve fitting and matching. Curves of logarithms of apparent
resistivities (ρ α ) are normally plotted on the Y-axis against the logarithms of AB 2 on X-axis. The results of
curve fitting and matching showed a rough estimate of layer resistivities and thickness of the aquifer.

The computer assisted interpretation is based on the algorithm which employ digital linear filters for the fast
computation of the resistivity function for a given set of layer parameters (Egbai, 1997). Typical smoothened field
curves were used for the iteration data. The six curves obtained are respectively shown in figures 1, 2, 3, 4, 5 and 6.

515
Pelagia Research Library
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________

6.2

15
6
14.5

14

13.5
5.8
13

12.5

12
5.6
11.5

11

10.5
5.4
10

9.5

5.2

5.4 5.6
Fig. 8. Contour map of the study Area

516
Pelagia Research Library
Fig. 1: VES Curve for Loc. 1 Fig. 2: VES Curve for Loc. 2
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________

Fig. 3: VES Curve for Loc. 3 Fig. 4: VES Curve for Loc. 4

Fig. 5: VES Curve for Loc. 5 Fig. 6: VES Curve for Loc. 6

517
Pelagia Research Library
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________
The six curves are H-type. The shapes of the curves help in formulating the model parameters got from the
theoretical curves using the kernel function. The summary of the results are shown in Table 1. The depth of the
aquifers are located within the second layer of the various formations and is within the range of 10 to 20m. The
various resistivities, thickness and depth all locations show in the model parameters of the Table 1. The curves of all
location have low percentage errors as shows in Table1 and figure 1 to 6.

The driller’s log results and hand-dug wells in these six locations at Ughelli confirm the results of surface vertical
resistivity sounding (VES) in terms of aquifer depth and thickness.

The driller’s log results are as shown in fig. 7 and 8 respectively. Fig. 7 shows the driller’s log for locations A, B
and C while fig. 8 is for locations D, E and F.

Table 1: Summary of Results (Model Parameter)

LOCATIONS 1 2 3 4 5 6
1st Layer model lα 742.50Ωm 940.00 Ωm 892.1Ωm 717.6.00Ωm 545.7Ωm 579.3Ωm
2nd Layer model lα 358.10Ωm 291.5Ωm 407.4Ωm 312.0Ωm 150.7Ωm 161.2Ωm
3rd Layer model lα 2102.7Ωm 1121.3Ωm 1634.4Ωm 731.6Ωm 298.7Ωm 534.4Ωm

1st Layer model thickness 2.0m 7.6m 4.8m 1.8m 2.1m 2.6m
2nd Layer model thickness 42.00m 39.3m 68.5m 43.4m 46.5m 25.6m

Percentage error between field data curves and

3.6 2.5 4.0 4.6 3.5 3.3
theoretical curves RMS (%)

CONCLUSION

The field data got from the work has good impression with the kernel function. The result obtained from the kernel
function are in agreement with well logs interpretation in Ughelli with low percentage error.

Acknowledgement
I wish to thank Professor M. B. Asokhai for his assistance in the theory aspect of this work and provision of
equipment used. Lady Egbai, F. O. is acknowledged for her moral and financial assistance.

REFERENCES

[1] Egbai, J. C. (1997), Correlation between resistivity survey and well logging in Delta State. Ph.D Thesis
submitted to Edo State University, Ekpoma.
[2] Egbai, J. C. and Ekpekpo Arthur (2003), J. Nig. Ass. Math. Phys. Vol.7, pp. 155 – 168.
[3] filters by the Wierner-Hopf least-squares method. Geophysical Prosp. Vol. 27, pp.245-250. Geophysical
Prospecting, Vol. 24 Issue 2 pp233-240
[4] Koefoed O. 2006, Progress in the Direct Interpretation of Resistivity soundings. An Algorithm
[5] Koefoed, O. (1965b), Geophys. Prosp. Vol. 13, 4, pp. 568 – 591.
[6] Koefoed, O. (1970), Geophys. Prosp. Vol. 18, No. 4, pp. 564 – 570.
[7] Koefoed, O. (1965a), Geophys. Prospecting. Vol. 13, No.2, pp. 259 – 282.
[8] Koefoed,O., and Dirks, F.J.H;1976,Determination of Resistivity sounding
[9] Kohlbeck,F. 2006, computing the kernel Function in Resistivity Sounding
[10] Langer, R. E. (1933), Bull. Amer. Math. Soc. Vol. 39, pp. 814 – 820.
[11] O’Neil, D.J. 1975 , Improved Linear filter coefficients for application in apparent resistivity computations. The
Bulletin of the Australian Society of Exploration Geophysists
[12] Slichter, L. B. (1933), The interpretation of the resistivity prospecting method for horizontal structures. Phys.
Vol. 4, pp. 307 – 322.
[13] Stefanesco, S., Schlumberger, C and Schlumberger, M. (1930). Sur la distribution electrique potentaille autour
d’ une de terre pontuelle dans in terrain a couches horizontals, homogene et isotopes. J. de Physique et le Radium,
Series 7, Vol.1, pp. 132 – 140.
[14] Straub, A. (1995a), Geophysics 60, 1656 – 1670.

518
Pelagia Research Library
J. C. EGBAI Adv. Appl. Sci. Res., 2012, 3(1):508-519
_____________________________________________________________________________
[15] Straub, A. (1995b), Geophysics 60, 1671 – 1681.

519
Pelagia Research Library