Beruflich Dokumente
Kultur Dokumente
INTERACTIVE
COMPUTER/GRAPHIC-DISPLAY-TERMINAL
SYSTEM
FOR INTERPRETATION
OF RESISTIVITY
SOUNDINGS
*
BY
H. K. JOHANSEN
**
ABSTRACT
JOHANSEN,
for Interpretation
Computer/Graphic-Display-Terminal
Geophysical Prospecting
23, 449-458.
System
A fast computer-procedure
giving the apparent resistivity
curve as well as the partial
derivatives
with respect to the layer-parameters
is presented. It is based on the linear
filter method developed by D. P. Ghosh in 1971.
The sampling frequency is IO points per decade, and 3 decades are covered. The maximum relative error is less than 10-3, and in most cases orders of magnitude
smaller.
The computation
time on a CDC 6400 for one curve given in 30 points ranges linearly
from .17s for a two-layer case to .36s for a ten-layer case.
The procedure is used to plot master curves interactively
on a graphic display terminal
(Tektronix
4010)
connected to the CDC 6400. By trial-and-error
adjustments
a set of
layer-parameters
is found, giving essentially the measured curve.
INTRODUCTION
The linear filter method developed by Ghosh (1971, 1,971) has opened a
new range of possibilities for the interpretation
of resistivity soundings. It
is well suited for computerization.
Because of the small number of operations
needed the execution time is almost negligible. One can afford to calculate
master curves based on purely guessed layer-sequences and just throw them
away if they do not fit the data.
In his paper Ghosh points out that a sampling frequency of three points
per decade is a good compromise when the tools are a desk calculator and
logarithmic paper, because a relative accuracy better than .02 is achieved using
only nine filter coefficients. However, when working with an electronic computer
there is no reason not to use about hundred coefficients, if this proves to be
advantageous.
The direct interpretation
method is to calculate the resistivity transform T
from the field data and then derive the layer-parameters
(resistivities pj and
* Presented at the 35th meeting of the European
Association
of Exploration
physicists, Brighton,
June 1973.
** Laboratory
of Geophysics, Finlandsgade
8, DK-8200 Arhus N, Denmark.
Geo-
H. K. JOHANSEN
450
thicknesses dj) from T. In order to obtain T by the filter method, one has
to extrapolate
the measured curve both to the left and to the right, thus
using data that were not really measured. It is also necessary to smooth the
data. Both operations possibly introduce errors in the interpretation,
and a
weak indication
of a nearly suppressed layer might get lost, -too. Since our
aim is to extract the maximum amount of information
from the measurements
it might be preferable to take the indirect approach:
A guess of the layer-parameters
z) T(pj, do) is calculated, and papp
coefficients,
3) Trial-and-error
adjustments
of
curve agrees with the measured
I)
the calculated
papp-
Notice that the tinprocessed data are used for comparison only. The influence
of the interpreters personal judgment is thus reduced to the absolute minimum :
to decide whether the model is acceptable or not.
Step 3 may be carried out by an iterative least-squares programme, but the
method remains indirect in principle.
The relative accuracy of the field measurements is about .03. We wouldlike
the inaccuracies of the calculated curve to be an order of magnitude
smaller
so that their influence on the interpretation
can be safely neglected. In order
to achieve this a smaller sampling interval and hence a larger set of filtercoefficients must be used.
SUMMARY
In the Schlumberger
OFTHE
arrangement
FILTERMETHOD
we have
pas(s) = s2 ; T(h)Jl(hs)Adh
0
where s is half the electrode spacing and T(A) is the resistivity
(Koefoed 1968, 1970, Ghosh 1971, 197Ia).
Introducing
transform
and substituting
(1)
(2)
= i T(YM-m
y)dy
(3)
(4)
where
b(x) = Jl(exp
INTERACTIVE
TERMINAL
SYSTEM
property
FOR
RESISTIVITY
INTERPRETATION
of the convolution
integral
Y-
F(f)
= G(f)
(3) :
WY
451
(5)
. H(f).
(6)
pair of functions
P(X) = exp (3X)/(1 + exp (2x))52
T(y) = Q ~XP (-
Y) =P (-
exp (-
(7)
~1)
satisfy (3).
Their Fourier transforms
are found by numerical integration,
modulus Qdf) and phase O(f) of the quotient F/G determine H.
For a given sampling interval Ax the Nyquist frequency is fp
and the corresponding sine-response function is given by
c(x)
's"Qcf,cm
I/fN
(a(f)
zxfx)df.
(A+)
@(fN)
+ S)C(jAx -
Prospecting,
Vol.
23
= ~/(zAx),
(8)
and the
S)
coefficients
(10)
with smaller or
29
H. K. JOHANSEN
4.52
TN-I(~)
pN-1
kN-
1 eXp
(-
I + kN.e I exp (-
is
(11)
ZdN- lh)
where
kN-1 =
layer
(PN-1
(12)
dl
91
+ PN).
PN)/(~N-1
/--T(h)
i
layeri
rTi+l
j-TN-1
layer
(A)
(A)
N-l
QN-1
dN-1
A
layer
(infinite
Fig.
I.
QN
substratum)
The transform Tj for layer (pj, dj) on top of the sequence (p~+l, . . . . . , pN;
d,+l, . . . . . dN-1) with transform TJ+I is given by
w,(h) + Tj+$)
T(A) = I + Wj(A)Tj+,(A)/p;
j=N-z,N--3
,...,
2,1;
(13)
where
W,(V
Recursive
application
I - exp (- zd+)
= Pj I + exp (- zd+) *
(14)
(15)
INTERACTIVE
TERMINAL
SYSTEM
FOR RESISTIVITY
INTERPRETATION
453
and hence
h-+co
A --f o .
Tj(h) --f pj
Tj +1(A)
(16)
The way in which these limits are approached depends on the specific set
of layer-parameters
considered. Let us look at the simple two-layer case:
For large A we have
TN-I(~)
The relative
M p~-1(1
departure
(17)
value
plv-1 I / plv-I
(1s)
E if only
ln(a/z)/(zdlv-
1).
(19)
with
A$ = exp (-
iAx)
on i. If E = 10-2,
(20)
= I and Ax = (In IO)/IO,
&,-I
Something similar holds in the general case for T = TI and pl, although
the expansion of T is more complicated than (17). Experience has shown that
we are on the safe side if i c - 9.
For small A we have
cases corresponding
TN-I(~)
M PN(I -
TN-I(~)
M ~N(I
to kN-1 w -l I :
(~N/~N-I}~~N-I)
for
PN
>
PN-1
(22)
and
+{p~-I/~N}~~N-I)
forp~<p~-l.
(23)
Since the resistivity ratios can take on all positive values, we have to compromise.
Let us consider (23) say. If PN-~/PN 5 10~ and dii-1 < Io3, then A 5 IOs2
if only A 5 10-l.
This happens if i > IOO.
Experience shows that i > IOO is sufficient also in multi-layer cases.
CUTOFF AND ACCURACY
Apart from the sampling frequency there are two factors affecting the
accuracy: the tails of the filter must be cut off at some point, and the coefficients can only be calculated with a limited precision.
H. K. JOHANSEN
454
The problem is how
at a relative accuracy
Let us first consider
For convenience we
in (10).
(24)
i
C(f) and C$~S)=
I--m
i C(5)
1=1%
(26)
form a transform
PI
pair resembling
(~2 -
PI) exp
a two-layer
(--
exp
(-
4)
(28)
case.
8
0.00,
0.01
0.1
10
100
1000
The filter
decade.
i
-99
-98
-97
-96
-95
-94
-93
-92
-91
-90
-89
-88
.-87
-86
-45
-84
-83
-82
-81
-80
-79
-78
-77
-76
-75
-74
-73
-72
-71
-70
-69
-68
-67
-66
-65
coefficients
to a san@ling frequency of
Ax = (ln IO)/IO,
j, = -100,
jz = 40, S = --I.py,g458,
6.174~~ - 5, C!) = 6.0goglo - 4
C(j) . 108
-12484
12726
-12975
13231
-13494
13765
-14043
14330
-14625
14930
-15244
15567
-15901
16246
6602
16971
--I7352
17746
-I
-18154
18577
-19015
19469
-19941
20429
-20936
21463
-22oog
22577
-23166
23779
-24416
25079
-25768
26487
-27235
corresponding
-64
-63
-62
-61
-60
-59
-58
Y-57
-56
-55
-54
-53
-52
-51
-50
-49
-48
-47
-46
-45
-44
-43
-42
-41
-40
-39
-38
-37
-36
-35
-34
-33
-32
-31
-30
C(1) . 108
28016
-28830
29680
-30568
31496
-32467
33484
-34549
35666
-36838
38069
-39363
40724
ccn . 108
-110741
117248
-27
-26
-25
-24
-23
-124303
132085
-140461
-159826
-22
-21
-20
171917
-182946
199955
---I9
-209469
239052
-234543
304916
-234124
453990
-106745
899282
550573
2442523
3250077
7926675
13023345
25610307
41150741
64zy8og
72803988
36118538
-18
-17
-42156
-16
43666
-45259
46940
-48717
50596
-52587
54697
--56936
59314
-61845
64540
-67414
70484
-73767
77284
-81057
85111
-89475
94183
-99267
I04775
---I5
-14
-13
-12
-11
-10
1;
1:
-5
-4
-3
-2
-1
-100406442
0
I
2
3
4
5
-242172543
20052460
444506381
-489348908
wG%m398
-137791072
points per
C!-loo) =
C(l) . 10s
-29
-28
149959
IO
IO
II
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
61285163
--29362w
15817356
-9504597
6226174
-435i505
3198475
-2441493
1920840
-1548505
1273595
-1065148
903512
-775750
673079
-589375
520264
-462558
413891
-372478
336951
-306251
279543
-256168
235594
-217394
201216
---I86773
I73826
-162176
151657
-142126
133463
-125568
H. K. JOHANSEN
456
TABLE
Exewtion
no. of layers
0.174
IO
0.217
0.234
0.271
0.364
CP time (set)
0.191
At our Laboratory
we have at our disposal a graphic display terminal
(Tektronix
4010) connected to the Aarhus University
computing
centers
CDC 6400 via telephone. We are using this facility in the following interactive way:
The measured resistivity values are displayed with error bars indicating
the estimated accuracy.
z) A (skilled) guess of the layer-parameters
is entered by the interpreter.
3) The corresponding p app-curve is computed and displayed.
4) Trial-and-error
adjustments
of the layer-parameters
are made until the
theoretical papp-curve agrees with the measured data.
I)
IO
100
AU2 1000 m
INTERACTIVE
TERMINAL
INTERPRETATION
457
curve
fig. 3.
layer
no.
I
2
3
4
27
I5
40
20
160
15
40
160
3.5
3.5
27
I.5
160
21
3.5
APPENDIX
THE
PARTIAL
DERIVATIVES
OF T
we difas pk,
we are
a, and
Then
~TN-I
3PJJ
~TN-1
~
+N-1
- R * pit-l/(pN-1
= TN-JPN-I-R
+ pN121
(29)
* PN/(PN-1 -~PIv)~~
(31)
Since the pk for k = N, N we have
JTj
apk
3Tf
-
3Tj+1
~JT~+I
ajk
I, . .: . . j + I enter
k=N,N-I,
TJ only through
. . . . . j +I.
T~+I,
(32)
458
JOHANSEN,
INTERACTIVE
TERMINAL
SYSTEM
Also
aTj
--3d3 -
3T3
3w3
4P3a
(I +a+
(33)
where
a = exp (Putting
zd31).
I + W3T3+,/p~ = c, we have
JT3
-=-3P3
3T3 w3
a w3 p3 + 2(w3
(34)
and
3T3
3T3+1
iJTi
__
aw3
= (I -
= (I -
w,/p;yc2
T;+Jp;)/c2
S).
REFERENCES
GHOSH, D. P., 1971, The Application
of Linear Filter Theory to the Direct Interpretation
of Geoelectrical
Resistivity
Sounding Measurements,
Geophysical
Prospecting
rg,
192-217
GHOSH, D. P., rg7ra, Inverse Filter Coefficients for the Computation
of Apparent Resistivity Standard Curves for a Horizontally
Stratified Earth, Geophysical Prospecting
19, 769-775.
KOEPOED, O., 1968, The Application
of the Kernel Function in Interpreting
Geoelectrical
Measurements,
Geoexploration
Monographs,
Series I, No. 2, Gebriider Borntraeger,
Berlin-Stuttgart.
KOEFOED, O., 1970, A Fast Method for Determining
the Layer Distribution
from the
Raised Kernel Function, Geophysical Prospecting
18, 564-570.
KOEFOED, O., Igp, A Note on the Linear Filter Method of Interpreting
Resistivity
Sounding Data, Geophysical
Prospecting 20, 403-405.
VAN DAM, J. C., 1965, A Simple Method for the Calculation
of Standard Graphs to be
Used in Geo-electrical
Prospecting,
Geophysical Prospecting
13, 37-65.