Vous êtes sur la page 1sur 6

GEOPHYSICS. VOL. 50, NO. 3 (MARCH 1985); P. 495-500, 5 FIGS.

Short Note

Use of integrated energy spectra for thin-layer recognition

Andrew Marangakis *, John K. Costaini, and Cahit Coruhg

INTRODUCTION

The detection and resolution of thin beds are important In this paper, the signature of a thin bed is studied in the
problems in reflection seismology. A thin bed is defined as one frequency domain. It is shown herein that changes in bed
for which the two-way traveltime thickness is less than the thickness of less than the tuning thickness can be detected.
tuning thickness for the incident wavelet. Reservoir thickness
and shape are of critical importance in estimating hydrocarbon GENERATION OF SYNTHETIC THIN-BED MODELS
reserves. Previous studies of thin-bed resolution have focused
attention on the time domain, i.e., on the properties of the An input zero-phase, sine wavelet similar to that of Vibroseis
seismic trace and wavelet. Widess (1973) defined a thin bed as data was derived from autocorrelation of a l&60 Hz linear
one whose thickness is less than h/8, where h is the dominant sweep tapered at each end used to model thin beds of varying
wavelength of the seismic wavelet in the thin bed. He based this thickness and number. The resulting synthetic trace, the reflec-
criterion on the observation that for thin beds of thickness less tivity function, and the wavelet were each separately trans-
than or equal to h/8, the reflected wavelet is essentially the formed to the frequency domain and their amplitude spectra
shape of the derivative of the incident wavelet. were examined in order to develop recognition criteria.
More recent studies on thin-bed resolution were given by The single-bed model, herein referred to as a dipole, con-
Widess (1973, 1980), Koefoed and de Voogd (1980) Kallweit sists of two reflection coefficients of equal amplitude and op-
and Wood (1982), and de Voogd and den Rooijen (1983). de posite polarity. The spacing between the reflection coefficients
Voogd and den Rooijen studied the effect of spectral bandwidth represents the thickness of the layer. For a reflectivity function
on interface resolution and thin-layer response. Kallweit and corresponding to a single dipole, r(t) = h(t) - h(t - T), where
Wood placed the practical limit of resolution at a greater 8(t) is the Dirac delta function. The Fourier transform S(o) of
thickness than Widess. Observing that apparent thickness, as the output trace is
defined by peak-to-trough time does not stabilize into the
S(0) = W(w)[ 1 - exp (- iwr)], (1)
derivative waveform except at the limiting value of zero thick-
ness, Kallweit and Wood concluded that the practical limit of where W(o) is the Fourier transform of the seismic wavelet and
resolution, for a bandwidth exceeding two octaves, is approxi- r is the dipole thickness in two-way traveltime.
mately 1/(1.4s,), where f, is the upper frequency bandwidth One of the parameters that affects the amplitude spectrum of
limit of the wavelet. This limit is approximately the tuning a reflection from a thin bed is the thickness of the bed. The
thickness T/2 for maximum constructive interference from the amplitude spectrum of a single dipole has a periodicity of l/z in
top and bottom of a thin bed (Sengbush et al., 1961). For a the frequency domain. At the tuning thickness, the dominant
wavelet with an upper frequency limit of 60 Hz, the Kallweit frequency is the same for both the wavelet used and the first
and Wood limit corresponds to l/(1.4 * 60) = 12 ms (two-way period of the reflectivity function. For a dipole of thickness less
traveltime) which is approximately equal to the tuning thick- than tuning thickness, the maximum amplitude in the first
ness T/2; however, because of a tapered Vibroseis@ sweep of period of the reflectivity function occurs at a higher frequency
l&60 Hz, the dominant frequency of our wavelet is 35 Hz. than the dominant frequency of the wavelet. For a thickness
Using Widess criterion, this corresponds to a wavelet breadth greater than tuning, it is at a frequency less than the dominant
T of 29 ms which results in approximately 14 ms for the tuning frequency of the wavelet (Marangakis, 1983).
thickness T/2. For a one-layer model, a progressively thinner bed corre-
sponds to a progressively higher peak frequency in the spec-
@Registered trademark of Conoco, Inc. trum of the wavelet reflected from the thin bed relative to the

Manuscript received by the Editor January 30, 1984; revised manuscript received September 17, 1984.
*Formerly Virginia Polytechnic Institute and State University, Blacksburg, VA; presently ARC0 Exploration Co., Denver, CO 80202.
IVirginia Polytechnic Institute and State University, Regional Geophysics Laboratory, Department of Geological Sciences,Blacksburg, VA 24061.
University of Istanbul, Jeofizik Bolumu, Vezniciler, Istanbul, Turkey.
0 1985 Society of Exploration Geophysicists. All rights reserved.

495
Marangakis et al.
DIPOLE THICKNESS: 16 #DIPOLES: I

100
90
80
70
60
50
40
30
20
10
l-l
uO 20 40 60 80 100

FREQ,HZ.
DIPOLE THICKNESS:8 #DIPOLES: I

100 I
90
80
70
60
50
40
30
20
10 .
0 I
DIPOLE THICKNESS: 14 #DIPOLES: I

100 I
90 .
80 .
70
60
50 .
40
30
20
10
0 I
0 20 60 80 100

FREQ,HZ.
FIG. 1. Integrated energy spectra of the one-layer models. (a) 16 ms, (b) 8 ms (tuning), (c) 14 ms dipole thickness. Solid lines are the
spectra for the synthetic traces, dashed lines are those for the incident wavelet.
Thin-layer Recognition 497

dominant frequency of the incident seismic wavelet. The shift to (high-slope areas) corresponding to peaks in the amplitude
higher frequencies for a reflection from a thin bed is a predict- spectrum at higher frequencies. Integrated energy spectra are
able consequence of differentiation of the wavelet. Thus, the affected not only by spectral amplitude, but by spectral shape as
signature of a thin layer in the frequency domain can help to well.
interpret data if time-domain waveforms are not diagnostic. In For two dipoles where the dipole separation is equal to the
order to accomplish this, the energy of the waveform by means dipole thickness, the frequency of the peak amplitude is un-
of Bessels, or Parsevals, equality (Papoulis, 1962) is used to changed at the tuning thickness (Marangakis, 1983); however, a
obtain integrated energy spectra as a function of frequency. It is narrower spectrum with side lobes is formed as a result of
shown that integrated partial energy plotted against frequency squaring the spectrum of the reflectivity function. The number
and herein defined as INTegrated ENergy Spectra (INTENS) and amplitude of the side lobes depend upon dipole thickness
can be used to detect changes in thickness of thin beds and and separation between the dipoles. These side lobes may mask
lenses that are not immediately recognizable in the time the high-frequency shift of the maximum amplitude of the
domain. reflected wavelet for dipoles thinner than the tuning thickness.
Further increase in the number of dipoles (thin beds) results in
INTEGRATED ENERGY SPECTRA additional lobes. INTENS for more than one thin dipole show
energy confined to a narrower frequency band than the incident
The normalized integrated partial energy, or percent inte- wavelet because of multiplication of the wavelet spectrum by an
grated energy spectrum E(f) is defined for the bandwidth of the increasing number of dipole spectra. Addition of side lobes to
wavelet used as (Marangakis, 1983), the spectrum of two or more thin beds may result in a spectrum
60
that is similar to one obtained from a thick bed because the
df A(f dL spectrum of a thick bed has a shorter period than that of a thin
E(f) = 100 W)A*(f) (4
s10 il 10 bed. Amplitude maxima occur at frequencies of n/2T Hz, n = 1,
where A( f ),f s 60, is the amplitude of the spectral component 3, 5, . . . . Differentiation between these two models, however,
and A* is its complex conjugate. becomes possible after inspection of INTENS curves
Integrated energy spectra for dipole thicknesses of 8, 16, and (Marangakis, 1983).
14 ms (tuning thickness) are shown in Figure 1. Comparison of
INTENS with conventional amplitude spectra confirms that a INTEGRATED ENERGY SPECTRA FOR MULTITRACE
high-amplitude peak in the amplitude spectrum transforms to a MODELS
steep slope in INTENS (Marangakis, t983). As expected, low-
amplitude valleys in amplitude spectra transform to plateaus of The above discussion has focused on interpretation of a
relatively low slope on INTENS. Compared with the incident single seismic trace in terms of integrated energy spectra. Con-
wavelet, INTENS for the thin dipole shows points of inflection ventional displays of seismic data are in the form of record
v, = 30 km/set, v, = 5.5 kmlsec, V3 = 4.0 km/set V, = 3.0 kmlsec, V, z 55 km/set, V3 = 4.0 kmlsec

60
I2

N 40
I
6 30
K
u_ 20

w____ I ! 0 1 I / 0.0 I ;
IO 20 30 2 IO 20 30

THICKNESS MS THICKNESS MS
(a) 04

FIG. 2. Integrated energy spectrum patterns for a single wedge: high-velocity wedge has monotonically increasing thickness which
increases to the right by 2 ms/trace after trace three, to a maximum of 34 ms. (a) Reflectivity function only which is shown above
the pattern plot, (b) after convolution of the reflectivity function with the incident wavelet.
498 Marangakis et al.

sections that reveal the two-dimensional (2-D) subsurface ge- corresponding to the transition zone. The reflectivity function
ometry of lenses, thin beds, sedimentary wedges, etc. In order to r(t) is defined in the usual way as r(t) = d/dt {ln [o(t)]} = con-
observe changes in the thickness of thin bed models in two stant, where In [r(r)] is the velocity function of Sengbush et
dimensions, a frequency-domain representation is chosen for al. (1961). As discussedby Sengbush et al. (1961) such a velocity
which the dipole (bed) thickness and frequency are the abcissa function returns two integrated wavelets, one from the top and
and ordinate, respectively, with the partial integrated energy one from the bottom of the transition layer. This holds for zone
spectrum contoured to give lines of equal percent of total thicknesses down to the tuning thickness T/2 where maximum
energy. Automatic contouring was done using computer pro- constructive interference of the two reflected integrated wave-
gram GPCP (1971). lets occurs. The amplitude spectrum of a wavelet reflected from
Figure 2a shows INTENS contours of reflectivity functions a transition zone has a lower dominant frequency than the
for a model of a single wedge with a monotonic increase in dominant frequency of the incident wavelet because of integra-
thickness from left to right. Velocities used to form the high- tion which corresponds to division by iw in the frequency
velocity wedge are 3.0, 5.5, and 4.0 km/s. The configurations of domain. The INTENS representation of the reflectivity func-
INTENS contours near the thin edge are distinctive in both tion of a monotonically thickening transition-wedge model is
Figure 2a (reflectivity function only), and Figure 2b (reflectivity shown in Figure 3a. A geologic model corresponding to this
function after convolution with the incident wavelet). The ex- might be that of a elastic wedge that becomes progressively
cursions of the contours of partial integrated energy are sub- coarser and more poorly sorted with depth. An equally impor-
dued in Figure 2b because of the smoothing effect of wavelet tant model is that of a structurally truncated (i.e., by a decolle-
filtering, but the edge of the thinning bed is clearly defined. The ment) linear transition interval. The contoured INTENS show
pattern formed by the minimum and maximum INTENS con- a shift of energy to lower frequencies as the zone thickens. The
tours for thicknesses greater than 2 ms at higher frequencies is same model after convolution with the incident wavelet is
also distinctive. The direction of thinning is clearly defined by shown in Figure 3b. The general pattern of contours is that the
the slope of the INTENS contours and by the shifting of INTENS for the transition wedge are enriched in lower fre-
frequency minima and maxima. A low- or high-velocity wedge quencies by the transition zone thickness (Figure 3b), and the
has a pattern different from that of a step-wise increasing direction of thinning is clearly indicated. Near the tuning thick-
velocity wedge (Marangakis, 1983). This suggeststhat the pres- nessthe higher percent INTENS contours are characteristically
ence and shape of a sedimentary wedge can be confirmed by inclined in the direction of thickening. The thin edge of the
transformation of amplitude spectra to INTENS, and that wedge is defined by stabilized contours of zero slope.
integrated energy spectra can be useful supplements in the INTENS patterns and reflectivity functions for thin lenses
interpretation and modeling of stratigraphic and structural are shown in Figures 4a, 4b and Figures 5a, 5b. In Figure 4a a
traps. step-wise increasing velocity lens is modeled and therefore no
A model of a velocity transition in which the velocity changes tuning effect is apparent, which makes recognition of the lens in
linearly with depth according to the equation u = u,, + az was the time domain almost impossible. Figure 4b shows the model
also examined. The model is represented in the time domain by of Figure 4a after convolution with the incident wavelet. Al-
a reflectivity function that is constant over the time interval though perturbations in the excursions of the INTENS con-

. i/i
2 IO

60 60

N 40
I
ti 30
E
Ll_ 20

10
I I

2 IO 20 30 2 IO 20 30

THICKNESS MS TH tCKNESS MS
(4 (b)

FIG. 3. Integrated energy spectra patterns for a transition zone: velocity transition-zone has monotonically increasing thickness. (a)
Reflectivity function only which is shown above the contour plot, (b) after convolution of the reflectivity function with the incident
wavelet.
Thin-layer Recognition 499

V, q 3.0 km/set, V, = 4.0 kmlsec. V3 = 5.5 km/set V, = 3.0 km/set, V, - 4.0 km/set, V, = 5.5 kmhec

0 12 12 0

60 60

50

7
2 40

ci 30
ii
IL 20
10
I------ 0.O
/
0 III0 I I
0 12 12 0 0 12 12 0

THICKNESS MS THICKNESS MS

(4 (b)

FIG. 4. Integrated energy spectra patterns for a lens model: the reflectivity function represents a step-wise increasing velocity where
maximum thickness of lens is 12 ms. (a) Reflectivity function only which is shown above the contour plot, (b) after convolution of
the reflectivity function with the incident wavelet.

V, = 3.0 km/set, V2 =5.5 km/src. V, = 4.0 km/sac V, = 3.0 kmlsec, V, = 55 kmlsec, V, = 4.0 kmlsec

60 60
Ill t
0

50 50

$l 40 y 40

ci 30 ci 30
trw z
IL 26 LL 20
10 c. c 10

0 I I I I -I__-.. I I I
0
0 20 20 0 0 20 20 0

THICKNESS MS THICKNESS MS
69 (b)

FIG. 5. Integrated energy spectra patterns for a lens model: the reflectivity function represents a high-velocity lens where maximum
thickness of lens is 20 ms. (a) Reflectivity function only which is shown above the contour plot, (b) after convolution of the
reflectivity function with the incident wavelet.
500 Marangakis et al.

tours are smaller because of the large radius of curvature of the different, but diagnostic, patterns. Although the discussion is
lens surfaces, it is still possible to detect the edges of the lens as based on the use of a sine wavelet, the methods of analysis used
well as the directions of thinning. and the conclusions reached are believed to be valid for any
A high-velocity lens model is also used and INTENS pat- seismic wavelet.
terns are given in Figures 5a and Sb. The direction of thinning
is clearly indicated by the slope of the contours, and the termi- ACKNOWLEDGMENTS
nation points of the lenses are clearly defined. Comparison of
Figure 4b and Figure 5b suggests that a high- or low-velocity This study was supported by an Atlantic Richfield Fellow-
lens can be discriminated from a step-wise increasing velocity ship to A. Marangakis while at Virginia Tech, by Nuclear
lens. In both figures the edges of the lenses and the direction of Regulatory Commission contract no. NRC-04-75-237, and Na-
the thinning are preserved in spite of the filtering effect of the tional Science Foundation grant no. EAR-8009549-2 to L.
wavelet. Glover. III and J. K. Costain.

CONCLUSIONS REFERENCES
de Voogd, N., and den Rooijen, H., 1983, Thin-layer response and
Reflectivity functions for models, i.e., the thick bed, thin bed, spectral bandwidth: Geophysics, 48,12-18.
and velocity transition, can be suitably combined to model any GPCP-A General Purpose Contouring Program, 1971, Users
manual : California Computer Products, Inc.
vertical and horizontal velocity distribution of arbitrary com- Kallweit, R. S., and Wood, L. C., 1982, The limits of resolution of
plexity. Two-dimensional integrated energy spectra can then be zero-phase wavelets: Geophysics, 47, 1035-1046.
Koefoed, O:, and de Voogd, N., 1980, The linear properties of thin
contoured for the synthetic data and compared with the inte-
layers, with an application to synthetic seismograms over coal
grated spectra of real seismic data. seams: Geophysics, 45,12541268.
Although further studies of INTENS patterns are underway Marangakis, A., 1983: Interpretation of reflection seismic data by
analysis of cumulative energy spectra: MS. thesis, Virginia Poly-
with real data and with noise added to synthetic data, at this technic Institute and State University, Blacksburg, VA, 80 p.
time we conclude that a display of integrated energy spectra Papoulis, A., 1962, The Fourier integral and its applications: McGraw-
can aid in the interpretation of seismic data where stratigraphic Hill Book Company.
Sengbush, R. L., Lawrence, P. L., and McDonal, F. J., 1961, Interpreta-
and structural traps are important. Contouring of integrated tion of synthetic seismograms: Geophysics, 26, 138-157.
energy spectra is independent of changes in relative amplitude Widess, M. B., 1973, How thin is a thin bed? Geophysics, 38, 1176
1180.
between adjacent traces, and, by the time-shifting theorem, is
- 1980, Generalized resolving power and system optimization:
unaffected by static shifts between traces. The method is sensi- Presented at the 50th Annual International SEG Meeting, Novem-
tive to spatial sampling, however, and for large distances be- ber 19, Houston.
tween common depth points INTENS contours may show