Reflection Seismology Concepts

CHAPTER 5. REFLECTION SEISMOLOGY CONCEPTS

(“SIMPLE SEISMICS”)
Overview Consider a sound wave sent downward into the
earth by an explosion or vibratory disturbance. In
Most of geophysical interpretation involves exploration seismology, these waves are usually
the use of surface reflection seismology. generated by dynamite or vibroseis sources.
Although other seismic techniques such as
refraction seismology, borehole seismology If these waves encounter a discontinuity in
(cross-borehole and VSPs), and teleseismic the acoustical impedance (product of rock
methods are sometimes used, reflection density and seismic velocity), this results in a
seismology remains the workhorse of reflected wave. The ratio of the reflection
exploration geophysics. There are many good amplitude to the incident wave amplitude is
textbooks that describe theoretical seismology given by the reflection coefficient. By solving
(Aki and Richards, 1980) exploration seismology the boundary conditions for layer discontinuities
(Sheriff and Geldart, 1995), and seismic data for an acoustical wave at vertical incidence (as
processing (Yilmaz, 2001). Our purpose is not to described in Robinson and Treitel, 1980), we can
duplicate the excellent information on derive the reflection coefficient as being the
exploration seismology in those books. Rather, following.
our purpose in this chapter is to explain some
basic principles of reflection seismology. These ρ 2 v 2 − ρ1v1
involve reflection amplitudes and traveltime. r=
They are necessary for seismic interpretation and ρ 2 v 2 + ρ1v1
we leave the details to these other textbooks.
where ρ1v1 represent the density and velocity of
Reflection Coefficients the first layer in which the wave travels and
ρ 2 v2 represent the density and seismic velocity
Reflections from within the earth arise due
of the second layer that the wave encounters.
to changes in seismic velocity and rock density.
(See Figure 5.1 which was taken from Robinson,
The simplest model involves the consideration of
1967.)
a set of reflections from a series of rock layers.

Figure 5.1: Reflected and transmitted waves from an incident wave (from Robinson, 1967).

5-1
Reflection Seismology Concepts

Convolutional Model y1 = w0 r1 + w1 r0 .
The convolutional model provides a
The third sample and fourth samples of y(t)
description of the primary reflections in a
seismic trace. It is basically represents a way to are given by y 2 = w1 r1 + w2 r0 and
do the bookkeeping on the set of reflected y 3 = w2 r1 .
seismic pulses in a seismogram. For the time
being, we will ignore multiple reflections and
save this discussion for later. We note that the seismic trace consists of
sums of products. Note that the sum of the
We know that our seismogram is not a series subscripts of terms in the series have the same
of knife-sharp arrivals. There is a finite response sum. This sum is equal to the index of the trace
time for any seismic disturbance since the earth sample index. We can generally write the time
does not respond instantaneously to an impulsive series for the seismic trace in terms of the source
disturbance. The earth’s displacement includes wavelet and reflectivity time series by the
compression, followed by a rebound or following
rarefaction. Since seismometers generally
respond to the velocity of this disturbance, the y t = ∑ wτ rt −τ .
source pulse is actually a time derivative of this
displacement. The source pulse or response of This equation expresses the discrete
the earth to an impulse is often termed the
convolution of a time series, wt , with a time
seismic source wavelet.
series, rt , and it is symbolically expressed as
The recorded seismic trace can be viewed as
a series of reflected source wavelets which reach
a receiver at different delay times. In other words y t = wt * rt .
the recorded trace is a sum of echoes. Let us
view this in physical terms and show that the The process of convolution can be described
mathematical description is a convolution. in words as “flip, shift, multiply, and add”. That
is, the order of one of the time sequences is
For the sake of simplicity, consider an earth flipped or reversed in time. It is then shifted by
with three layers of equal traveltimes, with two- some amount; t, the terms in the time series are
way traveltimes equal to the sample interval. multiplied and the products are added together.
That is, we consider two reflecting boundaries in The same process could be achieved by
which the reflection time-series is ( r0 , r1 ) . multiplication of polynomials and identifying the
coefficients.
Consider a three-term source wavelet sequence
whose amplitudes are described by a time series
The convolutional process can also be
( w0 , w1 , w2 ) which is sent down into the earth. achieved by polynomial multiplication by using
Let us consider the seismic trace, y(t), a time z-transforms. We can compute a convolution by
series of primary reflections recorded by a applying the above formula or we can achieve
geophone just above the top layer. the same result by use of polynomials known as
z-transform. The z-transform of a time series is
The first sample is given by the start of the simple to compute. For example, if we have a
wavelet reflected off the top layer. The first time series for the wavelet given by
sample of the trace is given by ( w0 , w1 , w2 ) , we write down its z-transform
by considering a polynomial whose coefficients
y 0 = w0 r0 . of powers of z are given by the time series
values. In other words, the z-transform for the
The second time sample is given by the sum wavelet time-series is given by
of two arrivals, the first wavelet value bouncing W ( z ) = w0 + w1 z + w2 z . Likewise, the z-
off the second reflector and the second wavelet
transform for the reflectivity series in our
value bouncing off the top reflector:
example is given by R ( z ) = r0 + r1 z . We note

5-2
 Reflection Seismology Concepts

that the z-transform of the seismic trace is given will be discussed further in later chapters where
by: it is shown that predictive deconvolution and
normal moveout followed by stacking can prove
Y ( z ) = W ( z ) R( z ). useful in multiple suppression

We can verify that this is the case by

multiplying these polynomials and comparing
the coefficients of the resulting z-transform to
those derived in the earlier paragraph.

− iω
If we set z = e in the previous
expression, the z-transforms become discrete
Fourier transforms. Therefore, it can be seen that
convolution in the time domain can be achieved
through multiplication of Fourier transforms in
the frequency domain.

We have developed a mathematical and

physical basis for convolutions and shown their
relationship to Fourier transforms.

Before we leave the convolutions, let us

briefly revisit the seismic trace model. In our
discussion, we considered only primary
reflections. However, we know that multiple
reflections exist in seismograms, and in fact, .
these can be quite troublesome. When can
multiples be safely ignored and when do we need Figure 5.2: Hyperbolic traveltime vs distance
to account for them or take steps to suppress curve (above) are shown for reflections from a
them? horizontal boundary. This figure is from
Lowrie (1997).
Since multiples involve two or more
reflections off layer discontinuities, their Traveltime, Velocity, and Normal
amplitudes are related to products of reflection Moveout
coefficients, one reflection coefficient for every
bounce off a reflector. For typical values of rock
It has been said that the most useful equation
densities and seismic velocities, most seismic
reflection coefficients are not greater than 0.1, in seismology is d = vt , or distance = velocity
which means that multiples would typically have * time. In exploration seismology problems, we
products of reflections coefficients which were wish to determine the distance of drilling targets
0.01 or less. Nevertheless, multiples can be by using seismograms whose responses are
troublesome, especially in the marine case where measured in time. Hence, velocity of seismic
the air-water interface has a reflection wave propagation provides the essential link
coeffiecient of nearly 1.0 and the water-rock between the desired objective and the time
interface which has reflection coefficients which measurements of seismic energy.
are much greater than 0.1. We can model the
response of a layered medium for primaries and It can be argued that seismic velocity
multiples by using the techniques outlined by estimation is the most important problem in
Robinson and Treitel (1980). exploration seismology since it impacts both
processing and interpretation of seismic data. In
Seismic processing can allow us to attenuate reflection seismology, the velocity is generally
multiples by using the fact that multiples estimated by a technique known as normal
generally have a predictable nature and they moveout.
generally have a slower apparent velocity than
primaries arriving from deeper reflectors. This

5-3
Reflection Seismology Concepts

To understand normal moveout requires a mathematical expression for NMO:

simple knowledge of Pythagoras’ Theorem. In
Figure 5.2 shown above (from Lowrie, 1997), we 1
X2
consider a primary reflection that travels from a ∆T = T0 (1 + 2 2 ) 2 − T0
seismic source down to a flat seismic reflector v T0
and back to a receiver at the surface. Let the
distance from the source to receiver be denoted The values of the NMO, ∆T , as a function
by X, the depth to the reflector be Z and the of offset and normal incidence time, can be used
distance from the source to the reflecting point to estimate the velocity, v . There is a convenient
be S. For a flat reflector, the downward path is
relationship between velocity and NMO that
the mirror image of the upward path. By
exists for values of offset which are small
Pythagoras’ Theorem, we note in the ray path
relative to depth. For such cases we can use the
figure that we have a right-angle triangle whose binomial approximation that
sides are related by
1
S 2 = Z 2 + ( X / 2) 2 . X2 2 X2
(1 + ) ≅ 1 + .
v 2T02 2v 2T02
On a seismogram, we measure the two-way,
reflection time T, for a given source-receiver Make this substitution into the previous
distance, X. For the case of zero-offset where expression for ∆T , we obtain
X=0, we measure the two-way normal-incidence
reflection time of T0 . The distances in the X2
Pythagorean relationship can be rewritten in ∆T ≅ .
2v 2T0
terms of traveltimes using S = vT / 2 and
Z = vT0 / 2 :
For such cases, the velocity can be directly
determined from the NMO, normal-incidence
(vT / 2) 2 = (vT0 / 2) 2 + ( X / 2) 2 . reflection time and offset from the following
expression:
This produces the well-known hyperbolic
relationship between time, distance, and velocity X
v= .
for a primary reflection. In other words, for flat 2∆TT0
reflectors, reflections will have the traveltime
trajectory whose shape is a hyperbola governed
by the following equation: At first glance, it might seem that a single
flat layer model would be too simplistic for use
in a real world situation of many layers.
T 2 = T02 + X 2 / v 2 . However, an analysis of the multilayered case by
Taner and Koehler (1969) showed that the
By inspection of the above equation, it is traveltime equation for several layers is in the
evident that the difference between the same form as a single layer, except for the fact
traveltime for non-zero offset, T , and the that velocity, v , is replaced by the root mean
normal incidence time, T0 , is governed by the square velocity, v rms . That is,
offset distance and the velocity. This difference
is termed the normal moveout (abbreviated as X2
NMO), and is given by ∆T = T − T0 . T 2 = T02 + 2
.
v rms
Much of classical velocity estimation is
given by NMO analysis. To understand this more Also, if we determine the root mean square
completely, let us examine the velocity for two successive reflectors, we can use
Dix’s equation to determine the interval velocity
of the layer between these reflectors. The reader
is referred to the classical paper by Dix (1955).

5-4
 Reflection Seismology Concepts

Seismic Data Acquisition References

In initial presentations of this interpretation Aki, K., and Richards, P.G., 1980, Quantitative
course, the authors developed a short chapter on Seismology: Theory and Methods: Freeman
data acquisition. However, during that same and Company, San Francisco, CA.
year, there was an SEG monograph published by Dix, H., 1955, Seismic velocities from surface
Evans (1997) which covered in more detail measurements: Geophysics, 20, 68-86.
topics in seismic acquisition. Evans, B.J., 1997, A Handbook for Seismic Data
Acquisition in Exploration: Society of
Seismic Data Processing Exploration Geophysicists, Tulsa, OK
Lowrie, W., 1997, Fundamentals of Geophysics:
While the publication by Evans (1997) Cambridge University Press, Cambridge,
covers seismic data acquisition, the topic of U.K.
seismic processing is covered in detail by Robinson, E.A., 1967, Multichannel Time Series
Yilmaz (2001). After the assignment of geometry Analysis with Digital Computer Programs:
and demultiplexing of field tapes, the steps of Holden-Day, San Francisco.
statics corrections, velocity analysis and NMO Robinson, E.A. and Treitel, S., 1980,
corrections, the summation or “stacking” of Geophysical Signal Analysis: Prentice-Hall,
common reflection point data is a very important New York, NY (reprinted by Society of
step for the reduction of noise. The steps of Exploration Geophysicists, Tulsa, OK).
deconvolution and migration basically improve Sheriff, R. and Geldart., L., 1995, Exploration
seismic resolution in the temporal and spatial Seismology: Cambridge University Press,
sense. These steps are covered in other sections Cambridge, UK.
in the course notes. Taner, T. and Koehler, F., 1969, Velocity
Spectra – Digital computer derivation and
applications of velocity functions,
Geophysics, 35, 551-573.
Yilmaz, O., 2001, Seismic Data Processing:
Society of Exploration Geophysicists,
Tulsa, OK.

5-5