Seismic Reflection
Seismic Reflection
Two Examples
Air gun record from the Gulf of Patras, Greece, showing
Holocene hemipelagic (h) and deltaic (d) sediments
overlying an irregular erosion surface (rockhead, RH) cut
into tectonized Mesozoic and Tertiary rocks of the
Hellenide (Alpine) orogenic belt. SB: sea bed reflection;
SBM1 and SBM2: first and second multiples of sea bed
reflection; RHM1: first multiple of rockhead reflection.
A seismic section from the northern Amadeus basin,
central Australia, illustrating a depositional sequence
bounded by major unconformities.
From Kearey, Philip and Micheal Brooks, An
Introduction to Geophysical Exploration. 2nd ed.
Blackwell Science: 1991.
In order to generate the previous images there are numerous operations that need to be applied to the
data. Much of the data processing is tied to the hypothesis that the earth's properties vary most
strongly in the vertical direction. As background we first review the principles of a reflection
seismogram and then consider travel time curves that arise from a horizontally layered earth.
An important concept not covered in these web-notes is the "Optimum-offset" method of seismic
reflection data acquisition.
The normal incidence seismic trace is obtained by the convolution of a seismic wavelet (input pulse)
with the reflectivity function. The amplitude of each spike on the reflectivity function is equal to the
value of the reflection coefficient that corresponds to a particular boundary. (In reality this value is
further altered by the transmission coefficients). The times for each reflection event are obtained by
knowing the layer thickness and velocities. Each impulse on the reflection function generates a scaled
replication of the seismic wavelet. The composite of all of the reflection events generates the seismic
trace.
The reflection seismogram viewed as
the convolved output of a reflectivity
Function with an input pulse.
From Kearey, Philip and
Micheal Brooks, An
Introduction to
Geophysical
Exploration. 2nd ed.
Blackwell Science:
1991.
Notice how the negative reflection coefficients chage the polarity of the signal recoreded at the
receiver.
A synthetic seismogram.
From Kearey, Philip and Micheal Brooks,
An Introduction to Geophysical Exploration.
2nd ed. Blackwell Science: 1991.
Vertical resolution
The vertical resolution of a seismic data sets describes how thin a layer can be before the reflections
from its top and its bottom become indistinguishable. This depends upon the shape of the seismic
source wavelet. The shape, or more particularly the signal wavelength depends upon the frequency
and the velocity of the materials. Recall that wavelength = velocity * time.
or
In fact, the theoretical minimum thickness is 1/4wavelength. In other words, careful processing can
extract the pulses if they ovelap by half a wavelength, but this theoretical minimum is never really
achieved in practice.
How does this thickness in terms of wavelength translate into real distances? As noted above,
wavelength depends upon frequency of the signal and on velocity of the medium. Therefore the
vertical resolution of a survey depends upon both the signals used and the materials themselves. The
following image with its table of velocities and frequencies illustrates.
Finally, recall that as energy propagates into the ground, higher frequencies attenuate faster than
lower frequencies. Since seimsic signals consist of a range of frequencies (i.e. they have a broad
spectrum), their spectral character will change as they travel. The result is that resolution of a survey
is poorer at greater depths.
There are several common geometries for the "instrument" - that is the configuration of sources and
receivers.
1. The simplest is a "common offset" array in which the source and receiver distance is always
the same. This is the way most GPR surveys are performed. Signals at one receiver are
recorded from the shot (or source), then the shot-receiver are moved to a new location and the
process is repeated.
2. A variation on this is the "optimum offset" array in which many receivers are recorded for
each shot. Then the shot and receiver string are moved to a new location and the process is
repeated. This works well if targets are relatively shallow (perhaps less than 100m or so) AND
if reflecting horizons are clear and distinct (ie if the acoustic properties vary significantly and
sharply across the boundaries).
3. The most common, though most expensive, form of surveying is the "multichannel reflection
survey". As for optimum offset surveys, many receivers are used for each shot. The difference
is that the survey system is carefully designed so that each reflecting point in the subsurface is
sampled more than once. In other words, the objective is to obtain several different echoes
(reflections) from identical subsurface points. This type of surveying involves some care in
setting up the field work, and some effort in the processing steps. Details will be covered next.
A draped seismic record of a shot gather from a split spread (courtesy Prakla-Seismos GMBH).
Sets of reflected arrivals from individual interfaces are recognizable by their characteristic
hyperbolic form. The late-arriving, high amplitude, low-frequency events, defining a
triangular-shaped central zone within which reflected arrivals are masked, represent surface
waves (ground roll).
Both from Kearey, Philip and Micheal Brooks, An Introduction to Geophysical Exploration.
2nd ed. Blackwell Science:1991.
A split spread seismic record is shown above right. The seismic traces all belong to a single source
and hence this is referred to as a "Common Source Gather". The first arrivals are direct or critically
refracted arrivals. Reflection hyperbolae from numerous boundaries are observed. The strong energy
in the triangular central portion is ground roll caused by surface waves. It masks the reflection events.
Fundamental procedure
In order to benefit from gathering several echoes from each reflecting point there are numerous
operations that need to be applied to the data. Much of the data processing is tied to the hypothesis
that the earth's properties vary most strongly in the vertical direction. The table shown next illustrates
the fundamental procedural concept underlying the creation of a final seismic reflection section:
Objective:
We want to characterize the earth using echo sounding with this geometry:
(1)
(2)
Logistics:
However, surveying with one shot and many geophones is more cost-effective:
(3)
Therefore:
Field work must be arranged as follows. Blue italics text refers to the figure below.
1. We gather data using the geometry of type (3).
Data from one shot into many geophones ("common shot data") are shown below under the
label Shot Record 14.
2. Next, sort many of these "common shot data" so that traces appear as if gathered using the
geometry of type (2).
All traces that reflected under one location are collected into a "common mid point gather",
one from each of many common shot data sets. See the panel under CMP loc. 27.
3. Stack these traces to produce one trace which represents measurements obtained using the
desired geometry of (1). This is the CMP (common mid point) trace.
This is the single trace next to the CMP panel.
4. Then many of these CMP traces are combined into one cross section of the earth's structure.
Traces are labelled CMP number, and the one trace shownis identified with arrows.
5. Interpretation in terms of geology is the final step.
The procedures to accomplish these steps will be explained in subsequent pages of the notes.
where n is the moveup rate in units of geophone spacing. "Moveup rate" is in fact (shot spacing)/
(geophone spacing). For example, if geophones are 2m apart and shots are employed every 4 metres,
then the moveup rate is n=4/2=2. This can be less than one if there are shots set more often than
geophone spacing, a practice that is sometimes done in marine seismology, especially ocean bottom
profiling.
The schematic below shows a single ended spread with 8 geophones and moveup rate of n=2.
We see that each point in the subsurface is sampled only twice. Notice that the distance between the
reflection points in the subsurface is half the geophone spacing.
In the above diagram to is the 2-way vertical travel time. It is the minimum time at which a reflection
will be recorded. The additional time taken for a signal to reach a receiver at offset x is called the
"Normal Moveout" time, T. This value is required for every trace in the common depth point data
set in order to shift echoes up so they align for stacking. How is it obtained? First let us find a way of
determining velocity and t .
0
For this simple earth structure the velocity and layer thickness can
readily be obtained from the hyperbola. Squaring both sides yields,
The sequence of algebra shown to the right has only one complicated step - a binomial
expansion must be applied to obtain a simple relation without square roots etc.
The approximation is valid so long as the source-receiver separation (or offset) is
"small" which means much less than the vertical depth to the reflecting layer
(i.e. x 0<< Vt ). The result is a simple expression for normal moveout.
Each echo can be shifted up to align with the t0 position, so long as the trace position
(x), the vertical incident travel time (t 0) and the velocity are known. Velocity can be
estimated using the slope of the t2 - x2 plot,or with several other methods, which we
will discuss in pages following.
(b)
As outlined in the figure above, the reflection curve for small offsets is still like a hyperbola, but the
associated velocity is Vrms , not a true interval velocity.
By fitting hyperbolas to each reflection event one can obtain (t n, Vnrms) for n = 1, 2, ... The interval
velocity and layer thickness of each layer can be found using the formula below.
These formulae for the interval velocity and thickness of the n th layer are directly obtainable from the
definition of Vnrms given above.
The RMS velocity for the nth layer is given by,
where vi is the velocity of the ith layer, and i is the one-way travel time through the ith layer.
From Kearey, Philip and Micheal Brooks, An Introduction to Geophysical Exploration. 2nd ed. Blackwell
Science: 1991.
4. Finally, stacking the normal moveout corrected traces generates a single trace. Each
trace corresponds to a zero-offset trace, that is, the seismic trace that would have been
recorded by a receiver that is coincident with the source.
As an example consider the two CMP gathers in the figure below left (click for a larger version). The
most prominent seismic wavelet at times between 50 and 70 ms is a bedrock reflection from about 9
m below the surface. The geophone offsets were 3.7 m (12 ft) for the nearest traces and 17 m (56 ft)
for the farthest trace with 1.22 m (4 ft) between geophones. The reflection even of interest is at about
55 msec.
In the figure above right, the CMP gather for point 988 has been moveout corrected using different
velocites. The moveout correction using a velocity of 1225 ft/s (328 m/s) causes the reflection events
to occur at about the same time for all traces. Stacking these signals will produce a high quality
reflection signal. Conversely, a velocity of 1075 ft/s is too small and produces too large a correction
at far offsets.
To show what happens if the wrong velocity is choosen for
carrying out the normal moveout correction, five 12-fold
CMP traces are shown (right) processed with three different
velocities. When the velocity is too low, the frequency of
the reflection wavelet is lowered and it is therefore depicted
too shallow on the seismic section. When the velocity is too
high the frequency dereases and the reflection wavelet is
depicted at too large a time on the seismic section. The
correct velocity gives the correct position for the wavelet
and preserves the high frequencies which allow best
resolution of small features and thin beds. Correct velocity
is about 373 m/s (1225 ft/s).
For each CMP gather, the velocity analysis is done at many times.
A common method of displaying this work can be seen in the example to the
right (click the image for a larger version). The velocity spectrum defines the
stacking velocity as a function of reflection time. The energy (cross-power )
function used to define the stacking velocity is calculated for a large number
of narrow time windows along a seismic section. The velocity spectrum is
typically displayed alongside the draped section. In the example shown here,
stacking velocities for all times less than about 3 seconds are reasonably
clear. Beyond 3 seconds there are multiple peaks that are caused by
overlapping hyperbolic trajectories. The low velocity arrivals are due to
multiple reflections in the near surface. They arrive at the same time as the
reflection events from deep horizons.
A stacking velocity is defined for each time sample. The normal moveout correction and stacking
can be carried out by summing along a different hyperbola for each time on the seismogram. Equivalently,
the CMP gather may first be NMO corrected and then stacked. Both approaches are done in practise.
Summary: Essential
Elements in CMP Processing
There are many different processing steps that could be performed.
An example from GS Baker, 1999, is shown in the flow chart image here. However, the essential
steps are summarized in the following short list.
1. Obtain CSP (Common Source Point) gathers.
2. Sort into CMP (Common Midpoint) gathers.
Reflection events (coming from approximately the same point in the earth)
appear as hyperbolic trajectories. The goal is to stack them into a single
trace.
3. For each event, perform a velocity analysis to find the stacking velocity.
4. Perform NMO correction and stack.
This yields a single trace corresponding to a coincident source and receiver.
5. Composite the traces into a CMP processed section.
These are the only steps we will be concerned with in these notes. Other steps may be used by
experienced contractors and they may be necessary to produce more useful sections for
interpretation, but the details are beyond the scope of this set of notes.
Static Corrections
Before carrying out the NMO correction it is usually necessary to perform a static correction, which
amounts to moving the entire seismic trace up or down in time.
There are two main reasons for applying static corrections.
i. Put shots and receivers on a flat datum plane.
ii. Correct for near surface velocity anomalies beneath the source or receiver.
Elevation Statics
Consider the need for (i). Common midpoint shot receiver pairs
acquire data on an irregular interface (right figures, top panel).
Time differences are caused because of extra travel time associated
with elevation of source and receiever. As a result reflections
observed on the CMP gather will not have a hyperbolic form and
they will not be amenable to normal CMP processing (bottom
panel).
The correction procedure involves establishing a datum on which
to locate source and receiver, and then adding or subtracting the
incremental time. The reference velocity will be that of the upper
layer.
The reflections of interest are usually coming from great depth and the upcoming energy is travelling
nearly vertical. So the static correction due to elevation expressed as a change in travel time is
, and this time is subtracted from the record. That is, the whole seismic record is
shifted in time by a value deltat.
After static correction, the subsurface events will look more like an hyperbola and they will be ready
for velocity analysis, NMO and stacking.
The first layer is often highly weathered and it has variable thickness and velocity. It is also usually
poorly consolidated and therefore is a poor transmitter of seismic energy. In exploration it is common
to drill through the weathered zone into the upper layer. Drill measurements establish the thickness at
the shot. By having a seismometer at the surface above the shot one can estimate the velocity of the
weathered zone. The diagram below shows the differences between reflection events on adjacent
seismograms due to the different elevations of shots and detectors and the presence of a weathered
layer. After static corrections the seismograms should show better alignment of reflection events.
From Kearey, Philip and Micheal Brooks, An Introduction to Geophysical Exploration. 2nd ed. Blackwell Science: 1991.
One can apply a frequency filter to remove all signals that have a frequency high than fH
and all frequencies lower than f L.
The figures to the right illustrate effects of low pass and high pass filters. First examin the
figure and try to identify aspects caused by high frequencies. Then try to identify aspects
caused by lower freqeuncy components in the signal.
No Filter
Low Pass
High Pass
2. Deconvolution
a) Deghosting.
In marine work a source is detonated slightly below the
water surface. The primary seismic wavelet propagates
downward but there is an upward propagating wave that
reflects from the water surface and then propagates
downward. The latter reflection is referred to as a 'ghost".
A seismic wavelet is composed of the orginal wavelet plus a closely spaced reflection.
This complicates the wavelet as shown in the figure below.
A deghosting filter can remove the effect of the ghost so each reflection looks like a
primary wavelet.
b) Wavelet or Signature Deconvolution
An "inverse filter" is designed so that the original source wavelet is contracted to a
narrower and symmetric form. e.g.
This can enhance the vertical resolution, though not beyond the "theoretical"
maximum which is controlled by wavelength.
c) Predictive Deconvolution
This type of processing can remove some multiples from a seismic section.
Reverberation (multiples of the ocean-bottom reflection) in marine surveys is a
common example of this type of problem.
The procedure for carrying out f-k filtering is provided in the flow chart, below
left. The final diagram shows four shallow marine records before and after f-k
filtering to remove coherent linear noise. Hyperbolic reflections are observed after
removal of the noise.
Migration
The ideal goal of seismic processing is to have a section which
consists of true reflections which arise from structure directly
beneath the source. The image should provide quantitative
information about geologic structure. The vertical axis ideally is
depth but generally the time-to-depth conversion is not carried out
and the axis remains time. A processing step which is important in
putting reflectors in their correct location is migration. The need for
migration is most evident when layers are not flat lying. Consider a
single dipping layer and seismic traces corresponding to a
coincident source and receiver.
The reflection time is given by 2l/vi. But when we plot the section we think of the reflections arriving
from directly beneath the shot. The result is that a sloping reflector in the ground will appear as a
sloping reflector in the seismic section, but the slopes will be different. The true dip will be greater
than that observed on the section. The relationship between the two dips is
sin(alphat) = tan(alphas).
A planar-dipping reflector surface and its associated
record surface derived from a non-migrated
seismic section.
Other artifacts are visible in non-migrated sections. Synclines will be observed as "bow ties." Their
width on the seismic section is narrower than reality and their structural depth is diminished.
Conversely, anticlines will appear wider than they really are. Finally a point scatterer will be
observed as a hyperbola.
The plots above show several synthetic geologic structures and their associated non-migrated seismic
sections.
Migration is the process of reconstructing a seismic section so that reflection events are repositioned
under their correct surface location and at a correct vertical reflection travel time. The effects of
migration are to:
collapse a diffraction hyperbola back to a point;
make dipping structures appear with correct dip angle;
remove the bow ties from synclines and
shorten anticlines.
The principles of migration are shown in the figure below left. A single point diffractor will generate
a hyperbolic trajectory in a travel time section. Migration collapses the hyperbolic event back to its
apex.
Principles of diffraction migration. (a) Reflection
paths from a point reflector. (b) Migration of
individual reflection events back to a position of
point reflector. (c) Use of wavefront chart and curve
of maximum convexity to migrate a specific
reflection event: the event is tangential to the
appropriate curve of maximum convexity, and the
migrated position of the event is at the intersection of
the wavefront with the apex of the curve.
From Kearey, Philip and Micheal Brooks, An
Introduction to Geophysical Exploration. 2nd ed.
Blackwell Science: 1991.
The example above right shows an unmigrated section (top). The section after migration is shown in
the bottom panel.
UBC Earth and Ocean Sciences, F. Jones.