THE MARINE SEISMIC SOURCE

SEISMOLOGY AND EXPLORATION GEOPHYSICS

 GREGG PARKES
and
LESHATTON
Merlin Profilers (Research) Ltd., U.K.

THE MARINE
SEISMIC SOURCE

SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.

library of Congress Cataloging in Publication Data

Parkes, Gregg, 1954-

The marine seismic souree.

(Seismology and exploration geophysics)

Bibliography : p.
Inc1udes index.
1. Submarine geology. 2. Seismology-Data proccssing. 3.
Seismic reflection method-Deconvolution. 1. Hatton, Les. 1948-
Il. Title. III. Series.
QE39.P37 1986 551.2'2 86-3278
ISBN 978-90-481-8425-5 ISBN 978-94-017-3385-4 (eBook)
DOI 10.1007/978-94-017-3385-4

AII Rights Reserved

© 1986 by Springer Science+Business Media Dordrecht
Origina1ly published by D. Reidel Publishing Company, Dordrecht, Holland in 1986
No part of thc material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanicaI
includ ing photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner
 TABLE OF CONTENTS

Preface vii
Acknowledgements ix

1. Underlying Physics and Concepts

1.1. ACOUSTIC WAVE GENERATION TECHNIQUES 3
1. 2. ACOUSTIC WAVE PROPAGATION IN WATER 4
1.3. THE RELATIONSHIP BETWEEN VELOCITY, PRESSURE AND ENERGY 7
1.4. BUBBLE MOTION 8
1.5. THE GHOST OR VIRTUAL IMAGE 10
1.6. THE NEAR AND FAR RADIATION FIELDS 15
1. 7. THE MEASUREMENT OF SIGNATURES 16
1.8. RELATIVE MOTION EFFECTS 17
1.9. SIGNATURE PARAMETERS 19
1.10. FACTORS DETERMINING CHOICE OF SOURCE 20

2. Source Arrays and Directivity 23

2.1. THE REASONS FOR USING ARRAYS 25
2.2. PULSE SHAPING 25
2.3. THE PRINCIPLES OF DIRECTIVITY 26
2.4. DIRECTIVITY DISPLAYS 29
2.5. BEAM STEERING 31
2.6. DIRECTIVITY OF LONG ARRAYS 33
2.7. DIRECTIVITY OF WIDE ARRAYS 34
2.8. THE USE OF DIRECTIVITY FOR NOISE SUPPRESSION 36
2.9. EFFECTS OF DIRECTIVITY ON PRIMARY DATA 39

3. Interaction and Wavefield Determination 43

3.l. CATEGORIES OF SOURCE INTERACTION 45
3.2. THE WAVEFIELD OF A LINEARLY INTERACTING ARRAY 50
3.3. ENERGY CONSIDERATIONS 59
3.4. THE AVOIDANCE OF INTERACTION EFFECTS 61
3.5. SUMMARY 62

4. Practical Aspects of Wavefield Stabilit~ 65

4.1. TIME SYNCHRONISATION 67
4.2. SOURCE GEOMETRY STABILITY 68
4.3. SYSTEM STABILITY 70
4.4. THE EFFECTS OF WEATHER 72
vi TABLE OF CONTENTS

4.5. MARINE GHOST VARIABILITY 74

4.6. SUMMARY 77

5. Source Signature Deconvolution 79

5.1. INTRODUCTION 81
5.2. DETERMINISTIC DECONVOLUTION 83
5.3. STATISTICAL DECONVOLUTION 93
5.4. PRACTICAL NOTES 94

Appendix - Technical Description of Main Sources 97

References 105
Index 111
 PREFACE

This book is about marine seismic sources, their history, their physical
principles and their deconvolution. It is particularly accented towards
the physical aspects rather than the mathematical principles of
signature generation in water as it is these aspects which the authors
have found to be somewhat neglected. A huge amount of research has been
carried out by both commercial and academic institutions over the years
and the resulting literature is a little daunting, to say the least. In
spite of this, the subject is intrinsically very simple and relies on a
very few fundamental physical principles, a somewhat larger number of
heuristic principles and a refreshingly small amount of blunderbuss
mathematics. As such it is still one of those subjects in which the
gifted practical engineer reigns supreme and from which many of the
important advances have originated.
In Chapter 1 of the book, the underlying physics and concepts are
discussed, including pressure and wave propagation, bubble motion,
virtual images and the factors determining choice of source. In marine
reflection seismology, almost all of the seismic data acquired currently
is done with either the airgun or the watergun, which rely on the
expulsion of air and water respectively to generate acoustic energy. As
a consequence, the discussion in this chapter is geared towards these
two sources, as is much of the rest of the book. Many noteworthy
sources have appeared in the past however, which used different physical
principles but which nevertheless obey the same basic laws, and these
sources are not enlarged upon here as they are now very much in the
minority.
Chapter 2 discusses directivity, a much misunderstood phenomenon.
The underlying principles are described, as are the directional
dependence of radiation for arrays of sources and the methods of
exploiting this directivity to improve the quality of seismic data.
Since all known sources are directional to some extent as a result of
the virtual image or ghost reflection in the sea surface, an
understanding of this basically simple subject is important.
The deconvolution of seismic sources is an important topic of this
book, as the real goal of marine seismic source studies is to produce
something that can promptly and simply be removed by the processing
geophysicist ! Such deconvolution techniques break naturally into two
categories, the statistical and the deterministic. Deterministic
methods require a prior knowledge of the source signature. In practical
arrays, such knowledge is available only if the phenomenon of
interaction is understood. The problem of interaction arises because
any source fired simultaneously with other sources does not behave the
same way as when fired in isolation. This rather confusing effect is
studied in considerable detail in Chapter 3, followed by a number of
proposed methods for determining the source signature, some of which are
outstandingly successful. In addition, the fundamental difference
between directivity and interaction is also re-emphasized.
viii PREFACE

The science of marine seismic sources is an eminently pragmatic

subject and Chapter 4 goes into considerable detail on the practical
aspects of the stability of the radiated acoustic energy field. The
effects of source synchronization, source geometry, weather and the
surface ghost response are all analysed.
Chapter 5 discusses how all this accumulated knowledge is used to
shape the predicted or estimated source radiation field into some ideal
response from the point of view of the processing geophysicist. After
studying the effects of both simple band-limiting and absorptive
band-limiting on the source signature, deterministic deconvolution as
well as a number of statistical deconvolution techniques are described,
including their advantages and disadvantages from a practical point of
view.
Finally, technical descriptions of the most common sources are
included as an appendix for reference purposes.
 ACKNOWLEDGEMENTS

This book was written after three exhilarating and fairly intense years
research into marine seismic sources and most notably the airgun.
During this period, many people influenced our understanding but most of
all became good friends emphasizing the tightly-knit nature of the
subject, and we would like to acknowledge their contributions here.
First of all, without Tor Haugland's practical engineering genius,
we (and a lot of other people) would have achieved little. Mind you his
suggestion of cod's tongues as a viable breakfast during airgun trials
didn't help. We are still good friends.
One name that frequently crops up in this and other published
material on seismic sources is that of another old friend Anton
Ziolkowski. Together we have shared numerous long evenings during trials
arguing about the pros and cons of various schemes disguising our real
purpose which was to improve the sales of Scottish malt whisky. As
Professor of Geophysics at Delft, he has been a constant and
distinguished presence in the development of marine seismic sources.
We would also like to acknowledge a number of rewarding
conversations with Svein Vaage, founder of the triple club, and with
Bjorn Ursin, both of whom have been in the thick of the action for a
number of years.
A number of other people helped in one way or another with a
comment here or an exquisite diatribe there, but we would especially
like to mention Mark Loveridge and Robert Laws, who know lots about this
sort of thing, and Steve Levey and the rest of the crews of the Liv and
Nina Profiler for engineering ingenuity and sleeplessness above and
beyond the call of duty.
We would like to thank the S.E.G. and the E.A.E.G. for permission
to use some of the diagrams published in Geophysics and First Break and
also Britoil plc for permission to publish some of their array trials in
the first place.
Finally, we would like to thank our families for putting up with
all this.

L.H.
G.P.
Woking, U.K., January 1986.
 CHAPTER 1

Underlying Physics and Concepts

1.1. ACOUSTIC WAVE GENERATION TECHNIQUES
The list of available marine seismic sources is extensive. Many have
stood the test of time, whilst others have fallen by the wayside, then
there are the countless variations on a theme. Technical details of the
sources which are currently used most are given in the appendix. This
will be updated periodically to keep it in line with the times. The
methods used to generate acoustic waves in marine seismic sources can be
grouped into four broad categories. These primary divisions are -
chemical, mechanical, pneumatic/hydraulic, and electrical.
Chemically based, or explosive sources, are perhaps the simplest in
concept. They are rarely used these days, so their importance is mainly
historical. In early days explosive sources were somewhat haphazard. It
was almost a case of chucking a stick of dynamite over the side and
hoping for the best! The uncertainty in the location of the charge
when it exploded usually necessitated the use of a second boat to tow
the receiver! These methods were superseded by a second generation of
more sophisticated and carefully controlled explosive based sources. The
methods of operation were quite varied. The most straightforward were
solid explosive capsules with appropriately timed fuses. The dispensing
of these charges was automated and relatively safe. In other solid
explosive sources, the charge was detonated in an enclosed perforated
cage, with the object of attenuating the secondary bubble oscillations.
The problem here was that the cage suffered cumulative damage and needed
frequent replacement. Then there were sources in which explosive
mixtures of gases were ignited in a rubber sleeve, an exhaust system
dispelling the waste gases. This latter type of source is still
occasionally used, but increasingly rarely. Nonetheless an enormous
volume of seismic data has been shot with explosive sources, and from
time to time some of it emerges for reconsideration.
Sources which fall into the category 'mechanical' rely on a
physical vibration, usually of a plate, to transmit seismic waves into
the water. A typical example would be the marine equivalent of
Vibroseis*, in which a plate is vibrated through a sweep of
frequencies. Although the signature of such a source is extremely long,
the bandwidth and spectrum can be controlled easily. Only recently have
such sources been introduced to the marine environment, Peacock et al.
(1982).
The next category is pneumatic/hydraulic. In these sources a slug
of gas (usually air) or fluid (usually water) is dispelled under
pressure into the water. The driving pressure can be pneumatic (e.g.
compressed air) or hydraulic, and in some cases is a combination of
both. The most notable example is the airgun, in which a controlled
volume of high pressure air is suddenly vented into the water. The
resultant rapid expansion and subsequent oscillation of the air bubble
generates seismic radiation. Another example is the watergun in which a
* Registered trademark of CONOCO
4 CHAPTER 1

slug of water is expelled under pressure. Cavitation occurs and the

resultant implosion produces the radiation. Other sources in this
general category rely on rather different principles, such as the
expulsion of a bubble of steam. Rapid condensation of the steam produces
a cavity. The resultant implosion of water generates the seismic
radiation.
The final category of electrically operated sources tend to be
special purpose. The principle is that a large electrical charge 1S
stored, and then discharged through the water. Vapourisation occurs and
a steam bubble is formed, this is followed by condensation and
implosion. This source is known as the 'sparker'.
The diversity of marine seismic sources discussed above produces an
equal diversity in radiation characteristics. However the physical
principles that govern the design and operation of all these sources are
the same. Therefore in the remainder of this book these principles will
be considered with as little reference to specific sources as possible.

1.2. ACOUSTIC WAVE PROPAGATION IN WATER

The only waves of concern when considering marine seismic sources are
'elastic' waves. That is waves which propagate with a velocity that
depends upon the elastic properties of the surrounding medium. Elastic
waves fall into two main categories, longitudinal or P waves in which
particle motion is in the same direction as wave propagation, and S
waves in which particle motion is perpendicular to the direction of wave
propagation. Fluids have no resistance to shear and so will not support
the propagation of S waves. Therefore only P waves need be considered.
A full discussion of the equations which describe the propagation
of these waves would require a book to itself. However the important
steps and physical concepts will be treated for completeness, and for
those with a mathematical bent. For a more detailed discussion se'e for
instance Batchelor (1967) or Aki and Richards (1980). In what follows
the formalism of Batchelor will be followed.
For convenience, the short hand notation of equation 1.1 will be
introduced :

D a ( 1.1)
Dt == at + u • V

in which u is the vector velocity of the fluid. The first fundamental

equation -which must be introduced is the 'continuity' or 'conservation
of mass' equation. The physical principle this imbodies is simply that
for each element of the fluid, the rate of change of volume must be
balanced by the rate of change of density. So,
 UNDERLYING PHYSICS AND CONCEPTS 5

(1. 2)

where P is the density. Equation 1.2 can be re-written as 1.3, by

introducing the quantity 'a' as defined by equation 1.4 :

_1_Q.E+'V u=O (1. 3)

pa2 Dt

and,

(l.4 )

where p is pressure, and the S represents constant entropy. The second

fundamental equation which must be included is the 'equation of motion',
which relates the acceleration of the fluid to the forces and stresses
acting upon it. The equation in its full form is called the
'Navier-Stokes equation of motion', and contains terms which account for
molecular transport effects. For the fluid flow conditions under
consideration it is legitimate to assume that the entropy of each
element remains constant. The flow field is then said to be 'isentropic'
and all molecular transport effects can be neglected. The resulting
simplified equation of motion is then given by 1.5 :

Du
P Dt = pE - 'Vp (1. 5)

where F is the vector body force experienced by the fluid.

A-further simplification that can be made at this stage is to
assume that the disturbance to the fluid is only slight as the wave
propagates. That is, the variation in density from the equilibrium
value, PI (= P - po). the variation in pressure, PI (= P - PO) .and the
velocity u are all small. With this simplification equations 1.3 and 1.5
can be approximated by equations 1.6 and 1.7 :
6 CHAPTER 1

apI
at + 'iJ • ~ = 0 (1. 6)

and,

au
Po at = PIt - VPI (1. 7)

in which an is the value of a at P : Po . The relationship of equation

1.8 may nO~1 be arrived at by differentiating equation 1.6 with respect
to time, substituting the expression for ( au/at ) from equation 1.7,
and using the relationship 1.4 : -

2
I a PI (1.8)
a 2 at 2
o
Now the body force E is due to the Earth's gravitational field, so E= ~
and hence the 'iJ • F term is zero and the final F term of equation 1.8 is
completely negligiole. We are left with the spherical wave equation
which is given in its more usual guise in equation 1.9

(1. 9)

The quantity aD turns out to be the phase velocity of wave propagation,

c, in equation 1.9. The main case of interest is the point source with a
spherically symmetric radiation field, so it is convenient to convert to
spherical polar coordinates. If the possibility of inward travelling
waves is ignored, then the solution of equation 1.9 is of the form:

p := -
I f(t - -)
r (1.10)
r c

where f is some unknown source dependent function. If K is the bulk

modulus of elasticity of the fluid, then the phase velocity or speed of
sound is given by equation 1.11 :
 UNDERLYING PHYSICS AND CONCEPTS 7

(1.11)

At seismic frequencies, the velocity is essentially independent of

frequency, however it is sensitive to temperature and pressure. The
insensitivity to frequency (or absence of dispersion) means that the
characteristic pulse shape of the radiation from a marine seismic source
is maintained as the pulse propagates through the water.
Typical sound velocities in water likely to be encountered in
marine seismic acquisition lie in the range 1450-1500 m/s. In general,
velocity increases with temperature, with pressure, and with salinity.
For example, at 3% salinity, the velocities at 5°C, 10°C, and 15°C, are
respectively 1462 mis, 1480 mis, and 1500 m/s. In general, in shallow
waters in which the temperature drops with depth, there will be a
corresponding fall in the acoustic velocity. However in water deeper
than about 1km the temperature is nearly constant, so the increase in
pressure results in a general increase of velocity with depth.

1.3 THE RELATIONSHIP BETWEEN VELOCITY, PRESSURE AND ENERGY

The relationship between velocity and pressure may be derived as
follows. Differentiate the expression for pressure of equation 1.10 with
respect to r and substitute the result into equation 1.7. If it assumed
that the body force F is zero and that the distance from the source is
large enough for I/r2 terms to be neglected, then the expression of
equation 1.12 results:

1 (1.12)
u = pc p

The kinitic energy of a particle of mass m, moving at a speed u, is

simply ~mu . The potential energy in a sound wave is equal to the
kinetic energy as shown for instance by Coulson (1941). E , the total
rate of energy flow through a shell at a radius r from the source can
now be calculated as the area of the shell times the speed of flow times
the total energy density (potential plus kinetic). E is therefore given
by equation 1.13

E = 41fr 2 c pu 2 (1.13 )
8 CHAPTER!

An expression for the total energy emission from the source, T, can now
be derived by combining equations 1.12 and 1.13, recognising that the
pressure p is in fact a function of time, p(t), and integrating over
time.

T = 4nr2 ooJ p(t)2 dt (1.14)

pc
o
Having derived the relationship between pressure and energy, it is
worth considering how these parameters vary with distance from the
source. If it is assumed that there is no absorption of energy as it
propagates through the water, then the law of conservation of energy
states that the total rate of energy transfer through any sphere of
radius r from the source, must be constant. The area of a sphere is
simply 4nr 2., so the energy transfer rate per unit area must decrease as
the inverse of the square of the distance from the source. At large
distance from the source (the far field), where equation 1.14 holds, the
pressure of the wave will decrease as the inverse of the distance from
the source.

1.4. BUBBLE MOTION

A substantial number of seismic sources produce their acoustic radiation
by the release of high pressure gases into the water. This may be
directly, as in airguns, for example, or indirectly, as in explosive
sources, in which gases are produced as a result of chemical reactions.
The generation of acoustic energy is closely coupled to the high speed
motion of these gas bubbles.
The behaviour of a bubble of high pressure gas when it is released
suddenly into water is illustrated in the top half of Figure 1.1, in
-which bubble radius is plotted against time. Initially the pressure
inside the bubble greatly exceeds the external or hydrostatic pressure,
so the gas bubble expands rapidly. The momentum of the expanding bubble
is sufficient to carry the expansion well beyond the point (when the
bubble radius = RO ) at which the internal and hydrostatic pressures are
equal. When the expansion ceases, the internal bubble pressure is below
the hydrostatic pressure, so the bubble starts to collapse. The collapse
overshoots the equilibrium position and the cycle starts once again. The
bubble continues to oscillate, with a period typically in the range of
tens to hundreds of milliseconds (the corresponding radii being in the
range 0.1 - 1 metre). The oscillation would continue indefinitely if it
were not for frictional losses, and the buoyancy of the bubble which
eventually causes it to break the surface. The cyclic compression and
rarefaction in the water surrounding the bubble as it oscillates
generates acoustic radiation. The form of this radiation as a pressure
wave is illustrated in the lower half of Figure 1.1, and is essentially
a periodic damped oscillation.
 UNDERLYING PHYSICS AND CONCEPTS 9

BUBBLE

~: f¥Vv~p Figure I.! :

The top graph shows the radi us of an
oscillating bubble as a function of
time. The lower graph is the
o corresponding radiated pressure. Ro is
the radius at which the bubble's
RADIATED internal pressure is equal to
PRESSURE hydrostat ic pressure.

To a first approximation the period of oscillation of the bubble is

described by equation 1.15 :

(1.15)

in which P and V are the initial internal pressure and volume of the gas
before it is released, and Po is the hydrostatic pressure. The constant
Cl depends upon the details of the source design. Detailed experimental
and theoretical work on airguns (see Vaage et al. (1983) and references
therein) has produced the following expressions for the primary
amplitude of the emitted pulse

A = C2 V1/ 3 (1.16)

when the pressure and depth are held constant, and

A C3 p3/4 (1.17)

when the volume and depth are held constant. Once again the constants of
proportionality C2 and C3 are functions of the source design.
10 CHAPTER 1

The detailed physics of seismic source bubbles is complex, and

beyond the scope of this work ( see Ziolkowski and Metselaar (1984) and
references therein ). The parameters of the bubble as a function of time
are really quite surprising. The temperature can drop to a hundred or so
degrees Celsius below zero, causing ice crystals to form in the bubble.
There is also evidence that the bubble is constituted of a foam of small
bubbles rather than being one large one.

1.5. THE GHOST OR VIRTUAL IMAGE

The time domain pressure wavelet which characterises the radiation
emitted by a seismic source is referred to as the source signature.
This is a misnomer in that there is no single wavelet which fully
describes the source radiation (as will be seen later). However the term
'the source signature' is commonly used, without specifying position in
the wavefield, to describe the vertically travelling pressure wavelet.
To avoid confusion, it is as well to be more specific, and always call
this special case the vertically travelling far field signature.
The reflection of the source radiation at the sea surface has an
inextricable effect on the radiation field, and as such is considered to
be an intrinsic feature of the source wavefield, from the point of view
of data processing. Source signatures, therefore, are usually presented
with this intrinsic reflection component included.

SEA
SURFACE
Figure 1.2 :
The marine ghost pulse, which results
from the sea surface reflection, appears
to originate from the virtual image of
the seismic source. It is delayed in
time by xlc with respect to the primary
pul se, where cis the speed of sound in
water.
GHOST
PULSE

PRIMAFlY
PULSE
 UNDERLYING PHYSICS AND CONCEPTS 11

At seismic frequencies, a planar water/air interface is generally

an almost perfect reflector. In some circumstances this is not the case
as will be seen in Chapter 4. However for the discussion of this section
a sea surface reflection coefficient of -1 will be assumed throughout.
The minus sign represents a polarity reversal of the reflected waves. In
essence, therefore, seismic energy generated beneath the surface of the
sea cannot escape into the air. Acting like a mirror, the sea surface
reflection produces a virtual image of the seismic radiation which is
generally called the 'marine ghost'.
Considering Figure 1.2, the ghost pulse or signature will be
delayed in time with respect to the primary pulse by x/c, where c is the
velocity of sound in water. The time delay, T , is therefore given by
equation 1.18 :

T =2d.cos(<p) ( 1.18)
c

where d is the source depth. The time delay is a function of the

anQle! ~. In fact, the majority of the seismic energy recorded in
selsmlC experiments will have originated at angles very close to the
vertical, giving this direction special significance. In the vertical
direction the ghost pulse will be delayed in time by 2d/c, where d is
the depth of the source. The composite wavelet of energy is therefore
the summation of a primary wavelet, and a ghost wavelet delayed ;n time
and reversed in polarity with respect to the primary, as shown in Figure
1.3.
~~~-------

PRIMARY WAVELET

Figure 1.3 :
GHOST The ghost wavelet is delayed in time,
and has opposite polarity to the primary
wavelet. The composite source signature
is the summation of primary wavelet and
ghost.

COMPOSITE
12 CHAPTER 1

Figure 1.3 demonstrates the effect of the ghost on the time domain
wavelet. The consequences of this in the frequency domain should also be
examined.

DIRECT PULSE

Figure 1.4 :
The ghost pulse is reversed in polarity
and has a time delay T with respect to
the direct pulse. The exact value
GHOST PULSE of T determines whether the
interference between the pulses is
constructive or destructive.

~\\"-

Consider Figure 1.4, in which the seismic source signature ;s a

sinusoid. The ghost pulse is reversed in polarity, and delayed in time
by ~ , where T is given by equation 1.18. The distance travelled in
time T will be TC = 2dcos\ <j». Clearly, if 2dcos( <j» is half a
wavelength, then the direct and ghost waves will add constructively,
whereas if 2dcos( <j» is a full wavelength then the two sinusoids will
add destructively. So in summary:
Destruction if 2dcos(<j» = O.OA, l.OA, 2.0A, .. .
Construction if 2dcos(<j» = O.5A. l.5A, 2.5A, .. .

However, of course, the typical seismic source signature will not be

sinusoidal, but will contain components at a range of frequencies. The
minimum and maximum interference conditions above will therefore
modulate the complete continuous amplitude spectrum with the function
shown in Figure 1.S.The frequencies, v ,of the ghost notches are
given by equation 1.19 : n

(1.19)
 UNDERLYING PHYSICS AND CONCEPTS 13

in which n is a positive integer, c is the speed of sound in water and d

is the source depth.

/
I
I

Figure 1.5: Amplitude template introduced by the source ghost.

Amplitude is plotted as a function of dh , where d is
the source depth and A the radiation wavelength.

The effect of the ghost on the frequency content of the seismic

signal is therefore substantial. Figure 1.6 shows the vertically
travelling far field signatures for the same source operated at a
variety of depths. The signatures are shown on the left of the figure
and the corresponding amplitude spectra (on a linear scale) are shown on
the right. The correct relative amplitudes have been preserved
throughout. The increased ghost delay is noticeable in the signatures as
the depth increases, however the effects on the amplitude spectra appear
much more dramatic. For instance, at 12m the ghost notches heavily
modulate the spectrum, compared to 3m, say. However the low frequency
amplitude (O-10Hz) at 3m is a factor of 2 or 3 below that at 12m. Choice
of depth is therefore a most important aspect of source design, and will
depend on the type of source being used and the overall aims of the
survey. For instance, if deep penetration is required, the low frequency
content of the signature should not be compromised (high frequencies are
quickly attenuated in the Earth, so penetration to depth can only be
achieved with low frequencies), and a deep source should be used.
Conversely, if high resolution in the shallow is required, then a
shallow source will generally be used to extend the high frequency
bandwidth. Other effects compete with this ghost effect. For instance,
from equation 1.15 it is clear that the greater hydrostatic pressure at
greater depth reduces the bubble oscillation period. This reduces low
frequencies with increasing depth in direct opposition to the ghost
14 CHAPTER 1

effect. This secondary effect is however completely dominated by the

ghost effect.

DEPTH

Ii 12m
Jlr--~-
V
~

,
it 9m
II
Jl, r-----'~
\i
Figure 1,6 :
Variation of the vertically travelling
far field signature and its spectrum as
a funct i on of source depth.
11 6 m

JV--~-
,
 UNDERLYING PHYSICS AND CONCEPTS 15

1.6. THE NEAR AND FAR RADIATION FIELDS

Unfortunately there is some confusion in the use of the terms 'near
field' and 'far field'. They do have quite precise definitions, however
it is common to find them used in a much looser sense. The precise
definitions relate to effects which are only important close to the
source. For the purposes of this discussion it will be convenient to
rewrite equation 1.10 as follows:

p = 1r f' (t - I)
c
(1. 20)

where f' is the time differential of some function f, p is pressure, r

is radial distance from the source and c is the speed of sound in water.
Now Newton's second law may be written as in equation 1.21 :

a :;: 12.£ (1. 21)

p ar

in which a is acceleration, and p is density. Differentiating equation

1.20 and substituting the result in equation 1.21 gives:

a :;: _1_ f' (t - !:.) + _1_ f"(t - !:.) (1.22)

pr 2 c pcr c

Finally the expression for the particle velocity is obtained by

integrating equation 1.22 with respect to time, to give

u :;: __1_ f(t - !:.) + __1_ f'(t - ~) (1.23)

pr2 c pcr c

Considering the above expression, the 'near field' may be defined as

that region in which the 1/r2 term dominates that is close to the
source. The 'far field' is then the region distant from the source in
which the Ifr term dominates. It is interesting to note that the shape
of the particle velocity wave is dependent on distance, because it is
the sum of two terms that depend differently on distance. On the other
hand there is only a single distance dependent term in the expression
 CHAPTER 1

for pressure (equation 1.20), so the shape of the pressure wave is

independent of distance.
A common rather different use of the term far field considers it to
start at that point at which the pressure signature is indistinguishable
(i.e. within some error tolerance) from the signature that would be
measured at infinity in the same direction. The exact position adopted
depends on many factors, not least opinion ! However more
scientifically, it depends upon the acceptable error tolerance, source
depth, frequency of interest, direction and source spatial extent. For
example, many would consider 100m to be in the far field of a point
source deployed at 5m depth. However at 100m the amplitudes of ghost and
direct arrivals will differ by 10 percent, whereas in the true far field
they are identical ! Whatever error criterion is used, the far field of
spatially extended sources starts much further away than that of point
sources.
The term 'far field signature' is often used with the implicit
understanding that it refers to the vertically travelling signature.
This is rather a misnomer. In its broader use far field signatures exist
for all directions. In addition note that there is no such thing as the
far field signature since all sources are directional, even so-called
'point sources' by virtue of the ghost.

1.7. THE MEASUREMENT OF SIGNATURES

A detailed knowledge of the radiation fields of seismic sources is
essential to deciding which source to use for a particular purpose, and
how to deploy it. The same detailed knowledge is required to enable the
wavefield to be deconvolved when processing the data. The source may be
directional and/or interacting, and this may need to be taken into
account in the processing. However, to a first approximation at least,
the source wavefield can be characterized by the vertically travelling
far field signature. An accurate knowledge of this signature is
therefore mandatory. Estimates of the vertically travelling far field
signature may be obtained by hydrophone measurements in the far field,
or by extrapolating near field measurements to the far field.
Far field measurements are notoriously difficult and expensive.
They must be carried out in deep water to ensure that the hydrophone is
indeed in the far field. The biggest problem is aligning the hydrophone
vertically beneath the source. A hydrophone cable of perhaps 150m long
will not hang vertically in the water, especially if there are cross
currents, or if the boat towing the hydrophone is moving. An alternative
to towing the hydrophone is to attach it to a stationary buoy, and steam
past it towing the source. Again it is impossible to tow the source to a
point vertically above the hydrophone especially if the source is
extended. Triangulation techniques can be used to locate the hydrophone.
For example, by using travel times from extreme sources on an extended
array, or by using acoustic 'pingers' outside the seismic band since
hydrophones have a much broader bandwidth than is required. So with
care, reasonable estimates of the vertically travelling far field
signature can be obtained.
 UNDERLYING PHYSICS AND CONCEPTS 17

Near field measurements also have their problems. The hydrophone

must be robust to survive so close to the source. Its sensitivity must
also be chosen carefully to avoid it being over- or under-driven.
However suitable equipment does exist and good near field measurements
are made routinely. The signature at any position in the wavefield can
be deduced from near field measurements as will be seen.
The law of superposition of wavefields states that two or more
waves can traverse the same space independently of each other, and that
the particle displacement at any point is simply the vector addition of
the displacements caused by the independent waves. So, assuming the
principle of superposition, the signature at any position in the
wavefield of a marine seismic source can be calculated as follows. The
direct wavelet from each element of the source array is calculated for
the required position by delaying the near field signature and including
amplitude scaling according to the l/r law. Similarly all the secondary
ghost wavelets are computed. The composite signature is simply the
direct sum of all these components. For marine seismic sources the
superposition principle has been shown to work well for single sources
and for non-interacting arrays of sources. The wavefield of an
interacting array cannot be obtained by superposing the near field
signatures of the individual array elements fired in isolation. However,
as will be shown in Chapter 3, it is possible to derive a set of
equivalent or 'notional' sources which can be used with the law of
superposition to give the correct result.

1.8. RELATIVE MOTION EFFECTS

If the signature of a single element seismic source is known for some
fixed distance, then the signature to be expected at some other fixed
distance can be computed using the l/r amplitude scaling law. For
example, the calculation of far field signatures from near field
measurements, as discussed in the preceding section. However there are
often relative motion effects between seismic source and hydrophone
which must be allowed for. This relative motion arises because the
hydrophone moves through the water at the speed of the towing vessel,
which is typically 2-3 m/s. However for many seismic sources, the source
of the radiation is released into the water and thereafter follows its
own peculiar motion. For example, in the case of an airgun, a bubble of
air is released into the water with an initial forward velocity which is
the same as that of the towing vessel. However the drag on the bubble
causes this forward motion to slow rapidly. Furthermore the buoyancy of
the bubble causes it to rise in the water. This upward motion is complex
and non-linear, in that the rapid expansion and contraction of the
bubble modulates the drag and hence the rise velocity. Without such
motions, the rise velocity should be constant in the manner of a
spherical cap bubble, (Batchelor (1967».
For a far field hydrophone, the relative motion between source and
receiver is usually negligibly small. However in the near field the
effects can be large. As an illustration consider Figure 1.7. In this
example it is assumed that the seismic source is a gas bubble which has
18 CHAPTER I

upward motion at constant velocity, but no forward motion. On the other

hand, the near field hydrophone has a constant forward velocity. Plotted
in the figure is r(t)/r(tO), where r(t) is the functional variation of
separation with time, and r(to) is the separation at time zero. In case
(a) the velocities are zero and there is no functional variation. In
cases (b) and (c) however, the variation is large. For example, in case
(b), at about 300 msec, the separation is less than half that at time
zero. The measured signal amplitude will therefore be more than double
the value that would be measured with no relative motion.

BUBBLE 1 BUBBLE2

BUBBLE rI'o) Vhy<lrophone Vbubb~

(0) 1m 0 0

(b) 1m 25m!. LOm!'

Figure 1. 7 :
(e) 2 10m 2·5""" I.DiM; Rel at i ve mot i on effects between
hydrophone and seismic source (bubble)
for three cases as shown. The
hydrophone mot i on is to the 1eft and
the bubble motion upwards. r(t O) is
(b) the hydrophone to bubble separation at
time zero, and r(t) is the functional
variation of separation with time .
•(t)
.(to)
(e)

(a)

·0 500.0 1000.0

TIME(m sed

To be of value, near field measurements must be referred to some

fixed distance, so the measured amplitudes must be corrected for the
separation effect illustrated in Figure 1.7. For seismic sources which
produce gas bubbles, a linear relative velocity model is an excellent
first approximation. However if even greater accuracy is required, then
acceleration terms must be included. Such accuracy has not been found
necessary in practice.
 UNDERLYING PHYSICS AND CONCEPTS 19

Figure 1.8 (a) is a measured near field signature of an airgun

source. The hydrophone was placed 1m from the source (above and behind
it). The relative motion effects are visible particularly in the second
and third negative excursions of the signature, for which the amplitude
is too large because of the reduced source/receiver separation at this
time. Figure 1.8 (b) shows the same signature corrected for relative
motion effects. The damped decaying oscillation is now as expected.

(al

Figure 1.8 :
(a) The measured near field signature
of an airgun.
(b) The same signature corrected for
(bl relative motion effects (hydrophone
forward velocity of 1.8 mls and
bubble rise velocity of 1.0 m/s).

1.9. SIGNATURE PARAMETERS

There are a number of useful measures commonly used to describe far
field signatures. The first is 'peak to peak amplitude'. This is simply
the amplitude difference between the peak of the primary event and the
peak of its ghost - the amplitude Al shown in Figure 1.9. Depending on
the type of source, the primary peak may not be at the start of the
signature. When specifying peak to peak amplitude, the bandpass filter
which has been applied to the signature must also be specified, since
there will clearly be a functional dependence on filter setting.
Another descriptive parameter is 'primary to bubble ratio'. The
bubble part of the signature is composed of the secondary oscillations
that occur after the primary event. The secondary peak to peak amplitude
is A2, as shown in Figure 1.9. The primary to bubble ratio is then
Al/A2. Again the filter setting must be specified.
In ~eophysics, the traditional unit for pressure is the bar (1 bar
= 10 5 N/m = 14.5 pSi, or approximately 1 atmosphere). The amplitude (or
20 CHAPTER 1

excess pressure) of a signature is normally measured in bar-metres

(b-m). Distance is usefully parameterised in these units. An amplitude
of A bar-metres means that at a distance of 1 metre from the source the
amplitude value would be A bars. So if the pressure measured at a
distance r from a source is P(r), then using the l/r scaling law, the
corresponding pressure in bar-metres will be P(r).r.

___ 1 __ _
~~--- ~--------
---T---
A2

Figure 1.9: Commonly used signature parameters -

(I) The peak to peak amplitude, AI.
(2) The primary to bubble ratio, Al/A2.

1.10. FACTORS DETERMINING CHOICE OF SOURCE

A wide range of factors may influence choices of source for specific
applications. Not least of these are practical considerations such as
which sources are available at the required time and place in the world,
and how expensive they are to use. In some geographical areas the choice
may be limited by operational factors. For instance, in shallow water
larger boats and source systems may be impossible to use. There may also
be environmental considerations, although modern sources are relatively
'pollution' free. There were more problems in the past with explosive
sources, for instance, which were sometimes banned because of their
destructive effect on the local fish population! Aside from such
practical considerations there are various design specifications that
the source must meet to fulfill the scientific objectives of the survey
in question.
In the time domain, the traditional criteria for assessing the
pulse are the pulse duration, the primary to bubble ratio, and the pulse
 UNDERLYING PHYSICS AND CONCEPTS 21

repeatability. It can be argued that the ideal profile is an impulse

response, so in practical terms short duration and high primary to
bubble ratio are considered desirable. Such an approach has the
advantage that the need to deconvolve a complex pulse shape diminishes.
Furthermore the source signature is often not known accurately, and may
even vary with time. Repeatability and stability of the pulse is
therefore another prime consideration. In source systems in which the
source is stable and the radiation field is well determined, the
signature deconvolution becomes much less of a problem and these
'classical' time domain constraints can and should be relaxed.
More modern thinking, coincident with much improved techniques for
predicting the far field radiation pattern, would place frequency domain
characteristics and energy output as the primary considerations. Since
higher frequencies are rapidly attenuated in the Earth, high energy, low
frequency sources are required if the objective is penetration to depth.
Conversely less powerful high frequency sources are used when the
objective is high resolution in shallow data. If it can be predicted,
the time domain response of the source is irrelevant as the
convolutional model (see later) is an excellent basis for source
signature correction. Note finally that it is even sometimes possible to
tune the source amplitude spectrum in detail so that the resultant
spectrum at the target (after ghosting and attenuation in the Earth,
etc.) is as desired.
 CHAPTER 2

Source Arrays and Directivity

2.1. THE REASONS FOR USING ARRAYS
It is common for marine seismic sources to be deployed in groups or
arrays. This is done for a variety of reasons. The most obvious of these
is simply to increase the power of the source. The alternative of using
increasingly larger sources is impractical because size increases of a
single source are inevitably accompanied by changes in the
characteristics (e.g. bandwidth) of the emitted pulse. In essence,
therefore, a source system of on' times the power of a single source can
be achieved by firing on' sources together as an array, whilst
maintaining the basic pulse characteristics of the single source.
The next progressive step is to use a variety of different sources
in the array. The possibilities here are endless. Low and high frequency
sources may be used together to improve the bandwidth. Sources which
have deep notches in their spectra may be combined with others that have
complementary peaks, to produce an overall flatter spectrum. Array
elements may be placed at different depths in the water to avoid large
ghost notches in the spectrum. There is also the possibility of firing
the array elements at different times to maximize energy emission in
directions other than the vertical (beam-steering techniques will be
treated later).
Unfortunately, life is never simple and a variety of other array
phenomena must be considered. Of these, the most important are
interference effects between the radiation fields of the individual
array elements, and interaction effects between the physical sources of
the radiation (this will be considered in the next chapter). Both these
phenomena, whilst introducing complications, can be of benefit.

2.2. PULSE SHAPING

One of the commonest applications of arrays is to control the shape of
the time domain pulse (see for example Nooteboom (1978}). The typical
marine source signature is composed of a primary pulse which is preceded
and/or followed by various other oscillations, with periods and
timescales dependent upon the type and size of the source. These
oscillations are often considered undesirable in that they detract from
the ideal situation of a short sharp impulse. By combining sources with
different secondary pulse characteristics in an array, these unwanted
oscillations can be attenuated. The composite pulse will then have a
more desirable shape (in the classical sense) than the individual
elements. Such arrays are called 'tuned arrays'. The principle is
illustrated in Figure 2.1, in which several airguns are combined such
that the primary pulses add constructively, whereas the periodic
secondary oscillations add up destructively.
26 CHAPTER 2

Figure2.! :
The bottom signature is the summation of
the top seven signatures. Notice how the
primary event adds constructively
whereas the secondary osci 11 at ions add
destructively. (Note : the true
amplitude of the composite signature is
three times that shown).

It is certainly true that a source signature closely resembling an

impulse response reduces the need of complex signature deconvolution.
This powerful argument has quite rightly gained almost universal
acceptance. However pulse shaping in an array does destroy potentially
useful energy, and it can be argued that provided the signature is known
accurately, it can be successfully deconvolved, no matter what it looks
1i ke.

2.3. THE PRINCIPLES OF DIRECTIVITY

By definition the radiation emitted by a point source in a homogeneous
medium has spherical symmetry. In such a medium, for example water, the
near field radiation of a single element seismic source is indeed close
to the spherically symmetric. However in the far field, about half the
energy in the propagating pulse originates from the sea surface
reflection, or ghost. Now the time delay between the primary and ghost
pulse is a function of angle. This is purely a geometrical effect. So in
the far field, even the simplest of marine seismic sources, the point
source, has a wavefield which is directional. More commonly this
directionality effect is called directivity.
 SOURCE ARRAYS AND DIRECTIVITY 27

Consider Figure 2.2, in which the radiation pulse emitted at an

angle cp from the vertical, is a combination of the direct pulse and the
secondary pulse caused by the sea surface reflection. The time delay
between the two pulses, T, will be the time taken to travel the
distance x in water, and is given by equation 1.9. Since T is a
function of cp, the emitted pulse will be a function of direction. The
minimum delay will be zero when ¢ 90 0 and the maximum delay,
corresponding to the travel time through twice the depth of the source,
occurs when cp = 0 0 •

SEA

SURFACE Figure 2.2 :

The ghost pulse from a single element
seismic source is delayed in time with
respect to the primary pulse by x/c,
where c is the speed of sound in water.
This delay depends on emission angle $ ,
9: SOURCE so this simplest of sources is
directive.
GHOST
PULSE

PRIMARY
PULSE

The spatial separation of two sources produces directivity in

precisely the same way as the separation of source and virtual image.
However the directivity of arrays can be much more extreme. The
variation of emitted signature with direction is simply a manifestation
of wavefield interference effects. Most readers will be familiar with
the fringe patterns produced by mono-chromatic light sources. The
directivity of seismic sources is precisely the same phenomenon.
However, in general, source configurations are 3-dimensional, and their
emission spectra are continuous, so the resultant fringe patterns can be
extremely complex. As a rule of thumb, these interference effects will
be strong at radiation wavelengths that are shorter than about the total
spatial extent of the source. For example, if the array size is 100m,
then interference effects will be visible at wavelengths less than 100m,
which corresponds to frequencies greater than 15 Hz in water.
Consider two sources separated by a distance, s, as shown in Figure
2.3. In some direction cp , there will be a phase delay, dt, between the
primary pulses of the two sources, where dt corresponds to the travel
time across the distance x; dt is given by equation 2.1 :
28 CHAPTER 2

dt =s sin(¢) (2.1)
c

in which c is the speed of sound in water. Again the time delay is a

funct i on of ¢, and will be zero when ¢ = 0°, and a maxi mum of s/ c (
which equals the travel time across the separation s ) when ¢ = 90°.
Clearly from equation 2.1, the potential maximum phase delay is directly
proportional to the source separation. Therefore at ¢ = 0° the pulses
add constructively, whereas at other angles destructive interference
takes place. In general, the greater the geometrical spread of the
array, the greater will be its directivity.

Figure 2.3 :
In arrays, the time "delay between the
pulses from neighbouring sources is x/co
This delay is a function of angle • ,
so the wavefield of the array is
directional.

It is clear from Figure 2.3 that when x is an exact number of

wavelengths, the energy at the corresponding frequency will add
constructively, so there will be a maximum in the energy emission.
Similarly, when x corresponds to an exact half number of wavelengths
(e.g. 1.5, 2.5, 3.5, etc) then there will be minima in the energy
emission. So, at any particular frequency, there will be a mainlobe of
emission in the vertical direction, and a family of sidelobes of
emission at angles as described above. The direction of any particular
order of sidelobe will be a function of frequency.
Perhaps the most important law that governs directivity is the
trusty old law of conservation of energy. Assuming interaction effects
do not come into play, then the total energy emission by a source array
will be independent of array configuration. The configuration determines
 SOURCE ARRAYS AND DIRECTIVITY 29

how this energy is distributed with angle. So if a particular

configuration attenuates energy in one direction, then the 'lost' energy
will inevitably appear in some other direction. The energy cannot be
destroyed, just redistributed. Furthermore the pressure wavelet
travelling in the vertical direction (assuming the sources are
synchronized in time) will also be independent of array configuration.
This follows directly from the law of superposition of wavefields.
However when the array elements are close enough to interact, the
situation is more complicated, and the vertically travelling pressure
wavelet can be a sensitive function of array geometry. In fact, in the
limiting case, the energy in the vertically travelling wavefield of an
interacting array can be a factor of two less than the vertical energy
from the same sources fired in a non-interacting configuration. These
effects will be discussed in more detail in the next chapter.

2.4. DIRECTIVITY DISPLAYS

Directivity functions are a graphical representation of the energy
emission characteristics of an array. In general, the wavefield of an
array will be a continuously varying function of spatial coordinate
(x,y,z). The wavefield itself can be represented in terms of amplitude
against time (i.e. the source signature), or perhaps more familiar to
the geophysicist, in terms of signal amplitude and phase against
frequency. In essence, therefore, full directivity functions are
6-dimensional, and extremely difficult to visualise.
Meaningful directivity displays can be produced by fixing the
values of several of the variables and displaying the resultant
functional variation of the others. For example, if space is considered
in terms of polar coordinates (r, e , ~ ), then r may be fixed at
infinity (i.e. the far field case), and the variation of the emitted
signature plotted as a function of ~, for fixed e.
Alternatively ~ may be fixed, or else signal ampl itude may be plotted
instead of the signature itself.

Figure 2.4 :
Definitions of the angles e and <I -
the azimuth and angle of dip.
 30 CHAPTER 2

Before continuing a suitable system of spherical polar coordinates

must be defined. In Figure 2.4 the direction of a vector in space is
defined by the angles ( e , ~ ). The angle of dip, or polar angle, $,
is the angle between the vector and the vertical. The azimuth, e , is
the angle between the projection of the vector on the sea surface and
the in-line direction.

(al

(bl

Figure 2.5 :
Alternative directivity function plots.
(a) The emitted signature versus
direction.
(b) The signal amplitude versus
frequency and direction.
(c) The signal ampl Hude versus
direction at fixed frequencies.

(el

Figure 2.5 shows three alternative ways of displaying directivity

information for the same array (in this case a 40m X 40m wide array).
All these plots consider the far field case of r infinity, fix the
azimuth at e 0°, and then consider the variation of the emitted
 SOURCE ARRAYS AND DIRECTIVITY 31

energy with angle of dip, $. In other words these functions show the
emitted energy distribution for the vertical plane that lies beneath the
line along which the boat is steaming (i.e. the in-line plane). Plot (a)
is of emitted signature versus angle of dip, and shows clearly how the
amplitude and phase of the signature varies with direction (and after
all it is these signatures that are convolved into our seismic
sections). Plot (b) is a 3-D representation in which the z-axis is
signal amplitude plotted against frequency and angle of dip. This plot
gives a feel for the overall emitted energy distribution. Plot (c) is
the well known polar directivity function in which a further variable,
the frequency, is fixed, in this case at 60 Hz. This polar plot is
essentially a slice through the 3-D plot above it at fixed frequency.
The 3-D plot of Figure 2.5 (b) is instructive in that it
illustrates some general characteristics of array directivity:
(1) The main beam of emission is vertically downwards (unless the array
is beam-steered - see next section).
(2) Directivity is a function of frequency. There is more at higher
frequencies.
(3) Sidelobes are evident at higher frequencies.
(4) Sidelobe direction is a function of frequency, so although a
sidelobe can be large at one particular frequency, over a range of
frequencies the effects smooth out. This is not true of the
mainlobe which is coherent.

2.5. BEAM STEERING

The elements of an array are usually fired simultaneously, producing an
energy field which has a maximum in the vertically downwards direction.
However in some (rare) circumstances it may be beneficial to adjust the
timings of the array elements to maximise the energy emission in some
other direction. This technique is called beam steering. For instance,
if the predominant dip of the survey is ~ degrees, and there is no
significant heterogeneity, then the maximum illumination of the
sub-surface reflectors will be_obtained if the source energy emission is
maximised at ~ degrees from the vertical.

ARRAY LENGTH

Figure 2.6 :
Beam steering at an angle B may be
achieved by introducing time delays
\~/
along AB such that the wavefront aligns
along CB. For example, the element at B
./

C is fi red as the pul se from A reaches C.

BEAM
\
DIRECTION

I
VERTICAL
32 CHAPTER 2

Consider Figure 2.6, in which AB represents the extent of the

array, and ~ is the required beam steering angle. The array element at
position A is fired first, and the element at B is fired last. The time
delay, dt, corresponds to the travel time in water between A and C. In
other words, B is fired as the pulse from A reaches position C. This
alignment of pulses along CB will maximise the energy emission (maximum
constructive interference) at angle ~ from the vertical. The required
time delay in seconds/metre is given by equation 2.2 for the general
case:

dt = 2.i!:WU (2.2)
dx c

where c is the speed of sound in water. Figure 2.7 shows directivity

functions for an array which has been beam-steered at 20°, corresponding
to dt/dx = 0.23 msec/m.

Figure 2.7 :
Directivity functions for an array which
has been beam steered at 20' . Thi s
corresponds to a 1 inear time delay along
the array of 0.23 msec/m.
 SOURCE ARRAYS AND D1RECfIVITY 33

2.6. DIRECTIVITY OF LONG ARRAYS

A long array is one in which the spatial extension is primarily in-line,
which corresponds to azimuth, e = 0°. An example of a long array and
its characteristic directivity functions is shown in Figure 2.8. This
configuration consists of six sub-arrays of sources, each separated by
30m, to give a total array length of 170m. Emitted signature directivity
functions are shown for the in-l ine (e = 00 ) and cross-l ine ( e = 90° )
planes. The in-line plane has special significance in that it contains
the source and receiver arrays. The cross-line plane shows the other
extreme of directivity, however it should be stressed that the
directivity will vary continuously at intermediate planes between these
two extremes.

LONG ARRAY

________ _LIN CROSS LINE

LINE

ARRAV=LONG
AZIMUTH=Oo
TIME (MSECI
"!!......!!:,o
Figure 2.8 :
Directivity functions for a typical long
array. Azimuth • 0° corresponds to the
in-line plane, azimuth", goe to the
cross-line plane.

ARRAV=LONG
AZIMUTH;9Qo

TIME (MSECI
.n·~o __
34 CHAPTER 2

The directivity functions are as expected. In the in-line plane,

where there is large geometrical dispersion, the emitted signature
varies dramatically with direction. In the cross-line direction the
array looks like a point source, so there is only a small variation
caused by the ghost. So a long array has a narrow vertical beam of
emission in the in-line plane, which gradually increases in size as the
azimuth changes from 0° to 90°. Cross-line the beam is very wide with
large amounts of energy being emitted at high angles of dip. Figure 2.9
further emphasises the directivity in the in-line plane. The first
50msec of the signatures are shown at high angular resolution around the
vertical direction. The signatures are plotted at 1° intervals out
to ± 15°. Clearly, the amplitude and phase of the signature is very
sensitive to angle.

ARRAY. LONG
AZIMUTH . o·

Figure 2. 9 :
Emitted signatures for the long array
configuration of Figure 2.8 . The
signatures are plotted at 1° intervals
for angles of dip close to the vertical,
in the in-line plane .

.. TIME(MSEC)

2. 1. DIRECTIVITY OF WIDE ARRAYS

From an operational point of view long arrays are relatively easy to
achieve in the field, in that a towed string of sources will naturally
align along the direction of motion. For wide arrays this becomes a
problem and special equipment must be used. The usual technique is to
attach the sources to some sort of rigid float. Using two towing points
the float is towed at an angle. This generates lateral lift which moves
the float out, wide of the towing line. The lateral position depends on
the towing angle and boat speed. Using such techniques, stable wide
configurations can be achieved. However, for practical reasons, these
wide arrays generally have length as well as breadth, and an overall
width of greater than about 100m is difficult to attain.
A typical wide array configuration is shown in Figure 2.10. This
array is 10m wide by 10m long and has a tapered shape. The corresponding
 SOURCE ARRAYS AND DIRECTIVITY 35

directivity functions for the in-line and cross-line planes are shown
underneath. In contrast to the long configuration of the previous
section, this array has geometrical dispersion at all azimuths. The
in-line and cross-line directivity functions are therefore quite
similar,and less extreme than those of the long array. In particular,
the wide array emits a great deal less energy out to the side at high
angles of dip. The exact configuration can be adjusted to give an
acceptable energy balance. Some of the applications of directivity will
be discussed in the next section.

WIDE ARRAY

~---
70m

Figure 2.10 :
Directivity functions for a typical wide
ARRAY=W1DE array. Azimuth = 0° corresponds to the
AZIMU'fH;; 0° TIME (MSEC) in-line plane, azimuth - 90° to the
'9'~O cross-1ine plane.

ARRAY=WIDE
AZIMUTH: 90"

\ :
36 CHAPTER 2

2.8. THE USE OF DIRECTIVITY FOR NOISE SUPPRESSION

The rather different directivity characteristics of long and wide arrays
have already been illustrated. In this section the noise problems that
can be tackled by careful array design will be considered. It is worth
reiterating here that directivity is frequency dependent. Figure 2.5 (c)
clearly shows how the beam of emission becomes narrower at higher
frequencies. Consequently the effectiveness of beam forming to attenuate
noise will also be frequency dependent.

Figure 2.11 :
SEA FlOOR For the field geometry illustrated, the
wedge of angles ¢1 to ¢2 can
contri bute to the fi rst order sea fl oor
multiple at the receiver.

f6,.02 MEASURED
FROM VERTICAL

As already mentioned the sea surface is a near perfect reflector of

seismic waves. In areas where the seabed is also a strong reflector,
energy can effectively become trapped, rebounding between sea surface
and floor. As a result useful primary seismic data becomes contaminated
by multiple images of the sea floor. This 'multiple' energy can be
partially suppressed in the data processing or by directivity of the
source array. Consider the scenario of Figure 2.11, in which the
receiving cable length is 3000m and the water depth is 750m. Energy
emitted between the angles Qll and Ql2 can contribute to the first order
multiple as shown. hand $2 are shown plotted against water depth in
Figure 2.12. It is clear that if the beam can be restricted to 10°-20°,
then a large proportion of the multiple energy will be attenuated for
most water depths shown. The source/receiver geometry dictates that
received multiple energy will have been emitted close to the in-line
plane, at azimuths around 0° . So multiple suppression can be achieved
with in-line directivity - in other words by using long arrays. The
example above considered the first order multiple. For higher order
multiples, the angles Qll and Ql2 become progressively smaller, so
progressively tighter in-line directivity is required.
 SOURCE ARRAYS AND DIRECTIVITY 37

1st Multiple
" .. AngleOf Energy Input
d=Water Depth

Figure 2.12 :
The range of source angles ( ~l to ~2 )
which can give rise to the first order
multiple, as a function of water depth.
This figure assumes a receiver cable
1ength of 3000m.

d(m)

There are a number of disadvantages to having a very restricted

in-line beam of emission:
(1) Reflections from in-line dipping reflectors require energy input at
an angle from the vertical, so such reflections will be attenuated.
(2) Reflections from shallow horizontal interfaces will also be
attenuated, since these too require energy input at an angle.
(3) The reduction of in-line emission produces enhanced emission at
other azimuths (conservation of energy). This can accentuate other
noise problems such as sideswipe.

SECTIONAL VIEW PLAN view

Figure 2.13 :
The phenomenon of sideswipe shown in
sectional and plan view. Energy emission
at any azimuth can contribute to
sideswi pe.
 38 CHAPTER 2

The phenomenon of sideswipe is illustrated in Figure 2.13. Energy

emitted out to the side can be reflected back from dipping reflectors,
or scattered back from an uneven sea floor. Energy emitted at any
azimuth can contribute. In fact at small azimuthal angles sideswipe is
an important contributor to the primary seismic data set. However as the
azimuthal angle increases, the returned data becomes spatially aliased
and can be considered as noise. Returns from shallow but distant side
scatterers will contaminate even the deeper seismic data. The emission
characteristics of wide arrays are suitable for reducing sideswipe, in
that the beam of emission is relatively narrow at all azimuths. Figure
2.14 illustrates this property by comparing mainlobe beamwidth versus
azimuth for the long and wide arrays of Figures 2.8 and 2.10. Again the
principle of conservation of energy is evident in that the reduced wide
array emission out to the side is balanced by increased emission around
00 (in-line). In some cicumstances, however excessive reduction of
sideswipe can be undesirable. For instance, in 3-D surveys and on strike
lines, a degree of sideswipe is required to obtain the necessary
information on cross-dipping reflectors.

, - - - - -------------------------,

BEAMWlDTH(I3) v. AZIMUTH (6)

AT 60Hz

90'

Figure 2.14 :

~(
LONG Mainlobe beamwidth (B) versus azimuth
60' ARRAY
( e) for typical wide and long array
configurat ions.

I 30'
yARRAY
WIDE)
I
I
0' I I

-I
-90' 0' 90'

"

Another category of noise that can be attenuated by array

directivity falls under the general heading of source generated noise.
There is an overlap between this type of noise and sideswipe. Source
generated noise embraces direct arrivals from the source, back scattered
noise from all azimuths, refracted waves, and other energy resulting
from a combination of reflections and refractions in shallow layers.
Lynn and Larner (1983) showed that wide arrays were indeed effective at
 SOURCE ARRAYS AND DIRECTIVITY 39

reducing sideswipe and source generated noise. However although the

improvement was visible in the raw data from single shots, in stacked
data there was no apparent improvement compared to data obtained using
compact arrays. They explained this by pointing out that the process of
stacking attenuated the same energy as the wide array. Nonetheless in
areas in which the side-scattered energy propagates at a velocity
similar to the stacking velocity, wide arrays will be effective where
stacking will fail.
Considering all the factors discussed above it is clear that there
is no universally optimal array design. A thorough understanding of the
regional geology is required before specialist long or wide arrays
should be used. In large surveys, considerable changes in geology are
likely to be encountered, and it is impractical to keep changing array
configuration. General purpose designs which have width and breadth, but
not too much of each, would appear to be the answer here.

2.9. EFFECTS OF DIRECTIVITY ON PRIMARY DATA

In previous sections it has been shown that directive arrays may be used
to attenuate unwanted noise in seismic records. A primary manifestation
of this directivity is that the emitted signature varies with direction.
If the array is very extended, then the signature can be a sensitive
function of direction even within a few degrees of the vertical
direction. Now the energy recorded in a seismic shot will have
originated from the source at a range of angles, so this signature
variation will have a profound effect on the recorded primary seismic
data.

Figure 2.15 :
Synthetic shot file for a point source.
The near and far offset source emission
angles are marked for the primary
events.

I!
I I
II:
!

lill
Ii

3080
40 CHAPTER 2

Figure 2.15 is a synthetic seismogram for a simple Earth model. A

single shot record is shown, the individual traces corresponding to
receiver positions 200m apart. In this example a point source was
assumed, so the signature shape is constant throughout. The amplitude
decay effect is due entirely to spherical spreading. On the left and
right hand side of each event the marked angles are the corresponding
source emission angles for each position. The angles are furthest from
the vertical in the shallow data and for the far offset receiver
positions .

. ,,.
Figure 2.16 :
2.0
Synthetic shot file for an array 110m
E
4>

2"
".
".
long. The varying signature with angle
result' from the array directivity.

2'0 3080
Offset (m)

Figure 2.16 is the same synthetic, this time assuming a directive

source array of 110m length. In this and the previous figure the traces
have been individually normalised. Nonetheless it is clear that the
amplitude behaviour of the events is quite different to the previous
example. The signature shape is also quite different particularly in the
shallow and at large offsets. Traditional signature deconvolution
assumes that there is a single source signature convolved into the data,
namely the vertically travelling far field wavelet. This is clearly not
the case for extended arrays as demonstrated by Figure 2.16. A
deconvolution of this data assuming the single wavelet model is shown in
Figure 2.17. As expected this deconvolution breaks down in the shallow
and at large offsets. However the severity of this effect is
dramatically reduced in fully processed data. For example, the data area
where directivity effects are strongest is just the area that is muted
out. Similarly, directivity effects are strongest at high frequencies,
so the worst data is often filtered out. Loveridge et al. (1984b) showed
that for array lengths of up to 100m, 4msec sampling, standard
processing and final high cut filters of 60-70 Hz, the effects of
 SOURCE ARRAYS AND DIRECTIVITY 41

directivity in the fully processed data are probably negligible except

in the shallowest data. Of course, if a greater bandwidth is required,
then the effects will become significant, and directionally dependent
deconvolution may have to be considered.

'",. Figure 2.17 :

Signature deconvolution of Figure 2.16
..... 2.0 using a single filter calculated from
2' the vertically travelling far field
wavelet. The 1imitations of this
approximation are clear.
n'

3080
olTset (m)
 CHAPTER 3

Interaction and Wavefield Determination

The previous chapters have concentrated on more general aspects of
seismic sources, such as their physical properties, their variety and
the basic principles of directivity. In this chapter, the subtleties of
source interaction and some practical aspects of determining signatures
in the presence of such interaction will be discussed.
Of course, it should not be forgotten that the real goal of seismic
source studies is to improve their deconvolution, thereby increasing the
resolution of the seismic method. Historically, deconvolution of such
signatures has been the domain of statistical methods as will be
discussed in Chapter 5. It is the belief of the authors that such
methods have little more to offer. Only by applying deterministic
methods can further progress be made. The theme of this chapter is
therefore determinism.

3.1. CATEGORIES OF SOURCE INTERACTION

Over the years two distinct categories of interaction have been
introduced into descriptions of seismic sources. Regrettably, there has
been little consistency of nomenclature with some attendant confusion.
The essential distinguishing physical characteristic is whether the
'immediate region of disturbance', for example, the bubble in airgun
sources, coalesces with the equivalent for a neighbouring source. This
coalescence is highly non-linear and its underlying physics is even
today a matter of some debate. In this book, the following terminology
will therefore be used:
(a) First order or linear pressure field interaction
This refers to neighbouring sources which do not coalesce. Such
sources interact by a coupling of the pressure radiation fields. When
arrays of sources started to become commonly used in reflection
seismology, it was widely believed that the far field signature of such
an array could be synthesized with arbitrary accuracy by the simple act
of linear superposition of the signatures of each of the contributing
source elements. That this is not the case can be seen clearly in
Figure 3.1 which compares the measured far field signature directly
beneath a seven-element airgun array with the signature calculated by
linearly superposing the seven individual signatures, all measured under
controlled conditions. Obvious differences in amplitude can be seen and
some indications of phase difference also. The differences become even
more pronounced when the same comparison is made after both comparison
signatures have been filtered with a 40 Hz high-cut filter as shown in
Figure 3.2, a more appropriate bandwidth from the point of view of the
processing geophysicist. Again the amplitude differences can be seen
but the phase differences have become much more obvious. Note
46 CHAPTER 3

Figure 3.1 :
Compari son between,
(a) signature measured 150m below an
array of seven airguns, and
(b) signature computed by superposing
the seven signatures obtained by
firing each gun in isolation.
b The filter setting was out - 360 Hz.

0·0 0·1 02
seconds

Figure 3.2 :
The same signatures as Figure 3.1, but
b with a 40 Hz high-cut filter applied.
There are obvi ous di fferences due to the
interaction effects impl icit in
signature (a).

0·0 0·1 0·2

seconds
 INTERACTION AND WAVEFIELD DETERMINATION 47

especially the variation in bubble period. In this case, the physical

reason for interaction can be understood in simple terms by considering
equation 1.15, which shows that the period of oscillation (amongst other
things) of a gas bubble in water depends upon the hydrostatic or water
pressure. If the airgun is fired deeper, hydrostatic pressure is
greater and hence the period is shorter. Similarly, if the airgun is
fired in the proximity of another airgun, then the external pressure it
sees is hydrostatic pressure plus a time varying pressure term induced
by the radiation of the other source. The period of oscillation and
other features of the emitted radiation will therefore change. This
effect is here termed linear pressure field interaction.
An important conclusion is that for airgun sources, linear
interaction effects are most important at lower frequencies. As has
already been discussed however, this frequency dependence is a physical
property of the source in use and may be entirely different for another
source. To emphasize this point and to understand some of the
subtleties of interaction, the following simple model is illuminating.
Consider a system consisting of N springs with weights, wrapped
around a circular frictionless cylinder in a damping medium as shown in
Figure 3.3. Now consider this system to be set in motion by displacing
one or more of the springs from the equilibrium position causing the
weights to oscillate. This system is a good model for interaction as
the movement of the ith weight is intimately associated with that of the
(i-1)th and (i+1)th weights. In other words, they interact.

Figure 3.3 :
A plan view of a model for interaction
consisting of N springs connected
together around a frictionless cylinder.

y; is the displacement of the ith

spring from equil ibrium.
T. is the tension in the ith spring.
k'
1
is the stiffness of the ith spring.
48 CHAPTER 3

Strictly speaking, they interact at two levels, one via the springs
which does not change with separation and one via the damping medium
which obeys the inverse square law. It will be further assumed that the
springs are long enough to neglect the latter, although this of course
is the normal mode of interaction in the seismic marine environment.
Referring to Figure 3.3, the equation of motion for the ith weight
is then
"
Y = -k ( Y - Y ) + k (y - y )
~i ~i ~i ~i-l ~i+l ~i+l ~i
(3.1)
I

-~~.Y ....
1 '1'1
where,
Y~. is the displacement of the ith weight from its equilibrium
1 position.
k~. is the stiffness of the ith spring.
~ 1 is the damping coefficient of the ith weight, dependent on the
~1' shape of the weight amongst other things.
~i = (i modulo N) + 1 because the 1st and Nth springs are connected
together.
and y denotes dy jdt.
The complete system can be written in matrix form as
o ... 0

o =0

o o
d2 d
o o -k N dt 2+( kN-k 1 )+~ dt YN

This system forms a coupled set of ordinary differential

equations. In its simplest form, consider N = 2. Then
,I I

Y1 = -k1(Yl- Y2) + k2 (Y 2 - Y1 ) - ~lYl

(3.2)
 INTERACfION AND WAVEFIELD DETERMINATION 49

Combining equations 3.2 yields

1111 III II I

Y1 + (~l + ~2)Yl + ~1~2Yl + (~2 - ~l)(kl - k2 )Yl

(3.3)

Searching for the presence of harmonic solutions suggests substituting

im1t
Y1 =e
for which case, equation 3.3 becomes

(3.4)

which gives a general solution of the form

eia.t ef:lt
(3.5)

Inspection of these solutions reveals the intricacy of the interacting

movement compared with the simplicity of the non-interacting movement of
one spring which is simple harmonic motion. Furthermore, the frequency
dependence is a funct i on of the phys i ca 1 parameters ki ' 1i
(b) Second order or non-linear pressure field interaction
Such interaction occurs when the 'regions of immediate disturbance' do
coalesce. It is fair to say at this stage that the physical
understanding of this situation is insufficient to be able to predict
exactly what effects this has on the resulting combined signature, and
accurate prediction, as described in the remainder of this chapter for
linearly interacting sources, is impossible. Be that as it may,
non-linear interaction has been deliberately induced by other workers in
the past, (c. f. Larner et al. (1982)) as a means of improving
primary-to-bubble ratios for airgun source arrays in the absence of a
modern understanding of linear interaction. Their rationale depended on
the property that two non-linearly interacting airguns appear as a
single gun of larger volume and consequently longer period than either
 CHAPTER 3

of the contributing smaller guns. This allowed them to achieve the

broad spread of primary-to-bubble ratios necessary for reasonable
statistical deconvolution using Wiener predictive techniques without
having to use a similarly broad spread of airgun volumes. In practice,
this increases the wear and tear of the non-linearly interacting guns
and it is not now widely used. A notable feature of this particular
paper is that it studies in some detail the effects of various band-pass
filters on the basic source array signature. This is in refreshing
contrast to the normal practice of displaying signatures as 'out-out' or
without any filtering, under which circumstance the signature is usually
nothing like its very band-limited appearance in normal seismic data.

3.2. THE WAVEFIELD OF A LINEARLY INTERACTING ARRAY

The problem of linear interaction between oscillating bubbles of
neighbouring airguns has been examined on many occasions with somewhat
different conclusions. Ziolkowski (1970) had considered the issue and
concluded that it was not a problem provided that at least three bubble
diameters separated the interacting bubbles. Giles and Johnston (1973)
followed by Nooteboom (1978) continued, with the latter evolving a
formula to calculate the minimum non-interacting distance. None of
these works however took into account the frequency range of interest.
Further experiments by Lugg (1979) showed that two identical 120
cubic inch (1.97 1) guns just exhibited interaction when separated by a
distance of 480 inches (12.2 m). Safar (1976) considered the
interaction between identical bubbles using another model based on
damped linear oscillators which although not modelling an airgun bubble
particularly well did include the important feature that interaction was
treated as a modulation of hydrostatic pressure, a fruitful avenue as
will be seen. Sinclair and Bhattacharya (1980) although dealing with
non-impulsive sources, concluded that interaction is a
frequency-dependent phenomenon and that it is a modulation of the
hydrostatic pressure field which is responsible.
More recently, Vaage, Ursin and Haugland (1984) reviewed previous
attempts to parametrize linear interaction, combining a discussion of
the various methods with a careful series of experiments including the
effects of a waveshape kit. (A waveshape kit is a baffle introduced
into the airgun chamber to reduce the bubble oscillation). Amongst
other things they concluded:
(a) Linear interaction effects between airguns should be taken into
account in airgun array design.
(b) The primary pulse shape is nearly unaffected by interaction.
(c) Interaction can be used to improve primary to bubble peak-to-peak
amplitude ratios although with an accompanying tendency to
destabilize the signature under field conditions.
 INTERACTION AND WAVE FIELD DETERMINATION 51

For reference, the minimum distance below which linear interaction takes
place is:
Safar (1976)

(3.6)

Nooteboom (1978)

(3.7)

or assuming isothermal expansion

= 8.2 Ro
Johnston (1978)

2.85 VII /3 (pOl·341/po.352) + d (3.8)

gun

where,
VI chamber volume of each gun
PI = chamber pressure of each gun
Po = hydrostatic pressure at airgun depth
dgun = airgun diameter at the exhaust ports
RO = equilibrium bubble radius.

(i) Prediction from near field measurements

Building on these results, the problem of linear interaction was first
solved by Ziolkowski et al. (1982) with further results presented by the
same authors in Parkes et al. (1984b) using airguns as the seismic
source, although the method is independent of the physical properties of
the source itself. Their method is as follows:
For a single airgun, the oscillations of the bubble are spherically
symmetric at seismic frequencies. This is true to at least 500 Hz. The
oscillations are driven by the pressure difference between the inside of
the bubble and the hydrostatic pressure. The hydrostatic pressure can
be taken in the first instance as constant since the rise of the buoyant
bubble is comparatively very slow and when stable, will be that of a
spherical-cap bubble as can be found in any text-book on fluid dynamics,
for example Batchelor (1967). Hence
52 CHAPTER 3

(3.9)

where,
Pd(t) is the driving pressure of the bubble,
P (t) is the internal pressure of the bubble and
PH is the hydrostatic pressure
If Pd(t) > 0, it tends to make the bubble expand or slow down collapse.
On the other hand, if Pd(t) < 0, it tends to make the bubble contract or
slow down expansion.
For n guns fired independently, the driving pressure at the ith gun
would be

(3.10)

Fired simultaneously, this behaviour is modified by interaction.

In particular, each gun sees a constant pressure field modulated by a
variable pressure field due to the effects of the remaining guns. This
may be written as

(3.11)

where,

PH.(t) = P~. + mi(t) (3.12)

1 1

where mi(t) is the modulating pressure.

Now equations 3.11 and 3.12 together give

(3.13 )
 INTERACTION AND WAVEFIELD DETERMINATION 53

Comparing equations 3.13 and 3.10, it may be seen that the

modulated case 3.13, can be considered as an unmodulated case 3.10,
where the internal pressure is modified to be

p.1 (t) - m. (t)

1
(3.14)

Hence, it may be seen that the case of modulating the background

pressure field is physically identical to that of modulating the inside
pressure of the bubble and keeping the background pressure constant.
The important point now is that when the background pressure is
constant, linear superposition can be used. Hence, interaction can be
modelled by linear superposition of the internally modulated bubbles.
Of course, such bubbles do not exist. They are a computational device
and will be referred to as 'notional sources'. However, their behaviour
can be calculated from a sufficient number of near field measurements.

HYDROPHONEj
Figure 3.4: The signature recorded at hydrophone j is the summation
of n 'notional' sources, each scaled and delayed in time
according to its distance. There will also be a
contribution from the n ghost sources.
54 CHAPTER 3

Consider the geometry of Figure 3.4 and suppose there are n

stationary bubbles beneath a free surface with a hydrophone 1m away from
each bubble, i.e. there are n hydrophones. Then the voltage output of
each hydrophone is given by

h . (t)
~:;:
Sj
n
L 1
i:;:1 r ij P,
I. (t _ ijc-1 )
r

(3.15)

where,
P'i (t) is the ith notional source at 1m.
rij is the distance from the ith bubble to the jth hydrophone.
(rg);j is the distance from the virtual image of the ith bubble to
the jth hydrophone.
R is the reflection coefficient at the free surface.
Sj is the sensitivity of the jth hydrophone in volts/bar.
Now if the hi(t) are measured, the notional sources P~ (t) can be
calculated from equation 3.13 by re-writing it as

h.(t) n(i;tj) 1 ,( r i .-1 )

p'J.(t):;: ~ - L - p. t - ~
Sj ;=1 r;j,l c
(3.16)
n
- L
i=l

and noting that r ~ 1m for i = j.

Now it can be seen why the hydrophones should be close to each
bubble. If this is the case, the first term of the right hand side of
equation 3.16, which is the measurement, is an excellent approximation
to the notional source allowing a simple iterative solution of 3.16.
Note however that it may also be solved directly owing to the nature of
the retarded time terms on the right hand side.
In practice, the bubbles are not stationary and Parkes et al.
(1984b) modified equations 3.16 by including a constant velocity such
that

r .. (t) r .. (0) + V .. t (3.17)

-'J -1J -1J
 INTERACTION AND WAVEFIELD DETERMINATION 55

where the v and r terms are vectors and Vij is the relative velocity
between hydrophones j and bubble i. As they showed, this provides an
excellent approximation and acceleration terms, although equally simple
to incorporate, are unnecessary. The end product is a predicted far
field signature which is arbitrarily close to a far field measurement as
is shown by Figure 3.5. It may be concluded that the problem is solved.

1390 cu in Sub Array at 5 m

Measured Signature

Figure 3.5 :
The top signature is a far field
measurement made vertically beneath an
array of seven airguns. The bottom
signature was calculated from near field
measurements using the method described
in the text.

(Signatures for depth of 100m)

1000 200.0 300 0

TIME iMSEC)

It should be noted that if 2n hydrophones are used, the virtual

image of the notional sources could also be calculated directly
eliminating the need to make any assumptions about the free surface.
In order to use the techniques in practice, it is important to
consider operational effects in more detail. First of all, inspection
of equation 3.16 shows that the calculation relies on the geometry being
precisely maintained and also on a knowledge of the hydrophone
sensitivities. Hydrophone sensitivity variation can be accommodated by
appropriate calibration. As far as the array geometry is concerned,
modern airgun suspension systems based on paravane technology (c.f.
Parkes et al. (1984a» provide excellent stability to the point where
the signature changes are negligible over very considerable distance
unless pathologic conditions such as the failure of an airgun occur.
This is illustrated by Figures 4.2 and 4.3. Figure 4.2 shows a sequence
of signatures measured for one sub-array of seven airguns in extreme
weather conditions over several km. Figure 4.3 shows a further sequence
 CHAPTER 3

from the same line and their deconvolution using a single spiking
deconvolution filter computed from their average. This stability is in
marked contrast to the shot-to-shot variations reported by Hargreaves
(1984). This method is considered in more detail next.

(ii) Extrapolation from near field measurements.

An alternative approach to the problem of predicting far field
signatures is described by Hargreaves (1984). This method has
considerable conceptual elegance and consists of towing a mini-streamer
a few metres beneath the source-arrays. A measurement at this mini- or
signature streamer can be extrapolated into the far field using
wave-field extrapolation techniques based on the Kirchhoff integral
solution of the scalar wave equation such that

Pfar(x, y, z, t) = F *J cos 8
(2~r)1/2 Pcable
(
xs ' zo' t - cr) ds (3.18)
cable

where,
F is the inverse transform of w1/ 2 e iTI/4 ,
* denotes convolution.
The author also does a careful analysis of error and concludes that both
the random and systematic error of the extrapolation technique are less
than the experimental error that arises out of a direct measurement of
the far field signature by deep-towed hydrophone.
One restriction of this extrapolation method is that the signature
streamer must be in the far field with respect to the cross-line
dimension of the array. In practice, this limits it to arrays with such
dimensions up to about 20m, which is somewhat less than many operational
arrays. As a result of the error analysis, the author concludes that
the shot-to-shot variations predicted result primarily from real
shot-to-shot variations which, as mentioned earlier, are in marked
contrast to the stability reported by Parkes et al. (1984a). It would
seem that this highlights deficiencies in the airgun suspension system
in use and it would be of interest to see how well this method would
fare with a more stable suspension system.

(iii) Scaling Law methods

Another method which may afford a direct treatment of the linear
interaction problem 1S that described in Ziolkowski (1980) and
Ziolkowski et al. (1980) whereby scaling laws are used to design arrays
of sources in such a way that a deconvolved seismogram may be obtained
directly.
 INTERACTION AND WAVEFIELD DETERMINATION 57

In essence, this scaling law states that for two different

explosive masses with two different equivalent cavities, the far field
wavelets, s' (t) and set), associated with the two sources are such that

s ' (t) = as ( t/ a) (3.19)

where a is the scale factor equal to the cube root of the corresponding
ratio of the source masses or source energies. The two seismograms
recorded by firing the two sources separately, together with the above
scaling law give a set of three equations in three unknowns, enabling a
direct solution. Unfortunately, there seem to be fundamental problems
with this method as reported for example in the excellent paper by
Davies et al. (1984) on the results of the Delft airgun experiment
(Ziolkowski (1984a» which may be due to the non-ideal nature of the
source (Vaage, personal communication), and there seems little further
to add at this stage until further research clarifies the issues.

(iv) Calibration techniques

This method is due to Newman (1985). In this method, a single
secondary source, in this case a watergun, is used to calibrate the
array by firing it a short time (about 2 seconds) before the main
array. This results in the normal seismogram being the concatenation of
two seismograms as shown schematically in Figure 3.6.

o 2.
__ t

Figure 3.6: The concatenated se; smog ram resul t i n9 from the fi ri ng of
a reference source 2 seconds before the main array.
58 CHAPTER 3

The convolutional model in the first two seconds yields

(3.20)

where,
sr(t) denotes the recorded seismogram
r (t) is the reflection series,
fr(t) is the far field of the reference source,
nr(t) is the additive noise appropriate to this seismogram.
After the primary array fires 1 seconds later, say, the seismogram
yielded is

(3.21)
+S(t+1)
r

The author then assumes that Sr (t+ 1) in equation 3.21 is

negligible and indicates various standard techniques for solving
equations 3.20 and 3.21 simultaneously to yield the desired result of
fr(t). Unfortunately, these techniques require the standard assumptions
about nr(t) and np(t) and their relationship and as the author shows no
examples it is difficult to comment further, although it seems to share
at least some of the problems of conventional statistical estimation
techniques as discussed in Chapter 5. One point made in this work on
which the present authors certainly agree is that the trend to source
arrays of greater and greater power is highly questionable to say the
1east.

(v) Numerical modelling.

So far, the theme of the above work is that the far field radiation can
be determined in some way from near field measurements. What, however
is to be done about the vast amount of data for which no measurements
are or were made ? Towards this end, the work of Dragoset (1984) is
particularly noteworthy. Building on the original work of Ziolkowski
(1970, 1982), the author attempts to predict the signature of an
interacting array of airguns based on the physical parameters of each
airgun, including the effects of interaction along the lines of
Ziolkowski et al. (1982).
The final model for an individual airgun has four important
adjustable parameters: the two damping coefficients, the port-open time
and the specific heat ratio of air.
By appropriate adjustment of these parameters, quite reasonable
agreement between predicted and measured signatures is obtainable, which
 INTERACTION AND WAVEFIELD DETERMINATION 59

the author theorises could be further improved by a more realistic model

of the behaviour of the gun-port as the gun is fired. Whilst the
accuracy of the model is currently inadequate for the demands of
signature deconvolution, it provides an admirable method for
investigating the gross effects of such parameters as depth and geometry
on arbitrary arrays of interacting airguns and contributes substantially
to an intuitive understanding of the physics of such arrays.

(vi) Marine Vibroseis

A further step up the ladder of determinism is to force the
temporal pressure signature to be of a certain behaviour. On land, this
method has been in use with outstanding success for many years and is
known as the Vibroseis method. In this method, a plate applied to the
medium is moved with a specified temporally varying force known as a
chirp in radar technology with a functional form:

A(t) =a sin (bt + ct 2 + d)

(3.22)
= a sin ([b + ct] t + d)

which corresponds to a time-variant frequency b + ct, a phase d, and an

amplitude a.
Hence A(t) is the source wavelet which can be removed by normal
deconvolution techniques, (usually based on correlation in this case).
Very recently, this technique has been applied with some success in the
marine environment. An added complication here is the fact that the
source can move several adjacent 'shot' points during the course of the
chirp or sweep as it is known. The effects of this on the deconvolution
and reduction of lateral resolution in the data are yet to be
investigated, but the method has a number of environmental and
operational advantages which may compensate. The development of this
source in the marine environment will be of considerable interest to
reflection seismologists over the next few years.

3.3 ENERGY CONSIDERATIONS

It is worthwhile to discuss the radiated energy and its
relationship to directivity and interaction. As has been described in
the foregoing, directivity is a geometric effect and interaction
pertains to the physical nature of the source in use. It is a property
of the scalar wave equation 1.9 that the law of linear superposition is
obeyed. In the case of directivity, this means that if p(x,y,z,t) and
pl(X,y,z,t) are individually solutions of this equation, then any linear
combination of the two solutions is itself a solution. When the sources
are fired in such a way as to interact, all that happens is that p and
pi change by vi rtue of thi s interaction to become P" and pi" , say, in
60 CHAPTER 3

a manner dependent on the physics of the sources. Neither linear

superposition nor the conservation of energy is breaking down here. The
problem has simply changed. To expand on this, seismic sources can be
categorized in two ways:
(1) Fixed energy sources, e.g. airguns. In these cases a fixed amount
of potential energy is imparted to the air bubble at the time it is
injected into the water. Thereafter this energy is released,
primarily as seismic radiation, although some of it may be lost in
heating up the water, etc. The key factor is that the potential
energy of the system is fixed at the instant of firing the gun.
(2) Fixed amplitude sources. Consider for example a vibrating plate
source in which energy is input into the system throughout the
period in which the source emits seismic radiation. The energy
available in such a system is not so obviously fixed. Suppose that
the system is designed such that the vibrating plate follows a
fixed amplitude pattern of vibration, whatever the external
pressure. Then in such a system more work must be done, i.e. more
energy input, when the external pressure is greater. So, for
instance, if the depth of the source varies during its sweep, then
the energy input to the source will also vary. For such sources
therefore the potential energy of the system is not fixed.
Consider now a single seismic source fired in a medium in which the
external pressure is constant with time. From equation 1.14, the total
wave energy emitted is proportional to the integral over time and space
of the wave pressure squared. For two identical sources, each of which
has a fixed potential energy, E, the pressure measured at any point in
the far field of each source is P, since the radiation emission is
spherically symmetric. In this case, E is simply proportional to the
integral of P squared over time and space. Now imagine that the two
sources have a horizontal separation of 0, and that they are fired
together. The law of superposition of wavefields suggests that the
combined pressure we would measure vertically beneath the two sources
would simply be 2P. We immediately hit a fundamental problem. Suppose
that the source separation, 0, is vanishingly small, so that the
radiation from the two sources together is spherically symmetric, then
the energy available appears to be proportional to (2P) squared, i.e.
°
4E, which is twice the energy available from the individual sources. The
problem does not arise when is large because the apparent quadrupling
of energy emission in some directions is balanced by a reduction of
energy in other directions in which the waves are out of phase and
cancel each other out. Therefore the integrated energy over direction
will be 2E, not 4E.
How can the energy conservation paradox apparently resulting when
the sources are very close together be resolved? In these circumstances
the two sources are clearly not operating in a medium of constant
pressure; rather the external pressure is heavily modulated by the
radiation emitted by the other source. At any particular time the
pressure may be more or less than hydrostatic pressure, but on average
 INTERACTION AND WAVEFIELD DETERMINATION 61

it will be greater. So to maintain an excess pressure wavelet which is

identical to the wavelet obtained when the source is fired in isolation,
additional energy input to the system is required. Because the external
pressure is greater, more work has to be done to increase the pressure
by some fixed amount. Now in the case of sources which fall into
category (1) above, there is no extra energy available, so the pressure
of the emitted wave must reduce to compensate. In fact, the average
pressure must be l/(root 2) less than when the source is fired in
isolation. This is the interaction effect. The detailed pressure wavelet
emitted by the sources fired together will be quite different to the
wavelet emitted by the sources fired in isolation. The form these
differences will take in terms of frequency, amplitude and phase will
depend entirely upon the physics and dynamics of the source in question.
Of course, for sources in category (2) above the situation will be quite
different. In these sources the wavelet amplitude is forced by the
system, so the interaction effects will force the system to input double
the energy, so the energy emitted by two sources will appear to be four
times that available from individual sources. Of course, our category
(2) sources are idealised and such perfection will be unlikely in
practice. Nonetheless the argument illustrates further that the
manifestation of interaction effects depends entirely upon the physics
of the source in question as was intimated by the springs model
introduced earlier.

From the above discussion the following may be concluded:

(a) The source separation at which interaction effects become
significant will depend entirely upon the physics of the source. In
particular, there is no physical reason to believe that this
interaction distance will be a function of frequency.
(b) In fixed energy systems, the energy emitted in the vertical
direction by an interacting array may be reduced by up to a factor
of 2, compared to the summation of the vertical energy emitted by
the same sources fired i~ isolation. However in energy systems
which are not fixed this reduction may be much less.
(c) Arrays may be interacting, or directive, or both. Directivity is
entirely a wavefield superposition or geometrical effect, whereas
interaction is a physical effect. The two effects are therefore
quite different and should not be considered as physically coupled.

3.4. THE AVOIDANCE OF INTERACTION EFFECTS

Given the problems associated with interaction, the reader may have
wondered whether it is possible to avoid such complications. A first
guess might be that the root of the problem lies in the fact that
multi-element sources are used and that if single very powerful sources
were used instead, the problem would cease to exist. Unfortunately,
quite apart from the operational difficulties associated with the
development of such a source, the directivity of multi-element source
62 CHAPTER 3

arrays can be advantageous as has already been discussed in the previous

chapter. Perhaps as important however, is the fact that there 1S no
such thing as a point source in use in marine seismology today due to
the presence of the virtual image or ghost. It may be called a ghost
but as far as the physics of the real source are concerned it is every
bit as real and interacts with the real source accordingly although this
is a secondary effect unless the source is particularly shallow. Hence
if multi-element sources must be used, the only two ways of minimizing
primary (i.e. not through the ghost) interaction are:

(a) Avoidance by spatial extension

A simple way of avoiding interaction effects is to spread the
sources out laterally so that they are too widely separated to
interact. The wavefield can then be determined in any direction by
linear superposition of the individual signatures provided of course
that they are deep enough so as not to interact with their virtual
images in the sea-surface. An unfortunate but inevitable concomitant of
spatial spreading is an increase in directivity, which being
frequency-dependent, can give a very uneven illumination of reflecting
horizons with varying dip, although as was discussed in Chapter 2,
directivity can have its uses.
In practice, lateral spreading of sources is done to achieve
directivity rather than to avoid interaction.

(b) Avoidance by temporal extension

An intriguing possibility of avoiding interaction by extension
whilst simultaneously avoiding unwanted directivity is to spread them
out temporally, as described by Stoffa and Ziolkowski (1983). In this
method, source n will only be fired when source n-1 has ceased to affect
the background pressure field significantly. This again allows linear
superposition to be used and appears to work well although the resulting
long duration temporal signature, (perhaps several seconds), will lead
to some reduction of lateral resolution just as described in Section
3.2. (vi). If the mechanism of interaction was not understood as well
as it now is, this would provide an attractive avenue. It is still a
reasonable solution in the sense that it requires no additional
acquisition hardware and is therefore as reliable operationally as
conventional techniques where the signature is parametrized in just two
ways; primary-to-bubble ratio and peak-to-peak amplitude.

3.5 SUMMARY
Determinism has a bright future. Currently the most tested method is
that descibed in 3.2. (a). In a number of experiments (c.f. Ziolkowski
(1984a), Davies et al. (1984», it appears to be much superior to any
known statistical technique and seems the best of the deterministic
 INTERACTION AND WAVEFIELD DETERMINATION 63

methods known to date in that the signature is predicted arbitrarily

accurately. Competing techniques may eventually achieve the same goal
without recourse to near field monitoring and its operational
implications.
 CHAPTER 4

Practical Aspects of Wavefield Stability

The previous chapters have described how the wavefie1d of a marine
seismlc source depends not only on the characteristics of the individual
elements of the source but also upon the details of the geometry of the
whole source system. Gross effects such as ghosting depend upon depth of
deployment, interaction effects depend sensitively upon the relative
positions of the elements, and directivity effects are a direct function
of geometry. It follows that if the 3-dimensiona1 wavefie1d of the array
is to be stable, then the 3-dimensiona1 geometry of the array must be
stable. So geometry plays a major role in wavefie1d stability, however
there are other factors which we can broadly group under the following
headings :
(1) The timing synchronisation accuracy of individual elements in an
array.
(2) The geometrical stability of the source deployment method.
(3) The intrinsic stability of the acoustic energy generation
technique.
(4) The weather.
The above headings are useful for the sake of discussion, however,
they are not independent. For instance, geometrical stability will
obviously deteriorate with the weather. These topics will be discussed
in the remainder of this chapter. Factor (1) is not a problem in
practice, factors (2) and (3) vary dramatically from source to source,
and factor (4) lies in the laps of the gods!

4.1. TIME SYNCHRONISATION

Correct synchronisation in time is an essential part of a marine source
array. The equipment for carrying out this task can be obtained 'off the
rack', nonetheless the necessity of synchronisation is so fundamental
that it deserves a brief discussion.
The required accuracy of the timing depends on the source
signature. As a general rule, the timing errors must be small compared
to the duration of the primary radiation pulses of the individual
sources. For most sources, the error tolerance is the order of 1
millisecond or less. In most cases, the aim of accurate timing is to
align the primary peaks in the emitted wavelet from each array element.
Some of the practical issues that the timing equipment must resolve to
achieve this are listed below:
(1) The primary radiation peak is not always at the start of the
signature (e.g. waterguns).
(2) The time between the start of radiation emission and the onset of
the primary peak may not be the same for each array element, even
for sources of the same type but different size (e.g. waterguns).
68 CHAPTER 4

(3) There is usually a time delay between triggering a source, and that
source firing. This systematic error may well be different for
sources of identical type and size.
(4) The time delay between triggering and firing may drift with time.
Modern timing systems encompass all the effects above in that they
detect the true start times of the primary peaks and synchronise
accordingly. Continuous monitoring and adjustment corrects for drift.

4.2. SOURCE GEOMETRY STABILITY

In this section some of the practicalities concerning source geometry
will be considered. There are two main problems in setting up
geometrically stable source arrays at sea. Firstly, the energy sources
must be maintained at known and constant depth as they are towed through
the water, especially in rough conditions. Achieving and maintaining
stable array configurations poses the second main problem. In
particular, it is difficult to accomplish stable lateral deviations from
the line of tow.
In the majority of marine acquisition systems the seismic sources
are hung from independent flotation systems which are towed behind the
boat (although some of the less common, more specialist sources are
deployed directly from the boat using cranes which overhang the sides).
The traditional flotation system consists of Norwegian buoys, one per
source. In sub-arrays of buoys, the individual elements are coupled by
cables beneath the surface, and the towing points themselves are
commonly located on the superstructure near the sources. The main
disadvantage of such a system is that the buoys can move independently
on the surface and are therefore sensitive to surface waves and swell.
The uneven drag on the buoys produced by the waves causes the source
suspension cables to oscillate about the vertical. Consequently the
source depths also oscillate. The amplitude of this oscillation will
depend on the source system and the weather conditions, but a 50 percent
depth variation is not unknown!
More recently, a variety of larger, more stable flotation systems
have emerged. These are usually long (about 20m), narrow profile floats,
which may be flexible or rigid. The rigid type, sometimes called
paravanes, are the most stable. The narrow profile of these paravanes
permit them to be towed through the water with minimal resistance,
cutting through surface waves rather than bobbing up and down as
Norwegian buoys are prone to. Furthermore the paravane itself can be
towed rather than the sources which hang down freely. These
characteristics ensure that the depth (and hence hydrostatic pressure)
at each seismic source is maintained accurately. Since emitted waveforms
are extremely sensitive to depth, array signature stability can only be
achieved with precise depth control. In paravane systems it has been
found that even in rough conditions, the gun depths do not vary by more
than a few tens of centimetres for source depths of 5-10m (see for
instance Parkes et al. (1984a) ).
 PRACTICAL ASPECTS OF WAVEFIELD STABILITY 69

Defining the geometry in the other two dimensions can also present
a problem. In particular, obtaining stable lateral deviation from the
line of tow requires that the source sub-arrays be equipped with some
sort of rudder. In the case of paravanes, the body of the float itself
may be used. The principle is illustrated in Figure 4.1. By varying the
cable lengths A and B, the amount of lateral 'lift' or Magnus force on
the paravane may be adjusted to achieve the desired lateral deviation.
This 'wide' deployment method is extremely stable, and lateral
deviations of about 50m can be obtained. The actual deviation achieved
can be predicted from thin aerofoil theory - see for instance Batchelor
(1967) .

WIDE DEPLOYMENT TOWING METHOD

Figure 4.1 ;
Source sub-arrays may be deployed with
lateral deviations from the 1 ine of tow
using the method shown. By varying the
cable lengths A and B. the amount of
lateral 1 ift on the paravane can be
adjusted to give the desired lateral
deviation.

The radiation emitted by marine seismic sources is in general very

sensitive to the external pressure, which in turn is a function of
depth. These effects have been studied sytematically and in detail for
airgun sources, however there has been very little information published
for other source types. From equation 1.15 it is clear that the period
of oscillation of an air bubble will get smaller as the depth and hence
hydrostatic pressure increases. This effect is non-linear and the exact
value depends upon the initial volume and pressure of the gun. As an
illustration, for a 100 cubic inch airgun fired at 2000 psi and at a
depth of 10m, a 10% variation in depth produces an approximately 3%
variation in bubble period. For more details of these effects see for
instance Ziolkowski (1971), Ziolkowski (1970), Vaage et al. (1983) and
Dragoset (1984).
70 CHAPTER 4

4.3. SYSTEM STABILITY

Stability considerations must relate to individual systems, so it is
d i ffi cult to general i se, nonethel ess we can ill ustrate the stabil ity
that can be obtained with a system that ranks among the best of those
that are available. Figure 4.2 shows signatures of a seven gun sub-array
of airguns for a consecutive sequence of about 50 shots. The sub-array
float was of the solid, narrow profile type, which appears to be the
most stable available. The data were obtained whilst shooting a seismic
line in the North Sea, with the airguns deployed at a depth of 5m. At
the time the weather conditions were extremely poor (force 8 to 9). The
signatures shown in the figure were obtained by directly summing the
measurements from seven near field hydrophones which were placed 1m from
each airgun. The contribution of the ghost and the second order terms of
equations 3.12 and 3.13 have therefore been ignored. Nonetheless the
direct summation of the hydrophone measurements is a good measure of
signature stability.

Figure 4.2: Shot to shot variations of the signature of a 7-gun

sub-array of airguns. Even in the bad weather conditions
of this test, signature stability is extremely good.
 PRACTICAL ASPECTS OF WAVE FIELD STABILITY 71

Clearly in this example the signatures are remarkably constant from shot
to shot, especially considering the extreme conditions of this test. The
most notable shot to shot variations occur at about 1.2 - 1.3 seconds.
The similarity of the signatures can be more critically assessed by
trying to deconvolve a sequence similar to Figure 4.2 using a single
filter. Such an analysis is shown in Figure 4.3. The spiking filter was
derived from an average signature calculated from the sequence. The
quality of the deconvolution is illustrated on the right. Clearly, this
particular source sub-array is extremely stable.

.1 .1

.2 .2

.3 .3

.4 .4

.5 .5
,. I r
I!! I I

SPIKING DECONVOLUTION

Figure 4.3: The sequence of signatures on the left have been

deconvolved with a single filter to demonstrate their
similarity.

The data of the previous two figures were part of a full lOkm test
line. The correlation of the individual signatures with an average
signature over the complete line is illustrated in Figure 4.4. The
cross-correlation coefficient is plotted against distance along the
line. The remarkable stability of this particular system is once again
proven (the perfect case is a constant level of unity). The few
mis-fires along the line are only slight, and in fact the largest spike
at about 5.5km was caused by a faulty data recording, rather than a
source fault.
72 CHAPTER 4

1.0

.8
1 ~I T

.6
X-CORR.
COEFF.
.4

.2

.0
0 2.5 5 7.5
DISTANCE (km)

Figure 4.4: Signature stability in terms of the cross-correlation

coefficient between the signature from individual shots
along the lOkm line and an average for the line.

4.4. THE EFFECTS OF WEATHER

Deterioration in sea state is the main weather effect of concern.
Increasing size of sea surface waves and swell has two main effects, (1)
source geometry stability deteriorates, and (2) the planar surface
approximation that the sea I air interface reflection coefficient = -1
no longer holds. Geometrical stability performance is extremely system
dependent. The level of stability that can be achieved in practice has
been illustrated in the previous section. Geometry variation will also
affect directivity characteristics of arrays. Idealised functions such
as those in Chapter 2 will be smoothed out in bad weather. In
particular, holes between side-lobes will fill in, and high resolution
features will tend to disappear.
In this section the effect of surface waves on the reflection
coefficient will be considered. According to Jovanovich et al. (1983),
the RMS amplitude of a sea wave distribution is given by equation 4.1 :

( 4.1 )
 PRACTICAL ASPECTS OF WAVEFIELD STABILITY 73

where Hobs is the observed wave height. This observed height is

approximately equal to the average height of the largest one third of
the waves (called the significant height).

Figure 4.5: A simple model of a wa'Vy surface in which the path

difference between rays A and B equals 2xcos( e).

The effect a rough surface has on the reflection coefficient itself

will be considered by following the formalism of Clay and Medwin (1977).
In Figure 4.5 the height of the wave at any point is x, where x=O is the
mean level. Now consider the two rays, A and B, as shown. Ray A reflects
from the wavy surface, whereas ray B reflects from a planar surface at
the mean level. From the geometry it can be seen that the travel path
difference between the two rays is 2xcos( e). The phase difference, <jl ,
is simply the path difference divided by the wavelength and is given by
equation 4.2 :

<jl = -2xvcos(e) (4.2)

c

where v and c are the frequency and velocity respectively. Now the
wave elevation, x, will have some distribution, p(x), which will be
assumed to be Gaussian and given by equation 4.3 :
74 CHAPTER 4

p(x) (4.3)

where cr is the RMS amplitude of the sea wave distribution given by

equation 4.1. The wave reflected from the surface will be the summation
over all elevations, x. An average reflection coefficient, Rav , can
therefore be obtained by integrating the effect of the phase change over
all x, as described by equation 4.4 :

+00

Rav = RO f exp (-2ix~COS(e) ) 1 /2 exp (-2:22 ) dx

cr(2'IT) 1 u
(4.4)

The evaluation of this integral is given in equation 4.5

(4.5)

The average reflection coefficient is therefore a function of angle and

frequency, as well as of wave height. Of course, the simplifying
assumptions of the above model must be borne in mind. In particular, it
assumes that each point on the surface wave acts like a horizontal
reflector. The approximation is therefore only valid for small
wave-heights and/or long incident wavelengths. For moderate wave heights
the relationship should hold for frequencies up to about 250 Hz (see
Loveridge (1985)).

4.5. MARINE GHOST VARIABILITY

The variation in the reflection coefficient discussed above will have a
profound effect on the wavefields of marine seismic sources. In a
vertically travelling far field wavelet, about half the energy results
from a coherent sea surface reflection. The effect of a rough surface
will be to produce a composite, smoothed out ghost wavelet in the time
domain, and in the frequency domain the narrow deep 'ghost notches' will
become shallower and broader (see for instance Jovanovich et al.
(1983)). An example of this weather effect is shown in Figure 4.6, in
which the top signature is a far field measurement of an airgun in good
weather conditions, and the bottom signature is a similar measurement in
 PRACTICAL ASPECTS OF WAVEFIELD STABILITY 7S

poor weather. The broadening out of the ghost wavelet is quite evident.
The high frequencies in the ghost wavelet are removed, so in the
frequency domain this 'weather effect' acts like a low pass filter.

FAR-FIELD MEASUREMENTS

Figure 4.6 :
The top far field measurement was made
in good weather and the bottom
measurement in poor weather. The reduced
ghost size due to surface roughness is
evident.

There is another rather different mechanism which may affect the

ghost wavelet in certain circumstances (see Loveridge et al. (1984a) and
Loveridge (1985)). This mechanism is called the 'shot effect'.
Especially with the more powerful sources, it is common to see a surface
disturbance immediately above the source at the instant of firing. A
fine spray is thrown from the surface which becomes agitated and appears
to seeth. For sources such as airguns, this disturbance has nothing to
do with the air bubbles which do not reach the surface for several
seconds. A particularly demonstrative example of this can be seen in
historical pictures of anti-submarine warfare, where the surface is
ruptured by a depth-charge almost instantly and several seconds before
the expanding gases reach the surface producing a huge fountain of
water. The effect is illustrated in Figure 4.7. It can be understood as
follows. The atmospheric pressure and the cohesion forces in the water
combine to give the water surface a certain tensile strength. Now if the
excess pressure in the seismic wave is greater than some critical value,
Pc , this breaking strength will be exceeded and the water will be
lifted up. A very simple model of this mechanism is illustrated in
Figure 4.8. Here it is assumed that all the energy above Pc is removed
from the reflected wavelet because it is expended in lifting and
vapourlslng the water and as heat. Therefore any part of the wavelet
that exceeds Pc will be clipped at this level.
76 CHAPTER 4

SHOT EFFECT

Spray and wOller V.1pour

/ thrown oU surlLlcc

\ II 1111111/,
-~~
---~"-~
Figure 4.7 :
' " Area 01 surface
dlslurt>ancc
The 'shot effect'. The radi at i on from
the sei smi c source produces an area of
agitated water and spray at the surface.

r:::~
\::J

(a)

lime Figure 4.8 :

In a simple model of the shot effect,
the peaks of the incident wavelet (a)
which exceed some critical pressure,
Pc ' are removed to give wavelet (b).
'b)
 PRACTICAL ASPECTS OF WAVEFIELD STABILITY 77

The above model is too simplistic in that some of the 'lost' energy
will be re-radiated later. There will also be secondary effects in that
the roughening of the surface will in itself produce a change in the
reflection coefficient in a manner analogous to the weather effect
discussed previously. There seems to be very considerable debate as to
the exact value of the tensile strength of water, with estimates ranging
from less than one bar up to several hundred bars (e.g. Weston (1960),
Nyborg et al. (1972». Nonetheless it is clear that the shot effect does
take place. Like the weather effect it will act primarily as a low pass
filter and will be most important for very powerful sources and/or for
sources which are fired at shallow depths. A full understanding of the
mechanism must await detailed experimental and theoretical work.

4.6. SUMMARY
The modern seismic source is extremely robust, predictable and reliable,
if a single element is considered under 'laboratory' conditions. However
in production acquisition 3-dimensional arrays are towed from boats in
the open ocean, and often in far from ideal conditions. In normal
operation there are numerous practical problems which must be resolved.
The main issues are as follows:
(1) The radiation emission characteristics of the source must be
suitable for the job. This is really an issue of source and array
design. In general, the emission spectrum must be smooth, without
deep notches, and the balance between low and high frequencies must
be tuned to the application. As previously mentioned the shape of
the time domain wavelet is less important than the spectrum, but
must be known with accuracy !
(2) The whole source system must be stable. This includes flotation
systems, towing systems and adaptive timing.
(3) The radiation field of the system must be known so that it can be
deconvolved from the recorded data.
Point 3 above should cover the possibility that the radiation field
will vary from shot to shot. If the system stability is good then shot
to shot variations can be minimised. Nonetheless, they will always be
present, even in the best of systems. There have been two main
approaches to signature deconvolution in the past. One has been to tow a
'far field' hydrophone under the source during acquisition. This goes
some way towards including shot to shot variations, but has some major
drawbacks, in that the hydrophone is not in the true far field, and it
only samples one point in a spatially variant wavefield. The more usual
approach has been to measure the vertically travelling far field
signature in a well controlled trial, and to use this single signature
to deconvolve the wavefield thereafter, assuming that there is no time
or space variability of the wavefield. This second technique has become
the industry standard, and sometimes works reasonably well, however it
has severe and obvious limitations.
 CHAPTER 4

Another possibility which has emerged recently has been to use

entirely synthetic signatures. Indeed models of some sources are now
extremely good and produce signatures that agree very closely with field
measurements. Furthermore such synthetics can be used to model the
spatial variation of the wavefield. However once again the inability to
cope with shot to shot variations is a severe limitation. Nonetheless
this method does offer some hope for the reprocessing of old data for
which signature measurements were not made.
Perhaps the most promising approach for the future lies with
techniques like that described in Chapter 3, in which near field
hydrophone measurements are recorded each shot. In theory, such a method
allows the full time and space variability of the wavefield to be
calculated and later deconvolved. If extra hydrophones are used, then
the ghost contribution to the wavefield can also be assessed
independently, so that the weather and other ghost effects previously
discussed can be included.
A primary aim of the seismic method is to extend the useful
bandwidth to higher frequencies and hence improve the spatial resolution
of the data. In the final analysis, attaining this goal rests upon
achieving increased stability of source systems and/or developing
sophisticated techniques for measuring shot to shot wavefield variations
and later deconvolving these variations.
 CHAPTER 5

Source Signature Deconvolution

5.1. INTRODUCTION
The deconvolution of the seismic source signature has retained the
interest of reflection seismologists from the earliest days of the
science. Today, with the ever-increasing emphasis on temporal
resolution, more is expected of the deconvolution process. The issue of
whether such resolution is attainable with the information at hand is
unanswered and quite often unaddressed. This is however beyond the
scope of this textbook and will not be considered further.
Historically, the merits of the deterministic approach to
deconvolution have always been recognized as desirable, but until
recently, have not been realized as has already been discussed in
Chapter 3. An early noteworthy attempt at signature deconvolution was
the Maxipulse* source which achieved considerable success in the early
1970's. This source involved the discharge of a 1/2 lb can of dynamite
timed to explode a metre or so away from a signature hydrophone.
Variations in the charges themselves necessitated a shot-by-shot
deconvolution and similar variation in the shot/hydrophone geometry
affected the quality of the result. Nevertheless, the resulting
deconvolution was often capable of achieving a signal-to-residual
convolution noise ratio of better than 20 db, even though the signature
itself was highly oscillatory and in excess of 1 second in duration.
The source was eventually displaced by non-explosive sources such as
airguns for reasons of economy and ecology as much as quality. However,
the deterministic method always places extra demands on an often
over-stretched acquisition system and so the growth of statistical
methods, whereby 'the' signature was extracted from the recorded data
itself, was inevitable.
Before discussing the various merits of the statistical methods, it
is worthwhile quickly reviewing just what exactly is the perceived
objective of seismic source deconvolution.
Inevitably, the underlying model used is the venerable and highly
successful convolutional model, represented by

s(t) = w(t) * r(t) + n(t) (5.1)

where,
s(t) is the recorded seismogram,
w(t) is the wavelet,
r(t) is the earth's reflection series
and
n(t) is additive noise.

* Registered trademark of Western Geophysical

81
82 CHAPTERS

See Ziolkowski (1985b) for a discussion of this model.

In practice, w(t) is itself composed of a number of convolutional
responses such as the source and receiver ghosts, the instrument impulse
response and so on but these will not be explicitly introduced here.
In an ideal deterministic approach, w(t) is known and a Wiener
shaping filter f(t) can be designed such that

w(t) * f(t} = d(t} (5.2)

where d(t) is the desired output and is usually of short duration and
pleasing shape, c.f. Hatton et al. (1986).
Convolving both sides of equation 5.1 with f(t) gives

f(t) * s(t) = d(t) * r(t) + f(t) * n(t) (5.3)

where the first term on the right hand side corresponds to the desired
seismogram and the second term to filtered noise.
Even if w(t) is known, it must be practically invertible, i.e.
N( v )/W( v) is small in some sense, where W( v) is the Fourier transform
of w(t), N( v) is the Fourier transform of the noise and v is the
frequency.
This follows from noting that, in the frequency domain, equation
5.2 can be written

Qi:U (5.4 )
F(v) = W(v)

and equation 5.3 can be written

F(v)S(v) = D(v)R(v) + F(v)N(v} (5.5)

Now, if F( v) N( v) » D( v) R( v), noise dominates the inversion.

This is equivalent to N( v) » W( v) R( v). Hence, practical
invertibility demands:
(a) A knowledge of W( v) for v in the frequency band of
interest.
(b) N( v) « W( v ) R( v), for v in the frequency band of
interest.
 SOURCE SIGNATURE DECONVOLUTION 83

In particular, note that notches in the spectrum are only important

if the signal-to-noise ratio is poor there, (as it usually is).

5.2 DETERMINISTIC DECONVOLUTION

Having introduced the convolutional model, this section will show some
examples of real source signatures and their deconvolution in ideal
circumstances. Note that the signatures are shown as the processing
geophysicist sees them. This is normally the opposite polarity to that
of the acquisition geophysicist.
The first point to make is to re-emphasize the point that when the
signature is known by any of the methods described in Chapter 3, for
example, the only additional requirement is that it be practically
invertible in the sense defined above. The actual temporal shape is
irrelevant.
To emphasize this fact, a regal but somewhat extreme source,
Windsorseis, is shown in Figure 5.1, with its truly awful amplitude
spectrum shown as Figure 5.2. Awful though it may look, this source can
still be inverted.

Figure 5.1 :
A hypothetical source, Windsorseis,
which has an improbable and extremely
undesirable signature (in the classical
sense) !

,_ _ J

TIME (MSECl

Figure 5.2 :
The ampl itude spectrum of the signature
of Figure 5.1.

FREOuENCY (HZ)
84 CHAPTERS

Figures 5.3 and 5.4 show a synthetic reflection series and an

actual seismogram computed using this source and the convolutional
model, equation 5.1. Using standard Wiener inverse filtering techniques
as discussed, for example, by Hatton et al. (1986), the deconvolved
seismogram is shown in Figure 5.5. The result is an excellent
reconstruction in spite of the pathological nature of the source.

"-1

Figure 5.3 :
A synthetic reflection series.

Figure 5.4 :
A seismogram obtained by convolving the
reflection series of Figure 5.3 with the
source Signature of Figure 5.1.

.".~. j

Figure 5.5 :
The seismogram of Figure 5.4 after the
source signature has been deconvolved
us i ng standard Wi ener inverse fi 1 ter; n9
techniques. Clearly the deconvolved
se; smogram ;s an excell ent
reconstruction of Figure 5.3, despite
the pathological nature of the source.
 SOURCE SIGNATURE DECONVOLUTION 85

Hence a knowledge of the exact temporal response of the source and a

belief in the convolutional model produces a reconstructed reflection
series limited only by the effective bandwidth of the source.
Such knowledge is particularly comforting in reality, as the
temporal shape of any practical source varies dramatically as a function
of bandwidth, amongst other things. Figure 5.6 shows the far field
signature of an airgun array of seven airguns with sizes ranging in size
from 50 to 305 cu. in., each at a depth of 5m and sampled at 2 msec
intervals. Figure 5.7 shows its amplitude spectrum which can be
inverted in the approximate bandwidth 4-140 Hz.

"'1
Figure 5.6 :
The vertically travelling far field
Signature of a typical airgun array with
seven guns, deployed at a depth of Sm.

-lze.0

TIME (MSEC)

~\
"\\
Figure 5.7 :
The ampl itude spectrum of the airgun
array signature shown in Figure 5.6.

.000 -----

FREQUENCY (HZ)
86 CHAPTERS

TIME CrtSECl

F
I

Figure 5.8 :
The Signature of Figure 5.6 after
filtering with zero-phase bandpass
filters as follows,
TIME (I'ISEC) (a) 10 - 80 Hz
(b) 10 - 60 Hz
(c) 10 - 40 Hz
(d) 8-30Hz.
The filter slopes on the low and hig~
side were respectively 12 and 36 db per
octave.

liME (MS[C)
 SOURCE SIGNATURE DECONVOLUTION R7

Figures 5.S a-d show the signature of Figure 5.6 filtered to

bandwidths 10-S0, 10-60, 10-40 and S-30 Hz using a zero-phase bandpass
filter with slopes of 12 and 36 db per octave on the low and high sides
respectively. This models the time-variant filtered appearance such a
signature might have on a typical seismic section as two-way travel time
increases from 0 to about 6 seconds. The change is considerable and the
rapidly decreasing primary-to-bubble ratio shows that the strong low
frequency content of this source is associated with the bubble period.
In reality, other effects combine to change the temporal signature
even further, particularly absorption. Absorption is often modelled
using the assumption that such losses are a constant fraction for each
wavelength traversed. This assumption leads to a decay proportional to
exp( - TIV t/Q ), where Q depends on the transmitting medium. Figure
5.9 shows the signature of Figure 5.6 after absorption losses
corresponding to a Q of 100 at a two-way travel time of 2 seconds, a
reasonable value for the North Sea, for example.

120.0

.0

-120.0

.0 200.0 300.0 400.0 500.0

TIME (M5EC)

Figure 5.9: The signature of Figure 5.6 corrected for frequency

dependent absorption in the Earth, corresponding to a Q
of 100 at a two-way travel time of 2 seconds.
88 CHAPTERS

As a comparison with the airgun behaviour described above, Figures

5.10-5.13 show exactly the same suite of examples for a single watergun.

15.00

.,.

-15.00

.. 100.0 200.0 300.0 400.0 500.0

TIME (MSEC)

Figure 5.10: The vertically travelling far field signature of a

typi cal watergun array.

I.Me

!)
r ~
i'---.---_~--~_----'
.000

.. 100.0 200.0

FREQUENCV (HZ)

Figure 5.11: The amplitude spectrum of the watergun array signature

shown in Figure 5.10.
 SOURCE SIGNATURE DECONVOLUTION 89

(a)

TIME lM5ECl

~----~~---~

I (b) I

Figure 5.12:
The signature of Figure 5.10 after
filtering with zero-phase bandpass
filters as follows,
(a) 10 - 80 Hz
(b) 10 - 60 Hz
(c) 10 - 40 Hz
(d) 8 - 30 Hz.
(e) The filter slopes on the low and high
side were respectively 12 and 36 db per
octave.
I'·

-----v
fI
i rj
!\\ .II'\/-..'-'----..
------~--------~-

~J V

TIME (MSEC)

"MGl ~

, -J\Iv \ r~<--~----v-II
- --- ~~_ _ r-- _ _ .~~~~_,_J
I
90 CHAPTERS

ts.ell

Figure 5.13:
The signature of Figure 5.10 corrected
for frequency dependent absorption in
the Earth, corresponding to a Q of 100
at a two-way travel time of 2 seconds.

TIME (MSECl

As has been stated previously, both airguns and waterguns are

eminently invertible as is shown in Figures 5.14-5.18.
Figure 5.14 shows a typical desired output of a Wiener signature
shaping filter.

Figure 5.14 :
A Wi ener fil ter is used to transform the
source signature into a more 'desirable'
wavelet. The wavelet shown is a typical
desired output; a minimum phase filter
with a passband of 10 - 100 Hz, and low
and high slopes of 12 and 36 db per
octave respectively.

TIME (MSEC)
 SOURCE SIGNATURE DECONVOLUTION 91

Figures 5.15 and 5.16 show the derived filters for the airgun array
and watergun respectively.

Figure 5.15 :
The Wiener fi lter that transforms the
airgun signature of Figure 5.6 into the
desired output of Figure 5.14.

TIME (M5EC)

I I -
. I ~_I
j~~~
Figure 5.16 :
The Wiener filter that transforms the
watergun signature of Figure 5.10 into
the desired output of Figure 5.14.
I

L ... I'--r------,----~__________r___'I
TIME (MSEC)
92 CHAPTER 5

One point worthy of note is that the strong low frequency content of the
watergun shaping filter is to compensate for the relative paucity of
such frequencies in the watergun source, Figure 5.10, compared with the
desired output, Figure 5.14. However, in both cases, an excellent
shaping is obtained with a 250 point filter with anticipation component
of length 125 points and a white light percentage of 0.2 as can be seen
in Figures 5.17 and 5.18, the actual outputs of the Wiener filtering for
the airgun array and watergun respectively. As a final comment, note
that in this case from Figure 5.15, a considerably shorter filter would
have performed just as well for the airgun shaping.

12tl.t

Figure 5.17 :
The resultant shaped airgun array
signature (wavelet 5.6 convolved with
wavelet 5.15).

120.0

TIME (MSEC)

lS.ee

Figure 5.18 :
The resultant shaped watergun signature
(wavelet 5.10 convolved with wavelet
5.16) .

nME (MSEC)
 SOURCE SIGNATURE DECONVOLUTION 93

5.3 STATISTICAL DECONVOLUTION

When looking at the literature, the first thing that impresses is the
ingenuity displayed. There are many such methods, each with their own
strengths and weaknesses. For more details, the work of Lines and
Ulrych (1977) or Jurkevics and Wiggins (1984) should be consulted.
Here, only a series of 'thumbnail' sketches will be given, illustrating
the most popular statistical methods along with some of their drawbacks.
(i) Predictive deconvolution
This is the standard deconvolution method in seismology, c.f.
Peacock and Treitel (1969), and its assumptions have dominated the area
of source signature deconvolution for many years. These assumptions
are:
(a) Stationarity
(b) Minimum phase condition
(c) Statistically white reflection series.
Under these assumptions, it is possible to reconstruct the wavelet
by assuming that the autocorrelation of the recorded seismogram is a
good measure of the autocorrelation of the wavelet using (a) and (c)
above, and then constructing the unique minimum phase wavelet with this
autocorrelation using (b). The classical way of satisfying assumption
(b) has been to achieve a primary-to-bubble ratio of some reasonable
amount, say 8 to 1 in a broadband recording, and then to state that such
a wavelet is minimum phase in the seismic band.
Considering the chain of assumptions, this has served reasonably
well. A comparison of this technique against the very accurate notional
source deterministic method described in section 3.2 (i) showing its
considerable practical limitations can be found in Ziolkowski (1985b).
The method can be improved a little by averaging traces from the same
shot prior to autocorrelation, but things are unlikely to improve much.

(ii) Forced minimum phase.

This approach has considerable intuitive appeal and is due to Taner
(1980). In essence, it is derived from the method described in Section
5.3 (i) with the additional step that the recorded seismogram has an
exponential taper applied to it before autocorrelation is applied.
After the wavelet is extracted, the inverse of this taper is applied to
the resulting wavelet. The rationale is that an exponential taper
guards against a mixed phase wavelet in the sense that it forces minimum
phase, if a severe enough taper is applied to force all poles of the
equivalent z-transform of the wavelet outside the unit circle, c.f.
 CHAPTER 5

Robinson and Treitel (1980). The non-minimum phase wavelet is then

extracted as in Section 5.3 (i) and the taper is re-applied. The taper
is assumed to commute with the process of extraction and therefore the
resulting wavelet should be independent of the taper, providing it is
severe enough to achieve the desired minimum phase condition. In
practice, this is not the case, although it is reasonably satisfactory
for wavelets with a small mixed-phase component. See Jurkevics and
Wiggins (1984) for a detailed analysis.

(iii) Minimum entropy deconvolution

This method is due to Wiggins (1978). It replaces the assumptions
of Section 5.3 (i) with one of sparseness of reflection coefficients in
that the recorded seismogram is assumed to be made up of few major
changes in acoustic impedance, separated by relatively homogeneous
regions. In practice, the method can produce wavelets which are highly
variable shot to shot and which are consequently suspect. Again, see
Jurkevics and Wiggins (loc.cit.) for more detail.

(iv) Dominant reflection deconvolution

In this approach, a dominant horizon is chosen, perhaps the
water-bottom in marine data, and a wavelet extracted directly by gating
or windowing the recorded seismogram at the appropriate time. This
method can work reasonable in practice, especially if considerable
lateral averaging is performed, however, it has the fundamental drawback
that the geophysicist is never sure if a dominant horizon is an isolated
'pure' reflection or a composite reflection from a number of closely
adjacent layers, as is often the case.

(v) Homomorphic methods

These methods derive originally from speech and radar processing.
In essence, they work by multiplicative filtering (liftering) of the
logarithmic Fourier domain. Although very appealing in theory, they
have proved disappointing on real seismic data. Buttkus (1975) and
Clayton and Wiggins (1976) show that the real problem is that of
unwrapping the phase in excessively noisy data. In very high quality
data such as occurs in speech processing, this is achievable. Very high
quality seismic data has yet to be acquired.

5.4. PRACTICAL NOTES

Before closing, it is worthwhile discussing another effect which
may become important when the major hurdle of determining the temporal
response of the source is overcome. As has already been noted, even a
single source is directional by virtue of its ghost. Arrays of sources
 SOURCE SIGNATURE DECONVOLUTION 95

are even more directional. In practical deconvolution, this

directionality means that the signature varies according to the
direction of emission. Ideally, this geometric effect should be
incorporated when deconvolution operators are designed.
As far as statistical methods are concerned, one of the most
important contributing factors to their continuing failure is the poor
quality of seismic data. Whilst this problem has been diminishing
gradually as acquisition methods improve, more is expected of the
resulting data by way of compensation. It would seem that until the
basic quality improves without correspondingly higher aspirations, the
future looks bleak for statistical methods. This is unfortunate as such
methods are of course ideal in the sense that no further demands are
made on the acquisition system as is not the case for most of the
deterministic methods discussed in Chapter 3. However, a lesson learned
time and again with seismic data is that progress is made only by
considering the acquisition and processing of seismic data as merely
different sides of the same coin. With this in mind, deterministic
approaches seem to offer the only way forward today.
 APPENDIX

Technical Description of Main Sources

There are a large number of different seismic sources presently
available and being used. However at the time of going to press 99% of
the world production of marine seismic data is using either airguns or
waterguns as the seismic source. Technical information on these two
sources is therefore given in this appendix.
100 APPENDIX

SOURCE TYPE AIRGUN

DESCRI PTION The airgun has become the most widely used of all
marine sources because of its simplicity, robustness
and reliability. The principle of operation, as the
name suggests, is the explosive release of a
'charge' of high pressure air into the water. The
traditional design has two air chambers, a control
chamber and a firing chamber. The two are connected
by a moving shuttle which is kept closed by the
pressure of air in the control chamber. At firing
time a charge of air is forced under the shuttle
pressure plate in the control chamber. The shuttle
triggers, opening the firing chamber and venting the
air. Some designs have four venting holes, others
have a sleeve which allows full 360 0 venting. On
release the bubble of air oscillates producing an
essentially periodic decaying signal (see Chapter 1
for more details). The amplitude of the primary
pulse and period of oscillation increase with
chamber volume. Special modifications to the airguns
called waveshape kits are available to attenuate the
bubble oscillations if desired.

VARIETIES Standard pressure, high pressure, sleeve guns, and

many variations.

TECHNICAL DATA Reservoir volume range 0.5 10000 cubic inches

(more restricted for high pressure and sleeve guns).
Operating pressure:
400 5000 psi for standard pressure guns
(typically 2000 psi).
4000 - 6000 psi for high pressure guns.
Cycle time typically 3 - 6 seconds.
Energy released per shot 200 - 11000000 Joules.
Weight 18 - 4000 lb.
Primary to bubble ratios of single guns typically
2:1 (0-250 Hz).
Power output examples (peak to zero, 0-250 Hz):
10 cubic inch - 0.5 b-m.
200 cubic inch - 3.0 b-m.

ARRAYS Airguns are usually fired in arrays in which a

mixture of chamber sizes are used to increase the
power, increase the primary to bubble ratio, and
produce a broad, relatively smooth spectrum. The
 APPENDIX 101

total airgun volume in typical arrays lies in the

range 100 - 8000 cubic inches. The peak to peak
output of a 1000 cubic inch array (a-125Hz) would
lie in the range 15 - 30 bar-metres, with primary to
bubble ratios of about 8:1.

OTHER INFORMATION Figure Al shows a representative unfiltered near

field airgun signature from a 200 cubic inch gun
fired at 2000 psi and 5m depth. Figure A2 shows a
representative array signature (vertical far field)
from a 910 cubic inch, 7-gun array fired at 2000 psi
and Sm depth. Both of the amplitude spectra on the
right are normalised and are on a linear amplitude
scale.

Figure AI: Near field signature and ampl itude spectrum of a 200
cubi cinch ai rgun fi red at Sm depth.

Figure A2: Vertically travelling far field signature and amplitude

spectrum of a 910 cubic inch array of airguns.
102 APPENDIX

REFERENCES The following list is a selection of papers on

airguns. These references and references therein
should provide a useful starting point for a more
definitive sample of the available literature.
Dragoset (1984)
Giles and Johnson (1973)
Johnson (1978)
Johnson (1980)
Johnson (1982)
Nooteboom (1978)
Parkes et al. (1984a)
Parkes et al. (1984b)
Safar (1980)
Vaage et al. (1983)
Vaage et al. (1984)
Ziol kowski (1970)
Ziol kowski (1971)
Ziolkowski and Metselaar (1984)
 APPENDIX !O3

SOURCE TYPE WATERGUN

DESCRIPTION The watergun is basically an implosive source. In

the traditional design there are two chambers, an
upper air chamber and a lower water chamber which is
open to the sea. A moving shuttle divides the
chambers. High pressure air is fed into the top
chamber. At firing time the shuttle is released, and
the high pressure of the air forces the shuttle
downwards. A slug of water is thereby violently
ejected from the lower chamber. As the shuttle slows
down and the water continues to expand a vacuum
cavity is formed between them. The expansion slows,
stops, and is followed by an implosion of water into
the cavity. The primary peak in the emitted
signature is caused by this implosion. This is
preceded by a lower frequency pre-cursor produced at
the water expansion stage. Since there is no gas
bubble, the radiation pulse is short and without
oscillations. In general, the emitted radiation is
broad band and extends to high frequencies, although
some waterguns suffer from deep notches in their
spectra at low frequencies.

VARIETIES Pneumatic, hydraulic and hybrid models available,

also combination models which can be operated as
airguns or waterguns.

TECHNICAL DATA Volume range 15 - 500 cubic inches (water emitted).

Water pressure 500 - 3000 pSi.
Air pressure 140 - 3000 pSi.
Weight 29 - 375 lb.
Some air driven models require air exhaust line to
surface.
Firing cycle 0.5 - 8 seconds.
Minimum and maximum firing depths are 1ft to 1800ft.
Power example: a 400 cubic inch gun delivers a peak
to zero power of about 3 bar-metres in the 0-125 Hz
band, with a primary to secondary ratio of about
4: 1.

ARRAYS Waterguns are frequently deployed in arrays,

although unlike airguns it is common for the array
elements to be the same size.
104 APPENDIX

OTHER INFORMATION Figure A3 shows a representative near field

signature from a large watergun. Figure A4 shows the
vertically travelling far field signature of the
same gun. Both of the amplitude spectra on the right
are normalised and are on a linear amplitude scale.

w
C

...
::J
I-
:::::i 1.0

:I'
w
;:,
5w
a:

Figure A3: Near field signature and amplitude spectrum of a large

watergun.

I!J
..
~,
:i
:I'
w
>
...w~
a:

Figure A4 : Vertically travell ing far field signature and ampl itude
spectrum for the watergun of Figure A3.

REFERENCES Newman et al. (1977)

Safar (1984)
Safar (1985a)
Safar (l985b)
Tree et al. (1982)
REFERENCES
Aki, K., and Richards, P.G., 1980, Quantitative Seismology - Theory and
Methods, Vol. I, W.H. Freeman and Company, San Francisco, U.S.A.
Batchelor, G.K., 1967, An introduction to Fluid Dynamics, Cambridge
University Press, Cambridge, England.
Buttkus, B., 1975, Homomorphic filtering - theory and practice,
Geophysical Prospecting, v. 23, p. 712-748.
Clay, C.S., and Medwin, H., 1977, Acoustical Oceanography, Wiley
Interscience Publications, New York, p. 544.
Clayton, R.W. and Wiggins, R.A., 1976, Source shape estimation and
deconvolution of teleseismic bodywaves, Geophys. J. R. astr. Soc. v. 47,
p. 151-177.
Coulson, C.A., 1941, Waves, Oliver and Boyd Ltd, Edinburgh, Scotland.
Davies, T.J., Raikes, S.A. and White R.E., 1984, Analysis of some
wavelet estimation trials using marine sources, Presented at the 46th
annual meeting of the European Association of Exploration Geophysicists,
London, England, June 1984.
Dragoset, W.H., 1984, A comprehensive method for evaluating the design
of air guns and air gun arrays, Geophysics : The Leading Edge of
Exploration, v. 3, No. 10, p. 52-61.
Giles, B.F., and Johnston, R.C., 1973, System approach to air-gun array
design, Geophysical Prospecting, v. 21, p. 77-101.
Hargreaves, N.D., 1984, Far-field signatures by wavefield extrapolation,
Presented at the 46th annual meeting of the European Association of
Exploration Geophysicists, London, England, June 1984.
Hatton, L., Worthington, M.H. and Makin, J., 1986, The theory and
practice of seismic data processing, Blackwell's, Oxford, England.
Johnston, R.C., 1978, The performance of marine airgun arrays of various
lengths and sizes, 48th S.E.G. meeting, San Francisco, California.
Johnston, R.C., 1980, Performance of 2000 and 6000 psi air guns: theory
and experiment, Geophysical Prospecting, v. 28, p. 700-715.
Johnston, R.C., 1982, Development of more efficient airgun arrays
theory and experiment, Geophysical Prospecting, v. 30, p. 752-773.

107
ID8 REFERENCES

Jovanovich, D.B., Sumner, D., and Akins-Easterlin, S.H., 1983, Ghosting

and marine signature deconvolution: a prerequisite for detailed seismic
interpretation, Geophysics, v. 48, p. 1468-1485.
Jurkevics, A., and Wiggins, R.A., 1984, A critique of seismic
deconvolution methods, Geophysics, v. 49, p. 2109-2116.
Larner, K.L., Hale, D., Misener Zinkham, S. and Hewlett, C.J.M., 1982,
Desired seismic characteristics of an airgun source, Geophysics, v. 47,
p. 1273-1284.
Lines, L.R. and Ulrych, T.J., 1977, The old and the new ain seismic
deconvolution and wavelet estimation, Geophysical Prospecting, v. 25, p.
512-540.
Loveridge, M.M., 1985, Marine Seismic Source Signatures, Directivity and
the Ghost, D. Phil. thesis, University of Oxford.
Loveridge, M.M., Parkes, G.E., and Hatton, L., 1984a, A study of the
marine ghost, Presented at the 46th Annual Meeting of the European
Association of Exploration Geophysicists, London, England, June 1984.
Loveridge, M.M, Parkes, G.E., Hatton, L., and Worthington, M.H., 1984b,
Effects of marine source array directivity on seismic data and source
signature deconvolution, First Break, v. 2, No.7, p. 16-22.
Lugg, R., 1979, Marine seismic sources, in Developments in geophysical
exploration methods, Ed. A.A. Fitch, Appl. Sci. Publ., London.
Lynn, W., and Larner, K., 1983, Effectiveness of wide marine seismic
source arrays, Presented at the 45th Annual Meeting of the European
Association of Exploration Geophysicists, Oslo, Norway, May 1983.
Newman, P., 1985, Continuous calibration of marine seismic sources,
Geophysical Prospecting, v. 33, p. 224-232.
Nooteboom, J.J., 1978, Signature and amplitude of linear airgun arrays,
Geophysical Prospecting, v. 26, p. 194-201.
Nyborg, W.L., Scott, A.F., and Ayres, F.D., 1972, Tensile strength and
surface tension of liquids, in American Institute of Physics Handbook,
3rd. edition, Ed. Gray, D.E., Mcgraw-Hill, New York, p. 2312.
Parkes, G.E., Hatton, L., and Haugland, T., 1984a, Marine source array
directivity : a new wide airgun array system, First Break, v. 2, No.7,
p. 9-15.
Parkes, G.E., Ziolkowski, A., Hatton, L., and Haugland, T., 1984b, The
signature of an air gun array : computation from near-field measurements
including interactions - practical considerations, Geophysics, v. 49, p.
105-111.
 REFERENCES 109

Peacock, K.L. and Treitel, S., 1969, Predictive deconvolution: Theory

and practice, Geophysics, v. 34, p. 155-.
Newman, P., Haskey, P., Small, J.O., and Waites, J.D., 1977, Theory and
application of water gun arrays in marine seismic exploration, Presented
at the 47nd Annual International Meeting of the Society of Exploration
Geophysicists, Calgary, Canada, October 1977.
Peacock, J.H., Peardon, L.G., Lerwill, W.E., and Wisotsky, S., 1982, An
evaluation of a wide-band marine vibratory system, Presented at the 52nd
Annual International Meeting of the Society of Exploration
Geophysicists, Dallas, USA, October 1982.
Robinson, E. and Treitel, S., 1980, Geophysical Signal Analysis,
Prentice-Hall, New Jersey, U.S.A.
Safar, M.H., 1976, The radiation of acoustic waves from an airgun,
Geophysical Prospecting, v. 24, p. 756-772.
Safar, M.H., 1980, An efficient method of operating the air-gun,
Geophysical Prospecting, v. 28, p. 85-94.
Safar, M.H., 1984, On the S80 and P400 water guns: a performance
comparison, First Break, v. 2, No.2, p. 20-24.
Safar, M.H., 1985a, On the calibration of the water gun pressure
signature, Geophysical Prospecting, v. 33, p. 97-109.
Safar, M.H., 1985b, Single water gun far field pressure signatures
estimated from near-field measurements, Geophysics, v. 50, No.2, p.
257 -261.
Sinclair, J.E. and Bhattacharya, G., 1980, Interaction effects in marine
seismic source arrays, Geophysical Prospecting, v. 28, p. 323-332.
Stoffa, P.L. and Ziolkowski, A.M., 1983, Seismic source decomposition,
Geophysics, v. 48, p. 1-11.
Taner, M.T., 1980, Wavelet processing through all stages of seismic data
processing, 42nd E.A.E.G. meeting, Istanbul, Turkey.
Tree, E.L., Lugg, R.D., and Brummitt, J.G., 1982, Why water guns ?,
Presented at the 52nd Annual International Meeting of the Society of
Exploration Geophysicists, Dallas, USA, October 1982.
Vaage, S., Haugland, K., and Utheim, T., 1983, Signatures from single
airguns, Geophysical Prospecting, v. 31, p. 87-97.
Vaage, S., Ursin, B., and Haugland K., 1984, Interaction between
airguns, Geophysical Prospecting, v. 32, P. 676-689.
110 REFERENCES

Weston, D.E., 1960, Underwater explosions as acoustic sources, Proc.

Phys. Soc., v. 76, p. 233-249.
Wiggins, R.A., 1978, Minimum entropy deconvolution, Geoexpl., v. 16,
p.21-.
Ziolkowski, A., 1970, A method for calculating the output waveform from
an air gun, Geophys. J. R. astr. Soc., v. 21, p. 137-161.
Ziolkowski, A., 1971, Design of a marine seismic reflection profiling
system using air guns as a sound source, Geophys. J. R. astr. Soc., v.
23, p. 499-530.
Ziolkowski, A.M., Lerwill, W.E., March, D.W., and Peardon, L.G., 1980,
Wavelet deconvolution using a source scaling law, Geophysical
Prospecting, v. 28, p. 872-901.
Ziolkowski, A.M., 1980, Source array scaling for wavelet deconvolution,
Geophysical Prospecting, v. 28, p. 902-918.
Ziolkowski, A., 1982, An airgun model which includes heat transfer and
bubble interactions, 52nd Annual S.E.G. meeting, Dallas, Texas.
Ziolkowski, A., Parkes G.E., Hatton L. and Haugland T., 1982, The
signature of an air gun array: Computation from near-field measurements
including interactions, Geophysics, v. 47, p. 1413-1421.
Ziolkowski, A., and Metselaar, G., 1984, The pressure wavefield of an
airgun array, Presented at the 46th Annual Meeting of the European
Association of Exploration Geophysicists, London, England, June 1984.
Ziolkowski, A., 1984a, The Delft airgun experiment, First Break, v. 2,
p. 9-18.
Ziolkowski, A., 1984b, Deconvolution, IHRDC, Boston.
INDEX
Absorption 87 Energy 7, 29
Acoustic wave generation 3 - conservation of 8
Acoustic wave propagation 4 - - law of 28
Airgun array 85 Energy conservation paradox 60
Airgun(s) 3,47,50,58,60,100 Equation of motion 5
Arrays 25 Exponential taper 93
- long 33
- tuned 25 'Far field' 15
- wide 34 'Far field' hydrophone 77
Average reflection coefficient 74 'Far field' measurements 16
'Far field' signature 10, 55
Bandwidth 45 Flotation 68
- effective 85 Forced minimum phase 93
Bar 19 Fringe patterns 27
Bar-metres 20
Beam forming 36 Ghost 10
Beam-steering 25, 31 Ghost notches 12, 13, 74
Bubble motion 8
Bubble period 69, 87 Interaction 45, 58, 61
- linear 45, 50
Chirp 59 - non-linear 49
Coalescence 45 Interaction distance 51
Convolutional model 21,58,81 Interference 12, 28
Isentropic 5
Determinism 45, 62
Deterministic deconvolution 83 Law of conservation of energy,
Deconvolution see Energy
- directionally dependent 41 LaWiOf superposition of wavefields,
Directivity see Superposition
- effects of 39 Linear interaction,
Directivity functions 29 see Interaction
Dominant horizon 94 Linear relative velocity model 18
Dominant reflection deconvolu- Linear superposition,
tion 94 see Superposition
Lo'iig-arrays,
Effective bandwidth, see Arrays
see Bandwidth
Efrects of directivity Mainlobe 38
see Directivity Marine ghost variability 74
Elasticity Marine vibroseis 59
- bulk modulus of 6 Mass equation
Elastic \'iaves 4 - conservation of 4

113
114 INDEX

Measurement of signatures 16 Spherical wave equation 6

Minimum entropy deconvolution 94 Stability 68
Mis-fires 71 Statistical deconvolution 93
Multiple images 36 Superposition
- law of 60
Navier-Stokes equation of motion - - of wavefields 17
5 - linear 53, 62
Near field 15 System stability 70
- measurements 18, 53
Noise suppression 36 Temporal extension 62
Non-linear interaction, Time synchronisation 67
see Interaction Timing errors 67
Norwegian buoys 68 Total energy emission 8
Notional source(s) 17,53,54 Tuned arrays,
Numerical modelling 58 see Arrays
Para vane 68 Velocity 7
Paravane technology 55 Virtual image 10
Particle velocity 15
'Peak to peak amplitude' 19 Watergun(s) 3, 57, 67. 88, 103
Period of oscillation 9 Wave height 73
Phase delay 27 Waveshape kit 50
Point sources 16 Weather 55,67,72,74
Predictive deconvolution 93 - effect 74
Pressure 7 Wide arrays,
Primary to bubble ratio(s) 19, see Arrays
20, 50, 87, 93 Wioe-deployment 69
Pulse duration 20
Pulse repeatability 20, 21
Pulse shaping 25
Reflection coefficient 73
Relative motion 17
Rough surface 73
Scal ing law 56
Sea wave distribution 72
Seismic source deconvolution 81
Shot effect 75
Shot to shot variations 71, 77
Sidelobes 28, 31, 72
Signature deconvolution 81,93
Significant height 73
Sideswipe 37, 38
Sound velocities 7
Source generated noise 38
Sparker 4
Spatial extension 62
Speed of sound 6
Spherical-cap bubble 51
Spherical spreading 40