٢٠١٣/
٢/
٣ Introduction to Array Design
Introduction to Array Design
In the early days of seismic exploration, the Common Mid Point (CMP) method did not exist, nor did
many data processing techniques. As a result, the geophysicist was obliged to interpret directly from
the field records. This meant that the field records needed to be as noise-free as possible, and
arrays were an important method to achieve this. Later, emphasis shifted to ensuring that the
acquired data were properly conditioned for the CMP stack and other processes. Typically, arrays
became simpler. Now the source and receiver spacing and arrays are chosen in conjunction with the
processing steps to be applied to the data. In modern acquisition, arrays are always selected taking
account of the effect of stacking and migrating the data.
The Several Functions of a Geophone Group
The functions of a group of geophones at each geophone station are to
Average variations in the coupling of the geophones to the ground,
Suppress ambient noise
Reduce the number of channels to be recorded
Of these functions the last is the most important and , as geophysicists, we look forward to the day
that instrumentation has developed sufficiently that individual geophones can be economically
recorded and geophone groups are no longer required. If the geophones are distributed over a
distance calculated to attenuate surface waves, the group is called an array and then performs the
third function of suppressing source-generated noise (in particular, surface waves).
Averaging the Geophone Coupling
A long-spiked geophone, pushed into firm soil has a good coupling to the earth (a good plant): the
geophone case effectively moves with the earth. However, ground conditions such as gravel, rock or
matted vegetation may inhibit the planting of the geophone. This causes variations in response for
different geophones. Using a number of geophones in a group provides some degree of averaging
among types of surface material, as well as some protection against unavoidably poor plants. It
cannot overcome the harmful effect of really bad plants, so attention must be paid to the way the
geophones are planted. This means that the vegetation should be cleared from around the geophone,
the wires should not be taut, the spike length should be appropriate for the ground conditions and, in
windy conditions the geophone should be buried. Ten or twelve geophones at each station are
typically used to ensure that each recorded channel has a similar response to all the others.
However, each individual geophone is part of the total sampling of the wavefield. Any deviation from
perfect coupling leads to the introduction of noise in that sampled wavefield.
Suppressing the Ambient Noise
In the discussion on Ambient Noise Reduction in Introduction to Field Work, it was shown that a
group of geophones, spaced apart, provides an improvement of the reflection signal relative to
random noise. It was assumed that the reflection signal, coming up from below, is common to all the
geophones of the group and the noise is different at all the geophones of the group.
If the noise is indeed different for N geophones (uncorrelated) from geophone to geophone, then the
improvement of the signal-to-noise ratio is . For this to be true, the spacing between the
geophones must exceed the correlation distance of the noise. For noise due to the blowing of leaves
or sand against the geophone case, the correlation distance of the noise is small; the geophones
may be placed as close as possible without touching. For noise due to the motion of a streamer
through the water, the correlation distance may be less than a meter. For noise due to distant
storms, the correlation distance may be tens or hundreds of meters. The correlation distance is likely
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 1/12
٢٠١٣/
٢/
٣ Introduction to Array Design
to be larger at low frequency than at high frequency.
The implications of this for the layout of geophone groups are very simple:
There is no point in placing geophones closer than the correlation distance of the
dominant type (and dominant frequency) of noise; to do so would waste time and
money.
To maximize the signal-to-noise ratio, the number N of geophones should be large.
If the number N is large, and if the geophones have a minimum useful spacing, it follows that the
group dimensions must be large, and that the best arrangement is areal, perhaps like Figure 1 (Areal
geophone layout with equal distances between adjacent geophones).
Figure 1
For suppression of ambient noise, then, the rule for the layout of a group is simple - The
Bigger, The Better, provided that the signal still arrives at each geophone in the group at the
same time. This means that there cannot be large elevation differences across the array nor NMO
nor weathering statics.
Suppressing the Source-Generated Noise
Any seismic source, and especially a surface source, generates a pattern of surface waves, or
ground roll. These waves spread out circularly on the surface, typically at a velocity in the range
180-750 m/s (600-2500 ft/s) but sometimes as high as 1500 m/s. Further, they are often
concentrated at the lower seismic frequencies, typically in the range 5-25 Hz. In some areas these
surface waves are very severe; where the near-surface is composed of young unconsolidated silts
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 2/12
٢٠١٣/
٢/
٣ Introduction to Array Design
and clays, saturated with water, the amplitude of the surface waves may be many times larger than
the reflected signal arriving at a geophone at the same time. In such areas, the dominant
consideration in selecting the group dimensions may cease to be the suppression of ambient noise,
and may become the suppression of surface waves.
A problem that may produce a similar effect is the generation of unwanted shear waves at or near
the source, and the propagation of these waves in the near surface. Depending on the mechanism of
propagation, shear waves travel along the spread at velocities typically in the range 600-1000 m/s
(2000-3300 ft/s). The shear frequencies are generally higher than those of surface waves. Again,
the amplitudes may be many times greater than the reflected signal arriving at the same time.
The broad term 'source-generated noise' will be used to embrace both surface waves and unwanted
shear waves (and, indeed, all other problems derived from the shot).
The present problem is the suppression of source-generated noise, of whatever type, that
propagates outwards from the source. As shown in Figure 2 (Principal of suppression of source-
generated noise by an array),
Figure 2
if we dispose a number of geophones (ten, in this illustration) over one wavelength of the surface
wave, the outputs of the geophones in one half-cycle cancel the outputs of those in the other.
In this context, the objective is to select a group dimension that suppresses the source-generated
noise. This is a more specific task than the suppression of ambient noise. For ambient noise, the
group is made as large as possible; for source-generated noise with well-defined frequency and
velocity, and thus a well-defined wavelength, we wish to sharply attenuate that wavelength.
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 3/12
٢٠١٣/
٢/
٣ Introduction to Array Design
When the group dimensions are selected based on a specific criterion like this, the group is defined to
be an array, and the array is directional (Figure 3: Concept of a directional array).
Figure 3
It should have maximum response to reflected signals (of prescribed frequency range and direction)
coming up from below, but to have minimum response to source-generated noise (of prescribed
frequency range and velocity) arriving along the surface.
From Figure 2 (Principal of suppression of source-generated noise by an array), the effective length
of the array, which corresponds to the wavelength for which there is zero response, is longer than
the actual length (the walking length); for ten geophones, the walking length is nine times the
geophone spacing, but the effective length is ten times the spacing.
In general, the effective length is N times the geophone spacing, where N is the number of
geophones.
Where source-generated noise is a problem, and with surface sources in particular, source arrays
may be used as well as geophone arrays. The principle is, of course, the same; the ten geophones of
this illustration could just as well be ten sources or be ten successive positions of one source, with
the corresponding records vertically stacked. Later, arrays will be defined as being formed by
elements, where the elements can be either geophones or sources. As far as source-generated noise
is concerned, the benefit obtainable from SOURCE arrays is of the same kind and the same degree
as that obtainable from GEOPHONE arrays. The two benefits can be compounded, by the use of
both SOURCE and GEOPHONE arrays.
The same is true of the benefit obtained by averaging the ground conditions.
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 4/12
٢٠١٣/
٢/
٣ Introduction to Array Design
A geophone array averages the geophone plants.
A source array averages the coupling of the source to the ground.
(The same is not true of the benefit obtained by suppressing the ambient noise; the geophones must
be spaced apart, the sources need not.)
Suppressing Scattered Noise
Scattered noise (see the discussion on Scattered Noise Reduction in INTRODUCTION TO FIELD
METHODS) is a special type of source-generated noise in which the noise can arrive at the geophone
group from any azimuth.
In the marine environment, energy from the source may be scattered back to the receiver by sand
dunes (the presence of long thin "banks" on the hydro-graphic charts is a warning to expect them.)
The scatterers are frequently referred to as secondary sources. Where the sea floor contains
outcrops of older rocks, these are certain to act as secondary sources. Other such sources include
sea-floor channels and glacial kettles, inundated reefs, and faults in (or approaching) the sea floor.
In the North Sea, glacial till is a common source of scattered noise. In the Gulf of Mexico, salt domes
may penetrate the sea floor or come close to it, slump faults may extend to the sea floor and gas
may seep upwards. Islands and shorelines are other obvious examples. In working through such
areas, we are often exposed to many of these secondary sources simultaneously, so that the
situation is that of Figure 4 (Profusion of secondary sources, each scattering the source-generated
noise back to the streamer; after Larner, et al, 1983).
Figure 4
Usually, the secondary sources responsible for most scattered noise lie within 200 mS of the sea
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 5/12
٢٠١٣/
٢/
٣ Introduction to Array Design
floor. The mode of propagation will likely be a trapped wave traveling between the sea surface and
the first hard interface at about 1600 m/S or, in shallow water, a slower Scholty wave, which is the
marine equivalent of ground roll.
In the land environment, there are many possible near surface scatterers, such as erosional features,
volcanic deposits and Karst limestone. The problem is most severe when there are strong surface
waves such as ground roll or refracted shear waves. Secondary sources are a major problem,
because stacking can enhance the backscattered arrivals.
The reason for this is evident from Figure 5 (Paths from source to sea-floor scatterer and spread, as
the ship moves past the scatterer).
Figure 5
In Frame (a) of this graphic, we see, in plan view, the seismic line (dashed) and a secondary source
at O, off the line. For illustration, we assume that the scattering body is 1200 m off the line, and the
far-group offset is 2400 m. We consider the gather formed at midpoint M1. For present purposes we
ignore the near-group offset, so that the near trace of the gather is represented by path AOA; this is
recorded when the ship has just passed source position A. The far trace of the gather is represented
by BOC; this is recorded when the ship has just passed the scatterer.
In Frame (b) of this graphic, we consider the gather formed at midpoint M2. The near trace
represents the path DOD; the far trace the path EOF. In Frame (c), we consider the gather at the
point M3, nearest to the secondary source; now the gather represents symmetrical ray paths-just
like the gather we get from a reflector of zero dip. In Frames d and e, the gathers for midpoints M4
and MS represent mirror images of M2 and M1. We have a symmetrical suite of gathers. In Figure 6
(The gathered events corresponding to the scattered paths seen in the previous graphic), we see
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 6/12
٢٠١٣/
٢/
٣ Introduction to Array Design
the alignments of the backscattered event on these gathers.
Figure 6
Starting with Frame (c) of this graphic, we see the gathered event as an alignment that looks in
every way like a reflection from the sea floor in a water depth of 1200 m. Its normal move out of 0.7
s is that corresponding to the low velocity of water, and is significantly greater than we would
expect for a reflection at a time of 1.6 s from within the earth. Therefore, in later stacking, when we
apply a normal-moveout correction appropriate to reflections from within the earth, this scattered
event at midpoint M3 will not be properly corrected, and will be at least partially suppressed.
Next we consider the gathered event shown in Frame (d), corresponding to midpoint M4; there is
significantly less normal moveout (0.5 s) at only slightly greater time (1.8 s). The gather looks more
like that of a reflection within the earth. And when we go to Frame (e), the distinction weakens still
further; a normal moveout correction appropriate to a reflection at this time may well be appropriate
for this gather, and thus cause it to stack well.
The total effect to be expected is that of Figure 7 (The effect of the scatterer shown in graphic 5,
as observed on the stacked section), which represents the stacked section from this portion of line.
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 7/12
٢٠١٣/
٢/
٣ Introduction to Array Design
Figure 7
A bold hyperbolic pattern cuts across the genuine reflections, with its apex at a time proportional to
the distance of the scatterer from the line. The pattern is weak and broken near the apex, where it
is substantially suppressed by the stack. On its flanks, however, the pattern is very strong, having
been effectively enhanced by the stack. Figure 8 (Section illustrating the effect of sea-floor
scatterers,
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 8/12
٢٠١٣/
٢/
٣ Introduction to Array Design
Figure 8
where differences in the wave trains are caused by the shapes of the scatterers and by their
varying distances from the line; from Newman, 1984, reprinted by permission of the EAGE and
Blackwell Scientific Publications, Ltd.) shows a section from an area plagued by a profusion of such
scatterers. Clearly, the effect is serious; the continuity of the reflections is impaired, and doubt is
introduced into the picking. The effect can be quantified by reference to Figure 9 (Scatterer
velocity) which shows a scatterer located near the surface.
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 9/12
٢٠١٣/
٢/
٣ Introduction to Array Design
Figure 9
A seismic line is being acquired in the dip direction along the front face of the cube and a CMP gather
is indicated as a thick line. A reflector, of dip , is also shown. If the rays to the reflector are rotated
through 900 to the horizontal plane, they will overlay the rays to the scatterer. Thus the ray
geometry for the CMP to the scatterer is the same as that for the reflector and Normal Move Out
can be applied to the scatterer in just the same manner as to the reflector. When NMO is applied,
the same correction for the dipping reflection will also be appropriate for the scatterer, if the
velocities are identical. Stacking will then enhance the scattered energy as well as the reflected
energy.
The velocity to be used for NMO correction for a dipping reflector is faster than that for a flat
reflector by the factor . In practice, the dip on a reflector is likely to be small and will not be
the same as the angle to the scatterer, nor will the velocities be the same. However, at some time in
the section, it is likely that
will equal
Where
= the reflector
= the scatterer,
= velocity,
= dip.
Under these conditions the scatterer will be stacked as well as the reflector. If the reflector is
assumed to be flat, a common assumption for NMO application, then the scatterer will be stacked
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 10/12
٢٠١٣/
٢/
٣ Introduction to Array Design
when
Figure 10 (Optimal scatterer locations for stacking) showed the locations of the scatterers from
which energy will be enhanced by the stacking process.
Web-Fonts not available -- using image fonts instead
Figure 10
If this energy is to be attenuated by arrays, then the ray geometry must be studied for this CMP
gather. The source-receiver pair S1-R1 can be excluded, since a mute for NMO stretch will be
employed in processing to eliminate these from the stack. The pair S2-R2 will require a wide array at
the source to attenuate energy traveling to the scatterers at A. However, a long array will be
required to attenuate energy traveling to scatterers at B. Similar considerations apply at the
receivers. The angles to the pair S3-R3 are different again. Thus, arrays are required to give
attenuation for all azimuths. This is impractical in many circumstances.
Two strategies have been applied in the marine environment. The first (Newman, 1977) was to use a
long array of waterguns as a source to eliminate the inline noise directly. This assumed that the
majority of the noise energy would be traveling in the inline direction. It had the disadvantage that it
caused some loss of high frequency information. The second approach (Tree, 1982) used a wide
array of waterguns as the source. Waterguns were used because these provided identical elements
for the array, whereas airguns must use different size guns to provide attenuation of the bubble. GI
guns can also be used now. The array provided the attenuation needed in the crossline direction and
f-k filtering was used to attenuate the noise in the inline direction.
On land, the most successful method has been to deploy several parallel seismic lines. These were
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 11/12
٢٠١٣/
٢/
٣ Introduction to Array Design
then combined to achieve the array dimensions needed. This is clearly very expensive but sometimes
is the only way to achieve satisfactory results in very bad data areas. (Regone, 1990).
There is one more form of source-generated noise to be considered. This arises in land-locked seas
and inlets with rocky shorelines (such as the Lower Cook Inlet in Alaska). The surface wave and the
guided wave may be reflected and scattered back from the rocky shoreline very strongly, in a way
that would not happen on a shelving beach. After crossing the inlet, it is similarly returned from the
opposite shore. Repeated pops every 9 or 10 seconds build up a virulent noise field, through which
the streamer is towed.
The effect can be clearly seen on the record from a misfire, which shows a complex herringbone
pattern of noise events with near-linear moveout. Long hydrophone arrays help, of course, but the
best solution is clearly an omni-directional source array.
[Link]/data/gp21/[Link]?UserID=34981&Code=6107 12/12