Académique Documents
Professionnel Documents
Culture Documents
To reduce the unwanted “ringing” that results from this, these airguns (the “seismic
source array”) each release different amounts of air, creating different sized bubbles, each
ringing with a different periodicity, creating a “tuned” seismic signal:
Depending on its size, and a few other factors, including towing depth and compressed
air pressure, each airgun’s bubble creates a different amount of pressure in the water:
The pressure and energy created any marine seismic source dies off rapidly with distance,
according to simple mathematical rules. Two principal physical facts, however, conspire
to complicate the result. Taking EWING’s largest seismic source array as an example, we
can quickly show how the operating environment and operational requirements affect the
acoustic “footprints” created by airgun arrays. If all EWING’s 20 airguns could be fired in
exactly the same place, and in a world where the air and water had the same acoustic
properties [impedances] this would be the result:
It’s impossible to have all twenty airguns in exactly the same place, however, and they
don’t work right if they are fired too close to each other, so the EWING’s array is spread
out, to cover a horizontal area of 5 x 36 meters. This simple fact changes the way
pressure is created in the surrounding ocean:
Directly beneath (and above), and not too close to the array, distances between a point of
observation and each airgun are not very much different. Therefore, the pulses arrive at
nearly the same time, and add together as shown in the “tuning” picture, above. Off to
the side, however, the distances differ enough that each pulse arrives at a slightly
different time. Individual peaks don’t line up, and the maximum sound level actually
produced is much lower than it is the same distance directly below the array.
The second physical fact affecting sound production is a very powerful and inescapable
one – the sea surface, or air-water interface. Since sound propagation in air and water are
so different, when sound waves in the water bounce off the sea surface, the reflections are
negative images of the original, and almost as loud.
This difference of 18 dB between the peak-to-peak and RMS dB levels for the same
signal are typical for impulsive airgun signals [W.J. Richardson, pers. com., 2002.]
Despite this apparent underestimation, published studies using this RMS determination
still showed that whale avoidance behavior only occurred with airgun sounds at higher
dB levels than for other, more constant noise types such as rig and shipping noise.
Therefore, it is correct to use this measure of sound pressure levels [SPL] for calculating
the safety radiuses to be observed for the various EWING seismic source arrays. Over the
past few years, most IHA mitigation programs approved by NMFS have used an SPL of
180 dB RMS to define a safety radius. Any cetacean observed within this radius is
defined to have been “taken” under the MMPA, and the IHA strictly limits the number of
such “takes” that will be allowed during a particular activity. It is important to note that
180 dB “takes” do not involve any physical harm to cetaceans [and it is generally agreed
that pinniped and fish species have even higher permissible levels.] Physical harm is not
likely for SPLs below 220 dB, though it is generally agreed that exposures of 180 dB or
above are to be avoided.
Airgun arrays are rated according to several measures, primarily the peak-to-peak
pressure levels, observed far from the array itself, but amplitude-corrected or back
projected back to a standard reference point, one meter from the array’s center. Array
output, measured thus in the “far field” and at a point within this primary downward
The very high pressure gradients above the array are due to the free surface or “Lloyd’s
mirror” effect discussed previously. Directly below the array, its downward directivity
begins to show in the form of largest separation of contour lines.
Contouring the overall pressure field in terms of RMS signal power produces a now-
familiar pattern:
Below, we chart the maximum horizontal extent of the isopleths, and for comparison
include the levels of some more familiar sounds that have been carefully measured and
presented in published literature.
When assessing seismic source arrays, either for their ability to penetrate the earth, or to
determine their impact on marine life, peak pressures are poor measure by which to
compare. It is far better to see how energy is produced as a function of frequency. Since
lowest frequency sound does the best job of returning echoes from deep within the earth,
marine airgun arrays, which put most of their power into low frequencies, have been
chosen to form EWING’s seismic source. Here is a modified view of the EWING test
signature presented at the beginning of this paper:
Here, the surface ghost has been [nearly] removed. On the left is the near field signal, and
on the right, its spectrum. Maximum energy is found at around 12 cycles per second [Hz.]
This frequency happens to correspond to the period of the bubble pulse, but the overall
trend of highest energy at low frequencies, and a steady falloff with increasing frequency
is quite clear. As we described above, considerable effort is made in “tuning” airgun
arrays to reduce the effects of the bubble pulses.
Typical airgun array signature from “Airgun Arrays and Marine Mammals”, IAGC, 9/2002.
In the spectral domain, maximum energy is around 214dB, at 60Hz. The average level
falls off steadily thereafter. The regularly spaced notches and repeated arch shapes
correlate to a towing depth of 6 meters. Increased falloff at frequencies above 875 Hz is
due to analog filtering. The authors have marked the overall spectral trend for us. This
trend is defined by two values: 210 dB at zero Hz; 152 dB at 1000 Hz. Each is much
lower than the corresponding signature’s peak-to-peak 264 dB.
The connection between the 264 dB peak-to-peak measurement and the much lower
levels seen here can be made by integrating the spectrum – calculating the area
underneath the curve. The higher the frequencies included, the greater the integration
sum, and the greater the peaks, which contain virtually all of the small amount of high
frequency energy in an Airgun signal.
As we show next, most of the energy produced by a seismic airgun array is found in its
signal’s low frequencies. The signal’s general “shape” at higher frequencies is very
similar for all 2000 psi airgun arrays, with fine scale variations introduced by differences
in tuning and towing depth. Since virtually all available seismic signal recordings include
the low frequency components and vary only in the extent that higher frequencies are
Peak intensity for this array in the frequency domain is approximately 221 dB, around 20
Hz, compared with the approx. 214 dB peak for the industry array example given earlier.
Although the EWING spectrum is only accurate up to 250 Hz or so, the rate of falloff can
be seen to be nearly identical, and we expect its relative amplitudes at high frequency to
be about the same.
The right hand plot above measures the EWING spectrum in a standard way – again,
integrating the spectrum from low to high frequency, but now showing the cumulative
result in terms of energy flux. The curve climbs steeply, but then flattens out markedly at
about 80 Hz. It is clear that most of the energy created by the array travels at subaudible
frequencies.
The predominance of low frequencies in all seismic airgun array signals and the falloff of
energy at higher frequencies can be quantified by examining additional examples.
Returning to the particularly clear and finely-sampled IAGC airgun array spectrum
presented earlier, we can begin to quantify the trends in spectral falloff. There is clearly
a steep initial falloff of energy across frequencies below 400 Hz. Thereafter, the falloff
In this example Goold & Fish (1998) have presented broadband spectra of a different
industry airgun array, measured at distances between 750 and 8000 meters (four traces at
the bottom). Due to the limitations of the hardware used to make the measurements, the
spectra are not usable at low [less than 400 – 500 Hz] frequencies and the signal levels of
the 8 km example fell below the usable threshold of their system above 10kHz. At the
top, I have corrected the spectra for spreading loss and superimposed the results. Spectral
falloff, as observed in the previous graph, continues though less steeply with increasing
frequency, and can be divided into three zones; 800 Hz – 2 kHz, 2 – 6.5 kHz, and all
frequencies about 6.5 kHz. Given a good, low-frequency spectrum of a particular seismic
airgun array, these results allow us to predict its output over the remaining frequency
range, up to 20kHz or higher.
Of great importance is the fact that different mammalian species communicate at, and
are sensitive to, quite different ranges of frequency. True assessment of the likelihood of
behavioral disturbance by seismic sources can only be made by comparing the
characteristics of source spectra with the acoustic sensitivity of particular species. A very
useful spectral-domain summary of various sound levels is presented by Ketten (1998):
Here acoustic levels produced by a variety of man-made systems are compared to the
The seismic airgun signal is “broadband” in that its energy is spread over a broad range
of frequencies, corresponding to a very short, impulsive time signal. This is especially
apparent when comparing the seismic spectrum to that of the kind of sonar used for
determining ocean depths and for detecting and tracking submarines. These sonars are
designed to emit a signal which is nearly monochromatic – the energy is confined to a
narrow range of frequencies, and usually a range which is quite different from that
present in the seismic source spectrum.
Despite the limitations imposed by the sampling interval and filtering in the example
above, the airgun array spectrum has energy at frequencies above 1000 Hz, although the
energy continues to diminish as frequency increases. The exact peak-to-peak values
measured, therefore, depend a lot on the range of frequencies included. The few
The Navy’s report on the Bahamas incident indicates that the sonars involved were
operating at 3.3 kHz, issuing two kinds of “pings” – constant frequency and “chirp”,
sweeping between 3250 and 3350 Hz. Both kinds of pings were issued in sets of three,
over a two-second period. Individual “pings” were as long as half a second, and for
several (more than three) hours, one of the sonars was operating at levels above the
nominal 235 dB [RMS] rating. Anecdotal information from ONR indicates that this level
may have been 240 dB, RMS. A 3.3 kHz, 240 dB RMS half-second “ping” contains
significantly more total energy than EWING’s 20 airgun seismic array, and the rise time,
described in the Navy report as “one or two cycles” is quite rapid – half a millisecond or
less. In contrast, the rise times for airgun signals are typically 2 – 4 milliseconds –
significantly less impulsive, and thus less likely to cause injury. For comparison, I have
superimposed, on a single plot and at the same scale, three curves; the IAGC (1002)
published example reproduced in detail earlier, The combined, decreasing falloff curves
for the IAGC spectrum and that matching the observations of Good & Fish (1998) and
lastly, the spectrum of a synthetic sonar “ping” exactly matching the description given in
the Navy’s report on the Bahamas strandings (2002). This “ping” is 0.5 seconds long,
rises to full 235 dB RMS amplitude in two cycles, and sweeps a short band of frequencies
centered on 3.3kHz. The gradual falloff plotted for this spectrum below 170 dB is largely
due to limits in the numerical resolution of the spectral calculation.
It is clear that at frequencies above 1.5 kHz, the industry airgun array signal, even back-
corrected to a nominal 1 meter distance, has fallen to beow 160 dB, and at 3.3kHz, it is
60dB below the sonar operating at the officially stated 235 dB RMS level. For mammals
who are sensitive to kHz-and-higher frequencies, airgun arrays, even large ones used by
exploration industry 3D seismic vessels, and research ships such as R/V EWING, are
highly unlikely to cause injury or distress.
The changes in amplitude due to cancellation by free surface ghosting were briefly
addressed at the end of the first section, and it is useful to examine here the associated
changes in the frequency domain. For practical purposes, the ocean surface is a perfect
reflector for upcoming sound waves in the water. Since the speed of sound is far less in
air than it is in water, a negative, or reversed phase reflection is generated. If the
upcoming wave has a peak strength of +1, the downgoing reflected wave has a strength
of –1, and at a receiver somewhere beneath the sound source, the time delay between
primary and reflected pulses will depend on distance of the source beneath the sea
surface. The resulting effect is a “plus one – minus one” time domain operator which is
convolved with the source’s near field signature to form the far field signature. The
reason this relationship is so useful is that both the operator and the near field signature
have characteristic frequency spectrums, which can be multiplied together to produce the
same result as convolution in the time domain.
While changing the towing depth of a tuned seismic source will not change its near field
signature greatly, the change in the ghosting operator can be very significant, though at
the same time very predictable, as indicated by these simple relationships.
This shows why marine geophysicists seeking high temporal resolution (requiring a
broadband source) tow sources and receivers more shallowly than researchers who are
trying to image deep structure (best done with low frequencies). The most important
observation, however, is that the shorter the time between primary and ghost signals, the
less low frequency energy is preserved.
As the horizontal offset between a source and a receiver increases, the interval time
between direct and surface-reflected arrivals decreases, with no change in tow depth. The
resulting “ghosting” operator is illustrated above for a 60-meter deep receiver [or
mammal] moving away from a 6-meter deep source. As offset increases, low frequency
For the first time, we consider the interaction of seismic waves with the seafloor:
Depending on water depth and seafloor velocity, postcritical reflection will occur for all
waves first reflecting beyond some “critical” distance (XC in the figures above) from the
source. When the seafloor is horizontal, the multiple arrivals will be received in order,
from 1 to N, and it is the nature of the hyperbolic geometry involved that the time
between arrivals will increase as the arrivals come in:
Although the model for these arrivals is simple, the arrivals are not. As just described, the
single-bounce N = 1 arrival is the earliest, and time between each sunsequent arrival
increases. Arrivals from 1 to 14 are postcritical, and much stronger than the later arrivals,
which have lost most of their energy into the seafloor. Although N = 1 has the shortest
path, it is most weakened by the interference between “direct” and surface-reflected
arrivals described in the first section of this paper. The same interference also causes
subsequent variations, seen as “beating” in the signature. This seismogram was created
with a point source – array directionality, if included, would diminish “low N” arrivals
even further.
In the realm of marine mammal mitigation, two very important features are illustrated in
this example. First, a simple, impulsive signal has given rise to a reverberating, longer
signal. As a result, its RMS measure will be close to the peak value, not reduced by
~16dB as in the simple deep water models seen earlier. Second, there arises the only
potential seen so far for the creation of high frequency energy, through the shortened