You are on page 1of 7

MARGO 2709

Marine Geology 164 (2000) 29–35


www.elsevier.nl/locate/margeo

A single-channel seismic reflection method for quantifying lateral


variations in BSR reflectivity
T.M. McGee*
Marine Minerals Technology Center, University of Mississippi, 220 Old Chemistry Building, University, MS 38677, USA
Received 9 July 1998; received in revised form 15 December 1998; accepted 5 February 1999

Abstract
Results of seismic inversion techniques and logs of deep-sea bore holes indicate that bottom simulating reflectors (BSRs)
which exhibit high reflection amplitudes are underlain by a thin layer of free gas. Often, however, BSRs exhibit relatively low
amplitudes and display significant lateral variability. In these cases the structure is not well understood and remains a topic of
research.
Waveform inversion has been used to investigate the distribution of propagation speeds in the vicinity of BSRs, but the
technique is not practical in some situations because it requires multi-channel data sets that include large offset distances
between sources and receivers. Such data are not available in many instances, so it has become attractive to consider other
methods of achieving the same end.
A method that is applicable to single-channel, short-offset data is discussed here. It was originally developed to help
characterize shallow submarine sediments for engineering and environmental purposes. Of course, no single-channel method
can provide information concerning speeds of propagation such as is available from multi-channel methods. In this case the
single-channel method has an advantage, however, in that it is self-calibrating. That allows it to provide, after correction for
wave-front divergence, true reflection amplitudes without considering source characteristics or referencing to a known, or
inferred, propagation parameter such as speed or density. These true amplitudes then yield reflection coefficients by correcting
for transmission losses.
Use of the method is illustrated with the help of synthetic data. It is demonstrated that the accuracy of results is improved by
using a rapid digitizing rate during data acquisition. The method is then applied to a set of real data that previously had been
analyzed by full-waveform inversion. The results are noisy, largely due to the data having been digitized at a rather slow rate
and the length of recording being too short; however, average values of reflection coefficients at the sea floor and the BSR
compare well with average values obtained by the inversion procedure. It is concluded that the single-channel method provides
reasonable values for reflection coefficients. This suggests that, with judicious constraints on density variations, single-channel
data could provide information on the structure of propagation speed in the vicinity of BSRs. Moreover, it would involve
substantially less effort than is required for full-waveform inversion. q 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Gas hydrates; Bottom simulating reflector; Reflection coefficient; Deconvolution

1. Introduction

Under proper conditions, the base of the gas hydrate


* Tel.: 11-662-915-7320; fax: 11-662-915-5625. stability zone in marine sediments is marked by a
E-mail address: tmm@mmri.olemiss.edu (T.M. McGee). bottom simulating reflector (BSR). For this reason,
0025-3227/00/$ - see front matter q 2000 Elsevier Science Ltd. All rights reserved.
PII: S0025-322 7(99)00124-3
30 T.M. McGee / Marine Geology 164 (2000) 29–35

BSRs are commonly relied upon as indicators of gas correction for wave front divergence, a measure of the
hydrates in the sea floor. The distribution of propaga- true amplitude of each reflected arrival. These true
tion speeds in the vicinity of BSRs, as revealed by amplitudes then yield reflection coefficients when
seismic reflectivity, has been investigated at several corrected for transmission losses.
locations worldwide by applying full-waveform
inversion to multi-offset reflection data. Results
from several locations indicate that BSRs character-
2. The single-channel method
ized by high reflection amplitudes are underlain by a
thin layer of free gas (Singh et al., 1993). This has If the signal-to-noise ratio is high, the frequency
been confirmed by drilling (MacKay et al., 1994; domain representation of the primary sequence,
Holbrook et al., 1996). P… f †; may be approximated as a multiplication of
Reflections from BSRs are not always high ampli- the spectral response of the system, S… f †; with that
tude, however, and they commonly display strong of the geologic layering, G… f †; i.e.:
lateral variability. The reasons for this are not well
understood. The use of waveform inversion to investi- P… f † ˆ S… f †G… f † …1†
gate BSRs is limited due to the requirement of
relatively large source/receiver offset distances. The Under calm conditions, the surface of the water is a
method also involves a significant level of effort for nearly perfect (negative) reflector which reflects the
processing and interpretation. upward traveling primary sequence back toward the
This paper discusses a method for determining sea floor to produce what is called a “water-layer
variations in BSR reflectivity that is fast and applic- multiple”. Each time the sequence of propagating
able to single-channel data. The method was origin- wavelets is reflected from the water surface, the effect
ally developed for environmental and engineering of the sea floor and subbottom layers can be repre-
applications as part of the European Communities sented in the frequency domain as an additional multi-
MAST-1 Project GISP (Theilen et al., 1993). It plication by 2G… f †; the minus sign accounting for the
requires that the source/receiver offset distance be polarity reversal that accompanies reflection from the
small compared to the water depth and that reflector underside of the water surface. Thus the spectrum of
dips not exceed a few degrees; prerequisites which are the first multiple sequence can be written:
usually satisfied in the context of BSRs. It also
requires that the data be recorded digitally until M… f † ˆ 2S… f †G… f †G… f † …2†
the onset of the second water-layer multiple; a pre-
requisite that is satisfied less often. At frequencies for which P… f † is not zero, Eq. (2)
The method produces best results if the water is may be divided by Eq. (1) to obtain the negative of the
deep enough that the so-called “primary sequence”, spectral response of the geologic layering:
i.e. the water-bottom reflection and the coda that
follows it (including shallow subsurface reflections, M… f †=P… f † ˆ 2G… f † …3†
the BSR and attendant interbed multiples) decay to
insignificance prior to the onset of the first water- which may be inversely transformed to obtain a time
layer multiple. Also, results are improved if digitiza- series containing the negative of the true amplitudes
tion is done at a rapid rate, i.e. fast enough to place at of primary reflections from the sea floor and sub-
least 10 samples within the dominant wavelength of bottom interfaces. This time series also contains
the seismic signal. This improves the coherence multiple reflections internal to the geologic layers,
between the primary sequence and the analogous so some interpretation is required to identify the
“first multiple sequence” that begins with the first primary reflections. After that is done, the reflection
water-layer multiple of the water-bottom reflection. coefficient of any layer interface may be determined
If the coherence is high enough, a deconvolution of by dividing the true amplitude of the primary
the first multiple sequence to the primary sequence reflection from that interface by the product of the
compensates for the source pulse and provides, after transmission coefficients of all interfaces above it.
T.M. McGee / Marine Geology 164 (2000) 29–35 31

Fig. 1. (a) Three portions of the two-layer synthetic reflection seismogram described in the text. (b)–(e) Convolutions of (a) with a real seismic
source signature digitized at intervals of 1, 2, 3 and 4 ms, respectively. Note the progressive degradation of wavelet shape as the sampling
becomes coarser.

3. Illustration using synthetic data calculated for two layers between half-spaces. Both
source and receiver are located immediately below the
The trace in Fig. 1(a) shows three intervals of a surface of the upper layer. Two-way travel time
one-dimensional synthetic reflection seismogram through the upper layer is 3000 ms and that through

Fig. 2. Results of applying the single-channel method to the traces in Fig. 1. Note the negative polarity due to reflection from the underside of the
sea surface. Also note the progressive degradation of times of peak amplitudes as the sampling becomes coarser.
32 T.M. McGee / Marine Geology 164 (2000) 29–35

Fig. 3. Subset of short-offset traces from a profile offshore Costa Rica. This profile has been analyzed using full-waveform inversion by Pecher et
al. (1998). The sea floor reflection, the BSR and the first water-layer multiple are indicated.

the lower is 449 ms. Reflection coefficients used in the The traces of Fig. 1(b)–(e) were created by
calculations are 1.0 at the top of the upper layer (to convolving Fig. 1(a) with a real seismic source sig-
simulate an air–water interface), 0.2 between the nature digitized at sampling intervals of 1, 2, 3, and
layers (to simulate a water–sediment interface) and 4 ms, respectively. It may be observed that the shape
20.5 at the bottom of the lower layer (to simulate a of the signature becomes progressively more
BSR). The magnitudes of the coefficients have been degraded as the sampling becomes coarser. The
defined a bit larger than what would usually be realis- trace of Fig. 1(d) is particularly noteworthy in this
tic in order that the internal multiple between the two regard. The maximum (negative) excursion of the
interfaces be clearly visible. Theoretical values of the sea-floor reflection appears truncated because no digi-
true amplitudes of the synthetic primary reflections tal sample happens to coincide with it. The maximum
are calculated to be 0.20 for the sea-floor reflection (positive) excursion of the BSR is also distorted, but
and 20.48 for the BSR. in a different way because the time between the BSR
T.M. McGee / Marine Geology 164 (2000) 29–35 33

excursions are shifted to erroneous times and their


amplitudes are as much as 20% in error.

4. Application to real data

In order to compare results obtained with real data,


the single-channel method described above was
applied to short-offset traces from a multi-channel
data set previously analyzed by Pecher et al. (1998)
using full-waveform inversion.
The data were acquired offshore Costa Rica in
about 3 s of water at a location where a BSR occurs
about 0.5 s below the sea floor. The source was a
tuned airgun array and the received signals were
digitized at a 4 ms sampling interval over a total
recording time of 8 s.
A gray-scale image of 57 short-offset traces from
the data set is shown in Fig. 3. The sea floor reflection
Fig. 4. Enlarged portion of Fig. 3 plotted as wiggle traces.
occurs slightly below 3 s, the BSR at about 3.6 s and
the first water-layer multiple slightly below 6 s.
The primary sequence extends from 3 to 6 s,
and the sea-floor reflection is not an integral multiple approximately, and the first multiple sequence
of the sampling interval. Such differences between would extend from slightly below 6 s to more than
digital waveforms degrade their coherence and consti- 9 s had the recording time been that long. Since
tute a source of processing error that is entirely due to calculation of the spectral quotient of Eq. (3) is greatly
the digitizing rate being inadequate. simplified when the numerator and denominator are of
The traces in Fig. 2 display the result of applying the same length, one second of zero values was
the spectral division of Eq. (3) and inverse trans- appended to the data. Undoubtedly, this is a source
formation to the traces in Fig. 1. Amplitudes on the of numerical noise and introduces some error into the
trace of Fig. 2(a) occur at the correct times and are results. In an effort to reduce such noise, data values at
within 0.1% of the theoretical values except that they times greater than 7.8 s were linearly tapered to
are negative due to the effect of the water surface. smooth the transition from non-zero to zero values.
Amplitudes on traces of Fig. 2(b)–(e) describe wave- An enlarged portion of Fig. 3 is plotted in Fig. 4 as
lets that are approximately symmetrical about their wiggle-traces to illustrate the waveforms that
maximum excursions. The symmetry results from comprise the sea floor and BSR reflections. It can be
both primary and multiple wavelets having the same seen that the dominant period of the signal is about
structure, i.e. that of the source signature, and the 25 m. This indicates a dominant frequency of about
quotient in Eq. (3) being equivalent to a deconvolu- 40 Hz.
tion of that structure. The symmetry degrades at The data in Fig. 3 were corrected for spherical
coarser sampling rates due to reduced coherence spreading (at 1500 m/s which is close to the speed
between numerator and denominator of Eq. (3). of propagation in water) and the spectral quotient of
When the degradation is negligible, i.e. Fig. 2(b), Eq. (3) computed after appending zeros to each trace.
the maximum excursions occur near the correct An enlarged portion of the output, Fig. 5, illustrates
onset times of the reflected wavelets and the ampli- the waveforms from which true amplitudes were
tudes are within a few percent of the theoretical measured. The dominant period at the sea floor and
values. When the degradation is substantial, as it is the BSR remains 25 ms, indicating that the deconvo-
in Fig. 2(d) and (e), the times of the maximum lution was not able to collapse the waveform to any
34 T.M. McGee / Marine Geology 164 (2000) 29–35

Fig. 6. Profile of compressional speed versus depth estimated by


Fig. 5. Enlarged portion of the output of the single-channel method
Pecher et al. (1998) using full waveform inversion on multi-channel
applied to traces in Fig. 3.
seismic data.

great extent. This is likely due to the data having 5. Discussion and conclusions
been band-pass filtered with steep cut-off slopes to
facilitate the rather slow digitizing rate. Application of the single-channel method to
Comparing Fig. 5 with Fig. 4, it is clear that the synthetic data shows that, if the digitizing rate is
output is noisier than the input. This is typical of great enough, amplitudes and arrival times can be
situations in which the coherence between numerator within a few percent of the values used to generate
and denominator of a spectral quotient is low. Spectral the synthetic traces. It also shows that the accuracies
notches in the denominator are not balanced by those of both amplitudes and arrival times degrade rapidly
in the numerator and noise is generated at frequencies as the digitizing rate is decreased. This leads to a
where large values are divided by small values. This conclusion that the accuracy of results is substantially
could result from either the digitizing rate being too reduced if the digital sampling rate is too slow. The
slow or the recording length being too short. The fact reason is decreased coherency between individual.
that both circumstances occur in this case only serves Pecher et al. (1998) initiated their full-waveform
to exacerbate the problem. It is possible, however, to inversion of the multi-channel data set by determining
follow both the sea floor reflection and the BSR from the reflection coefficient at the sea floor. They did this
trace 1 to trace 40 in Fig. 5 and make amplitude by dividing the amplitude of the first water-layer
measurements. multiple by that of the water-bottom reflection
Amplitude measurements on traces 29–39 (the set (Hyndman and Spence, 1992; Andreassen et al.,
over which the full-waveform inversion had been 1995). It was done on the two shortest-offset traces
averaged) average to 20.114 for the sea floor reflec- in common-mid-point gathers whose locations
tion and 10.059 for the BSR. Taking into account the correspond to traces 29–39 of the data subset shown
polarity reversal and the transmission coefficient at in Fig. 3. The 22 values thus obtained averaged to
the sea floor, the average reflection coefficients are 10.140. By correcting for transmission loss only at
found to be 10.114 at the sea floor and 20.060 at the sea floor, Pecher et al. (1998) assumed that the sea-
the BSR. floor reflection is generated solely by a change in
T.M. McGee / Marine Geology 164 (2000) 29–35 35

density, i.e. that there is no change in speed of propa- involve a significantly lower level of effort than do
gation across the water bottom, in order to produce multi-channel methods.
profiles of compressional speed versus depth. Fig. 6
shows such a profile at the location of trace 39. They
Acknowledgements
then calculated a reflection coefficient from the speed
decrease at the BSR by assuming that there is no
The assistance and encouragement of Ingo Pecher
change in density across the BSR and obtained a
in performing this work and the suggestions of Karin
value of 20.057. The single-channel method produces
Andreassen concerning the final manuscript are
a reflection coefficient of 20.060 at the BSR by
gratefully acknowledged.
correcting for transmission loss at the sea floor. These
values are in good agreement. It should be noted,
however, that the spectral quotient value was obtained References
without any assumptions concerning speed or density
changes across either the sea floor or the BSR. Andreassen, K., Hart, P.E., Grantz, A., 1995. Seismic studies of a
It can be seen in Figs. 4 and 5 that some reflected bottom simulating reflector related to gas hydrate beneath the
events occur between the sea floor and the BSR. If continental margin of the Beaufort Sea. J. Geophys. Res. 100,
12659–12673.
they are primary reflections, they would represent Holbrook, W.S., Hoskins, H., Wood, W.T., Stephen, R.A.,
transmission losses that should be included in the Lizzarralde, D., 1996. Methane hydrate and free gas on the Blake
calculation and thereby increase the magnitude of Ridge from vertical seismic profiling. Science 273, 1840–1843.
the reflection coefficient obtained for the BSR. The Hyndman, R.D., Spence, G.D., 1992. A seismic study of methane
events in question are rather weak, so the increase hydrate marine bottom simulating reflectors. J. Geophys. Res.
97, 6683–6698.
would not be great, perhaps changing the quoted MacKay, M.E., Jarrard, R.D., Westbrook, G.K., Hyndman, R.D.,
value to about 20.062, and there would still be 1994. Origin of bottom simulating reflectors: geophysical
good agreement with the value obtained by full- evidence from the Cascadia accretionary prism. Geology 22,
waveform inversion. 459–462.
It is concluded that, given proper data acquisition, Pecher, I.A., Ranero, C.R., von Huene, R., Minshull, T.A., Singh,
S.C., 1998. The nature and distribution of bottom simulating
the spectral quotient method can produce reflection reflectors at the Costa Rican convergent cargin. Geophys. J.
coefficients that agree well with those produced by Int. 133, 219–229.
full-waveform inversion. Singh, S.C., Minshull, T.A., Spence, G.D., 1993. Velocity Structure
If reasonable constraints are placed on density of a Gas Hydrate Reflector. Science 260, 204–207.
variations, the single-channel reflection coefficients Theilen, Fr., Ollier, G., McGee, T.M., Brussaard, P., De Batist, M.,
Missiaen, T., Ori, G.G., Perini, L., Kogler, F.-Chr., Neben, S.,
can then be used to study the structure of propagation Omlin, A. 1993. GISP—Geophysical In-Situ Probe, Final
speed in the vicinity of BSRs. Moreover, since only Report for Contract CT90-0057 of the European Communities
single-channel data are involved, such studies would MAST-1 Programme.

You might also like