You are on page 1of 11

Positioning Ocean Bottom Seismic Cables

Simon Partridge, Sonardyne International Ltd.


Noel Zinn, Western Geophysical Division of Baker Hughes
Overview of Ocean-Bottom and Transition-Zone Seismic Surveying
Reflection seismology is the primary geophysical method employed in the search for hydrocarbon
accumulations. Seismic data are collected on land, at sea and in the transition zones in between land and
sea. On land, seismic geophone sensors embedded in the earth and energy sources under the earth or at the
surface are positioned conventionally or with the Global Positioning System (GPS). At sea, often called
deep-marine seismic, seismic energy sources (typically air guns) and seismic hydrophone sensors towed
behind a vessel are positioned by a variety of devices that include GPS, magnetic compasses, underwater
acoustics, lasers and optical shaft encoders. In the transition zone of shallow surf, mud and alligators, we
use any combination of these technologies that gets the job done.
In the marine environment, ocean bottom cable (OBC) surveying, in which the seismic cable is
laid on the bottom, not towed near the surface, is gaining popularity. Some of the advantages of OBC over
towed streamer surveys are the flexibility of acquisition geometry, greater surface consistency (i.e., more
combinations of source and detector at different azimuths for a given midpoint, useful for resolving static
delays and for amplitude compensation), more flexibility in working around obstructed zones, the use of
dual sensors to remove ghosts and layer reverberations, reduced noise by eliminating cable vibration and
strumming caused by towing and surface weather conditions, and better coverage due to the elimination of
cable feather caused by currents.
Figure 1 shows the important elements of an OBC survey. There are cables with dual seismic
sensors (hydrophones and geophones) connected to a recording vessel in the middle of the graphic. The
shooting vessel with one or more air gun arrays that produce the seismic energy is shown sailing a regular
pattern orthogonally to the cables. Orthogonal shooting has geophysical and positioning advantages, but
in-line shooting is also possible. The collected midpoints of all possible combinations of sources and
detectors comprise a swath of coverage.
OBC surveys are today limited to about 200 meters of water depth, but 1000 meters surveys may
be commonplace in a couple of years. OBC surveys that extend on the shallow end into the surf zone are
the subset of transition zone (TZ) surveys that we address in this paper. In Section 5 we examine two
Southeast Asian prospects that were surveyed during the summer of 1998: an OBC survey in 70 meters
depth in the open ocean and a TZ survey in 2 to 12 meters depth along a shoreline.

Figure 1

Positioning Methods. Seismic energy source positioning in TZ and OBC is similar in technique and
quality to source positioning in deep-water streamer surveys. It basically consists of GPS receivers on the
source array. On the other hand, detector positioning techniques are less-widely standardized in TZ and
OBC than in land or towed-streamer surveys. Three techniques are common in the industry: (1) recording
and using the drop or placement coordinates of the detectors, (2) deploying high-frequency acoustic sensors
attached to the detectors and positioned independently of the seismic survey and (3) using multiple
occasions of the onset of seismic energy (first breaks) as surveying observations in a positioning algorithm.
A combination of acoustics and first breaks is also possible.
Drop Coordinates. Since drop positions must be recorded to assure that the actual detector locations are
near the planned locations, drop positions are the cheapest and easiest to implement. In calm shallow water
(such as an inland bay where the detectors may be placed on or thrust into the muddy bottom), the detector
drop position can be close to the resting position. In deeper water or in agitated surf zones, this is unlikely
due to waves, currents and drop trajectories.
Acoustics. High-frequency acoustic systems, which are technologically similar to those used for years in
deep-marine surveys, are available for TZ and OBC applications. An acoustic transducer on a survey vessel
interrogates transponders attached near the seismic detectors to determine their positions. Figure 2 is a
simulated, underwater view of acoustic ranges being acquired. The acoustic system used in the case studies
cited in this paper is described in Section 3.
Acoustics provide a precise observable, essentially a time pick computed in hardware.
Consequently, acoustic surveys can be quite accurate within their budget of systematic errors that includes
uncertainties in detector depth, the velocity of sound in water and instrumental delay between the acoustic
interrogation and the transducer's GPS coordinates. Acoustic processing software must account for vessel
motion and successfully reject outlying responses caused by surface ghosts or vessel noise. Acoustics can
save time by providing rapid positions when they are needed, but this is accomplished at the expense of
dedicated equipment and personnel.

Figure 2

First Breaks. Seismic first-breaks can be picked by any number of automated methods that choose a
significant change in the amplitude or inflection of the arriving seismic energy. Figure 3 shows a raw
seismic trace with the first break noted. Band-pass filtering, deconvolution and neural networks are used to
improve the quality of the pick. First-break pick times can be processed in a positioning algorithm. The
algorithm used in the case studies cited in this paper is described in Section 4.

Figure 3
Because seismic processors and processing systems may already be on the crew, the marginal cost
of first-break positioning is low. Typically coarser than acoustic observations, first breaks are typically
more numerous. Thus they benefit from more random error cancellation. First-break positioning software
must correctly compensate for the systematic errors of uncertainties in detector depth and velocities of
sound in water and in one or more refractors, instrumental delay, picking delay (or anticipation), velocity
gradients in the refractors and a possibly-complex near-surface geology. It must successfully reject outlying
picks caused by noisy seismic data.
Combination of Acoustics and First Breaks. A combination of acoustics and first breaks provides
positioning redundancy and independent assurance that coordinates are adequate for seismic purposes.
Acoustics provide highly precise observations, but usually fewer of them. First breaks provide a relative
abundance of observations, albeit of lesser quality.
Sonardyne TZ/OBC Acoustic Positioning System
System Overview. The Sonardyne TZ/OBC acoustic positioning system was specifically designed to
position hydrophone and geophone detectors and withstand the harsh operational environments associated
with large scale OBC and TZ operations. The system operates in the 40kHz band and typically consists of
four main parts: seabed transponders, an RFID (Radio Frequency IDentification) system (optional), a vessel
mounted acoustic transceiver and data collection and positioning software.
The system offers many operational advantages and freedoms due to its independence from the
main seismic data recording system. The small size and rugged nature of the equipment enables any vessel,
however small, to position the cable, allowing access into extremely shallow water. The transponders are
autonomous, therefore positions can be verified prior to the cable being connected to the recording vessel
and prior to shooting. This can save valuable time in the event that a cable needs to be repositioned.
The system architecture and software allows data to be collected from multiple, synchronized
acoustic transceivers typically towed behind the vessel or even on separate vessels. This increases the
number of ranges collected, improves geometry and can reduce acquisition time. Note this feature was not
used in the case studies detailed in this paper in which a single surface transceiver acquired ranges.
Seabed Transponders. Once in the field, transponders are programmed with their individual acoustic
identities and arranged in groups of nine individual reply codes, by means of a Programming Box with noncontact coupling to the transponder. There are up to 400 individual groups allowing as many as 3500

unique identities. Groups of transponders are sequentially interrogated from the surface transceiver with
nine separate replies being received.
In TZ operations, transponders are attached to the receiver package just prior to deployment. With
squirter-based OBC operations, transponders are permanently attached to the cables using a specially
designed "Carapace" that provide protection and a snag-free, low-noise outline. Although small and
powered from an alkaline battery pack, transponders typically have an operational life of in excess of 12
months and a depth rating of 500 meters as standard.
RFID System. With large scale operations there can be more than 1500 different transponder identities to
track. The RFID system uses small passive devices, housed within the Carapace, that hold information
about the transponder and, if necessary, the cable section. As the cable sections are deployed all
transponder information is read from the RFID device as it passes over an antenna built into the squirter
and is logged together with GPS drop co-ordinates.
The RFID drop coordinates can be compared with the pre-plot coordinates to determine the
accuracy with which ground stations are deployed.
Surface Transceiver. Under software control the transceiver sequentially interrogates transponder groups
within a selectable acquisition range. The two-way travel times are measured to a resolution of 13
microseconds and logged by the software. Four levels of receiver gain and two levels of transmit power are
available, under software control, enabling optimum performance in different reverberant environments.
The electronics are shock-mounted and waterproofed in a standard underwater pressure housing rated to
500 meters. The transducer is typically mounted on a through-hull or on an over-the-side pole with a GPS
antenna co-located above as shown in Figure 4. Alternatively the whole assembly can be towed behind the
vessel.

Figure 4

Acquisition and Positioning Software. Sonardyne's GENIEAPP software can be used to acquire and
process the range data to each transponder or, alternatively, the system can be, and has been, integrated into
other navigation software packages.
GENIEAPP allows the acoustic data to be collected and processed into geodetic positions in realtime, enabling faster verification of ground station positions. Knowledge of the approximate drop-positions
(X, Y and depth) and associated transponder identities is obtained from the RFID system or pre-plot file.
The vessel position and a UTC time stamp are obtained from GPS. As the vessel surveys a fixed offset
from the cable, the groups of transponders within acquisition range are interrogated. All raw data are
logged for reprocessing. The two way travel times reported from the transceiver are converted into one-way
range observations (less a fixed offset) with knowledge of the local speed of sound in water.
Range observations are processed into positions using a standard extended Kalman filter. The
filter starts with the approximate position and then uses the sequential observations to improve those
positions. In addition, positional precision is also estimated for each transponder.
Erroneous observations are detected before they are passed to the filter. The "blunder detector"
compares each observation with the current predicted value. The predicted variance of the observation is
used in association with the comparison to compute a weight for each observation. Observations with small
variances influence the filtered position more than those with large variances.
Having collected ranges from both sides of the line, a calibration (results) file is immediately
produced containing quality control statistics and raw range data for external reprocessing.
SDCOORD First Break Processing System
Source-Detector Coordinate Recomputation (SDCOORD), whose development began in 1982, is a Western
Geophysical Seismic Function Module (SFM) that positions detectors with first-break picks. SDCOORD is
a sequential least-squares algorithm based upon the measurement model of an extended Kalman filter. A
global polynomial regression pre-processor relates pick time to distance in three dimensions and solves for
several systematic errors. Blunder rejection is implemented in the global polynomial regression stage with a
simple, but effective, difference tolerance. Computed source-specific and detector-specific velocity trends
model a lateral velocity gradient if it exists.
Global Polynomial Regression. In SDCOORD our observations are the seismic first-arrival picks. A pick
is associated with nominal coordinates and depths for the source and the detector. The distance (or offset)
between these nominal coordinates can be determined using the Pythagorean Theorem in three dimensions.
Using least squares, offsets are regressed against pick times to determine the coefficients of the best-fitting
polynomial of a user-selected order. Using this polynomial, a "regressed" distance in meters or feet is
determined for a specific pick in milliseconds by substitution into the polynomial. After some additional
steps (see velocity trends below), these regressed distances are processed by the measurement model.

Figure 5

Figure 5 shows computed distances in meters plotted against pick times in milliseconds for a
typical data set. A polynomial fitted to these points determines the relationship between pick times and
regressed distances that are processed by the measurement model. Notice that the polynomial does not
cross the origin of the plot. This static, Y-axis intercept at zero pick time absorbs the systematic errors of
instrumental delay and delay (or anticipation) in the mathematical definition of the onset of seismic energy
in the first-break picker. The shape of the polynomial at near offsets will correct for a global bias in
bathymetry by bending into an approximate hyperbola. The polynomial differentiated with respect to pick
time (i.e., its slope) quantifies the vertical velocity profile over the refracted offset range. Outlying picks
are shown on this plot. They are rejected if they exceed a difference tolerance with respect to the
polynomial. This improves coordinate results and predicted coordinate error (dRMS).
Velocity Trends. After the global polynomial regression of SDCOORD, one or two optional types of leastsquares regressions can be performed, i.e., one type for every source gather and one type for every detector
gather. There may be as many individual regressions as there are total sources and detectors, e.g., tens of
thousands for a large swath. These regressions produce source-specific velocity trends and detector-specific
velocity trends (scale factors near unity). When implemented, the geometric mean of the velocity trends for
the source and detector associated with each pick is multiplied by the globally regressed distance before
processing in the measurement model. By providing a varying field of scaling factors over all sources and
detectors in the prospect, velocity trends model a lateral velocity gradient, if it exists.
Measurement Model. The equations of the measurement model of a sequential, extended Kalman filter are
widely published in the literature. SDCOORD rigorously adheres to these equations for the processing of
regressed, velocity-trended distances from which modeled biases and blunders have been removed as
described above.
Iteration Until Convergence. All stages of the SDCOORD algorithm are iterated in sequence until
coordinate convergence to a user-defined tolerance is achieved.
Quality Control. SDCOORD provides a wealth of quality control statistics by which to evaluate results.
These include: unscaled coordinate variances in the grid axes, unit variance factors for all detectors (which
quantify the "fit" of the adjustment given pick uncertainty), dRMS (radial error) values scaled by the unit
variance factor, the number of rejected picks, the total number of used picks for each detector and the
number in each octant, heuristic "cross-correlations" that reduce the octant distribution of geometry to
single numbers for quick reference, the sum of all residuals in each quadrant, the coefficients of the bestfitting polynomial, all the velocity trends and the distance to the adjacent detector. Also, plots of residuals
as a function of pick time and the usual seismic quality control of linear moveout corrected first arrival
displays can be produced.
Case Studies
This paper exhibits sample results from two separate Southeast Asian seismic surveys conducted during the
summer of 1998. Since the surveys were shot in distinctly different environments, they provide interesting
comparisons between the acoustic and first-break results.
Ocean Bottom Cable. The OBC survey was carried out by a crew that consisted of several large vessels
working in the unobstructed open sea with approximately 70 meters of water depth. (A typical OBC crew
has one seismic energy source vessel, two cable-laying vessels, one recording/processing vessel and one
utility vessel that may be used for acoustic interrogation among other functions.) Seismic cable was laid on
the bottom by automated hydraulic handling equipment called squirters. The seismic shooting lines were
parallel to the detector cables. Approximately 250 acoustic transponders were used, each placed in the
center of a cable section that contains six seismic detectors, to confirm the first-break coordinates at regular
intervals. The RFID system was used to log the transponder drop positions. Acoustic acquisition was
carried out by sailing either cable vessel parallel to the cable with a vessel-to-cable offset of approximately
130 meters. In the data shown, eight 9-km lines are deployed in the north-south direction.

Figure 6 shows the cables with seismic detectors and acoustic transponders attached ready for
deployment on the back deck of a cable-laying vessel. Figure 7 shows the acoustic transponder locations
for eight lines of a subset of this prospect.

Figure 6

OBC - Detector Coupled With Acoutic Positions


(Eight x 9KM Lines North-South)
100000

99000
98000

Northings (m)

97000
96000

95000
94000

93000
92000

91000
90000
40000

41000

42000 43000

44000

45000

46000

Eastings (m)

Figure 7

47000

48000

49000

50000

Transition Zone. This survey was a hybrid operation consisting of a land survey down to the coastline, a
TZ component through the surf zone and an OBC component in shallow water. A land-seismic cable with
regularly-spaced, submersible battery boxes was deployed from the beach to a water depth of
approximately 12 meters. (By contrast, an OBC cable receives unlimited power from the recording vessel.)
Drilled explosives on the land side and marine air-gun seismic energy were recorded by both the land and
marine ends of the seismic cable.
Acoustic positions on the submersed cable were valuable for three major reasons. The first reason
had to do with the hybrid aspect of the survey. The marine sections of the cable had to be powered up
during the shooting of the land charges. This extended operation depleted the batteries, which then needed
to be replaced at regular intervals. Replacing the batteries always presented the risk of moving the detectors
on either side of the submersed batteries. Confirming detector positions with first breaks (into a powered
cable) would further deplete batteries, a vicious cycle. On the other hand, acoustic positioning can be done
on an unpowered cable. A second reason is that, due to a large tidal variation, seismic array shooting was
restricted to times of high water. Therefore, confirming detector positions with first breaks could not
always be accomplished when needed and it competed with seismic operations for valuable seismic source
vessel time. A third reason had to do with the geometry of the surf zone. First-break positioning benefits
from many, long-offset, refracted picks in a balanced azimuthal distribution to limit random errors and
solve for velocity gradients. Azimuthal balance is more difficult to achieve near shore. On the other hand,
acoustic interrogation requires less depth clearance and shorter offsets.
Figure 4 shows the acoustic transducer under the GPS antenna on the interrogation vessel in this
survey. Figure 8 shows the acoustic transponder locations for a two-line subset of this prospect. Notice that
the acoustic density is greatest near the shore (to the south) where first-break geometry is least balanced. In
the deeper water (to the north), where the first-break geometry is typical of OBC operations, transponders
are only placed on either side of the batteries to check rapidly for cable movement near the batteries.
TZ - Detector Coupled with Transponder Positions
(Two lines running NNW-SSE)

Northings (m)

16500

15500

14500

13500
19000

20000

21000

22000

Eastings (m)

Figure 8
Comparisons of Acoustic and First-Break Results
Error Quantification and Compensation. Like all measurement systems, both acoustics and first breaks
have sources of random, systematic and gross error that must be quantified, modeled or eliminated by the
positioning software. Both are affected by the geometrical distribution of observations. It is important that
observations are collected from all quadrants around the detectors or transponders. Unbalanced geometry
decreases the reliability of the solution if gross observational errors (blunders, spikes, outliers) occur (as
they will) and are undetected.

We quantify random error in this paper with a metric called "distance root mean square" (dRMS),
or radial error, the root sum square of the scaled uncertainties in the X and Y coordinates. Acoustics and
first breaks share some sources of systematic error (deterioration of GPS positioning, instrumental delays,
uncertainties in depth and the velocity of propagation in water), but differ in others. Acoustics are more
affected by surface ghosts and bottom obstructions. First breaks are more affected by variations in refractor
velocity and complex geology. Their respective adjustment software must model and solve for these
sources of systematic error. Gross errors must be identified and eliminated by both systems. A comparison
of acoustic and first-break coordinates is a test of how well all this has been accomplished.
Summary of Comparisons. Figure 9 is a bullet plot of the comparison of 247 coordinate pairs from the
OBC job that excludes three outliers between 6 and 16 meters, about 1% of the comparisons. These outliers
cannot be explained by poor quality data or geometry of either the acoustics or the first breaks. Of course, a
few outliers several standard deviations from the mean are an expected statistical variation. (This fact may
be a wake-up call for those expecting engineering performance from systems subject to random and
systematic environmental factors!) All data were processed in the field and not modified for this paper.

OBC - Acoustics vs First-Break

TZ - Acoustics vs First Break

(XY Difference for all lines)

(XY Difference for two lines)


6

6
Western Line

Eastern Line
5

Northings (m)

Northings (m)

Eastern Lines
5

0
-6

-5

-4

-3

-2

-1

-1

Western Line

0
-6

-5

-4

-3

-2

-1

-1

-2

-2

-3

-3

-4

-4

-5

-5

-6

-6

Eastings (m)

Eastings (m)

Figure 9

Figure 10

Figure 10 is a bullet plot of the comparison of 118 coordinate pairs from the TZ job that excludes
no outliers. Water depth varies for these two lines from 6 to 10 meters. Again, all data were processed in
the field and not modified for this paper.
Ocean Bottom Cable
Acoustics

First
Breaks

Transition Zone
Acoustics

First
Breaks

Num Obs

74 19

74 22

43 20

134 44

Est Obs SD

1.07m

1.50m

1.24m

6.14m

dRMS

0.25m

0.35m

0.38m

1.06m

Mean Delta
X
Mean Delta
Y
Random
Radial

- 0.71m

+ 1.19m

+ 0.53m

0.02m

2.47m

2.69m

Table 1: Summary of Predicted Errors and Actual Difference

Table 1 summarizes some important statistics for the OBC and TZ case studies. "Num Obs" is the
average number of observations (acoustic or first break) for the transponders and detectors plus/minus a
standard deviation cited to indicate the variation in the number of observations. "Est Obs SD" is the
estimated precision of the acoustic and first-break observations themselves determined from parameters of
their respective adjustments. "dRMS" is the radial precision predicted by the acoustic and first-break
adjustment software respectively. "Mean Delta X" and "Mean Delta Y" are the average differences in
coordinates between the acoustic and first-break results. "Random Radial" is the root sum squared of the
standard deviations of the actual differences in X and Y, a measure of scatter in two dimensions. Although
geometry is not quantified in the table, it was fair for the OBC first breaks (more observations N/S than
E/W), but well balanced in all other cases.
Comments
We first note that the average differences between the acoustic and first-break positions (Mean Delta X and
Y) are about a meter in both the OBC and TZ cases. Although we may hope for no average differences,
these amounts give us confidence that these very different systems are free from major systematic biases.
On the other hand the actual random radial differences are 2.47 and 2.69 meters, reflecting the scatter seen
in the bullet plots (Figures 9 and 10). According to statistical laws of variance-covariance propagation, the
root sum square of the acoustic and first-break dRMS values should, in the absence of systematic bias,
predict the actual random radial error of the differenced coordinates. However, in both the OBC and TZ
cases the dRMS are overly optimistic, thus indicating some residual systematic biases in one or both
systems.
One systematic bias not heretofore mentioned, nor accounted for, is the fact that the acoustic
transponders are not co-located with the seismic hydrophones that produce the first breaks. Seismic
detectors can mask acoustic signals into the transponders, reduce the number of observations and degrade
the geometry. Furthermore, the larger overall diameter of co-located sensors would cause problems for the
squirters during deployment. Therefore, transponders are typically placed about half a meter from the
hydrophone as can be seen in Figure 11 from the TZ survey. If the direction of this offset were consistent,
we could compensate for it in the comparisons. In fact, in the routine of production seismic surveying, bidirectional cables often get reversed and exchanged and they do not always lie straight on the seabed. A
planned consistency in offset size and orientation will soon disappear, thus contributing to the mean or
random radial differences, or both, that we see in these data. Interestingly, if the offsets were consistent in
azimuth, the mean differences would show an in-line (N/S) bias. In fact, the largest biases are cross-line
(E/W).

Figure 11

Notice the number of acoustic observations in both prospects. Their variation is caused by
operational differences in geometry caused by the different water depths, vessel speed and transponder
separation, though the numbers are typical of normal field procedures. On the other hand, the number of
first-break observations is less than that often achieved, especially in the OBC prospect. In the TZ prospect
coverage was limited by the shoreline and the first-break requirement for balanced geometry. In the OBC
prospect coverage was limited by operational exigencies, i.e., by the need to immediately process the first
available data to confirm that the actual lay adequately matched the plan. Fortunately, the quality of the
first-break picks in the OBC job was excellent and a less-than-normal number of them produced good
dRMS quality metrics.
Closer examination of the bullet plots reveal other small systematic differences. For example, in
the OBC bullet plot most of the points to the west are from 7 of the 8 lines (plotted darkly as diamonds) and
most of the points to the east are from the remaining line (plotted lightly as dots). This is seen more clearly
in Figure 12, which plots the difference in Easting against the average Easting of the eight lines. In the TZ
bullet plot most of the points to the west are from 1 of the 2 lines (plotted darkly as triangles) and most of
the points to the east are from the remaining line (plotted lightly as dots). These small systematic
differences are under investigation in ongoing efforts to improve OBC/TZ seismic positioning. But since
most differences are well within 3 meters for these unaltered field comparisons, we can be confident that
our seismic objectives are achieved.

OBC - Acoustic vs First-Break (Eastings Difference)


5
4
3

Eastings Difference (m)

2
1
0
-1
-2
-3
-4
-5
Lines 1-8

Figure 11
Conclusions
Acoustics and first breaks are complementary technologies for the positioning of seismic cables on the sea
bed in TZ and OBC prospects. Acoustics provide rapid positions whenever required (such as confirming
the position of a cable that may have moved) and can operate in tight quarters (such as shallow-water
shorelines). Acoustics can save time and therefore money. These benefits are purchased at the expense of
additional equipment, personnel and operational complexity. First-break positioning uses assets already on
the crew, viz., seismic data and seismic-data processors, so its marginal cost is low. First-break positions
are limited by the availability of the seismic data and a requirement for balanced geometry. It is reassuring
that these disparate technologies agree to within seismically-acceptable tolerances.

You might also like