You are on page 1of 5

SEISMIC IMAGE IMPROVEMENTS AFTER CORRECTION FOR ERRORS

IN OBN ACQUISITION PARAMETERS


E. Muijzert1, C. Bagaini1, A. Misbah1, N. Seymour1, M. Salgadoe1, B. Szydlik1
1
Schlumberger

Summary
Uncertainties in acquisition parameters of Ocean Bottom Node data impact image quality. In this study
we consider spatial and temporal variations in water velocity and source positions as well node positions
and clock drift. Our method inverts the direct arrival travel times using a joint linearized inversion for
these parameters. Modelling shows that our approach is valid for travel times picks including offsets
up-to an offset-to-depth ratio of 2.5. We find that spatial-temporal variations in the water velocity
observed in the seismic data are similar to those observed by PIES. We also observe that during part of
the survey the estimated water velocities are different between two simultaneous shooting vessel and
from PIES data. Spatial variations in the water velocity have to be considered for surveys acquired by
simultaneous shooting using multiple source vessels. We also show that our method can determine
survey wide improvements to the position of each source array. We find that source positions are stable
along straight saillines but variable source position corrections are found along trajectories with rapid
change in the vessel heading. Application of the improved acquisition parameters results in observable
differences in the shallow and observable deeper parts of the seismic image.

83rd EAGE Annual Conference & Exhibition


Seismic image improvements after correction for errors in OBN acquisition parameters

Introduction
Ocean-bottom node (OBN) data suffers from a number of acquisition-related perturbations including
temporal changes in the water velocity, source and node position, and node clock drift (Olofsson and
Woje, 2010). These errors, if left uncorrected, result in a sub-optimal seismic image. The common way
to correct these errors is through analysis of the travel time of the direct arrival. However, the inversion
of the travel time of the direct arrival has some noticeable correlations between parameters such as
source and node position as well as between vertical position, water velocity and clock drift that prevent
a stable solution (Doherty and Hays, 2012). As a result, the acquisition-related errors are commonly
determined through a sequential process (Amini et al., 2016, Bekara et al., 2021).

We have developed a method for the joint inversion of direct arrival travel times for changes in the
water velocity, source and node position and node clock drift (Seymour et al., 2021). Following our
initial experience, we found evidence for spatial-temporal changes in the water column. Simultaneous
shooting using multiple vessels is increasingly common and provides evidence that not only temporal
variations within the water column should be accounted for but also spatial variations. Furthermore, as
the spatial variations are derived from different source vessels, vessel-dependent source positions need
to also be considered.

Method
The direct arrival traveltime Tsr between source at location (𝑥𝑠 , 𝑦𝑠 , 𝑧𝑠 ) and receiver at (𝑥𝑟 , 𝑦𝑟 , 𝑧𝑟 ) along
its path l for a depth z and time-dependent ts water-velocity profile V is given by:
𝑟 𝑑𝑙
𝑇𝑠𝑟 = ∫𝑠 . (1)
𝑣(𝑧,𝑡𝑠 )
To obtain an accurate estimate of the travel time of the direct arrival between source and node, it is
important that the direct arrival is the first arrival and that its waveform is free of interference from other
waves such as multiples and refractions that have a similar arrival time. We have modelled, for a range
a seabeds, the critical angle of reflection beyond which refractions are generated. A soft seabed (1,600
m/s) was found to have a critical angle of 700 or offset-to-depth ratio of 2.7, while with increasing
hardness of the seabed the critical angle and offset-to-depth ratio decreases. In practice we find that in
areas with localized, fast, shallow-velocity layers, for example, we have to reduce the offset-to-depth
ratio even further, sometimes to a value close to 1 due to the presence of interfering refractions.

The acoustic velocity of the water is subject to daily and seasonal variations caused by changes in
temperature and salinity resulting in a predominantly vertically-stratified velocity profile. Sound
velocity profiles (SVPs) are routinely acquired during OBN surveys by the remotely-operated vehicle
system that places the nodes on the seafloor. We use these SVP profiles together with pressure-inverted
echo sounder (PIES) measurements to determine an average velocity profile for the survey and to
investigate the temporal and spatial variations in the water column (Bagaini et al., 2021). The vertical
stratification of the velocity in the water column causes ray bending of the seismic energy at large
incident angles. We conducted ray-tracing experiments through such models, which showed that for an
offset-to-depth ratio smaller than 2.5, the impact of ray bending on the travel time of the direct wave is
well below 1 ms and a straight-ray approximation is a valid assumption.

The direct arrival travel times are inverted using a linearized inversion in the form of:

𝐴𝑥 = [𝑇𝑠𝑟 − 𝑇𝑠𝑟 ]. (2)
Matrix A contains the partial derivatives of the modelled travel times with respect to the model
𝜕𝑇
parameters x, 𝜕𝑥𝑠𝑟. The vector x contains the changes in the model parameters:
𝑖
𝑥 = [𝑑𝑥𝑟 , 𝑑𝑦𝑟 , 𝑑𝑧𝑟 , 𝑑𝑐0,𝑟 , 𝑑𝑐1,𝑟 , 𝑑𝑐2,𝑟 , 𝑑𝑣( 𝑥𝑠 , 𝑦𝑠 , 𝑡𝑠 ), 𝑑𝑠𝛼↑ , 𝑑𝑠𝛼→ , 𝑑𝑧𝑠 ], (3)
where 𝑑𝑐0,𝑟 , 𝑑𝑐1,𝑟 , 𝑑𝑐2,𝑟 are the coefficients of the quadratic clock drift model used and 𝑑𝑣( 𝑥𝑠 , 𝑦𝑠 , 𝑡𝑠 )
is the spatial and temporal change in the water velocity. Changes in the horizontal source position are
expressed in vessel-oriented coordinates 𝑠𝛼↑ 𝑎𝑛𝑑 𝑠𝛼→ , the inline and transverse source position,
respectively, as shown in Figure 1.

83rd EAGE Annual Conference & Exhibition


Figure 1 (left) Diagram illustrating the relationship between the inline transverse source-coordinate
system with the node-coordinate system for a vessel sailing with heading α. (right) Diagram showing
the vessel is sailing in the opposite direction.
Using vessel-oriented coordinates for source position is insightful as its changes can be directly linked
to the source configuration. For instance, a systematic position error in the centre of energy of the source
will result in a survey-wide, constant inline and/or transverse source shift. Inversion for the position of
groups of sources along lines instead of each single source prevents so-called stretching of the
coordinate system as reported by Doherty and Hays (2012).

The right-hand side of this equation contains the differences between the observed 𝑇𝑠𝑟 and modelled 𝑇𝑠𝑟
travel times for each source-receiver pair. Each row of this set of linear equations relates the observed

𝑇𝑠𝑟 and modelled 𝑇𝑠𝑟 to its modelled travel time and the partial derivatives to its model parameters:
𝜕𝑇
∑𝑛𝑖=1:𝑛 𝑠𝑟 𝑑𝑥𝑖 = [𝑇𝑠𝑟′
− 𝑇𝑠𝑟 ]. (4)
𝜕𝑥𝑖
The resulting matrix A has, typically, many more rows (millions of travel times) than columns
(thousands of parameters) and is sparse. Equation 2 is solved using one of the many available solvers,
here, we use the LSQR algorithm (Paige and Saunders, 1982). The linearized inversion also allows us
to approximate errors in the estimated parameters, and the covariances between them, as well as to
calculate the rank of the matrix.

A common assumption is that the water velocity is spatially invariant but analysis from PIES
measurements indicates that, in particular with large surveys, spatial and temporal variations in the water
velocity have to be considered (Bagaini et al., 2021). Spatial-temporal variations may be weather-related
such as hurricanes whereas (changing) currents may also move water with different temperature and
salinity in the survey area and exist on top of the common daily variations in water velocity. Individual
source arrays from the same vessel are operated close to each other and it is assumed that they are
operated in water with similar properties. A similar assumption can be made when simultaneous-
shooting source vessels are operated close to each other. Often, however, source vessels, in order to
facilitate effective separation of the energy from the different shots present in the recording, do not
operate close together. A large separation between the source vessels increases the likelihood that the
assumption of a spatially-invariant velocity field breaks down. We therefore adapt the joint inversion
scheme such that the inverted velocity model depends on the shot time and location. It is expected that
the variations in the estimated velocities will now also depend on the distance between the vessels.

Example
The method was applied to an OBN survey in the Gulf of Mexico that was acquired using two source
vessels, each shooting with three sources. Analysis showed that the average source-vessel separation
throughout this very large survey was around 30 km. A total of 2,048 nodes were deployed with an
interval of 1,200 m in water depths between 300 m and 1,300 m. Independent water-velocity estimates
were available from four PIES positioned along the boundary of the node grid separated by over 20 km
and at depths between 280 m and 950 m. Direct arrivals were picked in the data up to an offset of 3 km
although this was greatly reduced in the shallow part of the survey. Offsets were further reduced in parts
of the survey area where refractions from shallow, fast salt layers arrived before the direct arrival. An
average water velocity model was derived from SVPs acquired during the survey. Joint inversion of the
direct wave picks for node position, clock drift, velocity, and source position reduced the average RMS
of the travel time difference from 0.85 ms to 0.23 ms.

83rd EAGE Annual Conference & Exhibition


Figure 2. Comparison of the seismic estimated average water velocity nearby both vessels and from
the four PIES, referenced to a constant depth of 250 m.

A comparison was made between the estimated water velocities from the joint inversion and the PIES
measurements, see Figure 2. To compensate for the large variations in bathymetry within the survey
area, we have mapped the seismic velocities to a common depth of 250 m using the reference velocity
profile. Since the PIES were located at very different depths, we have mapped the average velocity
estimated with the PIES to a depth of 250 m as well. We observed that the velocity ranges from PIES
and seismic are compatible without any upfront calibration or matching of the two types of
measurements. The PIES were retrieved when seismic acquisition in their area had finished. The
trends of the velocities around the two vessels are similar and consistent with PIES 2506, PIES 2502
and PIES 2503. Between days 195 – 225 the seismic velocities drop and up-to day 215 this decrease
is more evident around vessel A. After day 225 the seismic velocities pick up the trend of PIES 2506
again. Both seismic and PIES measurements highlight that spatial, in addition to temporal, variations
are not negligible for this survey.

The inline source shift was found to vary between 1 m and 2 m towards the vessel, depending on the
source array. The transverse source shift was found to have a predominantly positive sign on source
lines with a heading of predominantly 0o, while on source lines with an opposite heading (180o), the
transverse source shift has a predominantly negative sign of under a meter. This situation may be caused
by a misalignment between the GPS antennae and the subsurface centre of the seismic energy of the
source. Application of the parameter estimates obtained by the joint inversion results in improvements
in the seismic image, visible close to the seabed and also in some deeper parts of the section (Figure 3).

Figure 3. 18 Hz reverse-time migration with unresolved acquisition errors (left). After correction of
acquisition errors, we observe improved amplitude continuity is the subsalt section (right).

A separate detailed study showed that while inversion for survey-wide source inline and transverse for
each source array greatly reduced the data residuals, the source lines closest to obstructions showed
elevated data residuals (Figure 4a). As these source lines showed substantial changes in heading near
the obstructions, a new binning scheme was tested that reduced the bin size if a rapid change in heading
was detected along a source line. The new binning scheme reduced the data residuals on the source lines
near the obstructions to the level observed on the straight source lines (Figure 4b). Furthermore, it is
observed on the source lines close to the obstruction (Figure 4c) that when the source turns (as shown
by the change in its heading (Figure 4d)), the direction of the transverse source shift is inwards relative
to the reported source position. A negative transverse shift is observed for a clockwise turn, while a
positive value is observed for a counter-clockwise turn. This observation can be explained by the

83rd EAGE Annual Conference & Exhibition


centrifugal force that acts on the underwater air-gun array when the source vessel changes direction
rapidly, and the flexible source rigging.

Figure 4. (a) Close-up of the travel time residual in milliseconds after joint inversion using source-line-
long bins, and (b) using the modified source binning based on heading. (c) The transverse source shift
in meters obtained by the joint inversion result shown in (b). The arrows in (c) indicate the turn
directions. (d) The source line heading in degrees where zero degrees is north.
Conclusions
We propose a method for the joint inversion of the direct arrival travel times of OBN data for source
and receiver positions, clock drift and water velocity. We show that spatial and temporal variations in
the water velocity can be observed by inverting data originating from simultaneously shooting vessels
at different locations. Furthermore, perturbations in the source positions require inversion for each
source array separately. Examples show that the water velocity estimates can be related to water velocity
estimates obtained by PIES, whereas improvements in the source position are stable along straight lines
throughout the survey, except in locations where the sailing direction of the source vessel changes
rapidly, such as around obstructions. Application of these corrections to the seismic data results in
improvements in the seismic image.
Acknowledgements
We thank WesternGeco Multiclient and TGS for use of these data.
References
Amini, A., Peng, H., Zhang, Z., Huang, R., and Yang, J. [2016] Joint inversion of water velocity and
node position for ocean-bottom node data. 86th Annual International Meeting, SEG, Expanded
Abstracts, 5490-5494.

Bagaini, C. Muyzert, E., Kapllani, K., Salgadoe, M., and N. Seymour, [2021]. On the usage of direct
measurements of water velocity in deep water OBN surveys. First International Meeting for Applied
Geoscience & Energy SEG, 21-24.

Bekara, M., Davision, C. and R. Djebbi, 2021. Parametric inversion of water column velocity for cold
water statics correction in Ocean Bottom seismic surveys. 82nd EAGE Annual Conference & Exhibition,
Extended abstracts.

Docherty, P., and Hays, D. [2012] Ambiguities in direct arrival time inversion of ocean bottom nodes:
74th EAGE Conference & Exhibition, Extended Abstracts.

Olofsson, B. and Woje, G., [2010] Ensuring correct clock timing in ocean bottom node acquisition. 80th
Annual International Meeting, SEG, Expanded Abstracts, 172-176.

Paige, C. C., and Saunders, M. A., 1982, LSQR: An algorithm for sparselinear equations and sparse
least squares. ACM Trans. Math. Software, 8, 43-71.

Seymour, N., Bagaini, C., Muyzert, E, Kapllan, K. and M. Salgaloe, [2021]. Sparse OBN image
improvements by correction for acquisition errors. First International Meeting for Applied Geoscience
& Energy, SEG, 81-84.

83rd EAGE Annual Conference & Exhibition

You might also like