In 2007, Shell acquired its first ocean bottom seismometer (OBS) survey in the deepwater Gulf of Mexico (GoM) Mars basin to obtain improved illumination of structurally complicated targets. This was a success and led to the discovery of the Boreas reservoir a couple of years later. In 2010, building on the demonstrated value of OBS, Shell embarked on a phased campaign of OBS acquisition at several locations in the deepwater GoM for exploration, development, and 4-D analysis.

Acquisition

The 2007 survey was acquired with roughly 800 Z3000 Fairfield Nodal nodes deployed on a hexagonal 400-m (1,300-ft) grid. The source was towed 12 m (40 ft) deep to ensure that low frequencies were generated strongly enough to illuminate the subsalt targets. The full extent of the receiver patch in 2010 was different than the 2007 survey and covered additional exploration objectives to the northeast and south of the Mars tension-leg platform (TLP). The source depth of the 2010 survey was decreased to 10 m (30 ft) to obtain higher frequencies for the shallow above-salt reservoirs.

The depth difference between the base and monitor surveys is shown before (left) and after (right) coldwater statics correction Clear striping parallel to the acquisition direction can be seen before the static correction. The differences reduce to nearly 0 after the static correction has been applied. Note that the static correction is applied pre-migration. (Images courtesy of Shell)

As the number of nodes to be deployed was larger than the number of physical units available, nodes were rolled from one side to the other of the survey, which was a first in the GoM. For both surveys, a dual-source airgun boat was used. For the 2010 reshoot, differences in sea current conditions made it unrealistic to attempt to repeat the actual 2007 shot positions. Instead, the 2010 survey was designed to acquire the nominal 2007 shot grid.

Standard OBS processing

Each OBS node can be processed completely independently from the other nodes (apart from the coldwater statics step). Processing can start as soon as the first nodes are picked up from the seafloor, and partially migrated images in the base and monitor survey can be identically processed using the following sequence: 1. De-signature to a common wavelet whose maximum frequency lies below the notch of the 2007 survey; 2. Random noise attenuation; 3. Shear wave leakage removal; 4. Hydrophone-to-geophone calibration; and 5. Up- and down-going wave separation.

4-D specific OBS processing

The GoM is subject to large and rapidly varying water temperature changes that impact the speed of sound in the water. These variations create so-called coldwater statics on the data. If not corrected, coldwater statics will contaminate the time/depth shifts between base and monitor, obscuring the link between these shifts and geomechanics. It is thus essential to account for this effect.

However, the task is complicated by the node internal clock drift, and methods have been used to improve the confidence of node timing corrections. To determine the coldwater statics, the first arrival was picked on the hydrophone data, and misfits with a ray-traced first arrival in a reference water velocity were computed. A least-square inversion of the misfits provided quality control (QC) and an update of both the node and shot locations as well as their static corrections. The node corrections and shot statics were applied once for the up-going energy and, after multiplication by a factor of three, for the down-going energy. This procedure was applied independently for the base and monitor surveys after setting the node depth for the two vintages to the same reference depth based on a bathymetry map measured by an AUV survey from 2008. To QC the statics, a mirror migration of the down-going wave energy for both vintages was performed. The image of the water bottom for the base and monitor was then picked.

With streamer data, 4-D binning and trace selection is an important step to get to an optimal 4-D response. This comes as a direct consequence of the poor sampling of the wavefield in a narrow-azimuth survey. With the richness of offsets and azimuths intrinsic to OBS data, the 4-D binning and trace selection is simply reduced to selecting the shots that are in both base and monitor shot outline for every repeated node. This ensures that the number of shots and the offset and azimuth distributions are the same for both vintages.

Mars OBS 4-D time lapse results

The depth shifts required to align the monitor to the base survey were derived on both the down-going and up-going images. The up-going wave-field travels only once through the water column and is hence not as sensitive to the coldwater statics as the down-going wavefield. The depth shifts derived from the up-going energy are thus more reliable, and linear discontinuity between the compacting basin and the nearly unproduced area can be clearly seen. The production-related effect from a single well that started production in 2008 also is better seen on the depth shifts measured from the up-going wave.

Both base and monitor up-going and down-going waves were migrated using a vertical transverse isotropy reverse time migration algorithm up to a maximum frequency of 45 Hz. After alignment of the base and monitor, the difference between the two was generated, and a normalized root mean square (NRMS) was computed. A value of NRMS of 6% in zones where no time-lapse signal is expected was achieved, even close to the Mars TLP, an area where streamer 4-D would be problematic. Clear 4-D signal has been confirmed to be related to water injection.

When comparing the amplitude map (left) of the H1 reservoir for the 2010 data and corresponding 4-D difference map (right) between the 2010 and the 2007 surveys, a clear 4-D signal can be seen near the downdip well where water coning is occurring. Note the clear correspondence between the 4-D signal and the depth contour.

Further study

Proper correction of the shot statics related to water temperature variations is essential if depth shifts are used to infer geomechanical effects. Further analysis (e.g. generating difference maps for every reservoir) and integration with production information as well as reservoir simulation is under way to fully harness all the valuable information contained in the data.

Acknowledgements

The authors wish to thank Shell International E&P, Shell Upstream Americas, and BP Exploration and Production Inc. for permission to publish this paper. The information in this article was originally presented at the 2011 annual meeting of the Society of Geophysicists and has been reprinted with the authors' permission. References available upon request.