|TIME-LAPSE ACQUISITION WITH A DUAL-SENSOR STREAMER OVER A CONVENTIONAL BASELINE SURVEY|
A time-lapse 3-D project involves repeat 3-D seismic surveys over a producing hydrocarbon reservoir targeted towards identifying changes in the physical state of the reservoir arising from production. Such time-lapse surveys are used to optimize the planning of producer and injector well placement. The first survey is the baseline survey, and each successive survey is a monitor survey. Ideally, the acquisition geometry and hardware is repeated exactly from survey to survey, as are environmental conditions, such that any observable difference is derived entirely from physical changes in the reservoir state. Variations in acquisition geometry naturally invoke changes in wave propagation through the Earth, variations in target illumination, and variations in wavefield sampling, and will contribute errors to the time-lapse signal. Variations in acquisition hardware and/or environmental conditions will naturally invoke changes in the signal-to-noise content, signal fidelity, and the seismic wavelet, thus also contributing errors to the time-lapse signal.
For conventional marine acquisition employing towed streamers that record the pressure wavefield, the requirement to repeat the acquisition geometry means that all monitor surveys must be acquired at the same acquisition depth as the base survey. However, if a dual-sensor streamer is used this requirement can be relaxed. Dual-sensor streamer technology allows the wavefield to be separated into up- and down-going parts, which may then be independently redatumed to simulate acquisition at any recording depth. Thus the streamer may be towed at any depth, which usually means that we choose a deeper towing depth than would be used for conventional acquisition in order to take advantage of the quieter recording environment and increase the low-frequency signal-to-noise ratio. In order to validate this approach, a time-lapse experiment was conducted in the North Sea whereby dual-sensor streamer data were acquired over five adjacent lines that had been acquired earlier in the year using a conventional streamer. The data were then processed and analyzed using a conventional time-lapse processing sequence. Given that only a few months had elapsed between baseline and monitor surveys and there had been no production in the area, we expect to observe minimal differences between the two datasets. This paper describes the results of this time-lapse experiment.
|ACQUISITION AND PROCESSING|
The time-lapse repeatability trial was carried out in Quad 26 of the Norwegian North Sea. A 3-D survey was acquired using a conventional streamer that recorded the pressure wavefield at a depth of 8m in April 2009. Although this survey was not specifically designed as a baseline survey for a time-lapse experiment, five adjacent sail lines were selected as the “baseline” survey for the dual-sensor repeatability trial. These five sail lines were then reacquired using a dual-sensor streamer towed at a depth of 15m in June 2009. This acquisition represents the “monitor” survey. The source parameters were identical for the baseline and monitor surveys; 3090in3 source arrays towed at 6m depth in dual-source shooting mode. Both the baseline and monitor surveys were acquired with 6 x 5100m streamers at 100m streamer separation. Source positions were matched quite well in the monitor survey, though time-sharing constraints did not permit optimal feather matching. Nevertheless, the baseline survey geometry was repeated sufficiently closely to permit meaningful time-lapse analysis. The resulting fold of coverage is shown in Figure 1.
A dual-sensor streamer contains collocated sensors that measure the pressure and vertical component of particle velocity. These measurements can be combined to separate the wavefield into up- and down-going parts (Carlson et al., 2007). If we consider only the up-going wavefield, the bandwidth is improved compared to conventional total pressure field data due to the removal of the receiver ghost, which is clearly desirable. Although it is possible to remove the receiver ghost for total pressure field data in isolation, subject to certain assumptions about the sea surface state, the results tend to suffer from poor signal-to-noise in the vicinity of the notch frequencies characteristic of the receiver ghost (Tabti et al., 2009). Therefore, we prefer to process the dual-sensor data acquired at 15m depth to determine the total pressure field at the acquisition depth of the conventional streamer (8m). This procedure is more robust and relies on fewer assumptions than removing the ghost from the conventional data, and is therefore likely to produce a superior time-lapse result.
Figure 1: Mid-offset fold coverage maps for the baseline (left) and monitor (right) surveys. The result in the middle is the matching baseline and monitor traces after 4D binning. The bold rectangle in the middle is the full fold area used for various analyses.
After wavefield separation, the up- and down-going wavefields were independently redatumed from 15m to 8m depth, then summed to simulate the total pressure field recorded by the conventional cable. The wavefield separation and redatuming approach correctly handles the amplitudes for all emergence angles and takes account of deviations from the nominal recording depth (Söllner et al., 2008). The redatuming procedure requires that the up- and down-going wavefields are treated differently: when redatuming from a deep to a shallow depth, the up-going wavefield must be propagated forwards in time whilst the down-going wavefield must be propagated backwards in time. Hence, wavefield separation is a necessary prerequisite for this matching operation, which is straightforward for dual-sensor data but very difficult for conventional pressure data, especially in the vicinity of the receiver ghost notches. This limitation means that it is not practical to tow a conventional cable at a greater depth than that used for the baseline survey in order to take advantage of the quieter noise regime.
After reconstructing the total pressure field at 8m acquisition depth, a deterministic matching filter was applied to correct for the differences in the instrument response of the recording filters used for the dual-sensor and conventional streamers. Both the baseline and monitor datasets were then taken through a state-of-the-art time-lapse processing sequence, including tidal static correction, amplitude-preserving multiple removal, 4-D binning, 3-D wavefield regularization, pre-stack time migration, and post-migration global matching with a long design window. The 4-D binning step was used to replicate coverage holes in each of the two surveys. Detailed QC was pursued at each step in the time-lapse processing flow. The QC attributes used include Normalized RMS Difference (NRMSD) in amplitude, Normalized Cross-Correlation, Time shift, and Phase Rotation.
Figure 2 summarizes the progressive improvement in NRMSD through the processing sequence for a window centered on a typical target of interest in this region. The final NRMSD result was 11%, a result that is fully compatible with industry repeatability requirements. Figure 3 demonstrates that there is negligible residual energy remaining after differencing the final images for monitor and baseline surveys. Figure 4shows an excellent match in all details of the amplitude spectrum within the target window, including at 50Hz (arrowed) where the majority of the energy for the monitor survey is derived from the particle velocity sensor due to a notch in the hydrophone spectrum at this frequency. This result verifies that the dual-sensor streamer data is fundamentally sound and that all the processing steps are robust and amplitude preserving since such a small difference could not be obtained if any of these conditions were not met.
Note that in order to obtain these results we have re-introduced a receiver ghost in the dual-sensor streamer data. This procedure is necessary to permit time-lapse comparison with the conventional streamer data, but an undesirable effect is to limit the bandwidth which is detrimental to any subsequent inversion for reservoir attributes. For optimal time-lapse processing, it would be preferable to compare data without the receiver ghost, which could be achieved by using a dual-sensor streamer for both base and monitor surveys. Figure 5 illustrates the bandwidth improvements that are obtained when the receiver ghost is removed.
A time-lapse experiment was carried out in the North Sea whereby a “monitor” survey was acquired using dual-sensor streamers at 15m depth over a “baseline” survey that had been acquired a few months earlier using conventional streamers at 8m depth. Wavefield separation was performed for the dual-sensor streamer data and the up- and down-going pressure fields independently redatumed from 15m to 8m depth and summed to simulate the total pressure field acquired using the conventional streamer. The data were then processed using a standard time-lapse processing flow. The differences between the two datasets were found to be minimal as expected given that there was no production in the survey area during the months between the two surveys. This result demonstrates that a dual-sensor streamer can be used to acquire time-lapse monitor surveys over baseline surveys acquired using conventional streamers with high repeatability, and further serves to validate the integrity of the wavefield separation and redatuming steps applied to the dual-sensor streamer data. However, the full benefit of the increased bandwidth for time-lapse surveys can only be realised when both baseline and monitor surveys are acquired with dual-sensor streamer technology.
We thank the crews of the Ocean Explorer and Atlantic Explorer who acquired the data used in this study, Berit Mattson and Zbigniew Greplowski for their contributions to processing the data, and also PGS for permission to publish this paper.
Carlson, D., Long, A., Söllner, W., Tabti, H., Tenghamn, R. and Lunde, N., 2007, Increased resolution and penetration from a towed dual-sensor streamer, First Break, 25(12), 71-77.
Söllner, W., Day, A. and Tabti, H., 2008, Space-frequency domain processing of irregular dual-sensor towed streamer data, 78th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1078-1082.
Tabti, H., Day, A., Schade, T., Lesnes, M., and Høy, T., 2009, Conventional versus dual-sensor streamer de-ghosting: a case study from a Haltenbanken survey, First Break, 27(8), 101-108.