Motivation to include velocity sensors in a seismic streamer
A pressure sensor in a towed streamer always records two wavefields that interfere with each other. The two wavefields are the up-going pressure wavefield propagating directly to the pressure sensor from the earth below, and the down-going pressure wavefield reflected downwards from the free (sea) surface immediately above the streamer. Thus, every recorded reflection wavelet from conventional marine streamers is accompanied by a “ghost” reflection from the ocean’s surface. The reflection wavelet is undesirably elongated, reducing temporal resolution. The consequence, as illustrated in in the left side of Figure 1, is that a series of receiver ghost notches are introduced into the frequency spectra. There has historically been a forced trade-off between towing shallow to record high frequency (but noisy) data at the cost of reduced lower frequencies, or towing deep to record low frequencies at the cost of reduced higher frequencies.
It has long been understood that by recording seismic data from collocated hydrophones and velocity sensors, and by properly combining their signals, ghost reflections can be cancelled. The technology described here represents the first successful implementation of a dual-sensor towed marine streamer system (Tenghamn et al., 2007). An efficient architecture uses densely sampled groups of collocated pressure and velocity sensors in a low-noise solid-fill streamer with distributed electronics and Ethernet telemetry. For each type of measurement, the sea-surface reflections (ghosts) impose a filter on the data (the right side of Figure 1). In contrast to pressure sensors, however, velocity sensors are directional, so the down-going velocity wavefield is measured as having equal polarity to the up-going velocity wavefield. Consequently, as observed in the right side of Figure 1, receiver ghost notches for the pressure and velocity sensors are separated in the frequency domain by integer multiples of, but are offset by from each other
(Vw is the velocity of sound in water, and d is the streamer depth). Appropriate processing of the measured pressure and velocity data will completely cancel the amplitude of the ghost event (down-going wavefield from the sea surface) trailing each primary event, and the notches in the amplitude spectra will be removed. This is the case for all angles of incidence, i.e. for all source-receiver offsets.
Figure 2 is a simple synthetic example that conceptually demonstrates the summation of zero-offset stacks for pressure and velocity data from a dual-sensor streamer. The receiver ghost that complicates interpretation of relatively thick intervals is removed by summation, and a clear image results. Amplitude polarity on the right side of Figure 2 corresponds to the impedance contrast across each interface. Note that the simple summation used in Figure 2 is not applicable to finite offsets or dipping data, so a fully angle-dependent and frequencydependent processing solution is used in practice for real dual-sensor streamer data. Overall, the system described here provides a combination of geophysical and operational flexibility.
Acquisition and processing
Acquisition uses conventional streamer deployment, towing, and retrieval. Towing depth is typically 15 m, which has many operational benefits, and invokes no penalty to higher frequency content during processing. In fact, the pressure and velocity data are better behaved than would be the case for shallower towing, enabling accurate signal conditioning during processing.
Pre-processing of the dual-sensor data is relatively straightforward to yield the deghosted pressure and velocity wavefields. First, the impulse response of the velocity sensor (which has a non-flat spectrum) is matched to the flat zero phase hydrophone spectrum. This step takes into account the difference in the sensitivities between the two sensors. Signal conditioning is then applied to the velocity sensor data over the range of 0 - 20 Hz, constrained by the pressure sensor data using an exact mathematical formulation (Fokkema and van den Berg, 1993). This is desirable because velocity sensors record strong low-frequency mechanical noise. The conditioning of the low-end frequency range is mathematically robust for scattered wavefields (the direct arrivals are removed). After being separated into up-going and down-going components during data processing (Fokkema and van den Berg, 1993), both pressure data and particle velocity data can be extrapolated to any desired recording depth. The up-going pressure wavefield represents the “deghosted” pressure result, and is completely free of the receiver-side ghost. Note that the wavefield separation processing inherently incorporates all angle-dependent effects, and is not a simple “summation” process. Both the low and high frequency content is significantly boosted after deghosting. Thereafter, the data are passed on to a “conventional” processing flow, modified of course to exploit and preserve the improved frequency and signal-to-noise content of the signal. The data examples shown in this paper correspond to 2D acquisition, and consequently a 2D assumption has been made in data processing, however the extension to 3D processing is straightforward.
Figures 3 to 6 present dual-sensor data examples from two different survey locations in the North Sea. In both cases a conventional hydrophone-only streamer was towed at a depth of 8 m, and a dual-sensor streamer was towed below the conventional streamer at a depth of 15 m. This allowed a spatial correlation and verification between the two recorded datasets, but is not intended to be standard practice during production surveys. Data processing is deliberately simplistic, in an effort to highlight the fundamental benefits of dual-sensor streamer acquisition and processing. Figure 3 presents a shallow data window comparison. Note the complete removal of the receiver-side ghost on the right side of Figure 3.
Figure 4 is a processing QC/verification that compares a large data window and superimposed amplitude spectra for the following:
- Conventional pressure-only streamer towed at 8 m depth, and
Dual-sensor streamer towed directly below at 15 m depth; low frequency velocity data conditioning over 0 - 20 Hz, wave-field separation into up-going and down-going pressure wave-fields, extrapolation of both (separated) wave-fields to 8 m depth, and summation to yield the total pressure wave-field at 8 m depth.
- The results in Figure 4 show that the reconstructed total pressure wavefield from the dual-sensor data and the recorded total pressure wavefield are essentially identical (as expected), including the frequency spectra at very low frequencies. This (consistent) observation supports the accuracy of the low frequency velocity data conditioning, and the integrity of the other pre-processing steps.
Figure 5 demonstrates the benefit of deghosting to the amplitude spectra, both in terms of boosting low and higher frequencies normally lost because of the receiver ghost, and in terms of also boosting the low frequency content because of towing deep at 15 m. In a dual-sensor streamer, towing deep invokes no penalty to the higher frequencies. Furthermore, signal-to-noise content on both the pressure and velocity wavefields is excellent, as is the deghosted up-going pressure wavefield derived during data processing.
Figure 6 presents a data window over the first 8 s TWT below the water bottom (record length was 10 s). Note the improvement in deeper data event amplitude and character with the dual-sensor data. Hence, the data resolution and character is improved for all depths and locations
A further advantage arising from the ability to separate wavefields is that an advanced implementation of Surface Related Multiple Elimination (SRME) is possible (Söllner et al., 2007). Multiple prediction is based on the up-going pressure wavefield and the down-going velocity wavefield extrapolated to the source level so that the kinematics of surface related multiples are accurately represented. The key advantage of using the down-going velocity wavefield is that any variations in the sea surface level and reflection coefficient are implicitly included in the prediction (in contrast to conventional SRME). In addition, the use of a particle velocity field automatically incorporates a necessary angle-dependent scaling into the prediction process. Figure 7 presents a comparison of 2D data with conventional vs. dual-sensor acquisition and SRME processing. Note that 3D acquisition and processing are required to completely remove the residual multiple energy still evident on the right side of Figure 7.
A new dual-sensor streamer records both pressure and the vertical component of the particle velocity with collocated sensors towed at the same depth. Separation into up-going and down-going pressure and velocity wavefields is straightforward. Thus, the receiver ghost can be completely removed from the pressure wavefield, increasing data resolution. The related seismic data contains greater frequency bandwidth and greater signal-to-noise content than can be obtained using conventional streamers. Deep streamer towing facilitated by the dual-sensor technology increases the operational weather window, reduces noise, and increases signal penetration. Deeper events are stronger and more continuous. The dual-sensor streamer technology described here creates several new opportunities for advanced multiple removal, seismic inversion, and data interpretation.
Fokkema, J.T., and van den Berg, P.M., 1993, Seismic applications of acoustic reciprocity: Elsevier, Amsterdam, 350 p.
Söllner, W., Brox, E., Widmaier, M., and Vaage, S., 2007, Surface-related multiple suppression in dual-sensor towed-streamer data: Annual Meeting Expanded Abstracts, SEG, 2540-2544.
Tenghamn, R., Vaage, S., and Borresen, C., 2007, A dual-sensor, towed marine streamer; Its viable implementation and initial results: Annual Meeting Expanded Abstracts, SEG, 989-993.