CREWES, in conjunction with Husky Energy, Geokinetics, INOVA and Nanometrics, conducted a low-frequency 2D seismic experiment near Hussar, Alberta, Canada, in September of 2011. The purpose of the experiment was to study acquisition of low-frequency data in order to improve inversion results. Sources included three different Vibroseis units, and dynamite. Receivers on the ground were ION-sensor SM-7 10 Hz 3C geophones at 10 m station spacing, VectorSeis 3-C accelerometers at 10 m spacing, Sunfull 4.5 Hz 1C geophones at 20 m spacing, a partial line of SM-24 10 Hz high-sensitivity geophones at 20 m spacing, and Nanometrics compact broadband seismometers at 200 m spacing. Total receiver line length was 4.5 kilometers. On the last day of acquisition, a magnitude 6.3 earthquake occurred offshore Vancouver Island, British Columbia, Canada, approximately 1050 kilometers from the test line. The predominant frequency of earthquake arrivals was about 0.4 Hz, which is well out of the frequency range of 4.5 and 10 Hz geophones. However, the earthquake was recorded by all sensors that were part of the low-frequency experiment, and after correcting the data for geophone response, it is clear that data less than 1 Hz can be recorded on these geophones, for a sufficiently energetic source.
Many seismic datasets are recorded over geologic structures where lateral changes in the physical properties of the stratigraphic layers vary smoothly. For these situations, depth migration algorithms are not required and time migration imaging is known to provide a similar outcome and is more economic. In this paper, we discuss the implementation of the Full Waveform Inversion (FWI) algorithms for velocity inversion using Common Scatter Point (CSP) gathers. Since the formation of the CSP gathers are based on the Pre-Stack Kirchhoff Time Migration (PSTM), we reduce the computational effort commonly associated with depth migration.
We present a linear prestack amplitude inversion of PP data, collected through physical modeling, for the Thomsen anisotropy parameters (ε, δ, and γ) of a simulated fractured medium. 3D physically-modeled PP data were acquired along several azimuths over a phenolic layer using the Physical Seismic Modeling Facility at the University of Calgary. The PP amplitudes picked from the reflection off the top of the fractured layer for several azimuths were used as input for the inversion. A linearized PP reflection coefficient approximation for an HTI (horizontal transverse isotropy) medium was used to facilitate the least-squares AVAZ inversion. Some constraints on the vertical velocities and density were also incorporated in the inversion process. The results for all three anisotropy parameters from AVAZ inversion compared very favourably to those obtained previously by a traveltime inversion. This result makes it possible to compute the shear-wave splitting parameter, γ, (historically determined from shearwave data) directly related to fracture density from a quantitative analysis of the PP data.
A study was conducted to investigate the large variations in the hydraulic fracture response of a tight gas reservoir. The variability is hypothesized to be the result of faulting in the area that alters the mechanical properties of the rock mass. Laboratory experiments demonstrate the formation of aligned microcracks throughout the deformation process. Therefore, a penny-shaped crack model was used to investigate the associated properties of media containing aligned cracks that represent a pre-rupture fault. The reflectivity response for the detection of pre-rupture faults is discussed in addition to the presentation of an effective stress model for a medium containing aligned penny-shaped cracks. Using these results, the observed variations in the hydraulic fracture behavior can be understood from the response of the aligned cracks to a uniform normal traction applied by a pore fluid.
A seismic physical model experiment has been conducted to acquire multi-offset multi-azimuth P-wave 3D seismic data, and to verify the suitability of physically-modeled data for AVAZ (amplitude variation with azimuth) analysis. Our model consisted of an azimuthally anisotropic layer, phenolic layer simulating a vertically fractured medium, overlain by two isotropic layers with the top most layer being water. The amplitudes reflected from the top of the fractured layer have been picked from the primary reflection; acquisition was designed to avoid the overlapping of the primary and ghost events. The picked reflection amplitudes required corrections to make them suitable for an AVAZ study. In addition to amplitude corrections used for seismic field data, a directivity correction specific to the physical model transducers was needed. The corrected amplitudes from different azimuths showed a clear azimuthal variation caused by the fractured layer, and agreed with amplitudes predicted theoretically.
Knowledge of Q is desirable for improving seismic resolution through inverse Q filtering, facilitating amplitude analysis and seismic interpretation. However, the question of reliable Q estimation remains, especially in case of unfavorable signal-to-noise ratio (SNR). In addition, estimating Q from VSP data or even reflection data in presence of moderate noise with sufficient accuracy is still very challenging. To address this problem, a match-filter method for Q estimation is proposed and evaluated using synthetic 1D, 2D data and field data in this paper. Given two narrow time windows as might be used in the spectral-ratio method, we compute minimum phase equivalent wavelets for each window and then, by direct search over a broad Q range, we find the optimal forward Q filter that best matches the shallow wavelet to the deeper one. Testing results show that the proposed method is, compared to the spectral-ratio method, more robust to noise and more suitable for the Q estimation from reflection data, and has the potential to indentify a localized low Q zone of the subsurface, which can be used as a gas indicator.
We design a suite of surface-consistent matching filters for processing time-lapse seismic data in a surface-consistent manner. By matching filter we mean a convolutional filter that minimizes the sum-squared difference between two signals. Such filters are sometimes called shaping filters. The frequency-domain surface-consistent design equations are similar to those for surface-consistent deconvolution except that the data term is the spectral ratio of two surveys. We compute the spectral ratio in the time domain by first designing trace-sequential, least-squares matching filters, then Fourier transforming them. A subsequent least-squares solution then factors the trace-sequential matching filters into four surface-consistent operators: source, receiver, offset, and midpoint. We present a synthetic time-lapse example with nonrepeatable acquisition parameters and near-surface and subsurface model variability. Our matching filter algorithm significantly reduces the nonrepeatability often observed in time-lapse data sets.