Estimation of anisotropy parameters in application to Blackfoot seismic data

Anisotropy parameters are obtained by anisotropic velocity analysis performed on Blackfoot P-wave reflection-seismic data, in combination with sonic-log data. First, the line was processed with ProMAX seismic data-processing software using a sequence of conventional algorithms without taking anisotropy into account, in order to choose proper analysis points and horizons with confidence. Then, at selected analysis points, we correlated well logs, synthetic data, Blackfoot seismic data, and interpreted formation tops, to obtain the vertical interval velocities. Four seismic intervals were chosen for estimation of Thomsen’s anisotropy parameters. The results show that estimated values of ε and δ seem reasonable only if the time intervals are larger than about 200 ms. When the lower three layers are combined as one target layer, it exhibits relatively high values of ε and δ of about 0.37 and 0.20, respectively. INTRODUCTION We have demonstrated the viability of the joint inversion of P-wave reflection traveltimes and well data to give Thomsen’s anisotropy parameters, ε and δ, by applying it to synthetic data (Xiao et al., 2004). However, the practical application of this approach to real data is a more challenging task. Firstly, application of the algorithm requires the recovery of nonhyperbolic moveouts from long-spread CMP gathers. Secondly, the semblance search at high incidence angles is also hindered by phase shifts in postcritical reflections. Thirdly, we have to consider the influence of noise on semblance. This paper describes an application of this inversion procedure to some real data from Blackfoot. BLACKFOOT SEISMIC DATA PROCESSING Processing of seismic data for anisotropy parameter estimation is a challenge since there is a precarious balance between improvements to the signal-to-noise ratio and distortion of the curvature of the reflection traveltime curves. Ideally, processing should improve the continuity and resolution of events to facilitate horizon identification and allow traveltime picks to the largest offset range possible (nonhyperbolic moveout is only evident in the far offset). We are only interested in traveltime moveout information; conservation of frequency content and amplitudes is less important. Hence, the employed processing sequence starts with a mute, AGC, and bandpass filter. Two f-k filters are then applied in cascade to reduce the linear noise on the far offsets such that the picks can be extended to greater offsets. Then a predictive deconvolution filter is designed to further reduce the linear noise and improve the lateral continuity of reflectors. A second bandpass filter is applied to remove high-frequency noise introduced by the predictive deconvolution filter. Finally, adjacent CMPs are combined and similar offsets stacked. Xiao, Bancroft, and Brown 2 CREWES Research Report — Volume 17 (2005) These steps improve the continuity of reflections significantly. An extensive series of tests has been carried out to guarantee that the signal-to-noise ratio was improved and that events could be picked to large offsets without affecting the curvature of the reflections. Optimum input data for the application of this method is raw data with static corrections applied, but before top mutes are applied to remove any linear noise. Figure 1 is the CDP gather for estimating effective coefficients. FIG. 1. The CDP gather for estimating effective coefficients. Picking Figure 2 shows a seismic line from south-central Alberta acquired by the CREWES Project in 1997. The line was processed with ProMAX seismic data-processing software using a sequence of conventional algorithms without taking anisotropy into account. The processing sequence is outlined in the following list: (1) SEG-Y seismic data input (2) Preprocessing: Setup of field geometry Automatic gain control, bandpass filter Editing (kill bad traces or reversed traces) Picking first breaks for weathering statics calculation Elevation correction Weathering statics calculation (with GLI3D) (3) Surface-consistent deconvolution and weathering statics correction (4) CMP sorting and velocity analysis (5) First residual statics correction and velocity analysis Estimation of anisotropy parameters in VTI media CREWES Research Report — Volume 17 (2005) 3 (6) Second residual statics correction and velocity analysis (7) NMO correction, muting, and stacking (8) Deconvolution (9) Time-variant spectral whitening and filtering (10) CDP trim statics (11) Finite-difference migration (12) SEG-Y output FIG. 2. Post-stack migration using a sequence of conventional algorithms. We selected five different horizons (Figure 3) for analysis (free surface t = 0 as a horizon ), choosing those where t-x picks could be made with confidence. Conventional velocity analysis was carried out along the line before the horizons were selected in order to avoid picking multiples, and as a quality control on inversion results. The picking was done on the CMP with 4 km offset to each side. The picking was difficult in the target zone due to deterioration in the data quality (Figure 1). INTERVAL VERTICAL VELOCITY FROM SONIC LOG Although anisotropic moveout (AMO) analysis can provide information about horizontal velocity, conventional moveout analysis using either NMO or AMO equations cannot provide information about vertical velocity (Yang et al., 2002). How to obtain vertical and horizontal velocities is an important task in many applications such as AVO inversion, anisotropic imaging, and pore-pressure prediction. Vertical velocity is important information for the success of AVO inversion, anisotropic imaging, and porepressure prediction (Wright, 1987; Banik et al., 2003). Xiao, Bancroft, and Brown 4 CREWES Research Report — Volume 17 (2005) Figure 3(a) shows the correlation of well logs, synthetic data and Blackfoot seismic data as well as interpreted formation tops. The four seismic interfaces shown in Figure 3(b) are chosen for purposes of estimating Thomsen’s anisotropy parameters.