Stephen K. Chiu

Tomographic determination of three-dimensional seismic velocity structure using well logs, vertical seismic profiles, and surface seismic data

A tomographic technique (traveltime inversion) has been developed to obtain a two‐ or three‐dimensional velocity structure of the subsurface from well logs, vertical seismic profiles (VSP), and surface seismic measurements. The earth was modeled by continuous curved interfaces (polynomial or sinusoidal series), separating regions of constant velocity or transversely isotropic velocity. Ray tracing for each seismic source‐receiver pair was performed by solving a system of nonlinear equations which satisfy the generalized Snell’s law. Surface‐to‐borehole and surface‐to‐surface rays were included. A damped least‐squares formulation provided the updating of the earth model by minimizing the difference between the traveltimes picked from the real data and calculated traveltimes. Synthetic results indicated the following conclusions. For noise‐free cases, the inversion converged closely from the initial guess to the true model for either surface or VSP data. Adding random noise to the observations and performin...

Attenuation of Coherent Noise Using Localized-adaptive Eigenimage Filter

A new method uses eigenimages to construct a coherent noise model in a localized time-space window and performs the noise attenuation by adaptively subtracting the noise model from the input data. Advantages to this method include minimum spatial-amplitude smearing, effective attenuation on various types of coherent noise such as ground roll, air waves and near-surface scattered energy as well as handling both the aliased and non-aliased noise quite well. This new nonlinear filter significantly outperforms conventional techniques. We demonstrate the performance of this local-nonlinear filter with real data examples.

High Fidelity Vibratory Seismic (HFVS): Robust Inversion Using Generalized Inverse

One of key components of HFVS technology requires separating multiple sweeps and multi-vibrator gathers into a single source gather through a matrix inversion that involves solving a system of equations. Fast directequation solvers, such as LU decomposition, are often used to handle large volumes of 3D prestack data. However, the drawback of the fast direct-equation solvers does not reveal the uniqueness of the inverse problem and may fail to produce a satisfactory solution if the matrix is ill conditioned. Another better alternative is to use single value decomposition (SVD) to obtain a more robust leastsquares solution. But its high computational cost often limits its application to a small or moderate size of data volume. This paper shows that the use of SVD is well suited in the HFVS technology: requires minimum computational cost; provides diagnostic tools to analyze the uniqueness of the inverse problem; and produces a better source separation when the vibrator-sweep matrix is ill conditioned and a comparable result when the vibratorsweep matrix is relatively well conditioned. The synthetic and real data examples further illustrate that SVD is a preferable equation solver to be used in the HFVS system for the similar computational cost as a fast direct-equation solver.

F‐xy noise attenuation via Multichannel Singular Spectrum Analysis in randomized domain

The attenuation of coherent and random noise still poses technical challenges in seismic data processing, especially in an onshore environment. Multichannel Singular Spectrum Analysis (MSSA) is an existing and effective technique for random noise reduction. By incorporating a randomized operator into MSSA, this modification creates a new and powerful filtering method that can attenuate both coherent and random noise simultaneously. Application of this method on 3D synthetic data and cross-spread 3D prestack field data demonstrates its effectiveness in suppressing both types of noise as well as preserving primary signals.

Effect of near-surface anisotropy on a deep anisotropic target layer

Summary Near-surface anisotropy can distort P-wave traveltime and amplitude analysis from deep target layers. When the target layer is azimuthally anisotropic, the traveltime/velocity variation with azimuth (VVAZ) or amplitude variation with azimuth (AVAZ) from the target layer may show anomalous behavior due to the influence of the near-surface anisotropy. This study uses two synthetic cases to analyze the effect of near-surface anisotropy on a deep anisotropic target. Our results suggest that the traveltime data (or VVAZ signals) from the target layers can be distorted significantly due to the presence of near-surface anisotropy; but the near-surface anisotropy influence may be negligible on the AVAZ signals from the deep target layer.

Seismic Data Compression And Regularization Via Wave Packets

Seismic data compression and regularization are crucial, faced with large data volumes and incomplete data measurements in the oil industry. In this research, we present a comprehensive method of utilizing wave packets to perform seismic data compression and regularization. Wave packets, which can be regarded as symmetric “curvelets”, are viewed as optimally localized plane waves. With its intrinsic multi-scale and multiazimuth properties, we show that wave packet achieves a high compression rate for a sparse representation of seismic data volume. Besides, we demonstrate that via an iterative thresholding algorithm, irregular data can be regularized in wave packet domain without creating artifacts or losing original dipdependent information. Compared with regularization methods based on τ − p transform and wave atom transform, wave packet achieves the best result, demonstrated by some numerical examples.

A Decade of Field Trials in the Panhandle of Oklahoma and the Development of ZenSeis

Summary ConocoPhillips has spent the last decade in the field developing a high production Vibroseis acquisition technique based upon the concepts of point source, point receiver acquisition combined with high fold and simultaneous sourcing. This technology is the result of many field experiments and production 3D’s. The methodology has now been used to acquire over one thousand square miles of very high fold 3D seismic data. The purpose of our paper is to document some of the field experiments and touch upon the lessons and conclusions learned in the field. This paper will also set the stage for additional papers that will deal with the analysis of the test results and the production 3D’s that were acquired with this new method.

Vibroseis Source Signature Uncertainty and its Impact on Simultaneous Sourcing

Knowledge of the source wavelet in a controlled source seismic experiment significantly improves our ability to extract information from the resulting seismic data. The radiated source signature in Vibroseis field experiments is found to deviate from the pilot sweep and the ground force estimate (GFE) signal put out by the controller, especially at higher frequencies. The Vibroseis source signature uncertainty is a problem for simultaneous sweeping techniques that require reliable phase control and an accurate GFE in order to separate simultaneous sweeps.

Altering offsets and azimuths—in a computer

For years, seismic processing geophysicists used dip moveout (DMO) to alter the acquisition geometry. Our industry has found many uses for DMO because it converts the timing of nonzero-offset data to the timing of zero-offset data. More recently, theoreticians created data mapping as a generalization of the principle behind DMO. Data mapping converts data obtained at an observed offset and azimuth to data at a new offset and/or azimuth. The precise derivation of the data mapping resides in an arduous solution of the wave equation. Thus, data mapping transformations may appear magical. This article provides a simple, geometric understanding of the data mapping transformation. Before turning to the more general 3-D case, we first present the 2-D case. For the 2-D case, data mapping transforms data obtained at one offset distance into data “observed” at a second offset distance. To understand this procedure, we will use one principle, one requirement, and two assumptions. The principle is linear superposition and the requirement is physical invariance. For conceptual convenience, we assume reflection coefficients do not change with offset, and we also assume a constant velocity earth. The principle of linear superposition simplifies our task. As shown in Figure 1, it allows us to construct any input data from a linear superposition of spikes or impulses. In addition, data-mapped output is a linear superposition of data-mapped spike responses. Consequently, we need to understand only the data-mapping operation on a single spike at an arbitrary location. Figure 1. Linear superposition simplifies data mapping. The physical invariance restriction is obvious—the earth does not change as a result of altering the acquisition offset. Likewise, our image of that earth should not change. Depth migration should image the same subsurface using either the original data or the offset-transformed data. The constant velocity assumption …

Validating land data quality of simultaneous multiple vibrator acquisition

Summary In 2007, ConocoPhillips conducted field experiments designed to evaluate the data quality of multi-offset VSPs acquired by a single vibrator and simultaneous multiple vibrators. To check the repeatability of vibrator sources, we recorded 8 repeated sweeps at the same source location for both acquisitions. The data quality is consistent from sweep to sweep at the same source location showing good repeatability of vibrator sources. Inverting 8 repeated sweeps simultaneously by a least-squares approach produces a solution that is very comparable to an average solution derived from inverting each sweep separately. In some cases, the least-squares solution tends to handle the ambient noise better and gives a slightly better solution than the average solution. The analyses of downgoing and upgoing VSPs demonstrates that simultaneous multiple vibrator acquisition yields equivalent data quality when compared with a single vibrator and cross-talk artifacts generated by simultaneous multiple vibrators are minimal in this case.