A common task in seismic reflection imaging is to generate a structural image of the subsurface. This means that discontinuities in the recorded pre-stack data, the reflection events, have to be transformed into discontinuities of the medium properties, the reflectors, in the depth domain. The kinematic aspects of wave propagation provide the location of the subsurface reflectors. If the dynamic properties of wave propagation are also considered, the image additionally provides amplitudes that are related to the local reflection coefficients along these reflectors. A crucial problem of the imaging task is that the transformation from the pre-stack time domain to the depth domain requires the knowledge of the propagation velocities in the subsurface, at least at large scale: a smooth macro-velocity model is usually sufficient. However, such a model is, in general, not available a priori and has to be iteratively and interactively elaborated with significant effort.
In contrast to such model-based imaging methods, the imaging method presented in this thesis, the Common-Reflection-Surface (CRS) stack, is an entirely data-oriented approach that avoids the explicit parameterization of the depth model. Instead, the CRS stack directly makes use of the inherent redundancy in the pre-stack data and parameterizes the reflection events in the time domain. In this way, reflection energy stemming from one and the same reflector segment (or CRS) can be identified and summed: by means of coherence analysis in the pre-stack data, a second-order approximation of the kinematic reflection response of the CRS, the so-called CRS stacking operator, can be determined that fits best the actual reflection event. This approach is suited to simulate a zero-offset (ZO) section of high quality and high signal-to-noise ratio from the pre-stack data. Additionally, the parameters of the CRS stacking operators are available at any simulated ZO location. These kinematic CRS wavefield attributes are useful for a variety of applications, e. g., the estimation of the projected first Fresnel zone for ZO, the geometrical spreading factor, attribute-based time migration, and inversion, just to name a few.
In this thesis, I derive the CRS stacking operator for 2-D data acquisition based on the concepts of geometrical optics using object and image points. I introduce a new, extended strategy to determine the CRS wavefield attributes in an efficient way. This extended CRS stack strategy is able to handle more general situations compared to previously existing strategies: conflicting dip situations where two events intersect each other in the ZO section can now be handled such that each contributing event is detected and parameterized separately. A new attractive application of the wavefield attributes is introduced by merging concepts of Kirchhoff time migration and the CRS stack: by means of the approximate diffraction response of a hypothetical subsurface diffractor, the CRS stack results can be re-mapped to provide an attribute-based time migration result and, in principle, a conventional time migration velocity model. As an additional feature, I transfer the concept of operator tapering from Kirchhoff migration to the CRS stack approach.
I present an object-oriented implementation of the extended CRS stack strategy that includes several applications of the wavefield attributes. The implementation is designed to handle irregular acquisition geometries and uses a data format compatible to a common industry standard. I discuss the practical aspects of aperture sizes, tapering, and the implementation of the search strategies for the wavefield attributes.
This implementation is applied to three different marine data sets, one of them synthetic, the other two acquired offshore Chile and Costa Rica, respectively. The results for the real data examples are compared with the results obtained by means of conventional imaging methods, namely the normal moveout/dip moveout(NMO/DMO)/stack processing chain. On the one hand, this comparison reveals the superior ability of the CRS stack approach to image steep events and to detect and characterize reflection events in the pre-stack data, even in the case of poor data quality. On the other hand, the data examples also demonstrate the limitations of the CRS stack method: the method breaks down whenever the reflections can no longer be described with sufficient accuracy by means of an approximate operator of second-order or, even worse, the subsurface is so complex that no coherent reflection events can be found in the pre-stack data at all. The latter situation causes all usual imaging methods to fail, the former is problematic for most data-oriented methods.
For the different data examples, I present several useful applications of the CRS wavefield attributes: the identification of diffraction and reflection events, the estimation of the projected first Fresnel zone, attribute-based time migration, and the identification of multiples. Except the last one, such applications cannot be performed with the parameters obtained by conventional data-oriented imaging. For the deep-offshore data, I additionally discuss the results of post-stack depth migrations of the conventional result and the CRS result, respectively, as the time domain images do not fully reveal the actual improvement.
Logos Verlag, Berlin, Germany.
This excerpt appears with the implicit agreement of the copyright owner.