Diffraction tomography for crosswell imaging of nearly layered media

Abstract

Diffraction tomography is a high‐resolution imaging technique applicable to the mapping of formation velocities away from the borehole, achieving a spatial resolution of better than one acoustic wavelength when used to image synthetic model data. However, traditional filtered back‐propagation diffraction tomography algorithms are based on weak‐scattering and constant‐background velocity assumptions, which limits their applicability to models of realistic structural complexity. Results are obtained using a new, computationally efficient single‐mode (P‐wave) diffraction tomography algorithm that is applicable to models, including geologically realistic ones, whose strongest variations can be approximated as a set of horizontal layers. The algorithm starts with a layered model of the subsurface velocity structure, which may be constructed from well logs or by using traveltime tomography. Application of layered diffraction tomography updates this model to reveal 2-D structures such as faults, pinch‐outs, and dips. An overview of the layered diffraction tomography algorithm and information on its computational implementation are presented. Layered diffraction tomography can, given an accurate initial model, successfully image complex structures for which filtered back‐propagation fails, with a spatial resolution on the order of one‐third wavelength. Synthetic crosswell data are used to construct images of models ranging from an isolated simple target to geologically realistic structures containing faults, erosional surfaces, and dipping beds. The initial layered model must represent the actual subsurface structure with as much fidelity as possible; it is important to use all available a priori information in the construction of this model. The enhanced spatial resolution provided by diffraction tomography does not require the use of high‐angle reflections; in the presence of noise, image quality and resolution can be largely maintained by using only a short time window of (the high S/N) data immediately following the first break that is employed for traveltime imaging.