Seeing the invisible: The next generation in seismic depth imaging

Abstract

The recent advent of fast, inexpensive personal computers (PCs) is revolutionizing modern seismic depth imaging. Prices for supercomputer-style computation, graphical display, and data storage, halve every 18 months. Modern networking technologies couple inexpensive PCs into powerful clusters that easily solve problems once thought intractable on far more expensive machines. It is now feasible to utilize accurate, computationally intensive algorithms to produce subsurface images that are vastly superior to those produced with current Kirchhoff technology. Graphic capabilities provide a new interpretative processing environment for improved velocity analysis and earth model estimation. This article discusses these new advances and the technologies they make practical and argues that our current focus on identification and estimation of the depth-interval-velocity model must evolve to identification and estimation of the elastic-earth model . The emphasis is on comparisons between real and synthetic elastic data rather than on specific and detailed algorithms, but the conclusions and directions are clear. Exploration imaging technology, at commodity prices, can provide the necessary tools to enhance subsurface visibility and accurately position exploration targets. Physically unrealistic assumptions underlie many seismic imaging techniques. They are made to assure some measure of computational practicality. A flat, two-dimensional, constant-velocity earth provides an environment for tremendous algorithmic efficiencies. NMO correction, stacking, DMO, and the remarkable Fourier-domain Stolt ( f-k ) algorithm are direct consequences of such thinking. Under the computational constraints of the era, acceptable subsurface images were produced in what must have seemed to be record-breaking time. Membership in the Flat-Earth Society makes velocity estimation trivial but inaccurate. Poststack time domain Kirchhoff diffraction and finite-difference techniques handle variable velocity fields on modest computational platforms, but their straight-ray limitations mean that parts of the subsurface (steep dips, subsalt reflectors) are almost always invisible. Relaxation of the constant-velocity assumption for poststack zero-offset data improves subsurface visibility at modest computational …