Seismic Waveform Inversion: Practical aspects and Application to field seismic data

Abstract

Seismic waveform inversion involves estimation of elastic properties by iterative fitting of seismic reflection data with synthetic seismograms. Like many other geophysical techniques, the inversion is applied to common mid point gathers assuming locally one-dimensional plane layered earth model. The objective of this paper is two fold. First, we address the critical issues of robustness, stability and uncertainty in which both global and local optimization methods are developed. An adaptive regularization scheme is developed that is used both in local optimization using a truncated Gauss-Newton algorithm and a global optimization based on very fast simulated annealing (VFSA). The 'regularization weight' is analogous to 'model temperature' of VFSA. The VFSA and local algorithm are used to take advantage of the two approaches. We demonstrate that the adaptive regularization is essential to improve on stability and robustness. The algorithm is inherently parallel and we carry out computation in a cluster of personal computers where load balancing is done carefully. Second, we apply our techniques to two sets of data, one from the Gulf of Thailand and the other from offshore Oregon. We are able to obtain very clear indication of free gas zones from the impedance and Poisson's ratio maps. These are then transformed into maps of saturation and porosity using a rock-physics model for use in reservoir characterization.