Local Minima in Seismic Waveform Tomography

Wednesday, April 24, 2002 - 4:00pm - 4:20pm
Keller 3-180
R. Gerhard Pratt (Queen's University)
Seismic traveltime tomography has proven to be an effective imaging tool in many applications. More recently, the technique of waveform tomography has emerged, in which we use wave-theoretical methods and the direct arrival waveform.

Traveltime methods, while reasonably robust, are known to have a reduced resolution that scales with the Fresnel zone, while the superior resolution of waveform methods scales with the wavelength (Williamson, 1991; Schuster, 1996). Waveform methods, however, are far more likely to fail due to a lack of robustness.

Resolution is usually estimated by examining the impulse response of a given algorithm: This procedure can be carried out analytically; numerical studies generally confirm the predictions (Williamson and Worthington, 1993). However, studies based on spatial delta functions fail to account for non-linear effects created by distortions of the wavefield, or by high order scattering. In this paper we show that non-linear effects can dramatically affect the performance of high resolution of waveform tomography. We use chequerboard models in which the anomaly sizes vary from the dominant wavelength to the approximate size of the first Fresnel zone.