Dynamic Time Warping for coda wave interferometry studies

Abstract:

Accurate time-shift estimation between arrivals in two seismic traces before and after a small velocity change is crucial for estimating the location and amplitude of the velocity change. Windowed crosscorrelation and trace stretching are two time-domain techniques commonly used to estimate local time shifts between multiply scattered coda signals. These methods can both fail when the induced changes in the scattered wavefield are not simple time shifts. Cycle skipping is an example of one such obstacle. A common approach to mitigate such problems is to choose only part of the coda to analyze. In the work presented here, we implement Dynamic Time Warping (DTW) to search for the time shift at each time sample that globally minimizes the misfit between two seismic traces. We show that DTW is considerably less susceptible to errors in time-shift estimates caused by cycle skipping or disappearance of coda phases due to changes in the physical scattering properties. Our approach provides a new tool to estimate small time shifts in coda and has wide application across many disciplines of seismic monitoring and imaging.