Near‐surface imaging with traveltime and waveform inversion

Abstract

Due to acquisition related noise and ground roll only a limited time window of seismic data is available for depth imaging through conventional processing. This time window excludes coda near the direct arrivals and far offset refractions as they cannot be handled by the conventional processing methods. Moreover due to NMO stretch and mute, shallow reflection events have limited offsets and do not stack as well as deeper reflection events. As a result, in terrains with rough topography and rapidly varying near surface conditions, near surface images from conventional processing are often not appropriate. We present a case study of near-surface (< 1.5 km from topography) imaging using multiscale waveform inversion. We also show that multiscale waveform inversion not only yields an interpretable near surface image but also a velocity model that can be used in conventional processing for improved depth imaging. However, due to a high degree of nonlinearity inherent in waveform inversion, its success calls for a robust starting model and careful application. In this paper we estimate a suitable starting model for waveform inversion using regularized inversion of direct and reflection traveltimes. The seismic data used in this paper are from the Naga Thrust and Fold belt in Northeast India.