ABSTRACT
The irregular topography of the earth’s surface and low signal-to-noise ratios of land seismic data bring challenges to full-waveform inversion (FWI). We propose a robust method for waveform inversion (WI) to invert such land seismic data. The inversion uses finite-difference methods with rectangular meshes to simulate seismic wavefields efficiently. To accurately model irregular free surface topography, we use an improved immersed boundary method with an iterative symmetric interpolation. First-arrival signals including direct waves and refraction waves are used to estimate the P-wave velocity. To overcome the cycle skipping and dynamic inconsistency issues between the modeled data and the observed data, we create an intermediate data set by shifting the first arrivals of the predicted data toward that of the observed data within half a cycle. The intermediate data instead of the observed data are then inverted. Thus, the inversion essentially matches the traveltime information of first arrivals, which is the most reliable information contained in seismic data. Applications on the synthetic and field data sets demonstrate that the proposed WI algorithm is robust for recovering P-wave velocity from land seismic data. The resulting models have higher resolution and deeper support than that of ray-based traveltime tomography.