Blind Test of Methods for Obtaining 2-D Near-Surface Seismic Velocity Models from First-Arrival Traveltimes

Abstract Seismic refraction methods are used in environmental and engineering studies to image the shallow subsurface. We present a blind test of inversion and tomographic refraction analysis methods using a synthetic first-arrival-time dataset that was made available to the community in 2010. The data are realistic in terms of the near-surface velocity model, shot-receiver geometry and the data's frequency and added noise. Fourteen estimated models were determined by ten participants using eight different inversion algorithms, with the true model unknown to the participants until it was revealed at a session at the 2011 SAGEEP meeting. The estimated models are generally consistent in terms of their large-scale features, demonstrating the robustness of refraction data inversion in general, and the eight inversion algorithms in particular. When compared to the true model, all of the estimated models contain a smooth expression of its two main features: a large offset in the bedrock and the top of a steeply dipping low-velocity fault zone. The estimated models do not contain a subtle low-velocity zone and other fine-scale features, in accord with conventional wisdom. Together, the results support confidence in the reliability and robustness of modern refraction inversion and tomographic methods.