Evaluation of traveltime tomography in estimating near-surface velocity inversion

Near-surface velocity structures are generally inverted by ray-based traveltime tomography in which the objective function is usually constructed using raypath traveltime residuals. However, inversion accuracy in these situations remains low due to the asymptotic assumption. On the contrary, phase and instantaneous traveltime tomographies invert frequency-dependent traveltime rather than the asymptotic raypath version that is contained in the first-arrival waveform for near-surface velocities. They also take the finite-frequency effect of the seismic wave propagation into account by using wavepaths instead of raypaths as sensitivity kernels. In this context, we firstly compare synthetic results between raypath traveltime, phase traveltime and instantaneous traveltime in this study before comparing their corresponding sensitivity kernels. We then analyse three sets of inversion results to prove that it is still an acceptable way to construct the objective function using raypath traveltime for tomography within the seismic exploration frequency band. In this way, time-domain wave-equation forward simulation can be avoided and therefore the computational cost can be dramatically reduced. We also demonstrate that a sensitivity kernel that takes the finite-frequency effect into account is more important for inversion accuracy than traveltime used in tomography.