Seismic studies are characterized by constantly growing amount of data which have to be processed in a timely manner. Thus development of fast computational algorithms optimized for modern high-performance computational platforms is always an issue.
Here we present a fast algorithm for finite-difference (FD) computation of the first-arrival seismic waveforms. At every time step of the FD scheme the calculations are done only in a narrow strip following the first-arrival wave front. The first-arrival wavefront itself is precomputed by the numerical eikonal solver which is computationally inexpensive. Such approach leads to considerable speed-up compared to standard FD computations performed for the whole domain at every time step.
It is further used for subsurface velocity model building from seismic data using so-called wave-equation tomography or adjoint-state methods which require multiple waveform computations. We address several geophysical applications of our method: regional and local seismic tomography from earthquake seismologic data, near-surface characterization for engineering and exploration applications, controlling hydrofracturing from microseismic data, cross-well tomography for reservoir monitoring.