Accurate source location from waves scattered by surface topography

Accurate source locations of earthquakes and other seismic events are fundamental in seismology. The location accuracy is limited by several factors, including velocity models, which are often poorly known. In contrast, surface topography, the largest velocity contrast in the Earth, is often precisely mapped at the seismic wavelength (>100 m). In this study, we explore the use of P coda waves generated by scattering at surface topography to obtain high‐resolution locations of near‐surface seismic events. The Pacific Northwest region is chosen as an example to provide realistic topography. A grid search algorithm is combined with the 3‐D strain Green's tensor database to improve search efficiency as well as the quality of hypocenter solutions. The strain Green's tensor is calculated using a 3‐D collocated‐grid finite difference method on curvilinear grids. Solutions in the search volume are obtained based on the least squares misfit between the “observed” and predicted P and P coda waves. The 95% confidence interval of the solution is provided as an a posteriori error estimation. For shallow events tested in the study, scattering is mainly due to topography in comparison with stochastic lateral velocity heterogeneity. The incorporation of P coda significantly improves solution accuracy and reduces solution uncertainty. The solution remains robust with wide ranges of random noises in data, unmodeled random velocity heterogeneities, and uncertainties in moment tensors. The method can be extended to locate pairs of sources in close proximity by differential waveforms using source‐receiver reciprocity, further reducing errors caused by unmodeled velocity structures.