[S03-4] Imaging of heterogeneities in the Earth with seismic scattered waves and ambient noise IV

[S03-4-01] Quantifying the body-wave information retrieved from global earthquake coda correlation

^○Hsin-Hua Huang^{1, 2, 3}, Victor Tsai², Fan-Chi Lin³, Weitao Wang⁴, Julien Chaput⁵ (1.Institute of Earth Sciences, Academia Sinica, Taiwan, 2.Seismological Laboratory, California Institute of Technology, U.S.A., 3.Department of Geology and Geophysics, University of Utah, U.S.A., 4.Key Laboratory of Seismic Observation and Geophysical Imaging, Institute of Geophysics, China Earthquake Administration, China, 5.Department of Mathematics, Colorado State University, U.S.A.)

Body wave retrieval from noise or coda correlations provides sampling at greater depths and has advanced our ability to probe and monitor the deep Earth. However, to date the identification of body waves in correlations relies mainly qualitatively on the signal-to-noise ratio of phases and the matching of theoretical travel times and lacks a quantitative validation of their travel-time accuracy and spatial sensitivity. Based on recent observations that the late coda of large earthquakes is not diffuse and that the source locations of the large earthquakes used in coda correlations are well known, it becomes viable to approach the question through numerical simulation with spherically symmetric Earth models (no scattering) for coda correlation. By varying numerous parameters such as source duration, source distribution, and velocity/Q structure, 143 global large earthquakes (Mw7) and USArray stations are used to compare with and to validate the observations of recently-reported core phases (e.g. P'P') from coda correlations (Huang et al., 2015). Simulation results show that using only reverberations can replicate most of the features of the body-wave signals observed in correlation functions. The travel time bias present in the (zero-offset) autocorrelation functions is insignificant (within 0.5 s) as long as the selected coda windows are sufficiently late (after ~10000 s) while the bias in cross-correlation functions shows a systematic trend that increases with station separation regardless of source distribution. With the aid of ray-based modeling, we explore the mechanism of such bias, the emergence of spurious phases, and distinct frequency content between coda and noise correlation. Gaining a better understanding of such body-wave information improves our ability to use and validate correlation-based data, making it possible to reliably integrate such measurements with earthquake-based data in future applications.