[S03-3] Imaging of heterogeneities in the Earth with seismic scattered waves and ambient noise III

[S03-3-02] Approximate vector sensitivity kernels of coda waves to seismic velocity changes based on the scalar single isotropic scattering model

^○Hisashi Nakahara, Kentaro Emoto (Tohoku University, Sendai, Japan)

Coda-wave interferometry has been used to detect velocity changes in association with large earthquakes or volcanic eruptions. Sensitivity kernels of travel times of coda waves are necessary to determine the region of velocity changes correctly. These sensitivity kernels have been formulated so far for scalar waves based on different assumptions of two-dimensional single scattering or multiple scattering, three-dimensional single scattering or multiple scattering, or diffusion. However, no formulation has been made for vector waves as far as we know. Hence, we tackle this formulation and derive analytical approximate expressions for two-dimensional and three-dimensional cases. The key point in our simple extension to vector waves is the projection of seismic phonon energy into horizontal and vertical components by using the square of the direction cosine of the polarization direction. Thanks to this simple idea, we can derive analytical expressions of the approximate vector sensitivity kernels by using the single isotropic scattering model for scalar waves, though we can treat either P waves or S waves at a time. Our results show that the sensitivity kernels are different for different components, and accordingly different components show different travel time changes with respect to lapse time. These are theoretically shown for the first time by this study. The approximate vector sensitivity kernels have two clear peaks at a source and a receiver which are different from different components. Comparing with finite difference simulations of vector wave propagation, we find that our approximate vector sensitivity kernels are very good for two-dimensional cases, but are worse for three-dimensional cases. So far the reason is not clear yet. However, our approximate vector kernels are helpful to consider how to use seismograms of the different components simultaneously in coda-wave interferometry.