16:30 〜 16:45
[SSS09-05] Stochastic interpolation of subducted plate geometry at the Nankai trough
The geometries of subducting oceanic plates can be estimated with high-resolution by using reflected wave data of seismic reflection surveys and velocity models of refraction surveys. Sampled point distributions of such geometries are frequently irregular because the reflected waves are obscured or invisible in some places. Interpolation methods such as linear, bicubic, spline and Kriging are widely applied for interpolating topography data. However, these methods tend to generate smooth geometries because the power-law spectra of geometries cannot be considered. This study examined the applicability of stochastic interpolation based on fractional Brownian motion (fBm). The fBm is widely used to synthesize realistic earth topography of which power spectral density conforms to the power-law. Recently, a multipoint fractional Brownian bridge (mfBb) was proposed to generate a regularly sampled fBm trace passing through or close to multiple points that are irregularly distributed. We applied the mfBb for irregularly sampled subsurface plate geometries along six survey lines at the Nankai trough, Japan. The power spectral densities of the interpolated geometries for each survey line conformed to the power law certainly over a wide wavenumber range. This implies that mfBb generated more realistic geometries than other methods. The interpolated geometries showed random bending between the sampled points. The variation of geometries for different random seeds tends to be larger at wider gap areas. We also observed that this variation becomes smaller where the gradient of geometry was larger. Discussion of various geometries considering reflection profiles, velocity structures, and seismicity would help to select more realistic geometries and reduce variation in interpolated geometry. Such discussion may also become important to examine the reason why the reflected waves were not observed.