Transdimensional imaging of random velocity inhomogeneities in Nankai subduction zone

The Nankai trough in southwestern Japan is a convergent margin where the Philippine Sea plate is subducting beneath the Eurasian plate. We have conducted five seismic observations with ocean bottom seismograms (OBSs) from 2008 to 2012 to elucidate detailed seismic structures and its relations with fault segments of large earthquakes. These observations covered the entire area of the Nankai trough, but quantity and quality of data are not spatially uniform because of different observing lengths and various noises. Waveform data of OBSs suggests variously-sized anomalies of random velocity inhomogeneity (i.e., scattering strength) in this subduction zone. To clarify details of random inhomogeneity structures, we conducted a transdimensional imaging of random inhomogeneities by means of the reversible jump Markov Chain Monte Carlo (rjMCMC) without assuming smooth spatial distributions of unknown parameters.

We applied the rjMCMC for the inversion of peak delay times of S-wave envelopes at 4-8, 8-16, and 16-32 Hz, where the peak delay time is defined as the time lag from the S-wave onset to its maximal amplitude arrival. This delay time mainly reflects the accumulated multiple forward scattering effect due to random inhomogeneities. We assumed the von Karman type power spectral density function (PSDF) for random velocity fluctuation, and estimated two parameters related with the PSDF at large wavenumber. Study area is partitioned by discrete Voronoi cells of which number and spatial sizes are variable. Estimated random inhomogeneities show clear lateral variations along the Nankai trough. The strongest inhomogeneity on the Nankai trough was found near the subducted Kyushu-Palau ridge that is located at the western margin of the fault segments. We also find a horizontal variation of inhomogeneity along the non-volcanic tremor zone. Relatively strong inhomogeneities in this tremor zone were imaged beneath west Shikoku and Kii-Peninsula. These anomalies were not clearly imaged in our previous study because we had to impose smoothing constraints. Our result implies that transdimensional imaging can be used to reveal detailed seismic structures from spatially non-uniform data.