[S03-2] Imaging of heterogeneities in the Earth with seismic scattered waves and ambient noise II

[S03-2-02] Trans-dimensional imaging of scattering and intrinsic Q structures

^○Tsutomu Takahashi (Japan Agency for Marine-Earth Science and Technology)

Precise imaging of scattering and intrinsic Q structures is essential to describe wave propagation processes of high frequency seismic waves (>1Hz). The multiple lapse time window analysis (MLTWA) has been used to estimate these Q values by assuming constant Q in a study area (e.g., Hoshiba 1993). This study generalizes this MLTWA for a stable estimation of 2-D Q structures under the Bayesian framework in dimension variable space. Study area is partitioned into small areas by means of the Voronoi tessellation. Scattering and intrinsic Q in each small area are constant. We define a misfit function for spatiotemporal variations of wave energy as with the original MLTWA, and maximize the posterior probability with changing not only Q values but the number and spatial layout of the Voronoi cells. This maximization is conducted by using the reversible jump Markov chain Monte Carlo (rjMCMC) (Green 1995) since the dimension of posterior probability is variable. After a convergence to the maximum posterior, we estimate Q structures from the ensemble averages of MCMC samples around the maximum posterior probability. Synthetic tests showed stable reconstructions of input structures with reasonable error distributions. We applied this method for seismic waveform data of ocean bottom seismograms at the outer-rise area off Tohoku, and estimated Q values at 4-8Hz, 8-16Hz and 16-32Hz. Intrinsic Q are spatially uniform at all frequency bands, and scattering Q shows two distinct strong scattering regions at petit spot area and high seismicity area. These strong scattering are probably related to magma inclusions and fractured structure, respectively. While our generalization of MLTWA is still based on a classical waveform modeling in constant Q medium, this method can be a fundamental basis for tomographic imaging of Q structures in the crust.