P-12

Imaging Ground Motions in the Tokyo Metropolitan Area from MeSO-net Seismograms Based on LASSO

S. Mizusako (Earthquake Research Institute, The University of Tokyo), H. Nagao (Earthquake Research Institute, The University of Tokyo), K. Hirose (Graduate School of Engineering Science, Osaka University), M. Kano (Earthquake Research Institute, The University of Tokyo), M. Hori (Earthquake Research Institute, The University of Tokyo)

A rapid prediction of damage to constructions by a large earthquake is important for prompt and efficient restoration. Ground motion at the base of each construction, which is needed for seismic response analysis, is to be estimated from seismograms. We have been developing an algorithm based on sparse modeling to image the ground motions with spatially-high resolution from seismic array data, although the density of seismometers is much lower than that of constructions. Our target is the Tokyo metropolitan area, in which the seismic array "MeSO-net" (Metropolitan Seismic Observation network) is in operation.

The proposed algorithm adopts a regression model based on Taylor's theorem proposed in Mizusako[2013], and LASSO (Least Absolute Shrinkage and Selection Operator, Tibshirani[1996]), which automatically determines efficient coefficients owing to L1-norm regularization. A motivation to adopt LASSO is to objectively determine "truncation order" of Taylor expansion and groups of observatories called "cluster", under a tradeoff that a model with lower truncation order never accounts for ground motions with spatially-high resolution but a model with higher truncation order requires a larger cluster although Taylor's theorem premises a small cluster. Our results show that this algorithm increases estimation accuracy comparing with Mizusako[2013] in frequencies less than 0.15 Hz.