2018 Volume 18 Issue 2 Pages 2_95-2_114
In evaluating seismic intensity distribution maps for specified seismic source faults, several different cases of source fault model parameters are adopted in order to incorporate uncertainty associated with parameter settings. In this study, a method is proposed for evaluating spatial variation and correlation reflected in a set of seismic intensity distribution maps for different cases. The singular value decomposition technique is applied to all cases of seismic intensity distributions, through which spatial variations are derived as left-singular vectors in modal forms. Next, a simulation method for seismic intensity distributions preserving the spatial variation and correlation in original dataset is proposed. The vectors characterizing each individual seismic intensity distribution are replaced by vectors whose covariance matrix is equivalent to that for the original seismic intensity maps. A technique called "statistical preconditioning" is employed to efficiently perform such simulations. Numerical examples are shown for predicted earthquakes potentailly caused by Eastern Boundary Fault Zone of the Ishikari Lowland.
