Strong ground motions are composed of source terms, path effects, and site amplifications. These are well known characteristics for strong ground motions. Generalized inversion technique (GIT) proposed by Andrews (1982) is an effective tool to evaluate the characteristics. In this study, first we collected the waveforms distributed by NIED (K-NET, KiK-net), JMA (87-type and 95-type), and CEORKA from 1988 to 2016 and performed GIT by using the Fourier amplitudes of their waveforms. We confirmed that the characteristics estimated by GIT were stable and match the theoretical model as shown in the previous study. As for the site amplifications, we found that we could explain the S-wave site amplifications estimated by GIT with 1D-theory, and that the full-wave site amplifications have some correlation with AVS10 (average S-wave velocity of top 10m), AVS30 (average S-wave velocity of top 30m), AVSz (average S-wave velocity of the sediments), and Tz (travel time of the sediments). We also found that the site amplifications recalculated by using the whole duration of the waveforms could explain the predominant period in sedimentary basins in Kanto and Kinki areas.

Various tsunami source models of the 2011 Tohoku Earthquake (Mw9.0) have been proposed by using each set of observation data. However, although some of the observation data can be explained with these models, it could not well explain the observation data of GNSS and the coastal tsunami inundation height distribution together. In this study, we conducted tsunami source inversion analysis using offshore tsunami waveform, terrestrial / oceanographic geodetic data, tsunami inundation height data and GNSS data with a realistic plate shape model proposed in Yokota et al. (2017). The result suggests that the slip Miyagi off is started after 1 minute of generation and its reach Iwate off along the trench with slow motion until 4 minutes.

Study on tsunami inversion analysis by using Lasso regression, which is the representative approach of sparse modeling, was implemented. Then, characteristics of space-time distribution of fault slips were analyzed. The sparse modeling is characterized by the obtained solutions which meet the significant contribution to the output of the model. This feature is useful to investigate the development of prediction model. In this paper, tsunami inversion analysis based on sparse modeling was applied to the 2011 Tohoku earthquake tsunami and the characteristics of faults slips are analyzed.

Serious damage to nonstructural components such as window glass and suspended ceiling was reported after the 2011 earthquake off the Pacific coast of Tohoku. Strong-motion earthquake measurement was conducted at a low-rise steel-frame building with glass screens to examine its vibration characteristics that could generate force on the glass screens. From the analysis and an earthquake observation done, it was confirmed that the Glass Screen at the end of the building tends to be easily destroyed due to the influence of coupling with the torsional vibrations in addition to the large vibrations in out-of-plane direction.

A Mw 7.8 earthquake struck the Gorkha district of Kathmandu, Nepal on April 25, 2015. In Patan, vibrational characteristics of a two-story masonry building near Patan Durbar Square had been measured prior the earthquake. During the damage survey of the building after the earthquake, several new cracks were found and the natural frequencies after the earthquake were found to be smaller than those before the earthquake. In this study, microtremor observation was conducted at the building site as well as at ground motion observation sites, and the ground motion at the building site during the Gorkha earthquake was estimated. An analytical model of the structure representing the pre-earthquake condition was established so that the natural frequencies of the first five modes match the pre-earthquake measurements. Then, the dynamic analysis of the building using the estimated ground motion as the input is conducted using the refined distinct element method. The reduction of the first natural frequency was explained using the numerical model which allows the damage at the locations where the actual damages were observed.

The modal parameters of a torsionally coupled shear building are determined from Green's functions by a two-step determination process. At the first step, the natural frequencies and damping ratios are identified by genetic algorithms before reliable values are selected using the property of the Green's functions that the two parameters are dependent only on the modal order. The damping ratios are revealed to be less than 3.5 % for all the four modes. At the second step, the participation vectors are first identified by genetic algorithms, restricting to narrow search ranges for the natural frequencies and damping ratios. Reliable values of the participation vectors are then selected using another property of the Green's functions. This results in the participation vectors with predominant spectrum tending to be determined well.