The seismic response analysis results of a small building without footing girders (NF_GB) were shown in comparison with the general small building with footing girders. In addition, not only the response characteristics of NF_GB with parameters of shear wave velocity of free ground and the live load were shown, but also response characteristics of NF_GB with ground amplification removed were shown to establish a static structural design. Finally, the influence of the nonlinearity of structure, free ground and SSI spring of NF_GB on the seismic response reduction was discussed.

Concerning a petroleum refinery facility which consists of various structures, the shutdown trigger level should be evaluated by seismic intensity index correlated with actual facility damage. In addition, it is necessary to decide where to place seismometers in the site of facility. The optimal index and place are detected by the investigation of correlation between ground motion and response of structures according to various seismic waves. The above mentioned correlation is reflected in the seismic loss based on the probabilistic methodology. The shutdown trigger level is evaluated, comparing seismic loss between with shutdown and without one.

Probability of large earthquake such as the coming Nankai-trough earthquake, Japan, is extremely high, therefore enough attentions should be paid to the occurrence of severe accidents of vehicles on highways. When such severe accidents actually occur, not only the damage of the accident itself but also the trouble of transportation route cause a delay of emergency response. The influence of earthquakes on running vehicles and the countermeasures to avoid such accidents have to be considered in advance. This study focused on seismic response analysis of highway vehicles during an earthquake. Previous studies pointed out that when an earthquake occurs, collisions occurs between vehicles. We here develop car-collision model on the result of experiment about vehicle-to-vehicle and vehicle-to-wall collisions. The model can express collisions with non-linear spring, and it can reproduce the experiment results accurately.

We investigated the causes and remedies regarding the Fourier amplitude sags at intermediate frequencies in the source spectra of large earthquakes, which were synthesized from small earthquakes using the stochastic Green's function method. First, we summarized the basic equations and the similarity rules between the large and small earthquakes. And then we summarized the causes of the amplitude sags to (a) the sag at intermediate frequencies caused by the F function (the filtering function of the moment rate from the small to the large earthquake), (b) the sag at the connected frequencies caused by superimposing results based on different phase spectra, which are the coherent phases at lower frequencies and the random phases at higher frequencies, and (c) the sag at intermediate frequencies caused by approximating from the ω^-2 model to the ω^-3 model, as the number of fault divisions of large earthquakes increases. We investigated the causes of those sags and their improvement method. Furthermore, we theoretically investigated (d) the sag due to the superposition of the element earthquakes, and found that a large sag was caused at transition frequencies by the change in the amplitude of the square of the similarity ratio N at lower frequencies to N at higher frequencies. Finally, as a simple and practical improvement method for the sag, we proposed an amplitude correction method based on the theoretical ω^-2 model. We simulated the strong ground motions using the stochastic Green's function method, and compared the results with and without the amplitude correction, and demonstrated that the strong ground motions are greatly underestimated without the correction at the intermediate frequencies.

Satoh (2020) pointed out that some velocity time histories by integrating acceleration records of vertical components have linear trends to the downward direction then become almost flat with values “Vnon” and interpreted that Vnon was caused by negative dilatancy. In this study from many strong motion records with large Vnon, we show that velocity time histories of the horizontal records have linear trends and estimate the tilt angles and directions of the ground using the linear trends. Several records with large Vnon have been pointed out the increase of excess pore water pressure. The tilt direction in the near-fault regions is almost radial and the tilt starts at S-wave potions. The estimated tilt angle is much larger than that by crustal deformation calculated from fault models. We interpret from these results and Satoh (2020) that the tilt motions are caused by nonlinear site responses including the increase and dissipation of excess pore water pressure or the negative dilatancy accompanied by the settlement of the ground.

Keeping in mind for application to the performance-based design of buildings, near-fault strong motions which were recorded very frequently during active series of inland crustal earthquakes are investigated using sorted and ranked pseudo velocity response spectra and ranked response spectral ratios which are newly defined in this paper. In a near-fault region, it is confirmed that the almost largest responses of a structure might be generated twice or more times during a series of earthquakes. It is also confirmed that in shorter-period structures large responses occur much more times than the ones in longer-period structures during the series of earthquakes. Finally, ranked response spectral ratios of near-fault earthquake ground motions which could be considered in structural design are also proposed. The effects of strong motion repetition on building responses could be considered multiplying the response spectra of the design earthquake motions or the target spectra by the proposed ranked response spectral ratios.

Since the 1995 Southern Hyogo Prefecture Earthquake, the seismic intensity observation system in Japan has developed dramatically. This paper summarizes the transition from the start of seismic intensity observation in Japan to the present. Analysis of seismic intensity data from the JMA seismic catalog showed that as the density of seismic intensity observation points increased, the number of strong motion observations closer to the epicenter increased and the observed maximum seismic intensity increased. In addition, by considering the surface layer amplification and population distribution for each area obtained by dividing the whole country into voronoi regions based on seismic intensity observation points, it was shown that even in the cases of similar magnitude earthquake, the observed maximum seismic intensity increases when it occurs in a densely populated area with the soft ground and the dense observation points. This paper quantitatively shows the changes in the times and regional difference of the maximum seismic intensity observed, and clarifies the characteristics of the seismic intensity station arrangements of the Japan Meteorological Agency, National Research Institute for Earth Science and Disaster Resilience, and local governments.

It has been pointed out that earthquakes that occur in the Philippine Sea plate around Iyonada have inhomogeneous attenuation characteristics. One of the possible factors is the difference in attenuation between the continental plate and the Philippine Sea Plate and the heterogeneity of the horizontal attenuation structure. In this study, inhomogeneous attenuation structure, source and site amplification characteristics were separated from the observation records based on block inversion analysis that improves the method of Tomozawa et al. (2019) to estimate the inhomogeneous attenuation structure in the depth direction. As a result, the path region of the Philippine Sea plate was divided into the southwestern part of the Shikoku region and other regions, and became high-Q in the region where the upper surface depth of the Philippine Sea plate was deep. The short-period level of the 2001 Geiyo earthquake was estimated to be 3.12 × 10¹⁹ N m/s². It is also shown that the bias of the spatial distribution of the residuals can be improved by taking into account the horizontal inhomogeneity of each plate.

A new method for investigating the attenuation characteristics of strong ground motions is proposed. In the method, bedrock motions (BRM), which are obtained by dividing observed motions (OBS) by relative site factors (RSF), are examined in the study area. The RSF are evaluated as the ratios of site factors relative to a reference bedrock site by a regression analysis to satisfy all relative site factors of neighborhood site pairs involved in the area. This method was used to examine the attenuation characteristics of strong motions observed at K-NET and KiK-net sites in the Tohoku and Kanto regions caused by four plate boundary earthquakes that occurred in the Pacific subduction area. The attenuation distributions of BRM amplitudes obtained by this method for these earthquakes are characterized by smaller variations than those of OBS. A comparison of attenuation models evaluated by regression analyses of BRM and OBS using a polynomial formula of the logarithm of source distance shows that the difference of variations between the BRM and OBS attenuation models is consistent with the variations of the RSF. This indicates that the variations of attenuation model of BRM should be interpreted as variations of source radiation characteristics and/or wave propagation path characteristics.

After the Great East Japan Earthquake on March 11, 2011, the tsunami evacuation drill for residents and the disaster mitigation training are continuously conducted by the local government or the prefecture, because it is important for tsunami disaster mitigation to train about the tsunami disaster that is low frequency disaster. However, there are few opportunities for the local government in charge of disaster prevention to conduct the practical disaster exercise, because they do not have much knowledge and budget, information and experience of tsunami disaster. Therefore, we study on the conducting method of the tabletop exercise to respond a major tsunami disaster for the local governments in charge of disaster prevention. First, we ran a workshop to verify the disaster response issues in four municipalities. Second, we summarized the essential points of scenario creation about tsunami disaster and how to conduct four tabletop exercises. Third, we conducted four times tabletop exercises for disaster prevention staff in municipalities, following the scenario and the implementation method. In addition, the local government staff in charge of disaster prevention in Kamogawa City applied the scenario and the implementation method to their disaster mitigation training. Finally, we examined the scenario creation and the implementation method, and confirmed new issues of disaster response through the tabletop exercise.

We developed an automatic damage classification model for wooden houses using aerial photographs of plural earthquakes designed for supporting disaster responses. First, we visually classified all buildings into four damage levels, by utilizing six kinds of aerial images taken immediately after the 1995 Southern Hyogo earthquake, the 2011 off the Pacific coast of Tohoku earthquake, the foreshock and the main shock of the 2016 Kumamoto earthquake, and the 2018 Hokkaido Iburi eastern earthquake. Then we compared the visually classified result with the field survey result, which was conducted at Mashiki Town soon after the main shock of the Kumamoto earthquake. Then we clarified that an average recall is approximately 74 % under the condition of three damage levels. After that, we automatically cropped patch images of buildings from each aerial photograph, then constructed the training and test data set labelled as one of four damage levels. By using the training data set, we developed the CNN-based damage detection model, which shows higher performance than former model. Furthermore, we developed the program, which overlays a damage detection result and an aerial image. The program can be used for estimation of the damage concentrated area and the number of damaged buildings. Additionally, we confirmed that a cleansing of training data and a usage of training data of various conditions can be contributed to higher discrimination performance. Moreover, we performed a cross validation using image patches extracted from different image flames as training data and test data. Then confirmed that an average recall is approximately 77 % under the condition of three damage levels, which means that a spatial distribution of the training data in not affected by the generalization performance. Finally, for a validation of this detection model, we classified each damaged building into one of three damage levels by using six different aerial images as test data. As a result, an average recall is approximately 70 %, and we confirmed that this model is slightly affected by the earthquake motion and the building structure, however it shows an applicative generalization performance for estimating damage of wooden buildings in Japan.