Apart from seismic waveforms obtained by seismographs, one data source that records seismic motion and damage with a time axis is footage from shop cameras. The National Research Institute for Earth Science and Disaster Resilience (NIED) is continuing to archive video footage of customer-attracting facilities that have suffered significant earthquakes through a collaboration agreement with Aeon Co. By using this archive, we clarify the relationship between the images and the damage. It is hypothesized that this will facilitate more efficient and labor-saving responses and countermeasures based on the available evidence. In this study, a neural network model for discriminating the presence or absence of anomalies by machine learning was constructed and evaluated, using the videos taken by shop cameras during the shaking of an earthquake as the data source. The objective of this paper was to examine the fundamental principles of machine learning and to gain an understanding of a technique that can be employed to enhance accuracy. Consequently, the proposed method, which employed the concept of one-hot encoding and pre-processing, yielded an 84 % classification accuracy for earthquake images that were previously unknown to the model.

When earthquakes occur, running trains are quickly suspended if necessary and facilities are inspected based on the observed earthquake ground motions by seismometers installed along the railway. Addition to the discrete observed motions by the seismometers, continuous estimated motions along the railway are useful to determine the optimal inspection section. Although the estimated motions inevitably include estimation errors, the practical use of estimated motions has not progressed sufficiently because the handling of estimation errors has not been established. Therefore Iwata et al. (2023) proposed a practical method to estimate fused observed and estimated motions, after presenting the concept of handling the estimation errors. The method proposes distances to apply observations as a uniform. However, the distance should be established individually according to the degree of change in the subsurface ground amplification characteristics around the observation site. In this study, we propose the concept of the distances to apply observations using the spatial seismic amplification characteristics of subsurface ground by public institutes. Furthermore, we show an example of estimated motions using the proposed method for a virtual railway.

We proposed a new approach to estimate empirical site amplification factors through spectral inversion considering heterogeneous attenuation along the propagation path, and to inversely calculate response spectra at the free surface on seismic bedrock (Vs 3200 m/s) by dividing downhole records on hard soil by these amplification factors. In our spectral inversion process, we enhanced the existing methodology to model the simultaneous analysis of horizontal and vertical ground motions. We also selected two reference stations for constraint conditions based on quantitative indexes. Our amplification factors estimated by this method were compared with those estimated by an existing method based on a network of two adjacent points. Although results from both methods were generally consistent, our approach appeared more reliable in the high-frequency range (∼5-20 Hz) due to the optimal selection of reference stations and consideration of heterogeneous attenuation. Our response spectra at free surface on seismic bedrock were compared with those derived from a conventional method that integrates outcropping analysis based on 1-D multiple reflection theory with soil-velocity adjustment using an empirical equation. Both methods yielded generally consistent results at downhole stations on seismic bedrock equivalent (hard rock), where soil-velocity adjustment was unnecessary. On the other hand, larger discrepancies were appeared, particularly for periods longer than 0.2 s, at downhole stations on soft rock with significant depth difference from seismic bedrock. This seems to be the effect of our method, which estimates amplification factors considering not only surface but also deeper subsurface structures for each station.

In this paper, I propose a theory that transforms a shear beam model into a multi-degree of freedom system (MDOF). Furthermore, I validate this theory by numerical analysis. This validation shows findings as follows. First, the equivalent MDOF based on the proposed theory well agree with the original shear beam model if a deformation component which propagates as elastic wave has dominant influences on the total deformation. Second, shear springs in the equivalent MDOF have frequency dependence.

This study aimed to understand the 20-year recovery process in the Khao Lak area of Phang Nga Province in southern Thailand, which sustained severe damage from the Indian Ocean tsunami caused by the Sumatra earthquake on December 26, 2004, using optical satellite images and a total of seven field surveys. This area, where tourism and fishing are the main occupations, was still in the process of recovery when surveyed 10 years after the tsunami. However, in the latest survey 20 years later, the resort area was bustling with foreign tourists, and the number of hotels and buildings had increased significantly. Meanwhile, no hard measures such as seawalls were seen in terms of tsunami disaster prevention measures, and instead, progress had been made in installing tsunami warning sirens and evacuation towers, and setting up evacuation routes. Tsunami museums and memorial sites had also been established to pass on the memory of the disaster.