This study proposes a framework to assess the seismic risk by integrating city-scale numerical simulations with sensor data prediction. The study begins with advanced numerical simulations using two primary methods: the integrated earthquake simulator (IES) and the stochastic Green's function method. The stochastic Green function method is used to generate the motion from the fault model to the outcropped engineering bedrock. The IES is used to simulate the ground motion on the surface, considering different scenarios, and the sumulation results are seperated into training and validation datasets. In this study, all scenarios are generated from a single fault model, and variability is introduced by changing the rupture initiation point, representing a limited dataset but practically relevant setting. Proper orthogonal decomposition (POD) is then applied to the training datasets to extract spatial modes, and those modes are used for identifying the most suitable distributions for sensors. In this study, the sensors are strategically placed to maximize data collection efficiency while minimizing the overall number of sensors used. Optimizing sensor distribution is crucial owing to the challenges associated with deploying numerous sensors in actual situations. The distribution of sensors obtained by this method is validated using a test set generated from IES simualation. The case study focusing on Sendai city (Japan), for scenarios associated with the Nagamachi-Rifu fault, finds an optimal solution for the distribution of sensors and allows for an overall prediction of the expected accuracy. The results show that seismic risk prediction and assessment on a city scale can be achieved by this method while maintaining accuracy and the number of sensors. The present work therefore demonstrates the feasibility of the proposed rapid prediction framework under limited dataset conditions for a specific, high impact seismic source, while the extension to multiple faults and magnitudes is left for future work.
Abstract The 2024 Noto Peninsula earthquake (M w 7.5) in Japan generated strong long-period ground motions that were observed across the Kanto Plain, which includes the Tokyo metropolitan area. Advances in high-performance computing now allow rapid 3D simulations of long-period ground motions in sedimentary basins that can be used to provide both site-specific estimates and basin-wide reconstructions of the wavefield. We simulated long-period ground motions in the Kanto Plain using a 3D subsurface structure model and a high-degree-of-freedom seismic source process model released shortly after the earthquake, and then compared the simulated waveforms with observations. Despite some phase mismatches, the simulations reproduced the observed waveform amplitudes, durations, and pseudo-velocity response spectra. We then applied wave gradiometry analysis to the simulated long-period wavefield in the 6–10 s band and treated the north–south (NS) and east–west (EW) components separately. The estimated propagation direction of the NS component largely followed the epicentral azimuth, whereas that of the EW component progressively rotated southward along the Kanto Mountains toward central Tokyo. This component-dependent behavior indicates that the long-period ground motions in the Kanto Plain are best interpreted as the superposition of multiple packets of surface waves traveling along different paths rather than as a single plane wave. Graphical abstract
In mountain tunneling, it is important from the point of construction safety and effectiveness to obtain information on the weak zones such as fracture zone ahead of the face, which may pose problems during construction. To provide long-term outlook for tunnel work without disturbing excavation, the authors have developed a tunnel seismic reflection survey method known as the Taisei Blast Excavation Prospecting (T-BEP). This method extends the maximum surveyable distance beyond 350 m, doubling or tripling the capability of other conventional methods. However, complicated works are necessary in the preparation and measurement processes, and there are safety concerns. Therefore, we have improved the old T-BEP and developed a new version. In this paper, we present an overview of the new version and report the results of comparative validation experiments between the new and old T-BEPs.
Observation of tunnel excavation faces is one of the crucial tasks for ensuring safe construction. Photographs of the faces sometimes complement geological engineers’ visual investigation; however, they are insufficient for a comprehensive geological assessment. Instead of the conventional methods, we have developed a method for observing tunnel faces in virtual reality (VR), using visualization of 3D point cloud data with a customized software. Two types of manual strike and dip orientation measurement functions for discontinuities have been implemented in the software. To validate its effectiveness, we applied the method to tunnel sections where weathered and altered andesite and talus deposits were distributed. The method identified unconformities at the bottom of the talus, which were unclear from photographs alone. Moreover, the measured strike and dip data quantified the geological structure of the unconformity. From these results, we conclude that the VR observation is more effective for detailed geological evaluation than the conventional 2D photographic methods.
