Prediction of Elastic Wave Velocities for Shale in Earthquake-Prone Regions of Korea Using Machine Learning

Earthquakes are a worldwide natural disaster that can happen without warning. Earthquakes not only cause devastation to infrastructure but also change the soil characteristics. Recent large-scale magnitude earthquakes in the Gyeongju and Pohang regions of Korea led to soil liquefaction. To prevent damage in urban areas, it is crucial to predict and assess earthquake risk by deeply analyzing the geophysical properties in earthquake-prone areas. The dynamic properties of soil (e.g., P-wave velocity, S-wave velocity, and wave attenuation) are important parameters for analyzing seismic activities. It is important to build a unified database by collecting dispersed geophysical data for characteristic assessment of such regions. One of the important factors to reflect the seismic response spectrum of earthquake-prone regions is elastic wave velocities, which should undergo a detailed analysis. A database was established in this study by collecting data from 700 boreholes over two years in the Gyeongju and Pohang regions of Korea, including geospatial information (e.g., coordinates of borehole and depth of bedrock), dynamic properties (e.g., P-wave velocity and S-wave velocity), and physical properties (e.g., N-value and permeability). The geotechnical characteristics of the Gyeongju and Pohang regions in Korea were analyzed using an established database. The elastic wave velocities in these regions were predicted using machine learning. This research highlights the importance of analysis of the characteristics of earthquake-prone areas, Gyeongju and Pohang regions of Korea, using machine learning, and it will contribute to the earthquake risk prediction and assessment.