Enhanced Vs30 Prediction Models: Leveraging Geology and Terrain with Machine Learning

Understanding the contributors to ground motions at a specific site is essential for accurate hazard and ground motion estimations. Among these contributors, site response is recognized as a dominant factor. A comprehensive characterization of site effects requires careful consideration of the geological and mechanical conditions at a site, one of which is the shear wave velocity profile with depth. Its derivative—the time-averaged shear wave velocity of the upper 30 meters, Vs₃₀, has been the most commonly used proxy for site-effect predictions since the early 1990s, and is also incorporated into the Israeli building standard. In the case of very large engineering projects covering a wide geographical area, direct measurements of the shear wave velocity profiles becomes impractical. To address this, Vs₃₀ maps are developed using proxies such as terrain slope, geological information, or a combination of both. This study leverages machine learning (ML) models to generate a high-resolution Vs₃₀ map for Israel. ML models offer a robust framework for capturing complex, non-linear relationships between input parameters and Vs₃₀, surpassing traditional correlations. The model developed in this work was trained and validated using an extensive database of over 500 shear-wave velocity profile measurements. Additional parameters, including surface geology (lithology and age), soil type, and terrain-based features, were integrated to enhance predictive accuracy. The new model predictions demonstrate significant improvements compared to existing local and other global Vs₃₀ models. The new model is subsequently used for interpolation, to produce a state-wide Vs₃₀ map. This map will provide a valuable resource for national hazard assessments, seismic risk analysis, and engineering applications, offering improved spatial resolution and reliability compared to previous models. This study highlights the potential of integrating advanced ML techniques to enhance site-effect characterization and improve the accuracy of hazard assessments at regional and national scales.