Item

Abstract

The challenge of modeling site effects in complex geological environments remains a central topic in engineering seismology and is the focus of this session.

Addressing this issue begins with identifying "complex" sites where simple prediction models, based on 1D velocity profiles, fail to provide satisfactory results. This requires comparing actual site effects with predictions from physical models across large datasets. Recent advances now enable such analyses, thanks to the quantification of site effects through generalized inversion methods or spectral ratio calculations between deep and surface stations in regions equipped with borehole networks. Systematic tests using extensive data from Japan’s Kik-net and K-NET networks reveal that a significant proportion of sites deviate from 1D behavior, particularly at frequencies above 3 Hz.

To meet this challenge, we propose three complementary approaches to improve site effect predictions for complex environments:
- Enhanced High-Frequency Physical Modeling: Improving and calibrating attenuation models is essential and feasible, paving the way for more accurate high-frequency predictions.
- Increased Observation Density: Expanding observational coverage in urban areas through innovative methods, such as leveraging smartphone data, can significantly enhance datasets and support the development of high-resolution amplification maps.
- Machine Learning Applications: Developing machine learning models tailored to available site information—ranging from geological and geotechnical data to recorded seismic data—offers a flexible, novel, and testable framework for site effect prediction.

This presentation will discuss the methodologies and results of recent studies, highlighting how these strategies can advance our understanding and modeling of site effects in complex geological settings.