Machine Learning in Site Response Assessment

OpenSite

OpenSite is an open-source database for 1742 earthquake recording sites in the K-NET and KiK-net networks in Japan. This database contains site characterization parameters directly derived from available velocity profiles, including average wave velocities, bedrock depths, and velocity contrast. Meanwhile, it also consists of earthquake horizontal-to-vertical spectral ratio (HVSR) and peak parameters, for example, peak frequency, amplitude, width, and prominence. In addition, the site database also comprises topographic and geological proxies inferred from regional models or maps. Each parameter is derived in a consistent manner for all sites. This site database can benefit the application of machine learning techniques in studies on site amplification.

Besides, it can facilitate, for instance, the search for the optimal site parameter(s) for the prediction of site amplification, the development, and testing of ground-motion models or methodologies, as well as investigations on spatial or regional variability in site response. All resources (the site database, earthquake HVSR data at all sites, and the MATLAB script for peak identification) can be freely accessed via: https://doi.org/10.5880/GFZ.2.1.2020.006 

OpenAmp

OpenAmp is a global large-scale high-quality earthquake site amplification benchmark dataset. It contains the amplification at thousands of strong-motion stations in Europe, Japan, the US, China, and South Korea derived from earthquake recordings in Fourier and/or response spectral domains. Previously we have developed an open-source site database, OpenSite, which contains ample site metadata unparalleled by any other database. 

These two data sets (OpenSite and OpenAmp) could facilitate the methodological exploration of AI in multi-scale (from site-specific to regional) site amplification modeling in terms of model development and benchmarking. High-quality benchmark data set can enable the direct comparison of different AI models/algorithms such that the state-of-the-art approach can be readily identified and improved on, eventually advancing our specific research field. 

OpenAmp can help to unleash the full potential of AI in site amplification modelling, like ImageNet in visual recognition, and the Protein Data Bank in protein structure prediction. We adopt FAIR principles (findable, accessible, interoperable, and reusable) in database development. OpenAmp is an ongoing community-driven initiative, and we welcome contributions from the community.

SeismAmp

We developed a deep-learning model, SeismAmp, to disentangle site effects from source and path effects in single-station earthquake seismograms. We demonstrate that site response can be reliably separated from single-station seismograms in an end-to-end approach. When SeismAmp is tested at new sites in both Japan (in-domain) and Europe (cross-domain), it achieves the lowest standard deviation among all tested single-station techniques. We also find that horizontal-to-vertical spectral ratio (HVSR) is not the optimal use of single-station recordings. The individual components of each record carry salient information on site response, especially at high frequencies. However, part of the information is lost in HVSR.