← Back to Sessions

Estimating VS30 From Horizontal-to-Vertical Spectral Ratio Based on Supervised Machine Learning

Description: 

Average S-wave velocity (Vs) to 30 m depth (VS30) is an important proxy to estimate site amplification. Invasive and non-invasive methods, such as velocity loggings or active/passive surface wave methods, have been generally used to directly measure the VS30. Those methods are expensive and time consuming. The VS30 is also indirectly estimated by empirical methods based on geology, geomorphology, elevation, or slope angle etc. Those methods are inexpensive but not accurate. We intend to use a horizontal-to-vertical spectral ratio (H/V) to roughly estimate the VS30. The measurement of H/V is much easier and quicker compared with active surface wave methods (MASW) or microtremor array measurements (MAM). The inversion of the H/V is essentially non-unique and it is impossible to obtain unique Vs profiles only from H/V spectra. We apply supervised machine learning to roughly estimate the Vs profiles or VS30 from H/V spectra together with other available information, such as site location or geomorphology etc. Our machine learning consists of a neural network with one hidden layer. The pairs of the H/V spectra (input layer) and Vs profiles (output layer) are used as training data. Input layer consist of an observed H/V spectrum site coordinate, and geomorphological information. Output layer is a velocity profile obtained from the velocity loggings, surface wave measurements, or inversion of H/V. We have applied the machine learning to several different sites in U.S. and Japan. This presentation introduces a study at Napa Valley in California. We measured MASW, MAM and H/V at approximately 100 sites at the Napa Valley. The pairs of H/V spectra together with their coordinate and Vs profiles obtained from the inversion of dispersion curve compose the training data. Trained neural network predicts Vs profiles from observed H/V spectra. The VS30 calculated from predicted Vs profiles are reasonably consistent with those calculated from true Vs profiles obtained from the dispersion curves. The results implied that the machine learning could roughly estimate VS30 from H/V spectra together with available other information.

Session: Single-station Passive Exploration Methods: Status and Perspectives

Type: Oral

Date: 4/18/2023

Presentation Time: 05:15 PM (local time)

Presenting Author: Koichi Hayashi

Student Presenter: No

Invited Presentation: 

Authors