In disaster mitigation planning for future large earthquakes, predictions of seismic ground motions are a crucial part of early warning systems and earthquake hazard mapping. The way the ground moves depends on how the ground layers amplify the seismic waves (described in a mathematical site “amplification factor”). However, geophysical explorations to understand ground conditions are expensive, which to date limits the characterization of site amplification factors.

A new study by researchers from the University of Hiroshima published on April 5 in the Bulletin of the Seismological Society of America introduces a new artificial intelligence (AI)-based technique to estimate site amplification factors from ambient vibration or ground microtremor data.

Subsoil conditions, which determine how earthquakes affect a site, vary widely. Softer soils, for example, tend to amplify ground motion following an earthquake, while hard substrates can dampen it. Ambient ground vibrations or micro-tremors that occur over the entire surface of the Earth, caused by human or atmospheric disturbances, can be used to study ground conditions. Measuring microtremors provides valuable information about a site’s amplification factor (AF), hence its vulnerability to earthquake damage due to its tremor response.

The recent study by Hiroshima University researchers introduced a new way to estimate site effects from microtremor data. “The proposed method would contribute to more accurate and detailed seismic ground motion predictions for future earthquakes,” says lead author and associate professor Hiroyuki Miura of the Graduate School of Advanced Science and Engineering. The study investigated the relationship between microtremor data and site amplification factors using a deep neural network with the aim of developing a model that could be applied to any site in the world. .

The researchers looked at a common method known as horizontal-to-vertical spectral ratios (MHVR) that is typically used to estimate seismic ground resonance frequency. It can be generated from microtremor data; ambient seismic vibrations are analyzed in three dimensions to determine the resonant frequency of sediment layers above bedrock as they vibrate. Previous research has shown, however, that MHVR cannot be reliably used directly as a site-enhancing factor. Thus, this study proposed a deep neural network model to estimate site enhancement factors from MHVR data.

The study used 2012-2020 microtremor data from 105 sites in Chugoku District in western Japan. The sites are part of Japan’s National Seismograph Network which contains approximately 1700 observing stations spread out in a uniform grid at 20 km intervals across Japan. Using a generalized spectral inversion technique, which separates source, propagation and site parameters, the researchers analyzed site-specific amplifications.

The data from each site was divided into a training set, a validation set, and a test set. The training set was used to teach a deep neural network. The validation set was used in iterative network optimization of a model to describe the relationship between microtremor MHVRs and site amplification factors. The test data was a completely unknown set used to assess model performance.

The model performed well on test data, demonstrating its potential as a predictive tool for characterizing site amplification factors from microtremor data. However, notes Miura, “the number of training samples analyzed in this study (80) sites is still limited” and should be expanded before assuming that the neural network model applies nationally or globally. The researchers hope to further optimize the model with a larger data set.

Fast and cost-effective techniques are needed for more accurate prediction of seismic ground motion since the relationship is not always linear. Explains Miura, “By applying the proposed method, site amplification factors can be automatically and accurately estimated from microtremor data observed at an arbitrary site.” Going forward, the study authors aim to continue refining advanced AI techniques to assess nonlinear ground responses to earthquakes.

This research was funded by the National Research Institute for Earth Science and Disaster Prevention (NIED), Japan, and Neural Network Console provided by SONY (2021).

Source of the story:

Material provided by Hiroshima University. Note: Content may be edited for style and length.


The 2023 Sharjah Biennale Artist List, Curated by Okwui Enwezor, Unveiled


Bunnell Artist in Residence composes musical scores to reflect Homer's community and landscapes

Check Also