← Back to Gallery

A Machine Learning Approach to Developing Ground Motion Models From Simulated Ground Motions

The USGS is working towards incorporating regional-specific seismic hazard into the U.S. National Seismic Hazard Model. The large dataset of ground motions generated from simulations can serve to supplement empirical data in areas where observed ground motion data is lacking and help constrain trends in intensity where geologic structures complicate seismic wave propagation. Machine learning provides an avenue to incorporate the plethora of information from synthetic datasets, without enforcing specific mathematical formulations that ground motion trends must follow, potentially biasing ground motion models (GMMs).

Here, we use a machine learning approach to build a ground motion model (GMM) from a synthetic database of ground motions. First, we build a dataset of predictor variables and corresponding ground motion intensities extracted from the Southern California CyberShake study. An artificial neural network (ANN) is used to find the optimal weights and coefficients that best fit the target data (without overfitting), with predictor variables chosen to match that of state-of-the-art GMMs. We compare our synthetic-based GMM with empirically based GMMs derived from the globally based Next Generation Attenuation West2 dataset, finding near-zero median residuals and similar amplitude and trends (with period) of total variability. Additionally, we find that the ANN GMM has similar bias and variability to empirical GMMs from records of the recent 2019 Mw 7.1 Ridgecrest event, which neither GMM has included in its formulation. As simulations continue to more accurately model broadband ground motions and expand across a wider range of source, site, and path conditions, machine learning provides a way to utilize the vast amount of synthetically generated data to better constrain ground motions where empirical data is sparse and to identify the relationships and relative importance between parameters for future parameterization of GMMs.

Presenting Author: Kyle Withers

Additional Authors