Machine Learning based Estimator for ground Shaking maps applied to the Los Angeles basin region

Large earthquakes are among the most destructive natural phenomena. Fast estimation of the ground shaking intensities is a crucial task for hazard assessment after a large earthquake occurs.The Machine Learning Estimator for Ground Shaking maps (MLESmap) is proposed as a novel methodology that exploits the predictive power of Machine-Learning (ML) algorithms to estimate ground acceleration values a few seconds after a large earthquake occurs. The inferred information can produce shaking maps of the ground providing quasi-real-time affectation information to help us explore uncertainties quickly and reliably.MLESmap utilizes physics-based seismic scenarios to feed the algorithms.Moreover, as our synthetic catalogs will never contain all possible future events, we aim at having models capable of successfully interpolating non-trained events accurately.To set up the MLESmap technology, we used simulated ground motions from CyberShake Study 15.4, a physics-based Probabilistic Seismic Hazard model for Southern California at 1 Hz. In particular, this CyberShake study is focused on the Los Angeles basin region. Los Angeles sits on top of large sedimentary basins. These soft foundations can amplify the amount of damaging shaking these cities experience during an earthquake. Our approach (i.e. simulate, train, deploy) can help produce the next generation of ground shake maps, capturing physical information from wave propagation (directivity, topography, site effects) at the velocity of simple empirical GMPEs. In this work, we will present the MLESmap workflow and its application in the Los Angeles region as a use case.
The 28th IUGG General Assembly (IUGG2023) (Berlin 2023)