Informing Predictions from Above with Data from Below: AI-Driven Seismic Ground-Failure Model for Rapid Response and Scenario Planning
PI: Brett Maurer (UW), bwmaurer@uw.edu, ORCID: 0000-0001-9305-5745
Co PIs: none
AMOUNT & MATCH: $40,000 from PacTrans; $40,000 Match
PERFORMANCE PERIOD: 8/16/2020 – 8/15/2022
STATUS: Completed
CATEGORIES: Scenario Planning, Artificial Intelligence, Ground Failure
DESCRIPTION: The goal of this project is to develop an AI-informed, open source, high-resolution model to probabilistically predict liquefaction regionally – at no cost to the user – both in future scenario earthquakes (to inform mitigation and planning) or immediately following an event (to inform response and recovery). This model will: (i) predict subsurface soil properties using an array of predictor variables obtained from satellite remote sensing (i.e., predict below-ground traits using above-ground information; (ii) utilize machine- and/or deep-learning algorithms; (iii) be anchored to a mechanics-based framework for predicting liquefaction via subsurface soil properties, thus physically constraining the predictions; and (iv) have rapid capabilities, providing regional predictions minutes after an earthquake. The model will first be implemented in PacTrans Region 10 using Pacific Northwest data, but will be scalable to a larger study, and transferrable globally. In addition to providing the model to the transportation industry, the project will use the model to simulate Region 10 events. These will include ruptures on the Cascadia and Aleutian Subduction Zones, as well as crustal faults in the Puget-Willamette Lowlands.
DELIVERABLE | DUE DATE | DATE RECEIVED |
Research Project Progress Report #1 | 10/10/2021 | 10/5/2021 |
Research Project Progress Report #2 | 4/10/2022 | 4/14/2022 |
Draft Report | 6/15/2022 | 6/15/2022 |
Final Project Report | 8/15/2022 | 9/1/2022 |