← Back to Sessions

An Update to the Global Geospatial Liquefaction Model With Uncertainty Propagation

Session: How Well Can We Assess Site Effects So Far? [Poster]

Type: Poster

Date: 4/21/2021

Presentation Time: 03:45 PM Pacific

Description: 

This paper presents an update to the global liquefaction model developed by Zhu et al. (2015, 2017) using geospatial data from earthquakes around the world. An updated liquefaction inventory includes 54 earthquakes, with 7 earthquakes without liquefaction observations. We test 18 explanatory variables as proxies for soil saturation, soil density and the earthquake load including variables such as PGV, PGA, slope-derived Vs₃₀ and water table depth (Fan et al. 2013). Uncertainty in the explanatory variables is defined so that uncertainty can be propagated through the model. In order to study the correlation between the explanatory variables, interaction terms are included in some of the initial predictive models. The area under the receiver operating characteristic (ROC) curve is used as a metric to identify the models with the better performances. The identified candidate models are then studied in a probabilistic framework to estimate the uncertainties of the predictions. The estimated prediction uncertainties guide the choice of the optimal global geospatial liquefaction model and more importantly can help better define the level of confidence in the prediction which is a valuable asset in the decision-making process.

Presenting Author: Mehdi M. Akhlaghi

Student Presenter: Yes

Authors