← Back to Sessions

Geospatial Liquefaction Model Using Maximum Entropy (Maxent) Modeling for California

Description: 

Liquefaction modeling is a critical component of seismic hazard assessment. This study implements maximum entropy modeling to develop probability of liquefaction maps post-earthquake. Maximum entropy modeling is a data-driven methodology requiring only presence points of liquefaction along with geospatial and ground shaking intensity variables as predictors. This study builds on the current state of practice of using broadly accessible geospatial data as predictor variables for liquefaction models, while addressing current limitations of the traditionally used logistic regression model. The logistic regression model requires both presence and absence points of liquefaction; however, the absence points are often estimated and have high uncertainty because ground truth data for liquefaction absence following an earthquake is not generally collected. The maximum entropy model presented here requires only observed presence points of liquefaction, eliminating the added uncertainty in assuming absence points. A second advantage of the maximum entropy model is the ease of including categorical predictor variables such as surficial geology which is known to be a strong predictor of liquefaction occurrence. A dataset of observed liquefaction (e.g. presence) from nine earthquakes in California is used for model development and validation. The predictor variables include: PGV, distance to water, slope-based V_S30, water table depth, average annual precipitation, and surficial geology. The MaxEnt (Phillips et al., 2006) software is used to develop a geospatial liquefaction model based on the input geospatial and intensity data for each event. Then an ensemble model is created for use as a geospatial liquefaction model for the state of California. The model performance is evaluated based on a leave-one-out validation approach using the ensemble model. This research aims to improve the accuracy of map-based rapid estimation of liquefaction surface effects after earthquakes through the use of a novel modeling approach and the inclusion of surficial geology as an additional predictor variable.

Session: Data-Driven Advances in Liquefaction Hazard Analysis [Poster]

Type: Poster

Date: 4/17/2026

Presentation Time: 08:00 AM (local time)

Presenting Author: Maggie E. Roberts

Student Presenter: Yes

Invited Presentation: 

Poster Number: 21

Authors