Improved predictions of earthquake damage

No one knows exactly when or where an earthquake will strike. But computer simulations are helping scientists and engineers improve predictions for liquefaction — a sometimes deadly earthquake effect where the soil loses its stiffness, thus toppling buildings and more.

A dataset that provides key input for assessing earthquake-induced liquefaction triggering won a 2024 DesignSafe Dataset award, which recognized the dataset's diverse contributions to natural hazards research.

“The main objective of this project was to provide ground motion intensity measures for liquefaction triggering assessments,” said Renmin Pretell, an assistant professor in Civil & Environmental Engineering at The University of Nevada, Reno (UNR).

Scott Brandenberg (UCLA), Jonathan Stewart (UCLA) and Renmin Pretell (UNR) co-published the award-winning dataset PRJ-4022 | Consistently computed ground motion intensity measures for liquefaction triggering assessment. The dataset is publicly available on the NHERI DesignSafe cyberinfrastructure. The details of the study are provided in a recently published report.

Liquefaction occurs when the ground shifts from an earthquake, breaking up soil stiffness and shear strength as porewater pressure increases and the effective stress holding the soil particles together decreases. Soil liquefaction has been observed and documented in earthquake-prone areas such as California, Alaska, Japan, South America, Turkey, and other places around the world.

“Based on documentation of case histories, liquefaction triggering models were developed,” Pretell said. "One of the main components of these models is the peak ground acceleration at the sites where liquefaction occurred.”

“The goal of our project was to use a systematic technique to compute the peak ground acceleration at all of the case history sites,” Pretell said. The team used their improved technique uniformly across 565 case history sites based on the most recent data sets, which he said have been improving over time.

Pretell’s team not only estimated the peak ground accelerations also other ground motion intensity measures that can potentially advance liquefaction triggering prediction, such as peak ground velocity, Arias intensity, and the cumulative absolute velocity.

“The main finding of our dataset is that there are important discrepancies between the peak ground accelerations that were used in the past and the peak ground accelerations that we estimated based on our systematic approach,” Pretell said.

Liquefaction is one of the most devastating phenomena in geotechnical engineering. Many researchers over the years have devoted their lifetime to exploring and investigating liquefaction, building up datasets that decades later are still in use today.

“The concern is that we have semi-empirical models that rely heavily on the case histories of previous observations, and one of the components is the seismic demand,” Pretell said.

With these models, practicing engineers and researchers can assess the potential for liquefaction triggering for a given site, such as a future housing development. And because there are various techniques for estimating seismic demand, some more subjective than others, this motivated the team to seek improvements to the liquefaction triggering models using the latest dataset available and recent methods and statistical techniques.

The award-winning dataset PRJ-4022 benefitted from several previous efforts. It drew mainly from two big databases established to research earthquake engineering.

The Next Generation Liquefaction Database (NGL) has collected all of the case histories of liquefaction, assigning for each liquefaction site a site characterization of geotechnical parameters to help understand the ground conditions, geology, geotechnical properties of the soil, groundwater levels, and other characteristics that lead to surface liquefaction manifestation at or lack thereof.

The other main source of data used in the PRJ-4022 are from the Next Generation Attenuation Projects (NGA), which include the NGA-West2 for shallow crustal earthquakes in active tectonic regions; NGA-East for stable continental regions; and NGA-Subduction for earthquakes in subduction regions.

“Those are important projects that have collected ground motion records gathered from extensive recordings of earthquakes from stations all over the world,” Pretell said. “Our project benefited from those previous data collection efforts.”

The lifecycle of data typically goes from collection to documentation, and then storage. But the liquefaction dataset PRJ-4022 is different.

“We have used the NGL and the NGA databases, and we have developed spatial correlation models and estimated ground motion intensity measures. One can consider our dataset to be more of a simulation type of data set. Our dataset provides a collection of ground motion intensity measures at liquefaction case history sites for the most important earthquake events, as opposed to raw data,” Pretell said.

“The goal is to benefit any community, more commonly the research community. We have included a Python-based Jupyter notebook in our dataset repository that can be used to compute ground motion intensity measures at locations of interest for the earthquakes we included in our study. To this end, a user just needs to input the site latitude, longitude, and one soil parameter consisting of the time-averaged shear wave velocity,” Pretell added.

What’s more, Pretell developed and updates Python packages on GitHub to facilitate use of the Jupyter notebook.

The spatial correlation models developed as part of their project can also help researchers and agencies to conduct risk analysis on extended infrastructure, such as lifelines and transit.

Pretell’s team extended the dataset to generate spatial correlation models and ground motion intensity measures for the four events of the February 2023 Türkiye earthquake sequence, whose results were published in Earthquake Spectra.

“We provided estimates of the peak ground acceleration at dam, building, and hospital sites and for many different research teams that travelled to Turkey,” Pretell said. Researchers sought to understand, for example, why some hospitals suffered earthquake damage and others did not.

“We studied the spatial variability of peak ground acceleration, and we saw that in some zones there was an overall underestimation compared to the models that we use,” Pretell added.

Another study relevant to the PRJ-4022 dataset focused on the 1989 M6.9 Loma Prieta Earthquake, published February 2024 in Geo-Congress 2024, which estimated peak ground acceleration values at case history sites.

“In this publication, we showed the systematic technique that we are using to compute these ground motion intensity measures for the Loma Prieta earthquake,” Pretell said.

One of the big challenges the scientists had with the award-winning dataset is the mutability of the database, where errors are corrected, and records get added. Every time there is a change in the input databases, new calculations must be run to generate new results — making data versioning important for the project.

“DesignSafe makes the perfect platform because we update our dataset as more data are collected or results are improved, and we release a different version. Having the ability to update our results, while keeping the same name and Direct Object Identifier (DOI) — which DesignSafe helps provide —is tremendously helpful,” Pretell said.

“DesignSafe is also a reliable and long-term platform that the earthquake engineering community is very familiar with,” he added.

 “The key points that people should know is that we have a dataset of consistently estimated ground motion intensity measures at almost all liquefaction case history sites collected to date. This data can be used not only for liquefaction triggering models, but also liquefaction-induced lateral spreading and settlement estimation. This dataset can potentially benefit researchers that are developing methods or models for assessing the hazardous effects of earthquake-induced liquefaction,” said Pretell.

DesignSafe is a comprehensive cyberinfrastructure that is part of the NSF-funded Natural Hazard Engineering Research Infrastructure (NHERI) and provides cloud-based tools to manage, analyze, understand, and publish critical data for research to understand the impacts of natural hazards. The capabilities within the DesignSafe infrastructure are available at no-cost to all researchers working in natural hazards. The cyberinfrastructure and software development team is located at the Texas Advanced Computing Center (TACC) at The University of Texas at Austin, with a team of natural hazards researchers from the University of Texas, the Florida Institute of Technology, and Rice University comprising the senior management team.

NHERI is supported by multiple grants from the National Science Foundation, including the DesignSafe Cyberinfrastructure, Award #2022469.