Advanced Bay area earthquake simulations will be accessible to public

There is a life-or-death difference between an accurate earthquake model and a bad one. Our ability to accurately model earthquakes is limited due to the unpredictable nature of earthquakes, the complexity of the Earth's interior, limited seismic data, and high computational cost.

While high-fidelity intricate physics-based models to predict earthquakes are available, they can only be simulated using high-speed computing systems capable of performing billions of calculations per second. The United States accounts for almost 25 percent of the world's most powerful supercomputers, and access to these computing systems is extremely limited.

A team of scientists at the Berkeley Lab and Livermore Lab, in collaboration with the Pacific Earthquake Engineering Research Center (PEER), are trying to bridge the access gap by adapting the sophisticated multiscale ground motion simulations into agile datasets accessible on everyday computers.

The team used the extremely powerful scientific supercomputer, Summit, capable of performing as many as 200,000 trillion calculations per second, and Berkely Lab's Perlmutter, capable of 70,900 trillion calculations per second. These supercomputers have brought scientists to the threshold of solving a complex problem affecting millions of lives and livelihoods.

An end-to-end computation tool

Scientists are faced with a multidisciplinary problem that requires a joint effort of seismologists, mathematicians, geotechnical engineers, and structural engineers. To this end, the Earthquake Simulation (EQSIM) software has been developed, which embodies an unparalleled computational toolset and workflow for earthquake hazard and risk assessment.

Using an end-to-end approach, it simulates earthquake hazards, from fault ruptures to surface ground motion and earthquake risk, i.e., infrastructure response, capturing the complex interaction between incident seismic waves and infrastructure systems.

EQSIM has recently been used to simulate the effects of a magnitude 7.0 earthquake from a rupture on the Hayward fault line in the San Francisco Bay Area. Harnessing the power of exascale computing, EQSIM has advanced regional-scale ground motion simulation capabilities fivefold.

Publicly accessible computer-generated datasets

The tool can not only predict the effect of an earthquake on buildings but also create synthetic datasets which can enhance our understanding of the complex seismic phenomenon manifold. Instead of relying on limited data from single historical records measured by fixed instruments surveilling known fault lines, scientists can now assess motions everywhere in silico.

The team will also make the datasets publicly available to researchers and practicing engineers. Global collaborations, as a result, will facilitate our understanding of the hazard, performance, and overall resilience of risky areas. Such efforts will enable officials to identify vulnerable structures and efficiently allocate resources.