RESEARCH HIGHLIGHTS

In the last decade, more than 80,000 people globally have died from earthquakes. Advances in early warning systems and improved building codes have shrunk earthquakes’ death tolls when compared to the longer historical earthquake record, but scientists, engineers, and government officials all recognize there is more to be done to improve resilience against these abrupt, violent natural disasters. 

Dr. Alice-Agnes Gabriel, Professor at Ludwigs-Maximillians Universität München (LMU) and the University of California at San Diego (UCSD), dedicates her research efforts to understanding earthquakes at a fundamental level. Her and her collaborators at the Technical University of Munich (TUM), led by Dr. Michael Bader, use high-performance computing (HPC) at the Leibniz Supercomputing Centre (LRZ) to model earthquake dynamics and propagation under a variety of conditions, including coupled with subsequent tsunami risks. 

Recently, Gabriel and her colleagues ran a suite of simulations focused on understanding the differences in earthquake behavior and intensity based on fault size and how so-called “cascading” earthquakes can follow an initial large earthquake in unpredictable ways. “With access to larger HPC systems, we can go beyond these very large “hero” runs that model a certain historical or recent earthquake,” Gabriel said. “We can run more simulations that are focused on exploring uncertainties and include more complicated physics in the process. In the past, we’ve focused on combing earthquakes and tsunamis in our simulations, but we’ve recently focused heavily on fracture mechanics, which is more of an engineering field we are trying to better understand in the geophysical context.” 

Understanding earthquake dynamics at scale

Earthquakes begin when two or more tectonic plates below the Earth’s surface are unable to smoothly move against one another, building up pressure until rock layers eventually give way and the built-up pressure is rapidly released. Plates can grind against one another laterally (strike-slip fault earthquakes), or a portion of one plate can be pushed below the other (subduction earthquake). Subduction earthquakes usually result in more violent shaking at the surface, but that only tells one part of the story. 

Between simulations and sensors, scientists have a good understanding of how the powerful seismic waves unleashed by earthquakes propagate through Earth and at the surface that can shake and ultimately damage structures. Less understood, however, are the small-scale dynamics taking place deep below the Earth’s surface and how that energy can spread to other faults. 

While much of the tectonic energy released does travel as seismic waves to the surface, there is a significant portion of energy spent overcoming the subsurface friction between tectonic plates. That fracture energy can determine if earthquakes stop or jump between weak faults within the rock and result in a so-called “cascade” event. The 2023 earthquakes near the Turkish-Syrian border, for instance, cascaded out from the initial 7.8-magnitude event, triggering a 7.7 magnitude quake from a nearby fault line just 9 hours later and over 500 small and large aftershocks during the following day. 

Gabriel’s team wanted to update earthquake simulation approaches to better consider the distinct physics taking place in large earthquakes separately from those happening in small earthquakes and how those differences can impact an earthquake’s destructiveness. “In this field, we have been assuming that large and small earthquakes are fundamentally the same, so in studying these phenomena, we hypothesize with that assumption in mind,” she said. “By revisiting data that we’ve collected and with the help of modelling, we are starting to collect evidence that maybe they do not actually play by the same rules.” 

Gabriel added that researchers have access to far more data about small earthquakes because they are more common, and that even today, collecting earthquake data directly tens of kilometers deep under the Earth’s surface is impossible. To that end, the team compiled the data available to it, developed mathematical models that could more accurately express the physics happening at different magnitudes and scales of earthquakes, then wrote equations to update the physics behavior at different scales of a supercomputer simulation. With the help of the SuperMUC-NG supercomputer at LRZ, the team discovered a linear scaling relationship between the fracture energy released and the size of a fault that can improve the accuracy of future simulations and fundamental understanding of earthquakes, ultimately helping better inform city planners, engineers, and other stakeholders of the risks of cascading earthquakes and their effects.