It has been a scary July so far for many residents of California. A magnitude 6.4 quake struck on July 4 near Ridgecrest (about 200 kilometers northeast of Los Angeles), followed by a magnitude 7.1 quake in the on July 5. And there have been many aftershocks since. Scientists all across the state and the country are trying to understand how this quake unfolded – and what it could tell us about future earthquakes in this region. Although we have been lucky this time with no major loss of lives, the terrible headlines are all too familiar. A major earthquake strikes, with a devastating impact on lives, economies, and the environment. The initial event often triggers additional disasters such as fires or tsunamis that unleash substantial damages. In some cases, the impact of a major seismic event will continue for decades or longer, as we’ve seen in the Fukushima Daiichi nuclear disaster of 2011.

We can’t do anything to stop earthquakes, but we can take steps to help mitigate their impact. What if we could better understand the probabilities and paths of quakes, so we could gain the insight needed for better preparation before an event strikes—and provide faster, more effective responses after it occurs? Imagine how much more thorough our urban planning could be if we had the ability to more accurately predict how a major quake would impact buildings and infrastructure in urban areas. Gaining this level of insight could also better inform structural engineering to help minimize the possibilities of catastrophic failures in buildings and infrastructure – potentially saving lives, while bringing down economic and environmental costs.

In the immediate aftermath of a quake, it’s critical for public safety agencies to move emergency response teams and equipment to sites for search and rescue. Rapid response is essential to saving lives and preventing additional losses from fire, gas explosions, aftershocks, or other events that follow the initial shock. Consider how much more efficient first responders could be if they had more awareness about which structures and regions would be expected to suffer the most severe damage. Understanding which roadways and bridges are most likely to be damaged and unavailable could also inform a better understanding of the best response routes to expedite the delivery of emergency teams to areas where they are most urgently needed.

Seismic Modeling, the Foundation of Better Earthquake Safety

The first step in earthquake safety is gaining a better understanding of an event’s effects. To do it, scientists apply data from sophisticated models and simulations. Better insight into the probabilities and paths of quakes can help form the basis for seismic engineering to improve design and construction. It can also help organizations devise better emergency response plans.

For example, researchers could create a ground motion simulation that produces data that they might apply to probabilistic seismic hazard assessments. Seismic hazard analysis describes the potential for dangerous, earthquake-related natural phenomena such as ground shaking, fault rupture, or soil liquefaction. The output for this type of analysis might be, for example, “the earthquake hazard for this site is a peak ground acceleration of 0.28g, with a two percent probability of being exceeded in a 50-year period.”

A structural engineer could use this type of analysis as a starting point for additional studies, as they determine the best approach to designing or retrofitting a structure.

How Can We Make Our Models Accurate?

Seismic modeling is a robust method to develop insight that can be applied to earthquake safety initiatives. However, geologic systems are complex, so ensuring the accuracy and scope of seismic models is challenging.

First, seismic models must cover a broad frequency band. Different structures respond to various frequencies in different ways, so an effective model will need to encompass a wide range of possibilities. For example, tall buildings resonate with relatively low frequencies, while more rigid structures like houses are more sensitive to higher frequencies. Modeling seismic waves over a broad frequency band requires immense computing power.

Seismic modeling must also span four dimensions. It must cover the three spatial dimensions where waves propagate, as well as the dimension of time, which further increases computing demands.

Finally, accurate earthquake modeling must encompass systems of faults rather than a single, isolated fault. This makes it challenging to predict the future states of fault systems. Simulations must test as many scenarios as possible to approach a strong level of accuracy.

Public-Private Collaboration: A New Approach to Modeling

An innovative software package developed by researchers at the San Diego Supercomputer Center (SDSC) at UC San Diego and Intel can take seismic modeling capabilities a major step forward. New Extreme-Scale Discontinuous Galerkin Environment (EDGE) software is designed to take advantage of the latest generation of Intel processors. The research is the result of a joint initiative that also includes the Southern California Earthquake Center, which is one of the largest open research collaborations in geoscience. This sophisticated software offers seismic resources a new level of performance and scalability including:

• Record-Breaking Speed: EDGE is the fastest seismic simulation yet devised, and is capable of 10.4 PFLOPS (Peta Floating-point Operations Per Second, or one quadrillion calculations per second). This accelerated performance gives researchers the speed required for more simulations at the same frequency so they can explore scenarios more thoroughly.
• Multiple Events, One Execution: The EDGE initiative was built to simulate multiple events in one software execution. One of the most innovative features is its ability to apply similarities in setups for efficiency, such as sharing mountain topography in multiple simulations. What’s the benefit of improved seismic efficiency? Simulating multiple earthquake events in one execution leads to savings in time and cost. Researchers can run two to five times more simulations to make the most of their resources.
• Scalability is Fundamental: Among the most dramatic advances that EDGE software delivers is scalability. This is a powerful advantage compared to other modeling initiatives because it lets researchers increase modeling frequency range. But to take advantage of this capability, researchers need a compute infrastructure that can run at a large scale.

Realizing the Potential of EDGE by Scaling to the Cloud

Cloud offerings provide a combination of infrastructure flexibility and scalability that EDGE requires for its groundbreaking modeling. Amazon Web Services (AWS) can support application-tailored clusters that are capable of delivering maximum performance. Utilizing Amazon Elastic Compute Cloud (Amazon EC2), researchers can customize high-performance compute clusters to match EDGE demands.

The key to a good cloud solution for high performance computing i is its ability to offer elastic scalability, providing access to multi-petaflop machines capable of delivering extremely fast performance. It offers researchers access to virtually unlimited capacity and scale so they can grow the infrastructure they need to align perfectly with their workloads.

Just a few years ago, access to this level of computing power was only available via on-premises supercomputing centers. Today, the cloud offers virtually unlimited HPC infrastructure on-demand to researchers. The EDGE project is the first work of its kind to be engineered for such a large scale.

Making Better Modeling Available to More Researchers

Innovative seismic modeling has potential to improve our understanding of earthquakes and our ability to plan and mitigate their effects. To share these benefits, the challenge is putting the latest modeling approaches in the hands of as many professionals as possible.

Cloud offerings such as HPC on AWS offer a flexible, cost-effective solution to put supercomputing power in reach for more researchers. They provide the elastic scalability to take advantage of software that was created with scalability in mind, such as EDGE.

Making the latest, most innovative research tools available to more people in the scientific research community sets the stage for continuing innovation and new discoveries. Armed with the analysis and improved insights they need to make better decisions, engineers and public safety professionals will gain the ability to plan more proactively before a quake occurs. They will gain the understanding needed for improved responses after a seismic event to minimize damage and save lives.

If you’d like to learn more about how EDGE, paired with scalable HPC on AWS, is redefining the possibilities of seismic modeling, read this paper “Petaflop Seismic Simulations in the Public Cloud” that was recently presented at ISC High Performance 2019 in Frankfurt. You can find additional resources about using AWS for your HPC workloads here.