Faculty Feature: Camilla Cattania, Assistant Professor of Geophysics, finding what’s at fault in the Earth


Despite steady advances in the science of earthquake behaviors, we don’t know exactly when or where the next Big Ones will strike. In fact, “earthquake prediction may never be possible, if you define prediction as a very specific statement about the location, the time and the magnitude of an earthquake,” says MIT assistant professor Camilla Cattania. “What is possible and is routinely done is earthquake forecasting, in which you look at likelihood of an earthquake over a certain period of time.”

Using numerical modeling, earthquake physics, and statistical seismology techniques, Cattania and her colleagues have pushed forward the understanding of earthquake behaviors. Her basic research eventually may help to produce improved forecasts that let us better handle seismic events ranging from Big Ones down to the small quakes produced by industrial operations.

Considering complexity

Earthquakes begin in faults in the earth, created by the movements of the underlying tectonic plates. “Our research looks to provide fundamental physical understanding where we start from a description of friction on the fault, with some relatively simple laws that tell you how the friction changes with time,” Cattania says. “This allows you to start making sense of all these complex statistical patterns in nature.”

Much of her research examines the role of complexity in the geometry of faults, and how that complexity affects seismic events. “When you first come up with a model of a fault, you tend to think of a plane, because that’s the simplest representation,” she explains. “But in reality, if you look at faults on a map, you see geometric complexity, different parallel structures, what looks like fractal geometry.”

In research for her PhD at the German Research Center for Geosciences in Potsdam, Cattania worked on physical models of aftershock sequences that incorporated the geometric roughness of the faults. Large earthquakes produce stresses around their neighborhoods, which then trigger other faults, in a sequence of aftershocks. “We found that when you model all the geometrical complexity, you get a much better forecast for the evolution of the aftershock sequence,” she says.

More recently she’s been studying foreshocks–earthquakes that sometimes precede destructive earthquakes. “I find foreshock sequences particularly intriguing because they might have some potential in terms of earthquake forecasting,” she says. “Once again, geometrical complexity has a really important effect in the behavior of the fault.”

In one illustration, if you assume a large earthquake begins on a simple fault along a plane, a model will show a highly localized precursor signal before the earthquake itself that is very small, very fast and virtually impossible to detect, Cattania says. But if you add the geometrical complexity representing the fault’s roughness, you get a much wider and longer precursory signal, which also can include smaller earthquakes.

Cattania and her collaborators found their models also could match some of the typical features that are seen during foreshock sequences, including an increase in the number of events with time.

The models matched well with the foreshocks observed before some major earthquakes, including Japan’s 2011 Tōhoku earthquake. “So this is an encouraging result, even though we definitely need to think more about which cases in nature are well described by the model,” she says.

Modeling that incorporates fault complexity also may help to answer questions about scaling, which are among the fundamental puzzles in earthquake science. “How does the size of the source affect its behavior?” Cattania asks. For example, earthquakes of a given size may repeat themselves in periodic seismic cycles. Her studies showed that these highly periodic sequences only appear for seismic sources that are quite small–results corresponding well to what’s observed in nature.

Camilla Cattania breaks down her earthquake research in a short series of YouTube videos for MIT ILP

Shaking out answers at MIT

Cattania joined the Department of Earth, Atmospheric, and Planetary Sciences from Stanford University in the summer of 2020—in time to experience the small earthquake felt across southeastern New England in November.

One of MIT’s key attractions was its breadth of research. For instance, her department includes scientists with expertise in characterizing faults with high-precision geodetic data, studying rock behavior in the laboratory, or using seismic data to capture seismic sequences in great detail.

“Interacting with people with a broad range of expertise is very conducive to exciting research,” she says. “People like me who work on theory want to work as much as possible with people who collect the data, both to explain the observations and to come up with ways in which we can test our models.”

Earthquake science globally is benefiting from improved instrumentation and better seismic networks, she says. One striking example is the ability to use fiber optic telecommunications cables on the seafloor as networks of seismometers.

Machine learning methods also are contributing to earthquake studies, allowing investigators to observe seismic events that otherwise would go undetected, she says. Moreover, these technologies may identify new types of seismic signals that bring additional clues about earthquakes, informing theory and computational models.

Collecting insights for improved forecasts

As she pursues her basic research in earthquake physics, Cattania sees the potential for her modeling methods eventually to pay off in operational forecasts.

Over time, modeling opportunities may be found in “induced seismicity” events created by industrial operations, she says. One example is the injection of fluids during fracking or wastewater disposal, which can change the state of stress on faults and trigger earthquakes.

Other aspects of her work eventually may shed light on long-term seismic hazards. For instance, better models could aid in characterizing earthquake probabilities for insurance companies that want to determine the risks of property damage at a given site.

Additionally, understanding how earthquake behaviors change with their scale may have the potential to improve early warnings of major quakes. “If you look at the very beginning of an earthquake, can you in any way predict how large it will be?” she asks. “This might help you save a few seconds in sending out warnings.”

Over time, Cattania hopes that her work to understand all the processes that take place before a large earthquake will help to deliver more tailored forecasts. “It’s a very ambitious goal, but we have some reasons to hope,” she says.