In this month's "In Depth" we delve into the research of our colleague Thomas Bodin, a geophysicist specializing in earthquake studies and risk assessment.
How can we study an earthquake that happened centuries ago, long before the invention of seismographs? This is one of the questions that researcher Thomas Bodin, from the Institut de Ciències del Mar (ICM-CSIC), is trying to answer through a project that combines Bayesian statistics and seismology to improve our understanding of seismic hazard. His work includes an industrial PhD project developed in collaboration with EDF (Électricité de France), under the SIGMA3 research program focused on improving seismic hazard assessments in Europe. Bodin got a PhD in geophysics from the Australian National University and now develops methods for geophysical data interepretation. He is particularly interested in probabilisitic approaches and uncertainty quantification.
A journey into the past
When an earthquake occurs, the intensity of ground shaking varies depending on the distance from the epicenter, the event’s magnitude, and the characteristics of the local soil and geology. For earthquakes that occurred before the 20th century, there are no instrumental records of ground motion, and the information about intensity is only given by witness accounts, reports of damage, or chronicles describing the effects.
From these qualitative observations, seismologists can estimate the magnitude and location of these historical earthquakes. But as Bodin explains, “the resulting estimated magnitudes depend heavily on how well models are calibrated”.
Traditional models that relate the earthquake parameters (magnitude, depth, distance) to observed intensities are typically calibrated using modern, instrumentally recorded small earthquakes, and their accuracy decreases when applied to older or larger events. Moreover, observed intensities are inherently subjective and uncertain, adding another layer of complexity to the reconstruction process.
Bayesian statistics at the service of seismology
Bodin’s project introduces an innovative approach: a Bayesian inversion analysis that combines both instrumental and historical data to estimate earthquake magnitudes and locations of past earthquakes. Instead of first calibrating the model with modern events and then applying it to historical ones, this method treats all data—past and present—as part of a unified system.
“In a Bayesian framework, we don’t just look for the most likely value of each parameter,” Bodin explains. “We also quantify the uncertainty around it, which gives a much more realistic picture of the variability and possible errors in the estimated values.”
The work involves developing and testing algorithms that will first be validated on synthetic and real datasets, and later applied to the French earthquake catalogue FCAT-17, which compiles seismic data from the Middle Ages to the present. The goal is to update the magnitudes and locations of historical earthquakes, providing a more reliable foundation for seismic hazard assessment in France and beyond.
Combining different scales and regions
One of the major challenges of the project lies in the coexistence of different intensity scales used across time and regions. For instance, in France, the historical SisFrance dataset records intensities using the MSK scale, while the more recent database of the Bureau Central Sismologique Français (BCSF) relies on EMS-98. In border regions such as between France and Italy, both systems overlap, leading to inconsistencies.
The Bayesian framework developed by Bodin will allow these datasets to be harmonized through joint inversion, taking advantage of overlapping information between databases. This approach not only improves coherence across databases but also makes it possible to detect regional variations in the IPE parameters—an essential step toward adapting models to local geological conditions.
From the lab to the field
The project is carried out in collaboration with seismologists at EDF, with the support of SIGMA3 —an international program aimed at improving seismic hazard models used by both industry and public institutions.
The collaboration between ICM-CSIC and EDF is a prime example of how public research and private industry can converge for societal benefit. The CSIC contributes its expertise in advanced statistical modeling and inversion techniques, while EDF provides practical experience and access to seismic databases from across Europe.
“This kind of collaboration allows us to conduct cutting-edge research with concrete applications,” says Bodin. “It’s not just about developing theory—it’s about creating tools that can help protect infrastructure and communities from earthquakes.”
The value of industrial PhDs
Thomas Bodin’s work also highlights the growing importance of industrial PhD programs, which combine academic research with applied experience in companies or technology-based institutions. Unlike traditional doctoral programs, which tend to focus on fundamental science, industrial PhDs are designed to bridge the gap between research and real-world innovation.
Across Europe, this model has expanded rapidly as a way to bring science closer to industry, enhance researchers’ employability, and strengthen technological competitiveness.
“Industrial doctorates are an opportunity to train scientists with a broader vision of the innovation process,” says Bodin. “They push you to think not only about scientific quality but also about the practical usefulness of your work.”
Through this project, the ICM-CSIC and EDF team aims to contribute to a more accurate understanding of historical seismicity and to improve the models used to assess seismic hazard in Europe. In an era where resilience of critical infrastructure is increasingly vital, the combination of advanced statistics, historical data, and institutional collaboration could mark a turning point in how we evaluate and manage natural risk.
As Bodin concludes, “understanding the earthquakes of the past is one of the best ways to prepare for those of the future.”