Study of 600 California Earthquakes Backs Models That Forecast Next-Day Seismic Risk
University of Bristol researchers tested 27 forecasting models on more than 50,000 daily forecasts and found select ones stayed consistent with a decade of real quakes.
Scientists cannot say exactly when and where the next earthquake will strike, but a new study suggests they can say, with a record that stands up to testing, how likely one is on any given day. Nearly 600 California earthquakes have now helped validate the models behind that kind of forecast.
The University of Bristol-led study, published in Nature Communications, evaluated 27 earthquake forecasting models that over more than a decade automatically generated a database of more than 50,000 next-day forecasts for California. Researchers compared those forecasts with almost 600 earthquakes of magnitude 3.95 or greater recorded during the testing period.
The result was that select models described the location, timing and magnitude of earthquake activity consistently, as measured against observations gathered prospectively, meaning the forecasts were locked in before the earthquakes happened. That matters because a model can always be tuned to fit the past.
"No scientific method can currently predict exactly when and where the next earthquake will occur," said lead author Dr. José A. Bayona, a Leverhulme early-career research fellow in the University of Bristol's School of Earth Sciences. "However, probabilistic forecasting can provide useful information about where, when and how large earthquakes are more likely over a defined period."
Bayona said the type of models now used for operational earthquake forecasting in some countries can produce forecasts that remain consistent with observations over long periods. "This provides strong evidence for their use in public operational forecasting systems," he said.
The forecasts were produced automatically by the Collaboratory for the Study of Earthquake Predictability, or CSEP, a global community of scientists who test probabilistic seismicity models against real data. Earlier CSEP experiments evaluated forecasts in Japan, New Zealand and Italy. The California experiment is distinctive for the combination of models tested, the length of the experiment and the number of prospective earthquakes available to judge them.
The researchers also offer what they call a model "recipe": the elements that can be considered when building forecasts consistent with the observed number, locations and sizes of earthquakes. They released an open-source software package with the code, data and documentation needed to replicate the analysis.
The findings could matter to government agencies, civil protection authorities and scientific organizations that are developing, or considering, operational earthquake forecasting. The authors point to recent earthquakes in Colombia, Japan and Venezuela as reminders of why preparation matters. Forecasts of this kind do not tell anyone to evacuate, but they can help officials weigh risk during periods of heightened activity, such as the days after a large quake.
