Scientists’ Model Reliably Forecasted Complex Earthquake Sequence

Damaged buildings at Pescara del Tronto after the 2016 Amatrice event, with soft soils that collapsed during the earthquake. |Luigi Innocenzi

Scientists testing a newly created earthquake forecasting model said that the model reliably forecasted the time, location and strength of a complex Italian earthquake sequence. Their study, published in the September 13 issue of Science Advances, is the first to determine the accuracy of this model in a real experiment, and not just based on past data, according to author Warner Marzocchi of Istituto Nazionale di Geofisica e Vulcanologia in Rome, Italy.

The research examined the 2016-2017 Amatrice-Norcia earthquake sequence, which included 40 target earthquakes that severely damaged areas in Central Italy. "This kind of modeling represents a first generation of models that may establish a reference to measure future improvements in forecasting abilities," explained Marzocchi.

According to Marzocchi, the model could provide seismologists all around the world with the tools they need to accurately forecast complex earthquake sequences, which will provide reliable information for many different types of decision-makers in an earthquake zone. "This kind of forecast may help … plan rational actions for rescue teams after a damaging earthquake as well as provide information for the reopening of industrial activities in the aftermath of a destructive earthquake," he said.

The ability to accurately forecast earthquakes has remained a challenge, especially for seismic events that do not follow typical mainshock-aftershock sequences, where one large earthquake is followed by many smaller aftershocks. The most widely-accepted aftershock forecasting model, the Reasenberg and Jones (R&J) model, is used to predict these common earthquake sequences. However, the R&J model does not account for information regarding the specific location or even broader geographical area where the sequences take place, and assumes that earthquakes exceeding a specific magnitude will produce other earthquakes in addition to the mainshock.

The R&J model cannot reliably forecast earthquake sequences like the Amatrice-Norcia sequence, since it does not follow the typical mainshock-aftershock sequence. "This [Italian] sequence had several large earthquakes with several bursts of seismicity, and did not behave like classical aftershock sequences which produce smaller earthquakes with the number of aftershocks decaying continuously overtime," said Marzocchi.

Researchers have recently developed new models to account for characteristics of uncommon sequences like the recent events in Italy, but to date, the reliability of these models in predicting earthquakes has not been tested.

In an effort to gauge such models' statistical reliability, Marzocchi and his colleagues analyzed the weekly earthquake forecasts that occurred during the 2016-2017 Amatrice-Norcia sequence, to test the accuracy of the operational earthquake forecasting (OEF) system. The OEF system combines aspects of new models that account for important forecasting factors like location that are missing in commonly used models like the R&J model. The authors found that the OEF system provided statistically reliable forecasts when compared to the observed events of the largest earthquakes in central Italy during the sequence.

Marzocchi said that while the model is still in its pilot phase, "the results modify the way in which the earthquake predictability problem has been approached by scientists and perceived by society."

The next step for scientists is to collect more information concerning earthquake processes, like earthquake generation and fault interaction. This information will shed light on the distinct properties of earthquakes, which can be incorporated into the OEF system to help achieve complete knowledge to correctly forecast earthquakes.

[Credit for associated photo: Marco Anzidei]