Deep learning model boosts earthquake analysis

Known as RECAST (Recurrent Earthquake forecast), the programme was created by researchers at the University of California, Santa Cruz and Technical University of Munich. It’s claimed that RECAST outperformed the current model, known as the Epidemic Type Aftershock Sequence (ETAS) model, for earthquake catalogues of about 10,000 events and greater. The work is published in Geographical Physical Letters.

“The ETAS model approach was designed for the observations that we had in the 80s and 90s when we were trying to build reliable forecasts based on very few observations,” said lead author Kelian Dascher-Cousineau, who recently completed his PhD at UC Santa Cruz. “It’s a very different landscape today.” 

Due to the proliferation of more sensitive seismological equipment and an increase in data storage capabilities, earthquake catalogues have become larger and more detailed. According to study co-author Emily Brodsky, the traditional ETAS model was not built to handle these larger data sets.

“We’ve started to have million-earthquake catalogues, and the old model simply couldn’t handle that amount of data,” said Brodsky, a professor of earth and planetary sciences at UC Santa Cruz.

In order to demonstrate the capabilities of the RECAST model, the group first used an ETAS model to simulate an earthquake catalogue. After working with the synthetic data, the researchers tested the RECAST model using real earthquake data from Southern California. They found that RECAST performed slightly better than ETAS at forecasting aftershocks, particularly as the amount of data increased. The computational effort and time were also significantly better for larger catalogues.

The team believes the model’s flexibility could also open up new possibilities for earthquake forecasting. With the ability to adapt to large amounts of new data, models that use deep learning could potentially incorporate information from multiple regions at once to make better forecasts about poorly studied areas.

“We might be able to train on New Zealand, Japan, California and have a model that's actually quite good for forecasting somewhere where the data might not be as abundant,” said Dascher-Cousineau.