Earthquake forecasting worldwide is being significantly improved through the use of Deep Learning. An innovative model created by teams of researchers at the University of California and the Technical University of Munich benefits from Deep Learning’s ability to efficiently process large amounts of data. In their scientific article, the researchers present the Recurrent Earthquake Forecast (Recast), a model that plays a central role in this novel method. Thanks to the Recast model and the use of Deep Learning, the immense amount of data now available can be thoroughly analyzed. This enables a significant improvement in earthquake predictions.
The innovative Recast model is proving to be more efficient at processing large data sets compared to the older model. Using earthquake catalogs that included more than 10,000 events, the researchers have demonstrated that the Recast model performs better than the previously existing model. One of the biggest challenges was getting the older Etas model to process the large data volumes effectively to allow adequate comparison.
The Recast model’s flexibility with respect to earthquake prediction exceeds that of the older model. It has the ability to analyze data from regions with frequent earthquake activity, such as Japan, California, or New Zealand. The knowledge gained can then be applied to other areas that traditionally tend to generate less data to provide reliable earthquake predictions for these regions as well. In addition, the Recast model is capable of handling a wider range of data types. At the moment, it mainly uses data classified as earthquakes, however, the Recast model can also handle other seismic data. This allows for more accurate forecasts, which is a significant advance in earthquake prediction.
