New deep learning model shows promise for improved global earthquake forecasting

Scientists develop deep learning model RECAST that outperforms traditional models in predicting earthquake aftershocks, especially with larger datasets.

RECAST demonstrates superior adaptability and scalability compared to existing ETAS model when working with catalogs of 10,000+ events.

With ability to handle larger datasets, RECAST promises improved global earthquake forecasting by leveraging more comprehensive data.

RECAST slightly outperforms ETAS in test using real Southern California earthquake catalog data.

Model's flexibility could allow earthquake forecasts to incorporate multi-regional data to better predict areas with limited information.