Assessing the predicting power of GPS data for aftershocks forecasting

Forecasting large aftershocks is a challenge of great importance for human security. Today we dispose of statistical predictive models called Epidemic Type Aftershock Sequence (ETAS) tuned on the earthquake catalogue of the past seismicity. This catalogues contains basic information such as the location, the time and the magnitude of an earthquake. However we dispose of much richer data set about the crust dynamics, such as the daily displacement of the ground surface, that is nowadays measured by numerous GPS stations, devices that send their absolute position everyday to sattellites, thus telling us about how the ground deforms. In this study, we propose to forecast the Japanese aftershocks by means of a machine learning study of the GPS data alone. Our results show that this method is very promising and relies on the quality and the quantity of the available data.