The frontiers of distributed acoustic sensing for seismological applications

Distributed acoustic sensing (DAS) is proving to be an effective technology for seismological applications. Its success is due to the ability to deploy DAS instrumentation on the existing ever-growing telecommunication fiber networks across the globe. However, the benefits of DAS are hindered by the sheer volume of data commonly recorded from single-instrument deployments, which can easily reach tens of TBs. Additionally, since DAS measures along fiber strain, new data analysis paradigms are necessary to exhaustively exploit all the information contained within these large datasets. 

We showcase successful applications of DAS experiments using existing fiber cables located in different scenarios, from volcanic systems to densely populated urban environments. To harness the information within these novel datasets, we combine machine-learning tools with efficient algorithms running on high-performance computing architectures. For example, we showcase how the arrival times obtained from PhaseNet-DAS can provide real-time earthquake detection and localization, allowing for the inclusion of DAS data within earthquake early warning systems. Moreover, we demonstrate the capability of integrating real-time streamed DAS channels within seismic network operations. Our processing paradigm is proving to be an effective ground for discoveries and for creating the next generation of seismic monitoring frameworks.