Automatic Picking of Secondary Phases with DAS: The Case of Sedimentary Split Waves

Distributed Acoustic Sensing (DAS) has become a standard tool in seismology, proving capable of replicating most measurements traditionally obtained with conventional sensors. Beyond this, DAS’s ability to provide spatially continuous recordings introduces a novel observational paradigm, revealing features that were previously difficult to detect.

One of the strengths of DAS lies in its capacity to track faint secondary phases, such as converted and reflected waves at the different interfaces of the medium. In this study, we focus on sedimentary regions where seismic waves split at the bedrock-sediment interface, producing double arrivals. Identifying and accurately picking these split phases is essential for precise event localization and leveraging DAS’s dense spatial coverage for high-resolution imaging of active geological structures.

We analyzed one year of data from three DAS arrays deployed onshore along the Chilean margin as part of the ABYSS project. Using a custom tool for efficient 2D phase picking, we manually identified several hundred events, creating a robust training dataset for a deep learning picker. This specialized picker successfully distinguishes closely spaced secondary arrivals, improving automatic detection and processing.

Our results demonstrate that automated picking of these phases enhances localization accuracy and facilitates earthquake processing in sedimentary environments. This methodology can be extended to other types of secondary phases, such as reflection from the Moho, paving the way for fully automated seismic data analysis in the context of complex arrivals pattern using DAS.