Towards a widely applicable earthquake detection algorithm for fibreoptic and hybrid fibreoptic-seismometer networks

Distributed Acoustic Sensing (DAS) is a promising technology for providing dense (metre-scale) sampling of the seismic wavefield. However, harnessing this potential for earthquake detection with accurate phase picking and associated localisation remains challenging. Single-channel algorithms are limited by individual channel noise, while machine learning and semblance methods are typically limited to specific geological settings, have no physically-constrained phase association and/or require specific fibre geometries. Here, we present a method that seeks to detect seismicity for any geological setting, applicable for any fibre geometry, and combining both fibreoptic and conventional seismometer data to maximise the information used for detection and source localisation. This method adapts a proven back-migration detection method to also include DAS observations, propagating energy from many receivers back in time to search for localised peaks in energy, corresponding to seismic sources. The strengths of this method are capitalising on coherency over many channels to enhance detection sensitivity even in high-noise environments compared to single-channel algorithms, applicability to arbitrary fibre geometries, as well as built-in, physics-informed phase association and source localisation. We explore the performance of the method using three geologically and geometrically diverse settings: a glacier, a volcanic eruption and a geothermal borehole. Our results evidence the effect of spatial-sampling extent and non-optimal fibreoptic geometries, accounting for P and S wave sensitivity, coupling effects, and how the sensitivity of native fibreoptic strain measurements to shallow subsurface heterogeneities can affect detection. Finally, we attempt to also present a method-ambivalent overview of key challenges facing fibreoptic earthquake detection and possible avenues of future work to address them.