Leveraging Railway Fiber-Optic Networks with DAS: Multi-Scale Opportunities

Distributed Acoustic Sensing (DAS) enables the acquisition of seismic data with unrivalled spatio-temporal resolution over very large distances. Railway fiber-optic networks, originally deployed for telecommunications, offer cost-effective opportunities to monitor and characterize the subsurface at multiple scales. Here, we present a project conducted with the Swiss Federal Railways (SBB) involving the interrogation of dark fibers running along two perpendicular railway tracks, each approximately 40 km long. Data were acquired over three months using a dual-channel Sintela Onyx interrogator, with variable acquisition setups (spatial sampling, gauge length, and sampling frequency) tailored to different scientific objectives described below.

The primary objective was to assess the feasibility of using pre-existing telecommunications fibers for structural track-bed monitoring, specifically shallow subsurface Vs characterization through inversion of Rayleigh-wave dispersion curves (MASW). This requires high spatial sampling and short gauge length (3 m and 6 m, respectively) to capture short wavelengths. Several ambient noise interferometry strategies were tested, including stacking (1) all available time windows with various preprocessing schemes, (2) only time windows exhibiting strong directional wavefields, and (3) a coherent-source subsampling approach based on a Symmetric Variational Autoencoder to identify time windows contributing the most useful seismic energy. Unsurprisingly, trains constitute the most energetic and reliable seismic sources, from which dense Vs profiles can be derived, demonstrating the effectiveness of both the processing and inversion workflows.

Beyond shallow characterization, the experiment also yielded valuable data to complement dense nodal arrays deployed near Lavey-les-Bains, a site of significant geothermal interest and complex geological structure. The main objectives in this context are to (1) help characterizing the subsurface over the first kilometers, (2) investigate its relationship to geothermal circulation, (3) evaluate the joint use of dense nodal and DAS data for imaging, and (4) establish a high-quality, open-access dataset to support the development of next-generation passive imaging methodologies.

Finally, at an even larger scale, the experiment provided the opportunity to explore how DAS data can be leveraged within the operational Swiss Seismological Service (SED) network and to assess whether DAS can augment standard seismicity catalogues. Lower-resolution data (100 m spatial sampling, 200 Hz sampling frequency) were streamed and converted in real time into standard seismic formats (miniSEED and StationXML), demonstrating the feasibility of integrating DAS data into SeisComP for both automatic and manual processing.

We will present the dataset along with key results relevant to the three purposes outlined above.

We acknowledge Allianz Fahrbahn (grant agreement No. 100 072 202) for enabling this study.