Establishing continuous seismic monitoring by DAS interrogation of submarine telecommunication cables in Europe with the SUBMERSE consortium: tools and use cases

In the last years, fibre optic sensing methods, in particular Distributed Acoustic Sensing (DAS), have been experimentally demonstrated to be suitable for monitoring Earth System parameters in submarine cables. The SUBMERSE project (SUBMarinE cables for ReSearch and Exploration) aims to develop blueprints for using telecommunication fibre optic cables as sensors by attaching fibre optic interrogators at selected landing stations, also building a data infrastructure for both temporary storage of full resolution data and permanent archival of reduced data sets.

We analyse data from interrogating the East/West oriented Ionian Submarine System cable from both end points, i.e., Preveza, Greece, and Crotone, Italy, along the same fibre. This cable is located to the north of the Kefalonia Transform Zone. This fault zone marks the western termination of the Hellenic subduction system and is one of the most active seismic zones in Greece, with large damaging earthquakes above M > 6 occurring every few years on average.

In addition to acquiring the full large dataset, decimated channels (~100 in each case) acted as virtual seismic stations offshore, acquired at the NOA datacenter for realtime monitoring purposes. We explored various approaches to automated phase picking and magnitude determination on a reduced data set as well as the full resolution data. We also consider other test sites on the Ellalink cable branches extending from Sines in southern Portugal and from Madeira.

In order to support these and other acquisitions, we have developed a range of tools that can be deployed at future sites. We have enabled real-time streaming of DAS data following the standard Seedlink protocol, which allows straightforward integration into existing workflows at earthquake observatories. We have developed an automated, machine-learning-based algorithm for analysing earthquake waveforms and assembled a benchmark data set of earthquake recordings from DAS cables worldwide with labels of P and S arrival times that can serve to further refine machine learning and other automated analysis approaches. Finally, leveraging the Xdas platform (Trabattoni et al. 2025), we have extended the popular SeisBench platform (Woollam et al, 2022) for machine learning in seismology with the ability to efficiently process dense DAS datasets with algorithms/machine learning models operating across either single or multiple channels.