3D seismic model of Marathon area near-surface from cross-correlation of freight trains noise

Trains are powerful and persistent seismic sources of cultural noise that generate high-frequency energy that can be detectable up to 100 km away. Instead of blindly cross-correlating continuous seismic noise records, we suggest an alternative method that uses trains as an opportune noise source. The idea is to detect and characterize train signal segments as a function of the train position along the railway and cross-correlate only station pairs aligned with the train position. By doing so, body and surface-wave signals in the correlation functions are more easily retrieved. To illustrate the new method's potential, we show a case study in a mineral exploration context at the Marathon site in Canada, where we deployed a dense nodal array of 1020 sensors. First, we retrieve high-frequency surface and body waves using train signals only. Then, we use these arrivals in eikonal tomographic schemes to map their local phase velocities. Finally, following clues from numerical modeling, we perform a depth inversion of the surface wave velocities with body-wave constraints in the shallow subsurface. This way, the resulting 3D shear-wave velocity model exhibits a better resolved Vp/Vs ratio in the first layer.