Deciphering train-induced body waves in correlation functions to promote fault monitoring

Continuous monitoring of active faults at seismogenic depths is key for the detection of potential precursors and to advance our understanding of earthquakes. In the context of the FaultScan project, we use seismic waves generated by freight trains – energetic and repetitive sources – recorded by two dense seismic arrays on both sides of the San Jacinto fault to directly observe transient deformation. Computing correlation functions for specific time segments corresponding to the passage of the trains reveals clear signals around the expected P-wave arrival time. The principle of Green’s function retrieval cannot be invoked for their interpretation, since the required assumption of a homogeneous source distribution is far from being met. To address this problem, we rely on correlation seismology, an emergent field of research that acknowledges the fact that correlations are shaped by both the distribution of ambient noise sources and Earth structure. Correlations are interpreted as self-consistent observables.

We first give a general introduction to (i) the forward problem of modeling correlation functions for arbitrary noise source distributions in space and frequency in potentially 3D heterogeneous and attenuating media, and to (ii) the computation of sensitivity kernels for noise sources and Earth structure. We then present an application of this framework to study the described problem of reconstructed body waves in the context of the FaultScan project. Studying structure kernels illustrates how signals are formed by the interaction of different P-wave phases and which parts should be used to monitor the fault. Forward modelling different source scenarios and structural changes further contributes to the understanding of the observed signals and highlights source-structure trade-offs.

Combining new and creative data-driven experiments with methodological developments is a promising way forward and has the potential to accelerate new discoveries.