Seismic noise polarization analysis for unstable rock monitoring

In this work, we discuss three different approaches to investigate the polarization characteristics of ambient vibrations collected on unstable rock slopes for monitoring purposes. We consider the Principal Component Analysis (PCA) of the time-domain covariance matrix, the Horizontal-to-Vertical Spectral Ratio as a function of Azimuth (HVSRA), as well as the Singular Value decomposition (SVD) of the Hermitian spectral density matrix. A simple test taking into account seismic noise datasets collected in two acquisition sessions on a potentially unstable rock pillar suggests that the latter method is able to provide more information with respect to the PCA and HVSRA approaches. Processing of the spectral density matrix allows to estimate frequency-dependent polarization parameters, namely the degree of polarization and four angular quantities describing the particle motion in the 3-dimensional space. Provided that processing is properly tuned, frequency-dependent polarization analysis may be able to track subtle changes of unstable rock vibration modes and so to increase our rock failure forecasting capabilities.