Change-point analysis of Vp/Vs ratio time-series using a trans-dimensional McMC algorithm: applied to the Alto Tiberina Near Fault Observatory seismic network (Northern Apennines, Italy)

Time-series of VP/VS ratio have been used to track local changes in elastic properties of rock volumes. Identifying such variations can provide information on the geophysical processes taking place inside a rock volume during the seismic cycle. A value of VP/VS ratio can be computed from traveltime of P and S waves generated from a single local event and it is representative of the value of the VP/VS ratio for the rocks traversed by the seismic ray, between the source and the receiver. It is straightforward, during a seismic sequence, to generate time-series of VP/VS ratio for events located close together and a single station. Such time-series should be able to monitor temporal variations of elastic parameters in the rock volume. Due to the very small nature of the expected changes in P- and S-wave velocity, the evaluation of VP/VS ratio time-series has been problematic in the past, and subjective choices about, for example the time-averaging scheme applied or event selection for constructing the time-series, have been proven to strongly affect the outcomes of the analysis. In this contribution, we present the application of a new methodology for a statistical evaluation of changes in VP/VS ratio time-series. The new methodology belongs to the wide class of 'change-point analysis' algorithms and is developed in the framework of Bayesian inference. The posterior probability distribution (PPD) of the change-point locations is obtained using a trans-dimensional Markov chain Monte Carlo (trans-D McMC) algorithm, where the existence and number of change-points is directly dictated by the data themselves. We apply the new algorithm to the seismic catalogue produced by the Alto Tiberina Near Fault Observatory seismic network (Northern Apennines, Italy). Here the high rate of background seismic release and the dense seismic network allow for a robust statistical analysis. The occurrence of change-points in VP/VS time-series identified with the proposed procedure is represented in space and time. The space-time distributions of change-points in the study area shows a clear peak of change-points following the occurrence of local main events, clustered along the main fault system activated. The robustness of the proposed approach makes it appropriate as an automatic, real-time tool for monitoring rock property changes related to seismic activity.