GNGTS 2021 - Atti del 39° Convegno Nazionale

GNGTS 2021 S essione 3.2 426 Standard methods for ambient seismic noise cross-correlation were applied on vertical components, in the 2-20 Hz frequency range (Mainsant et al. , 2012). Hourly cross-correlograms were computed between site-reference stations (e.g. V2-V1) and filtered in 2-Hz bands. The hourly velocity change (dV/V) with respect to the average of all the considered correlograms was computed by the stretching technique. Short-duration high-energy events were extracted from the ambient seismic noise dataset with a STA/LTA (Short Time Average over Long Time Average) algorithm (STA window=0.3 s; LTA window=30 s; STA/LTA threshold=6). The classification of the detected events was then performed integrating visual analysis of the events spectrograms and k-means cluster analysis on salient time- and frequency-domain event parameters (Colombero et al. , 2018b). The temporal trend and released energy of the natural events possibly related to fracturing and/or slip movements was then analyzed in comparison with the meteorological parameters of the site and the available displacements measurements. Results Both ambient seismic noise and microseismicity analyses carried out on the quartzite tower emphasized seismic features related to the stability of the investigated volume. Ambient seismic noise spectral analysis highlighted amplification in two distinct frequency bands (f1 and f2, Figure 2b) at the stations located at the tower top, almost absent at the reference stations located outside the volume, and thus interpreted as the first two resonance frequencies of the tower. The vibration orientations at f1 and f2 (Fig. 2c) were found to be controlled by the two main fracture sets (K1 and K2) delimiting the potentially unstable volume. Differently from previous case studies on rock columns and prisms, f1 vibration was not found to be perpendicular to the open rear fracture of K1 set, but indicated bending perpendicular to the lateral fractures of K2 set. The main causes of this unusual bending orientation are attributed to the tower size, with a longer extension in the direction of K2 (12 m approximately) with respect to the width in K1 direction (10 m approximately) and the basal dip and dip direction towards the back slope (K3 set). The experimental findings were confirmed by 3D numerical modeling. The latter gave additional information on the tower rear constraints, partially hidden by the overburden on the tower sides. To reproduce the experimental vibration orientations, the tower needed to be modeled without any constraint at the rear fracture, thus probably indicating fracture opening down to the tower base and the absence of rock bridges between the tower and the stable cliff. Resonance frequencies variations at the seasonal scale are visibly controlled by temperature (Fig. 2a): f1 and f2 values increase as air temperature increases at the end of winter months. Velocity changes (dV/V, Fig. 2d) detected by ambient seismic noise correlation generally follow the same seasonal temperature-driven trend. This temperature-driven mechanisms inducing resonance frequency and velocity change reversible variations was previously identified in several case studies and classified as fracture effect (Colombero et al. , 2021). Air temperature fluctuations induce thermal expansion and contraction on the quartzite tower. With increasing temperature at the end of the winter period, tower thermal dilation causes progressive closing of fractures and microcrack, as testified by the decreasing displacements across the fractures at the tower top (C2 and C3 in Fig. 2h). A relative increase in fracture contact stiffness generates an increase in resonance frequency values and seismic velocity. The same temperature- driven fluctuations are depicted on the peak frequencies of the recorded microseismic events possibly related to fracturing processes (Fig. 2e). The peak frequencies interestingly cluster in two separated frequency ranges, probably suggesting the existence of two different and recurrent sources of these events, both sensitive to the air temperature variations. However, event source location is limited by the network geometry and by the absence of a detailed 3D