Classification of Microseismic Activity in an Unstable Rock Cliff

We installed a microseismic monitoring network on a 300 m high unstable rock face threatening the city of Lecco, in Northern Italy. The network is active since February 2013 and consists of 5 electromagnetic velocimeters, two of which are deployed in boreholes, two temperature sensors in air and in a shallow fracture, and a rain gauge. Regarding the detection of microseismic events, we decided to set the triggering algorithm in order to tolerate false alarms, and, as a consequence, the network has collected several thousands of events so far. Hence, it is necessary to develop an automatic processing scheme able to discard all the recorded events that are not related to the instability of the rock slope. According to the outcomes of previous studies presented in the scientific literature and to careful analysis of the collected data, we first focused on manual classification of recorded signals according to two main classes: a first one grouping events related to the stability conditions of the slope (referred to as MS and local events), and a second one clustering all disturbances (referred to as spikes, mixed event and unclassified noise). Then, we attempted to develop a classification routine in order to cluster possibly all the signals manually classified as MS events, and at the same time having few false positives. The development of classification algorithm involved analysis of parameters in both time and frequency domains, also supported by spectrograms and Radon transform computations, correlation with meteorological datasets, polarization assessment of the 3-component recordings along with principal component analysis. The algorithm we developed has proved to have a satisfactory success rate. We are now focusing on the last step of the microseismic monitoring activity that involves the localization of events related to the instability of the slope