Quantifying uncertainties in location and source mechanism for Long-Period events at Mt Etna

The manifestation of Long-Period (LP) events is documented at many volcanoes worldwide. However, the mechanism at their origin is still object of discussion. Models proposed so far involve (i) the resonance of fluid-filled cracks or conduits that are triggered by fluid instabilities or the brittle failure of high viscous magmas and (ii) the slow- rupture earthquakes in the shallow portion of volcanic edifices. Since LP activity usually precedes and accompanies volcanic eruption, the understanding of these sources is important in terms of hazard assessment and eruption early warning. This work thus mainly aimed at the assessment of the uncertainties in the determination of LP source properties as a consequence of poor knowledge of the velocity structure and location errors. We used data from temporary networks deployed on Mt Etna in 2005. During August 2005, about 13000 LP events were detected through a STA/LTA approach, and were classified into two families on the basis of waveform similarity. We located the most energetic source using three different approaches: (1) a single-station- location method based on the back-propagation of the polarization vector estimated from covariance analysis of three-component signals; (2) multi-channel analysis of data recorded by two seismic arrays; (3) relative locations based on inversion of differential times obtained through cross-correlation of similar waveforms. We then defined a volume of possible source location and performed a full-waveform, moment tensor (MT) inversion for the entire catalog of events. In this manner, we obtained a MT solution for each grid node of the investigated volume. The MT solutions are parameterised through the ratio of the MT eigenvalues. The spatial changes of these ratios allow gaining a complete picture of the uncertainties in the source mechanism/geometry associated with the location errors. We observed the λi / λ1 (i=2,3) ratios varying between (0.1-0.9) and (0.6-0,9), respectively. By the same token, the volumetric components of the retrieved mechanism were observed, varied by almost one order of magnitude throughout the investigated volume.