Stromboli is an active island volcano, belonging to the Aeolian Archipelago (Italy). In the last 13 ka it experienced four lateral collapses affecting its northwestern flank, with its most recent volcanic crisis (December 2002) associated with landslides and related tsunami events. This paper presents the first stability analysis of the Stromboli volcanic edifice. The main input is the geotechnical model of the volcano, defined on the basis of stratigraphical, lithological, material properties, and structural data, collected from in-situ surveys and laboratory tests. Two-dimensional stability analysis was performed by limit equilibrium methods (LEM) and finite difference modelling (FLAC 4.0 code), mainly focusing on the subaerial part of Stromboli's NW flank (Sciara del Fuoco). The variability of the Safety Factor was studied by deterministic, sensitivity and probabilistic analysis focusing on the effect of external forces, such as magma pressure and seismicity, as potential triggering mechanisms of lateral collapse. The LEM analyses were developed considering a maximum depth of 150-350 in for the sliding surface, to which correspond collapse volumes of 95,000-185,000 m(3)/m, respectively. The study shows that, without external forces, the investigated rock mass is stable and that the tectonic seismicity of the area alone does not destabilize the studied slope. On the contrary, magma pressure in dykes can represent a destabilizing factor. Numerical modelling results are concordant with those from LEM. In addition, the simulation reveals that deformations and superficial landslides, preannounce and contribute to retrogressive plasticization and maybe to failure surface deepening. Shallow submarine landslides represent a possible triggering mechanism, and the landslide events of 30 December 2002 consistently fit the simulated evolution. FLAC has revealed to be a useful tool for modelling such a complex system, it allowed to calibrate the response of the geotechnical model, test the validity of the assumptions, simulate the stress-strain evolution, and prepare a possible model for future scenarios. (c) 2004 Elsevier B.V. All rights reserved.
