On 30 December 2002 the coast of the volcanic island of Stromboli, in the Tyrrhenian sea, Italy, was attacked by two tsunamis generated by landslides that took place on the north-west flank of the volcano. The landslides and the tsunamis represented the most impressive and threatening episodes of a strong effusive eruption, started on 28 December from a new vent which opened close to the north-east crater of the volcano. In spite of the intensified monitoring carried out in response to the eruption, the landslides and the ensuing tsunamis were not foreseen, and the available instrumental data are insufficient to allow a precise reconstruction of the sequence of the events. The seismic network recorded two main landslides along the steep slope of Sciara del Fuoco, with onset around 13:15 and 13:23 local time (GMT+1). The tsunamis were the direct consequence of the mass movements. Three main post-event surveys helped make assessment on the wave impact on the coast. In this paper the attention is focussed on the accounts of the eye-witnesses, that help us clarify and understand what happened. People in the source area (Sciara del Fuoco) reported a small-volume subaerial slide taking place first, then a sharp cut forming in the sea water down to the sea floor (about 10–20 m deep) and propagating almost parallel to the coastline, be concomitantly associated with a sea retreat and a subsequent sea advance. It is suggested here that the cut was the effect of a large submarine landslide that detached from very close to the coast and produced the 13:15 signal in the recorded seismograms. The second, mostly subaerial, slump was observed to slide down 7–8 min later and to excite a train of waves some distance offshore. Not all the witnesses realised that two distinct tsunamis occurred. The tsunami period was probably in the order of 100 s, but shorter period crests were seen to travel on the top of the long-period waves by several persons. The duration of each tsunami was appreciated to be around 5–7 min. It is difficult to ascertain which tsunami was the largest, since there is no full agreement among the observers, but certainly by accounts both were characterised by large destructive waves.

South-eastern Sicily has been affected by large historical earthquakes, including the 11 January 1693 earthquake, considered the largest magnitude earthquake in the history of Italy (M_w = 7.4). This earthquake was accompanied by a large tsunami (tsunami magnitude 2.3 in the Murty-Loomis scale adopted in the Italian tsunami catalogue by Tinti et al., 2004), suggesting a source in the near offshore. The fault system of the eastern Sicily slope is characterised by NNW–SSE-trending east-dipping extensional faults active in the Quaternary. The geometry of a fault that appears currently active has been derived from the interpretation of seismic data, and has been used for modelling the tsunamigenic source. Synthetic tide-gauge records from modelling this fault source indicate that the first tsunami wave polarity is negative (sea retreat) in almost all the coastal nodes of eastern Sicily, in agreement with historical observations. The outcomes of the numerical simulations also indicate that the coastal stretch running from Catania to Siracusa suffered the strongest tsunami impact, and that the highest tsunami waves occurred in Augusta, aslo in agreement with the historical accounts. A large-size submarine slide (almost 5 km³) has also been identified along the slope, affecting the footwall of the active fault. Modelling indicates that this slide gives non-negligible tsunami signals along the coast; though not enough to match the historical observations for the 1693 tsunami event. The earthquake alone or a combination of earthquake faulting and slide can possibly account for the large run up waves reported for the 11 January 1693 event.

Stromboli is one of the most active volcanoes in the Aeolian island arc in south Tyrrhenian sea, Italy. In the last 100 years the most relevant volcanic eruptions have beenaccompanied by local tsunamis, that have caused damage and casualties. In some cases the direct mechanism of local tsunami generation is clear, i.e. pyroclastic flows entering the sea. In some others it is uncertain and some speculation concerning the collapse of the eruptive column on the sea surface or the failure of some underwater mass can be made. But the ordinary activity is unlikely to generate large regional tsunamis. These can be produced by the lateral collapse of the volcanic cone that geomorphological and volcanological investigations have proven to have occurred repeatedly in the recent history of the volcano, with return period in the order of some thousands of years. The last episode is dated to less than 5 ka BP, and left the Sciara del Fuoco scar on the north-west flank of Stromboli. Based on previous studies, the possible collapse of the nortwestern sector of Stromboli and the consequent generation and propagation of a tsunami are explored. The impact on Stromboli and on the other islands of the Aeolian archipelago is estimated, as well as the impact on the coast of Sicily and the Tyrrhenian coasts of Calabria. The simulation is carried out by means of a double model: a Lagrangian block model to compute the motion of the collapsing mass, and a finite-element hydrodynamic model to compute the evolution of the tsunami. Two distinct tsunami simulations are carried out, one on a very fine grid around the source region to evaluate the tsunami near Stromboli, and one utilising a coarser grid covering the whole south-east Tyrrhenian sea to compute the tsunami propagation toward Sicily and Calabria. It is found that a huge-volume collapse of the north-western flank of the Stromboli cone is capable of producing a regional tsunami which is catastrophic at the source and devastating on long stretches of Tyrrhenian coasts, but particularly in the neighbouring islands of Panarea and Salina, and along the Calabria coasts around Capo Vaticano.