Probabilistic hazard mapping of secondary pyroclastic density currents generated by paroxysm events at Stromboli (Italy)

Heraklion, Greece

Pyroclastic density currents (PDCs) at Stromboli are typically associated with Strombolian paroxysms. The main mechanisms originating the currents include the eruptive column collapse (i.e. primary PDC), and the remobilization and avalanching of fresh pyroclastic deposits (i.e. secondary PDC). Whereas primary PDCs are commonly confined into the Sciara del Fuoco (SdF), a steep depression on the northwestern side of the island, at least during the paroxysms of 1906, 1930 and 1944 secondary PDCs flowed out of it. These currents were mostly channelized in the valleys on the volcano flanks, and they inundated and buried regions at low elevation, finally reaching the sea and producing temporary deltas of pyroclastic material. The PDCs during the great paroxysm of 1930 reached San Vincenzo village, causing victims and damage to the buildings. In the two paroxysms of 2019 PDCs occurred in the SdF. Although the PDCs were confined inside the depression, they continued to flow over the water for several hundred meters. Moreover, during the paroxysm of July 3, 2019, tephra deposits emplaced on steep slopes right above the village of Ginostra produced small avalanches of incandescent rocks, which fortunately did not get further. Presently, PDCs represent one of the main dangerous manifestation of volcanic activity at Stromboli and therefore need quantitative hazard assessment. In this study we perform new simulations of various 2D depth averaged models of granular flows also considering the effect of some sources of uncertainty. Focusing on secondary PDCs, the source location of the flow was also assumed uncertain, and therefore we adopted a source zonation based on the analysis of topographic slopes and watersheds basins in the upper part of the island, where thick tephra deposits could more likely accumulate during a future paroxysm. Results highlight the areas more exposed to this phenomenon as a function of the model inputs.