Recent Mega-Thrust Tsunamigenic Earthquakes and PTHA

The occurrence of several mega-thrust tsunamigenic earthquakes in the last decade, including but not limited to the 2004 Sumatra-Andaman, the 2010 Maule, and 2011 Tohoku earthquakes, has been a dramatic reminder of the limitations in our capability of assessing earthquake and tsunami hazard and risk. However, the increasingly high-quality geophysical observational networks allowed the retrieval of most accurate than ever models of the rupture process of mega-thrust earthquakes, thus paving the way for future improved hazard assessments. Probabilistic Tsunami Hazard Analysis (PTHA) methodology, in particular, is less mature than its seismic counterpart, PSHA. Worldwide recent research efforts of the tsunami science community allowed to start filling this gap, and to define some best practices that are being progressively employed in PTHA for different regions and coasts at threat. In the first part of my talk, I will briefly review some rupture models of recent mega-thrust earthquakes, and highlight some of their surprising features that likely result in bigger error bars associated to PTHA results. More specifically, recent events of unexpected size at a given location, and with unexpected rupture process features, posed first-order open questions which prevent the definition of an heterogeneous rupture probability along a subduction zone, despite of several recent promising results on the subduction zone seismic cycle. In the second part of the talk, I will dig a bit more into a specific ongoing effort for improving PTHA methods, in particular as regards epistemic and aleatory uncertainties determination, and the computational PTHA feasibility when considering the full assumed source variability. Only logic trees are usually explicated in PTHA studies, accounting for different possible assumptions on the source zone properties and behavior. The selection of the earthquakes to be actually modelled is then in general made on a qualitative basis or remains implicit, despite different methods like event trees have been used for different applications. I will define a quite general PTHA framework, based on the mixed use of logic and event trees. I will first discuss a particular class of epistemic uncertainties, i.e. those related to the parametric fault characterization in terms of geometry, kinematics, and assessment of activity rates. A systematic classification in six justification levels of epistemic uncertainty related with the existence and behaviour of fault sources will be presented. Then, a particular branch of the logic tree is chosen in order to discuss just the aleatory variability of earthquake parameters, represented with an event tree. Even so, PTHA based on numerical scenarios is a too demanding computational task, particularly when probabilistic inundation maps are needed. For trying to reduce the computational burden without under-representing the source variability, the event tree is first constructed by taking care of densely (over-)sampling the earthquake parameter space, and then the earthquakes are filtered basing on their associated tsunami impact offshore, before calculating inundation maps. I'll describe this approach by means of a case study in the Mediterranean Sea, namely the PTHA for some locations of Eastern Sicily coasts and Southern Crete coast due to potential subduction earthquakes occurring on the Hellenic Arc.