The contribution of pattern recognition of seismic and morphostructural data to seismic hazard assessment
The reliable statistical characterization of the spatial and temporal properties of large earthquakes occurrence is one of the most debated issues in seismic hazard assessment, due to the unavoidably limited observations from past events. We show that pattern recognition techniques, which are designed in a formal and testable way, may provide significant space-time constraints about impending strong earthquakes. This information, when combined with physically sound methods for ground shaking computation, like the neo-deterministic approach (NDSHA), may produce effectively preventive seismic hazard maps. Pattern recognition analysis of morphostructural data provide quantitative and systematic criteria for identifying the areas prone to the largest events, taking into account a wide set of possible geophysical and geological data, whilst the formal identification of precursory seismicity patterns (by means of CN and M8S algorithms), duly validated by prospective testing, provides useful constraints about impending strong earthquakes at the intermediate space-time scale. According to a multi-scale approach, the information about the areas where a strong earthquake is likely to occur can be effectively integrated with different observations (e.g. geodetic and satellite data), including regional scale modeling of the stress field variations and of the seismic ground shaking, so as to identify a set of priority areas for detailed investigations of short-term precursors at local scale and for microzonation studies. Results from the pattern recognition of earthquake prone areas (M>=5.0) in the Po plain (Northern Italy), as well as from prospective testing and validation of the time-dependent NDSHA scenarios are presented.