The Sources of the two destructive Earthquakes of 1693 retrieved by Automatic Inversions

We automatically inverted the macroseismic intensity data set of the two destructive earthquakes of1693 in SE Sicily, and retrieved the principal geometric and kinematic source parameters of the events.The inversion results obtained by the grid-search method are printed in the Journal of Seismology(Sirovich and Pettenati, 1999), and in the BSSA (Sirovich and Pettenati, 2001). Here, we repeated andimprooved these inversions by: 1) using different intensity catalogs, and 2) by applying a NichingGenetic Algorithm, NGA, to speed up the automatic inversions (Levine, 1996). So doing, we obtainedsolutions which are consistent with those already published. Our technique uses Voronoi polygons andour KF kinematic model (Sirovich, 1996, 1997; Pettenati et al., 1999). It was possible to invert theregional intensity patterns of these two earthquakes because their data sets (one intesity datum per site;sites with heterogeneous geology) are sufficiently uncontaminated by local seismic responses, andretain regional traces of source effects. The Jan. 9 and Jan. 11, 1693 sources - constrained by ourinversions - form a NNE oriented segmented fault, approximately 60km long; then, according to thedifferent intensity data sets, the retrieved fault is steeply dipping towards ESE, or WNW, with rupturemechanism from pure strike-slip to 50% strike-slip and 50% dip-slip. This active structure would crossSE Sicily from the Hyblean Plateau to the coast of the Ionian Sea, south of the city of Catania. It isworth to be mentioned that this complex source best fits the regional distribution of damage observedin the field by various authors. At the limit of the negative error of the dip angle at depth, a value of 54°makes our line source for the Jan. 11 event compatible with the trace of the Scicli-Ragusa-Monte Lauroactive transcurrent fault found in the field, which outcrops 12-14km to the west (Grasso and Reuther,1988; see their Fig. 5). The retrieved complex source is compatible also with the orientation of themaximum horizontal geodynamical compressive stress, confidently measured in the area, as well aswith an authoritative tectonic interpretation, independent from this study. The combined use oftessellation, of our KF model, and of the NGA genetic algorithm is promising for automaticallyinverting intensity data sets of preinstrumental earthquakes. Our automatic technique was recentlyvalidated by treating the data of a recent and well documented earthquake in the Great Los AngelesRegion, California (Pettenati and Sirovich, 2003, in press by the BSSA).