GNGTS 2013 - Atti del 32° Convegno Nazionale

Recognition of nodes related to first rank lineaments. The results are summarized in Tab. 3. Two variants of recognition, A and B have been obtained, using different learning sets D 0 . In the variant A the set D 0 was formed using all the nodes situated most closely to the epicenters of past earthquakes M5+. In the variant B we excluded from D 0 three nodes 56, 75, 76 hosting events before 1500. The results of the recognition are presented in Tab. 4. In both variants, A and B, the same classification of the nodes into D and N classes is obtained. All the nodes hosting past earthquakes have been correctly recognized. Thus, each of the two variants can be considered as the main one. The reliability of the recognition results has been evaluated by a set of control tests relevant to the determination of earthquake-prone areas (Gorshkov et al. 2003). We define as main variants the classification III presented in Tab. 2 and the classification B presented in Tab. 3. Four tests have been performed to analyze the stability of the nodes classification obtained for each group of nodes, e.g. checking the recognition of specific nodes when they are excluded from the training set or looking for possible equivalent traits. The results of the tests performed support the validity of the main variants. Based on the described pattern recognition procedure the set of D nodes illustrated in Fig. 3 has been obtained, corresponding to the classification variant III for the Po plain and variant B for the nodes related to the first rank lineaments. Tab. 3 – Results of recognition for nodes located along the boundaries (1st rank lineaments) between the Po plain and surrounding mountains. The total number of nodes, both D and N, is 42. The total number of large past earthquakes occurred within the nodes is 53. Classification variant A All events from the UCI catalog are used to select D 0 B Only nodes hosting events after 1500 are used to select D 0 Total number of identified D nodes 26 (62%) 26 (62%) Total number of target earthquakes within the identified D nodes 52 (98%) 52 (98%) Discussion and conclusions. The identified earthquake prone areas provide first-order systematic information that may significantly contribute to seismic hazard assessment in the Italian territory, as shown by Zuccolo et al. (2011). The information about the possible location of strong earthquakes provided by the morphostructural analysis, in fact, can be directly incorporated in the neo-deterministic procedure for seismic hazard assessment, thus filling in possible gaps in known seismicity (Panza et al. , 2012). Moreover, the space information about earthquake prone areas can be fruitfully combined with the space-time information provided by the quantitative analysis of the seismic flow, so as to identify the priority areas (with linear dimensions of few tens kilometers), where the probability of a strong earthquake is relatively high (Peresan et al. , 2011; Panza et al. , 2011), for detailed local scale studies. The new indications about the seismogenic potential obtained from this study, although less accurate than detailed fault studies, have the advantage of being independent on past seismicity information, since they rely on the systematic and quantitative analysis of the available geological and morphostructural data. Thus, this analysis appears particularly useful in areas where historical information is scarce; special attention should be paid to seismogenic nodes that are not related with known active faults or past earthquakes. 92 GNGTS 2013 S essione 2.1