GNGTS 2018 - 37° Convegno Nazionale

GNGTS 2018 S essione 2.1 319 after 35 Italian earthquakes. Concerning peak ground motion parameters, PGV, PGA and PGD are considered. Such data are retrieved from the Italian Accelerometric Archive (ITACA ver. 2.2, see Luzi et al., 2017): ITACA contains data recorded by the National Accelerometric Network (RAN, operated by the Italian Civil Protection Department – DPC), the National Seismic Network (operated by Istituto Nazionale di Geofisica e Vulcanologia – INGV) and regional and international networks operated by several providers. This archive provides reliable evaluations of the above parameters, containing 32,271 three-component accelerometric waveforms generated by 1524 past Italian earthquakes, with magnitude greater than 3, in the time frame 1972-2016. Instead, macroseismic observations are collected from the DataBase of Macroseismic observations of Italy DBMI15 (Locati et al. , 2016), which is a large and homogeneous macroseismic collection of past Italian earthquakes covering a time period ranging from year 1005 to year 2014. A total number of 122,701 macroseismic records is contained in DBMI15, resulting from 3212 different earthquakes and revised from many studies, reports and bulletins. Part of the recorded seismic events follows the EMS98 scale, with intensities arranged in classes spaced by 0.5 intensity units (ranging from 1 to 12). However, non-numerical values may also be assigned to some localities (e.g. F-felt, HF-high felt, D-damage); in this case, corresponding intensities were assigned according to the convention described in the catalogue (Locati et al., 2016). By cross-matching these two types of data - macroseismic intensities and PGM parameters - an exhaustive and homogeneous database of 607 pairs of PGM/EMS98- intensity points, arising from 35 different Italian seismic events was derived. S eismic events contained in the dataset cover a time span between 1983 and 2016 (3.2 6.1), with I EMS 10; recent earthquakes of Emilia (2012) and Amatrice (2016) are included too. Figure 1 shows the detailed spatial distribution of the events considered. The association between macroseismic intensity with PGM parameters was carried out only for such sites located within 6 kilometers from an accelerometric station. This choice ensures a reasonable congruence between the observed macroseismic intensity and PGM measures, allowing maintaining a relatively high number of data pairs. The sufficient numerousness of the dataset allows deriving reliable regression equations. Within this distance there is a nearly uniform distribution of macroseismic intensities (see Fig. 2), with some scatters at the two extreme values (I EMS 2.5 and I EMS 8.5). Indeed, collected data are principally concentrated in an intensity interval between 2-8, even though the maximum value of I EMS covered by the database is 10. The same good spatial distribution was observed for PGM parameters. Data processing. Two approaches can be applied for the regression analysis, and depend on the method used for data processing. Indeed, it is possible to arrange the intensity observations in classes and, for each, the mean value of the recorded ground motion parameter is associated. Alternatively, the whole dataset can be used, without any grouping procedure, even if a robust statistic method for outliers detection and removal should be applied. In this work, the former approach is used, as done also in Faenza and Michelini 2010, i.e. the mean value of logarithm of PGMs μ g and their standard deviation values σ g are used, and EMS98 intensities are grouped in classes at 0.5 intensity bins. This choice is motivated by the distribution of PGMs data about the logarithmic means, which is in agreement with theoretical normal distribution curves. To verify the likelihood of the normal distributions, the 1-sample Kolmogorov-Smirnov normal test is carried out, confirming the above hypothesis. It is worth to recall that, for some intensity classes and ground motion measures it was not possible to compute mean values: indeed, the number of observations was considered insufficient to allow a reliable regression analysis of the data. These limits define the range of applicability of the proposed equations, as it will be further explained in the next Section. As a control tool, data points were also used for regression analysis without any data binning and averaging procedure. Results. The ODR regression (Boggs et al. , 1988), applied to pre-processed data, allows obtaining linear relationships that correlate EMS98 intensity (I EMS ) and base-10 logarithm of