GNGTS 2021 - Atti del 39° Convegno Nazionale

341 GNGTS 2021 S essione 2.2 (magnitude, distance, site condition), the time series simulation is done with a lognormal approximating Xs(t,f) : ~ X s ( t,f ) = Pa(t) f √ 2 π δ e - [ lnf-ln β(t) ] 2 /2 δ 2 (3) where  (t) and  are derived from F c (t) and F b (t): ln β ( t ) = ln F c ( t ) − δ 2 / 2 δ = √ ln [ 1 + F b 2 ( t ) / F c 2 ( t ) ] The four model parameters required for the simulation are obtained from regression analyses on motion dataset of shallow active crustal events in Italy (Lanzano et al. 2019) and are embedde simulation code: Arias intensity, significant duration, central frequency, and frequency bandwi frequency modulation is achieved with a lognormal function calibrated with the Brune’s  -squar (3) where β(t) and δ are derived from F c (t) and F b (t): approximating Xs(t,f) : ~ X s ( t,f ) = Pa(t) f √ 2 π δ e - [ lnf-ln β(t) ] 2 /2 δ 2 (3) where  (t) and  are derived from F c (t) and F b (t): ln β ( t ) = ln F c ( t ) − δ 2 / 2 δ = √ ln [ 1 + F b 2 ( t ) / F c 2 ( t ) ] The four model parameters required for the simulation are obtained from regression analyses on motion dataset of shallow active crustal events in Italy (Lanza o et al. 2019) and are embedde simulation code: Arias intensity, significant duration, central frequency, and frequency bandwi frequency modulation is achieved with a lognormal function calibrated with the Brune’s  -squar d The four model parameters required for the simulation are obtained from regression analyses on a strong motion dataset of shallow active crustal events in Italy (Lanzano et al. 2019) an are embedded in the simulation code: Ari s intensity, sig ificant duration, central freq ency, and frequency bandwidth. The frequency modulation is achiev d with a lognormal unctio calibrated with the Brun ’s ω-square model. The amplitude mo ulation in time is obtained using envelopes for the P, S, and coda waves through a lognormal function with the following characteristics: • a modal value correlated to the focal distance; • a standard deviation proportional to the strong-motion duration DV; • an area equal to the Arias intensity; • a total duration 30% greater than the value corresponding to the modal value plus 3DV. The simulation of the accelerograms is performed summing Fourier series with time- dependent coefficients derived from the function in eq (3) and random phases uniformly distributed between 0 and 2π. The computer code has been developed both in Fortran95 and Matlab and requires as input parameters Mw, distance, focal depth, V S30 , and style of faulting. It produces as output acceleration, velocity, and displacement time series, Fourier spectra, response spectra, and a summary statistic of various measures of ground motion. Results Fig. 2 shows a comparison of a simulation with the accelerogram recorded at the AQP station during an aftershock of the April 2009 L’Aquila earthquake. The variability included in the simulated time-series only by the introduction of the random phases, has been calculated for several scenarios differing in terms of magnitude distance and site conditions. The percentage variability of 100 simulations, calculated as (max-min)/min is around 60% for Arias Intensity, 120% for PGA, and 150% for PGV. The value Fig. 1 - Application of the S-transform to the EW component of the ground motion recorded at the station of CSC during the central Italy earthquake (Mw=6.5, R JB = 12.8 km, V S30 =698 m/s) of October 30, 2016: (a) Strong motion record; (b) 2D plot of S-Transform.