Ⓔ Stochastic Generation of Accelerograms for Subduction Earthquakes by Cristian Otarola and Sergio Ruiz Abstract The generation of accelerograms using stochastic methods has been a very useful methodology for solving the problem of the lack of appropriate strong- motion records for seismic design. Here, we propose the generation of synthetic strong motion for subduction earthquakes that present well-developed P waves and energetic arrivals of S waves associated with the main asperities of the source of these events. The first few seconds of these accelerograms are dominated by P waves; however, the strong motion is a mixture of S and P waves arriving at the same time. The traditional method considers only S waves. We propose to improve the stochastic generation of accelerograms taking into account a stratified velocity model, incident and azimuthal angles, free surface factors, and energy partition to incorporate the P and SV waves in the simulation. Finally, the simulated accelerograms are compared with the observed data recorded on rock by the Integrated Plate boundary Observatory Chile (IPOC) network during the 2007 Tocopilla and 2014 Iquique earthquakes. The use of P , SV , and SH waves in the stochastic simulation allowed us to generate three-component synthetic records. The early seconds are clearly associated with P waves, and the three components reproduce the shape and the amplitude in time and spectral domains for the observed and simulated records. Online Material: Figures showing fit between observed and simulated waveforms, maximum amplitude of acceleration response spectra, peak ground velocities, and peak ground accelerations. Introduction Recently, two large earthquakes occurred in northern Chile: (1) the 2007 Tocopilla (M w 7.8) and (2) the 2014 Iqui- que (M w 8.1). These earthquakes were well recorded by strong-motion instruments of the Integrated Plate boundary Observatory Chile (IPOC) network, deployed in northern Chile to survey a well-identified seismic gap where no mega- earthquakes have occurred since 1877 (Kelleher, 1972; Comte and Pardo, 1991; Peyrat et al., 2010; Ruiz et al., 2014). The accelerograms recorded during both earthquakes show well- developed P waves and energetic arrivals of S waves associ- ated with the main asperities of the source of these events (see Fig. 1). The first few seconds of the accelerograms are domi- nated by P waves; however, the strong motion is a mixture of S and P waves arriving at the same time. The generation of accelerograms using stochastic meth- ods (Boore, 1983, 2003) has been a very useful methodology for solving the problem of the lack of appropriate strong- motion records for seismic design. This idea was proposed by Hanks and McGuire (1981) who observed that the behav- ior of the accelerograms in the high-frequency range could be considered to be stochastic. Boore (1983), using some functional descriptions of the amplitude spectrum of ground motion (Aki, 1967; Brune, 1970), proposed to use a random phase spectrum, such that the simulated strong motion is distributed over a duration that depends on the earthquake magnitude and the hypocentral distance. The methodology of Boore (1983) has been improved by several authors: Beresnev and Atkinson (1997), Boore (2003), Motazedian and Atkinson (2005), among others. Accelerograms simu- lated using this method produced realistic results (Atkinson and Macias, 2009; Ugurhan et al., 2012; Yalcinkaya et al., 2012; Ghofrani et al., 2013). Here, we propose to improve the stochastic generation of accelerograms, taking into ac- count a stratified velocity model, incident and azimuthal an- gles, free surface factors, and energy partition, to incorporate the P and SV waves in the simulation. Finally, the simulated accelerograms are compared with the observed data recorded on rock by the IPOC network during the 2007 Tocopilla and 2014 Iquique earthquakes. Methodology Far-field displacement for P , SV , and SH waves is used to model the Fourier amplitude spectrum of acceleration in a BSSA Early Edition / 1 Bulletin of the Seismological Society of America, Vol. 106, No. 6, pp. –, December 2016, doi: 10.1785/0120150262