Spectrum of strong-motion records for large magnitude Chilean earthquakes

Abstract

SUMMARY We studied the broad-band spectra of the eight largest earthquakes that have occurred in Chile in the last 25 yr using strong-motion records and 1-Hz high-rate GNSS (cGNSS) data. To avoid the numerical instability problem with the double integration of the accelerograms, we computed velocity spectra integrating the acceleration time-series in the spectral domain and compared them to time-differentiated the cGNSS displacement records. To compute the velocity spectrum, we used a multitaper algorithm so as to provide stability over the entire spectral band. We found that the velocity spectra of records obtained close to the main rupture of the earthquakes are different from classical Aki and Brune spectra. The velocity spectrum of large events in Chile presents a flat trend at low frequencies produced by the near-field waves. This trend converges at low frequencies to the static displacement as determined from GNSS data. For different magnitude earthquakes, we observe a transition in the ground-velocity spectrum from a decay of ${f^{ - 1}}$ at high frequencies and a flat trend at low frequencies to a more classical model with a peak at the corner frequency. The source-station distance influences the shape of the velocity spectrum at low frequencies, but there is no simple rule for the records available at present. At intermediate frequencies, the spectra are controlled by surface waves and S waves. We found a transition in the velocity spectrum for the 2014 Iquique earthquake, which indicates a change in the decay of the spectrum for stations at distances greater than ∼200 km. Finally, we show that the flat low-frequency trend of the velocity spectra determined from accelerograms, and the peak ground-displacement (PGD) determined from GNSS data scales with the moment to the power 2/3.