Abstract
In this article we present a deterministic study to estimate seismic ground motions expected in urban areas located near active faults. The purpose was to generate bedrock synthetic time series to be used as seismic input into site effects evaluation analysis and loss estimates for the urban area and infrastructures of Thessaloniki (Northern Greece).
Two simulation techniques (a full wave method to generate low-frequency seismograms, <∼1 Hz, and a hybrid deterministic-stochastic technique to simulate high-frequency seismograms, >∼1 Hz) were used to compute time series associated with four different reference earthquakes having magnitudes from 5.9 to 6.5 and located within 30 km of Thessaloniki. The propagation medium and different source parameters were tested through the modeling of the 1978 Thessaloniki earthquake (M 6.5). Moreover two different nucleation points were considered for each fault in order to introduce additional variability in the ground-motion estimates. Between the two cases, the quasi-unilateral rupture propagation toward the city produces both higher median peak ground acceleration (PGA) and peak ground velocity (PGV) values and higher variability than bilateral ones. Conversely, the low-frequency ground motion, that is, peak ground displacement (PGD), is slightly influenced by the position of the nucleation point, and its variability is related to the final slip distribution on the faults of the reference earthquakes and to the location of the sites with respect to the nodal planes of the radiation pattern. To validate our deterministic shaking scenarios, we verified that the synthetic peak ground motions (PGA and PGV) and spectral ordinates are within one standard deviation of several ground-motion prediction equations valid for the region. At specific sites we combined the low- and high-frequency synthetics to obtain broadband time series that cover the entire frequency band of engineering interest (0–25 Hz). The use of synthetic seismograms instead of empirical equations in the hazard estimates provides a complete evaluation of the expected ground motions both in frequency and time domains, including predictions at short distances from the fault (0–10 km) and at periods larger than 2–3 sec.
