Skip to the content.

KAUST-IAMCS Workshop on Modeling and Simulation of Wave Propagation and Applications

Walter Imperatori, Swiss Federal Institute of Technology (Switzerland) and King Abdullah University of Science and Technology
Broadband Near-Field Ground Motion Simulations in 3D Scattering Media


The heterogeneous nature of Earth's crust is manifested in the scattering of propagating seismic waves. In recent years, different techniques have been developed to include such phenomenon in broadband ground-motion calculations, either considering scattering as a semi-stochastic or pure stochastic process. Here, using several 3D media characterized by Von Karman correlation functions with different correlation lengths and standard deviation values, we simulate broadband (0-10 Hz) ground motions using a 3D finite-difference wave propagation solver. Our goal is to study scattering characteristics and its influence on the wave-field at short and intermediate distances from the source in terms of PGV, PGA and SA. We also investigate other relevant scattering-related phenomena, such as the loss of radiation pattern and the directivity breakdown. We first simulate broadband ground motions for a point-source characterized by a classic omega-squared spectrum model. Fault finiteness is then introduced by means of a Haskell-type source model presenting both sub-shear and super-shear rupture speed. Results indicate that scattering plays an important role in ground motion even at short distances from the source, where source effects are thought to be dominating. In particular, peak ground motion parameters can be affected even at relatively low frequencies, implying that earthquake ground-motion simulations should include scattering also for PGV calculations. At the same time, we find a gradual loss of the source signature in the 2-5 Hz frequency range, together with a distortion of the Mach cones in case of super-shear rupture. For more complex source models and truly heterogeneous Earth, these effects may occur even at lower frequencies. Our simulations suggests that Von Karman correlation functions with correlation length between several hundred meters and a few kilometers, Hurst exponent around 0.3, and standard deviation in the 5-10% range better reproduce the available observations, although any unique estimate of a best model in this distance range is extremely challenging.