Kinematic source generation based on fractal fault geometries

William H. Savran, & Kim B. Olsen

Submitted August 15, 2016, SCEC Contribution #6864, 2016 SCEC Annual Meeting Poster #293

Simulating seismic wave propagation at frequencies of engineering interest (~10 Hz) requires a source description with extraordinary complexity, such as produced by geometrically rough faults. Although spontaneous ruptures on geometrically rough-faults have been shown to produce realistic far-field spectra and comparable fits with GMPEs, they are far too computationally demanding for projects involving a large-scale computational effort. Here, we present our implementation of a kinematic rupture generator that attempts to capture the processes of the rough-fault spontaneous rupture models, at least in a statistical sense, that are responsible for generating realistic source spectra to higher frequencies. Our method uses sequential Gaussian co-simulation to simulate slip, peak-slip velocity (PSV), and rupture velocity (vrup) based on two-point statistics defined by a linear model of co-regionalization. We incorporate 3 variogram basis functions at micro-scale (nugget), medium scale (several hundred meters), and large scale (several kilometers) to capture correlation effects at all scale lengths between all source parameters and the initial friction. In addition, we define a multivariate Gaussian model for one-point statistics (μ, σ) that accounts for correlations between slip, PSV, and vrup. We find strong correlations between (0.6-0.9) for all one-point statistics except σvrup, which shows negative correlations of approximately -0.5. This implies that ruptures that are energetically more favorable (i.e. resulting in larger average slip) tend to propagate at more similar vrup, and visa-versa. We condition the one-point model on average slip to control the M0 of the event. With the one-point and two-point statistics of the source parameters fully defined we parameterize an exponential source-time function at each sub-fault. This fully describes the spatiotemporal evolution of slip on the entire fault plane. Finally, we use the local sub-fault geometry and material properties to translate the slip-rate function into moment-rate tensor components used for ground motion simulation.

Key Words
high-f, kinematic source generator, geostatistics, ground motion prediction

Savran, W. H., & Olsen, K. B. (2016, 08). Kinematic source generation based on fractal fault geometries. Poster Presentation at 2016 SCEC Annual Meeting.

Related Projects & Working Groups
Ground Motion Simulation Validation (GMSV)