3D Velocity and Density Model of the Los Angeles Basin and Spectral Element Method Earthquake Simulations

We present a 3D velocity model and a 3D density model of the LA basin. The LA basin velocity model was constructed using sonic log and stacking velocity information, provided by oil industry sources and not previously incorporated into southern California velocity models. The density model is based upon a new database of approximately 300 oil industry density logs from across the Los Angeles basin. These logs use gamma ray emissions to determine formation density at samples of about one meter. We have developed an empirical relation between sonic velocity and density by comparing data from approximately 30 wells in which we have both sonic and density logs. For the remaining wells, we have derived relationships between depth and density, and characterized this relationship for the three main stratigraphic sub-divisions of the SCEC Phase 2 model (Quaternary to base Pico Fm., top Repetto Fm. to top Mohnian, and top Mohnian to basement). The density-depth and density-velocity relations will provide independent rules that can be employed to define density and velocity structure in areas where data does not exist, or in other areas with similar lithology to the Los Angeles basin. We use a spectral element method (SEM) for simulation of seismic wave propagation which is currently being implemented on a 156-node Pentium PC cluster at Cal Tech. Preliminary work shows that SEM results using a 1D velocity model for southern California compare very well to discrete-wavenumber results. Both the density structure and velocity structure must be defined in a 3D model for its use in simulations of seismic wave propagation with a spectral element method, to predict the distribution of hazardous ground shaking during large events. Previous work has typically used density values which were predicted by the sonic velocity values; use of our measured density values should provide more accurate ground shaking predictions, and comparison to previous results will provide a useful assessment of the importance of density models in such simulations.