| Ground motion recordings have been used widely in analyzing physical properties of the Earth materials and mechanism of earthquakes. The recent advances in computational technology and numerical methods allow seismologists to accurately simulate wave propagation in 3D strongly heterogeneous media and opened up the possibility of extracting more information from waveform recordings for seismic tomography and earthquake source inversion. Furthermore, body wave and surface wave phases can be all used to benefit seismic imaging and source inversion without additional assumptions in the full-wave method. To examine the advantages of full-wave method, we have implemented our tomogaphic and earthquake source inversion algorithms in Southern California where has complex geological structures and has well distributed seismic broadband arrays. We have developed a semi-automated waveform segmentation and selection algorithm and the tool has been applied to large amounts of waveforms for our full-wave earthquake source and tomographic inversions. In our earthquake source inversion algorithm, the receiver Green's tensors, which comprise the spatial-temporal displacements produced by the three orthogonal unit impulsive point forces acting at the receiver, are computed in our updated 3D velocity model. The receiver Green's tensors are stored in disk for efficiently generating synthetic seismograms by using reciprocity between stations and any spatial grid point in our model, and therefore our source inversion algorithm could rapidly invert earthquake source parameters and could provide physics-based ground motion predictions. In our tomographic inversion, the velocity model has been iteratively improved by using the scattering-integral and adjoin-wavefield methods. To accelerate iterations, the LSQR algorithm has been paralleled and optimized for solving our large tomographic inversion systems. The optimized LSQR algorithm reduces both communication cost and memory usage during inversions, so the calculation time and required computational resources are significantly reduced. After 26 iterations, the new crustal velocity model reveals a strongly heterogeneous crustal structure in Southern California. In addition, our tomography also reveals many features that are not shown in the initial model, but have been found in other independent studies. The new tomography enables more accurate physics-based seismic hazard applications and the tectonic reconstruction of Southern California. |
