Kinematic source modelling of normal‐faulting earthquakes using the finite element method

Summary Near‐field full elastic wave synthetic seismograms were computed for simulations of propagating normal faults using a 2‐D, finite element method. The kinematics of faulting were simulated using a modified application of the ‘split‐node’ technique of Melosh & Raefsky. Fault simulations using ‘split‐nodes’ require only a modification of the load vector and correctly models relative fault slip as a double‐couple force distribution without moment. Synthetic seismograms were computed for a suite of dip‐slip, normal fault models designed to illustrate waveform variations due to temporal and spatial changes of the relative slip and rupture velocity along the fault, in addition, to fault slip at the free surface. Results from these models show that distinct P ‐ and S ‐wave phases, observed in the near‐field, are produced by the initiation and termination of faulting. Primarily observed in the hanging wall, these phases are identified as the ‘starting’ and ‘stopping’ phases. Their general amplitude behaviour can be explained by appropriately orientated double‐couple point sources. For fault models with rapidly changing source‐property variations in the slip distribution, rupture velocity, and fault orientation, notable S ‐wave phases were produced and observed in the hanging wall. Comparison of seismograms show that changes in rupture velocity produce larger amplitude variations in this phase than corresponding changes in slip distribution. The amplitudes of the associated P ‐wave phase were generally too small, compared with the S ‐wave, to be observed on the seismograms. Development of the ‘split‐node’ technique for use in synthetic seismogram calculations represents an important new method permitting simple and accurate modelling of elastic waves due to fault propagation in a laterally varying velocity structure.