The ground motion from large earthquakes is often predicted based on finite-fault modeling, in which the fault plane is discretized into small independently rupturing subfaults; the radiation from all subfaults is summed at the observation point. Despite the success of the method in matching observed ground-motion characteristics, the physical interpretation of the subfaults has remained largely unclear, and a rationale for the choice of the subfault attributes has been lacking. Two key parameters—the subfault size and the maximum slip velocity on the fault—govern the amplitude of the source spectrum at intermediate and high frequencies, respectively. We determined these key source parameters, on an event-by-event basis, for all well-recorded moderate to large earthquakes in western North America (WNA) by fitting simulated to observed response spectra. We compare the values of these source parameters with those obtained previously for eastern North America (ENA) and the Michoacan, Mexico, earthquakes (a total of 26 modeled events).
We fin that the characteristic subevent size increases linearly with moment magnitude in an apparently deterministic manner. The subevent size relationship obtained for WNA is not statistically different from that obtained for ENA. In both regions, the subevent size follows the trend of log Δl = -2 + 0.4 M (4 ≤ M ≤ 8), where Δl is the subfault size in km. This trend agrees well with independent studies by Somerville et al. (1999) and Aki (1992), in which the characteristic size of the patches (“asperities” or “barriers”) on earthquake faults was determined. These results indicate that large earthquakes should be viewed as a sequence of smaller events that comprise the large rupture. Interestingly, the characteristic size of these constituent small events appears to be directly related to the size of the overall rupture.
The slip velocities determined for all 26 earthquakes vary in a narrow range from about 0.25 to 0.60 m/sec, with a mean of 0.40 m/sec and standard deviation of 0.09 m/sec. The slip velocities for the ENA events are distributed randomly over this range, while those for the WNA region appear to exhibit a decreasing trend with increasing magnitude. Our results indicate that a generic, region-independent earthquake source model for engineering prediction of strong ground motions can be developed.