Tsunami prediction is also limited by the complex hydrodynamics of tsunami propagation and runup. Considerable effort has been devoted to rigorously defining the effects of nonlinearity and dispersion for tsunamis that propagate over long distances (27). The 1992 Flores Island and the 1993 Okushiri Island tsunamis prompted investigation of the phenomenon of tsunami trapping near islands (28). Runup laws that relate the offshore waveform to maximum runup onshore have been derived theoretically, and these studies have been augmented by laboratory and numerical investigations of the runup associated with breaking waves and shallow beach slopes (29). Efforts are also under way to formulate site-specific inundation models near coastal population centers, which will require more accurate numerical methods that take into consideration topographic effects and bottom friction in the inundation region (30). Local tsunamis generated along continental margins pose a special problem for hazard mitigation (31). A particular issue is the excitation of edge waves, which can result in large-amplitude late arrivals, as observed in a tsunami generated by the 1992 Cape Mendocino, California, earthquake (32).
A major task for earthquake science is to characterize the geographical distribution of seismic hazards. At a particular location, the primary