Two aspects of the rupture initiation of earthquakes will be investigated: 1) the relation between the stress distribution on the fault surface and precursors and 2) the relationship between pre- existing fault geometry and the initiation of ruptures. Asperities and stress distribution on the fault surface associated with large earthquakes may be related to short-term precursors and predictability of such earthquakes. If an asperity gets stressed enough to produce dilantancy and then ruptures, short-term precursors are likely to be observed. If rupture originates in a low-stress ("weak") region of the fault and propagates and breaks asperities, there may be no stress-related precursors. Preliminary investigations of several large earthquakes support the above concept. It has been suggested that pre-existing fault geometry and the distribution of physical barriers may strongly influence the initiation and the total extent of earthquake ruptures. In particular, King and Nabelek (1985) pointed out the strong influence fault bends can have on the initiation and termination of rupture and showed several examples. If earthquakes initiate near bends (or other barriers) and if such behavior is sufficiently common, it can be used to identify areas for intensified monitoring. Similarly, if ruptures tend to be confined to regions between bends, the geometry of the fault-zone can be used in estimating earthquake hazard. We propose to test the above hypotheses. We shall determine the locations of asperities on the fault surfaces of large earthquakes by inversion of teleseismic waveforms. The locations of high-stress asperities will then be compared to the locations of observed precursors if any were observed. By inverting for the details of the faulting mechanisms and stress release, we will determine where the ruptures initiated and what segments of faults moved during the earthquakes. Detailed comparisons with geological observations will be made. We will analyze large events world-wide in well monitored areas for which good geological and/or precursory data are available. Initially, we will concentrate on nine large events in China for which we have already collected geological, seismological and precursory data. We will also analyze the most recent (1986) events in California and Alaska.
