Noncharacteristic behavior and complex recurrence of large subduction zoneearthquakes

The last few years have been remarkable with respect to the number of large underthrusting earthquakes in subduction zones that reruptured plate bound ary segments that failed in previous great events. Availability of modern s eismic data for two consecutive large earthquakes rupturing the same portio n of the plate interface provides the opportunity to compare the spatial di stribution of moment release for both events. Such comparisons have been ma de for the plate boundary segments that failed in (1) the 1957 (M-w=8.6), 1 986 (M-w=8.0), and 1996 (M-w=7.9) Aleutian Islands earthquakes; (2) the 196 3 (M-w=8.5) and 1995 (M-w=7.9) Kuril Islands earthquakes; (3) the 1971 (M-w =8.0) and 1995 (M-w=7.7) Solomon Islands earthquakes; and (4) the 1968 (M-w =8.2) and 1994 (M-w=7.7) northern Honshu earthquakes. The spatial distribut ion of moment release for all four of the initial great earthquakes and two of the repeat events has been determined in previous studies. Here, slip d istributions for the three most recent events are determined from inversion of source time functions, derived by empirical Greens function analysis of long-period surface waves and broadband body waves. Comparisons of the spa tial distribution of moment release for sequential earthquake ruptures reve al considerable differences in the pattern of recurrent fault slip. The 199 4 northern Honshu and 1995 Solomon Islands earthquakes primarily fill in ar eas of slip deficit left by their preceding events rather than rerupture id entical asperities. The 1995 Kuril Islands and the 1996 Aleutian Islands ea rthquakes both rerupture portions of an asperity distribution defined by pr eceding events but with variable amounts of slip. This study provides the f irst direct evidence that recurrence of large circum-Pacific plate boundary events is more complex than repeat rupture of dominant asperities. Recurre nt fault slip for the four plate boundary segments studied does not support characteristic slip models either where failure on an entire fault segment occurs repeatedly in events with nearly identical rupture lengths, locatio ns, and slip magnitudes or where failure of individual asperities occurs wi th identical slip functions through consecutive earthquake cycles. These se quential slip patterns are not consistent with physical models of earthquak e rupture where slip complexity is exclusively controlled by invariant geom etric and/or material heterogeneity but suggest that dynamic considerations are also important.