Variable Rupture Mode at Subduction Zones Around the Pacific

The enormity of the 2004 Sumatra-Andaman earthquake, in comparison with 19th- and 20th-century earthquakes in its rupture area, serves as a reminder that a subduction zone may produce earthquakes larger than those in recorded in the past. Historical record and paleoseismological data show that variability in rupture mode is characteristic of some subduction zones. Infrequent, gigantic earthquakes predominate in geologic records, while historic data tell of more frequent, smaller earthquakes. This implies that along the Cascadia subduction zone, great (M > 8) earthquake can occur more frequently than estimated from paleoseismological record. Like the 2004 Sumatra-Andaman earthquake, the giant 1960 Chilean earthquake (Mw 9.5) was unusually large. Historical predecessors of the 1960 earthquake occurred in 1837, 1737, and 1575. However, midway along the 1960 rupture, only the 1575 event produced geologic records of subsidence and tsunami as obvious as those of 1960. The 1837 and 1737 ruptures were probably small, at least at this latitude (Cisternas et al., 2005). Along the Nankai trough of southwest Japan, recurrence of semi-regular earthquakes has been documented in the 1300 years' written history, with an indication of some variability. The easternmost Suruga trough was ruptured in 1854 but not in 1944, leaving a seismic gap for the anticipated Tokai earthquake. The 1707 earthquake ruptured both Nankai and Tokai sources that ruptured separately in 1854 and in 1944 and 1946. The 1605 earthquake seems to be an unusual tsunami earthquake. Near Tokyo, along the Sagami trough, historical records and marine terraces show two types of large earthquakes (1923 type and 1703 type; Shishikura, 2003); their average recurrence intervals are estimated geologically as several hundred years and a few thousand years, respectively. Earthquakes larger than Mw 8.2 can happen along the southern Kuril trench even though they are unknown from the 200-year written history of Hokkaido. Plate-boundary earthquakes close to M 8, at intervals of 100 years or less, had been considered characteristic in this subduction zone. The 2003 Tokachi-oki earthquake (M 8.0), for instance, was preceded by similar earthquakes, from slightly different source areas, in 1952 and 1843. However, tsunami deposits show that unusually large tsunamis repeated at intervals averaging about 500 yr, with the most recent event in the 17th century (Hirakawa et al., 2000; Nanayama et al., 2003). The inferred inundation area is much wider than those typical earthquakes, and is best explained by earthquakes that broke more than one of the historical segments. Only these multi-segment earthquakes triggered deep postseismic creep that produced decimeters of coastal uplift (Sawai et al., 2004).