INDONESIAN OCEAN
Chapter 3. Paleogeodetic records from microatolls above the central Sumatran subduction zone
3.1. Introduction
3.1.1. Motivation
Understanding earthquakes and forecasting future major destructive events is largely hampered by limited data on how seismic strain accumulates and is released in an earthquake cycle [Zachariasen et al., 1999]. The validity and utility of theoretical models can only be rigorously examined if long and detailed records of large earthquakes are available. Unfortunately, well-documented seismic histories that extend through more than one earthquake cycle are uncommon. Thatcher [1989], for example, found only twelve large fault segments in the circum-pacific region for which the record was relatively complete for the past single earthquake cycle. For subduction earthquakes, in particular, evidence is generally sparser because the seismic sources are under water.
Moreover, even in some areas where the history of coseismic deformation is relatively well documented, such as in Japan [Ando, 1975; Kanamori, 1973] data on slow strain accumulation during the interseismic periods is largely lacking. Because, unlike earthquake events, which produce very large signals in a very short timespan, the interseismic periods are characterized by much smaller deformation that accrues over tens, hundreds, or even thousands of years. Thus, to gather evidence of this slow deformation we need a very stable and long-lived instrument that is able to monitor it continuously.
This challenge has motivated us to use massive corals or "microatolls" that are abundant in the Sumatran Equatorial waters. Coral microatolls not only record the magnitude of vertical deformation associated with earthquakes (paleoseismic data), but also continuously track the slow aseismic deformation during the intervals between earthquakes (paleogeodetic data) [Sieh et al., 1999; Zachariasen, 1998b; Zachariasen et al., 2000]. Furthermore, the islands off the west coast of Sumatra, only 70 to 150 km off the trench axis, are ideally located to record surface deformation related to seismic and aseismic subduction. Large extents of the coastlines of these islands are fringed by coral reefs, in which massive corals flourish.
Most of our coral sites are on these forearc islands. The islands, composed primarily of deformed accretionary-prism sediments and coralline limestone, are the sub- aerial expression of the crest of the outerarc rise [Budhitrisna and Andi Mangga, 1990].
All of the Mid-Holocene microatolls found in the islands of Pagai, Siberut, Tanamasa, and Tanabala were less than 2 m above their modern counterparts ([Zachariasen et al., 1999], this chapter). Model of Holocene global isostatic adjustment to deglaciation suggests that sea level in this area reached a maximum of about 2 to 3 m above the present level about 5,000 years ago [Peltier and Tushingham, 1991; Zachariasen et al., 1999]. Thus, we suggest that there has been little, if any, net vertical displacement in the Holocene. In other words, most of the subduction deformations are being accommodated by the elastic deformation of earthquake cycles.
3.1.2. Sumatran active subduction
The Sumatran active plate margin sits over the subducting Australian oceanic plate that obliquely converges about 50 to 70 mm/yr (Fig. 1.1). The plate convergence largely partitions into the frontal component that is accommodated by slip on the subduction interface, and the right-lateral strike-slip component that is mostly accommodated by the 1600 km onland Sumatran fault ([Fitch, 1972; Katili and Hehuwat, 1967; McCaffrey, 1991; Sieh and Natawidjaja, 2000; Sieh et al., 2000], Chapter 2).
The Sumatran subduction zone accommodates the largest part of the plate convergence, and has produced numerous very large earthquakes in the past two centuries (Fig. 1.2). Two very large earthquakes, the 1833 event (M~9) and the 1861 event (M~8.5), dominate the historical seismicity of the subduction interface. Historical records of shaking and tsunamis suggest that these two events involved rupture of all or most of the interface between about 20N and 50S [Newcomb and McCann, 1987]. In the Equatorial region, the principal strain release for the last century is the Mw7.7–1935 historical earthquake (Fig. 1.2) [Rivera et al., 2002]. This earthquake also involved rupture of the subduction interface, but sandwiched in between the locations of the 1861 and the 1833 ruptures. Recently, in June 2000, the Bengkulu earthquake (Mw7.9) occurred near the southern end of the 1833 rupture. Abercrombie et al. [submitted 2001]
have found that this recent event was complex: it involved rupture on the subduction interface, but was also coupled with a strike-slip fault within the down-going oceanic slab. This is consistent with activity on N-S oceanic fractures to the west [Deplus et al., 1998].
Recent upper crustal movements have been documented by the campaign-style GPS measurements that were obtained between 1989 and 1994 [Prawirodirdjo, 2000;
Prawirodirdjo et al., 1999]. These show that the large islands south of the Equator, in the area of the 1833 source rupture, moved in the direction of the relative plate-motion vector (Fig. 3.1). Models of these motions indicate that the subduction interface beneath these islands is currently fully locked [Prawirodirdjo et al., 1997]. Thus, this portion of the subduction interface is efficiently accumulating seismic strain for the next major earthquake. Zachariasen et al. [1999] have suggested, based on sparse evidence from microatolls, that the recurrence interval of the 1833-like events is about 230 years. By contrast, those stations on the islands north of the Equator, above the 1935 source, experienced motions nearly parallel to the trench. These surface movements indicate significant aseismic slip on the interface. Data from our coral “instruments’’ are able to address both the stationary and variability of the seismic and aseismic behavior of the subduction interface through the course of the earthquake cycle.
3.1.2.1. Seismicity of the central Sumatran subduction during the 20th century
In this chapter and in the following chapter, we will focus on the Equatorial region of the Sumatran subduction zone, in which the 1935 event occurred. This part of the subduction interface is unique. From the point of view of seismicity, it is in the intervening region between the 1833 and 1861 seismic sources. In a tectonic framework, this central region is a distinct middle tectonic domain that separates the simpler tectonic domain to the south and the more complex domain to the north [Sieh and Natawidjaja, 2000].
The past century’s seismicity in the region is shown in Figure 3.3. Prior to the 1935 event, almost none of the earthquakes had been recorded by teleseismic networks.
For about 25 years after the 1935 event, the region appears to have been seismically quiescent. However, the lack of recorded seismicity before 1960 could simply reflect the lack of a modern global seismic network. During the 1960s, most of the small earthquakes clustered in a period from 1961 to 1962, in which we found paleogeodetic evidence for a very large “silent earthquake.” During the 1970s, 1980s, and 1990s the occurrence of recorded small earthquakes increased slightly. A large earthquake also occurred during this period (1984, Mw7.1).
3.1.2.2. Coral microatolls of the central Sumatran subduction
In 1997, 1999, and 2000, we surveyed and collected 27 coral slabs from 15 sites in this Equatorial region (Fig. 3.2). Many of the slabs contain records of sea level for the past half-century. Eight of the bigger slabs enable reconstruction of the sea level history for almost the entire 20th century. Together, this paleogeodetic data suite illuminates the vertical deformation pattern of the 1935 event, as well as the pattern of continuous slow deformation before and after the event.
Our coral paleoseismic data for the vertical deformations associated with the 1935 event present a classical example of the upper crustal deformations that are produced by the seismic rupture on a subduction interface. The Tanabala Island that is closer to the trench rose by 90 cm. The eastern part of the region was sunk by as much as 35 cm.
Moreover, our coral paleogeodetic data reveals that in the western part, where coseismic emergence occurred, slow submergence occurred during the decades before and after the
event. In the eastern region, which experienced coseismic submergence, interseismic emergence occurred, slowly raising the islands. Furthermore, the data reveals that the interseismic emergence and submergence rates have varied both temporally and spatially.
This rich data set allows us to model the kinematics and other properties of the subduction interface in a fashion that can explain the observed vertical deformation history.
Among the most interesting phenomena we have discovered in the coral record is the occurrence of a large aseismic event or “silent earthquake” in 1962, 27 years after the 1935 event. This is especially interesting in the light of recent modern instrumental documentation of rapid aseismic events on subduction interfaces, such as in Cascadia [Dragert et al., 2001; Miller et al., 2002], South America [Lowry et al., 2001], and Japan [Heki et al., 1997; Hirose et al., 1999].