¹ Department of Earth and Planetary Sciences and Division of Engineering and Applied Sciences, Harvard University, Cambridge, MA 02138, USA

² Now at Bullard Laboratories, University of Cambridge, Cambridge, CB3 0EZ, UK

Abstract

Coseismic changes in Coulomb shear stress have been shown to correlate well with changes in seismicity following great earthquakes for both transform faults in continental regions and, in our work, for dip-slip faults at subduction zones [Dmowska et al., 1996, Taylor et al., JGR, in press, 1998]. These studies almost invariably adopt the approach of applying the inferred slip from one or more inversions of the main event to a 3D elastic half-space dislocation model to calculate the resulting Coulomb shear stress changes. There was no such inversion available immediately after the Biak, 1996 event, and so we took the alternative approach of using the spatial distribution of post-mainshock upper-plate seismicity to infer information about the distribution of slip in the main event. Our 3D subduction models with highly heterogeneous slip along-strike reveal distinct, characteristic patterns for the distribution of stresses in the upper-plate. The shear stress on arc-parallel strike-slip faults separates into two lobes, one of increased and the other decreased coseismic stress change. The extensional stress changes resolved onto normal faults with trace inclined at moderate to large angles to the trench likewise form two lobes of increased and decreased change. Based on this pattern and the positions and mechanisms of the first two upper-plate events (Feb. 17 and 18), the area of highest seismic slip was placed in the first week after the main event and confirmed by the subsequent year of seismicity [Taylor et al., 1998].

A subsequent inversion for the event by Kikuchi [1998] using the subevent deconvolution method of Kikuchi and Kanamori [1991] reveals a two phase source-time function. The rupture propagates roughly NW with the initial phase (M_w = 7.3, 19 s) followed by a second (M_w = 8.1, 32s) in which most of the moment is released. A comparison between this inversion and the position of highest slip inferred from coseismic stress changes shows they are essentially co-incident. The remarkable correlation between these results provides grounds for confidence in our method of approximately placing the position of highest moment release in a main event.
 
 

• The February 17, 1996 (M_w = 8.2) Biak earthquake ruptured at least 270 km along the New Guinea trench (Figures 1 and 2). The event was a thrust in a zone of very oblique convergence, and was followed by a number of events in the upper-plate, the two largest of which were the M_w = 6.4 (Feb. 18) and 6.5 (Feb. 17) events that occurred within 2 days, SW of the mainshock.

 
• Coseismic changes in Coulomb shear stress have been shown to correlate well with changes in seismicity following great earthquakes for both transform faults in continental regions and, in our work, for dip-slip faults at subduction zones [Dmowska et al., 1996, Taylor et al., 1998]. Our 3D subduction models with highly heterogeneous slip along-strike (Figure 3) reveal distinct, characteristic patterns for the distribution of both shear and extensional stresses in the upper-plate (Figures 4 and 5). Based on this pattern and the positions and mechanisms of the first two upper-plate events (Feb. 17 and 18,), the area of highest seismic slip was placed in the first week after the main event and confirmed by the subsequent year of seismicity [Taylor et al., 1998] (Figure 2).

 
• A subsequent inversion for the event by Kikuchi [1998] reveals a two phase source-time function (Figure 6). A comparison between this inversion and the position of highest slip inferred from coseismic stress changes shows they are essentially co-incident (Figure 7). The remarkable correlation between these results provides grounds for confidence in our method of approximately placing the position of highest moment release in a main event.

Coulomb shear stress on fault:

f = coefficient of friction

is the change in left lateral Coulomb shear stress, .

3D FEM model:

Use simple 3D finite element model of generic subduction zone [Dmowska et al., 1996, Taylor et al., 1998] (Figure 3) to calculate the extensional and shear stress changes due to oblique slip on an asperity (dark shaded region).

Slip, magnitude 'D', applied on asperity on thrust interface and modeled as "freely slipping" elsewhere (slip calculated to assure zero net coseismic change in shear stress).

Parameters in model:

= dip angle of subducting slab (~ 11° for Irian Jaya)

= angle of oblique slip from trench-normal (~ -13° for Irian Jaya)

W = down-dip width of seismogenic thrust interface (e.g. 100 km)

D = magnitude of imposed slip on asperity (e.g. 3 m)

= shear modulus (e.g. 30 GPa)

Calculate coseismic left-lateral Coulomb shear stress  and extensional stress  changes consistent with two largest (Feb. 17 and 18) upper-plate events (Figure 2):
 

 

 Figure 4: Coseismic change in left-lateral Coulomb shear stress  on arc-parallel faults for Irian Jaya-like parameters,  = 11°,  = -13°, and for f = 0.4.

Figure 5: Coseismic change in extensional normal stress  on faults oriented at 30° anti-clockwise from trench for Irian Jaya-like parameters,  = 11°,  = -13°.

Inferred position of highest slip:

Although most of upper-plate seismicity exhibits distinctly non double-couple solutions, still easily divisible into distinct groups relative to centroid of main event (Figure 2):

• Extensional events are to "right" (west) of centroid
• Strike-slip events are to "left" (east) of centroid

At time of analysis of the Biak earthquake no available inversion of the source-time function or distribution of slip from the main event. However, subsequently, an analysis by Kikuchi [1998] was carried out using the subevent deconvolution method of Kikuchi and Kanamori, [1991].

Figure 6 shows details of inversion:

• Hypocenter determined as 0.99S 137.43E depth 14 km.

• Earthquake consists of two phases, initial phase of M_w = 7.3 (duration 19 s) and main phase of M_w = 8.1 (32s).
• Main phase located 60 km NW of initial phase.
• If the rupture propagation time estimated 28s of 32s, with propagation velocity of 3 km/s, length of main phase fault is 80 km.

Right: Comparison of P arrivals at six stations
Bottom-left: CMT solution with station positions marked
Center-left: Source-time function
Top-left: Distribution of slip relative to epicenter.

• 3D FEM models of heterogeneous oblique slip along strike in subduction earthquakes produce characteristic patterns of coseismic change in extensional and Coulomb shear stress,  and .

 

• Comparison of patterns of stress change with upper-plate seismicity immediately following 1996 Biak earthquake allow us to suggest the position of highest slip (asperity) in the main event.

 

• Our inferred asperity position agrees remarkably well with that of a subsequent inversion [Kikuchi, 1998].

 

• This result provides grounds for confidence in the methodology of using upper-plate seismicity to infer regions of greatest moment release, something possible one day after the main event in this instance. However, this analysis was possible in Irian Jaya mainly because it is tectonically very complex - many pre-existing shallow faults, raising likelihood of seismicity being triggered by the relatively small stress changes induced in the upper-plate by the main event.

References:

Kikuchi, M., Inversion of the February 17 1992 Biak Earthquake, details available from ftp://ftp.eri.u-tokyo.ac.jp/pub/reports/YCU_News/Y51/txt.51, fig.2.ps, 1998.

Kikuchi, M. and H. Kanamori, Inversion of complex body waves, III, Bull. Seism. Soc. Am., 81, 2335-2350, 1991.

Puntodewo, S. S. O., R. McCaffrey, E. Calais, Y. Bock, J. Rais, C. Subarya, R. Poewariardi, C. Stevens, J. Genrich, Fauzi, P. Zwick, and S. Wdowinski, GPS measurements of crustal deformation within the Pacific-Australia plate boundary zone in Irian Jaya, Indonesia, Tectonophys., 237, 141-153, 1994.

Taylor, M. A. J., R. Dmowska, and J. R. Rice, Upper plate stressing and seismicity in the subduction earthquake cycle, J. Geophys. Res., 103, 24523-24542, 1998.
 

•  If you would like to print out a higher quality version of this than is possible simply from printing directly from this web-page, you may down-load a black & white postscript version from here (571K, tarred & gzipped).