Structure and evolution of lithospheric slab beneath the Sunda arc, Indonesia

Science

Volumn 271, Issue 5255, 1996, Pages 1566-1570

  Widiyantoro, Sri ^a,b   Van Der Hilst, Rob ^a,c  

^a AUSTRALIAN NATIONAL UNIVERSITY   (Australia)

^b BANDUNG INSTITUTE OF TECHNOLOGY   (Indonesia)

^c MASSACHUSETTS INSTITUTE OF TECHNOLOGY   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EVOLUTION; LITHOSPHERE; LITHOSPHERIC SLAB; SEISMIC ANOMALY; SLAB STRUCTURE; SUBDUCTION; TECTONICS; TERTIARY; TOMOGRAPHY;

INDONESIA, SUNDA ARC;

EID: 0029729893     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.271.5255.1566     Document Type: Article

References (54)

1. J. A. Katili. Tectonophysics 26, 165 (1975).
2. W. Hamilton, U.S. Geol Surv. Prof. Pap 1078 (1979), pp 37 and 81-83; in Island Arcs. Deep Sea Trenches, and Back-Arc Basins, M. Taiwani and W. Pitman III, Eds. (American Geophysical Union, Washington, DC, 1981), pp. 15-31.
3. W. Hamilton, U.S. Geol Surv. Prof. Pap 1078 (1979), pp 37 and 81-83; in Island Arcs. Deep Sea Trenches, and Back-Arc Basins, M. Taiwani and W. Pitman III, Eds. (American Geophysical Union, Washington, DC, 1981), pp. 15-31.
4. P. Molnar and P Tapponnier, Science 189, 419 (1975); P. Tapponnier and P. Molnar, Nature 264, 319 (1976); P. Patriat and J Achache, ibid. 311, 615 (1984).
5. P. Molnar and P Tapponnier, Science 189, 419 (1975); P. Tapponnier and P. Molnar, Nature 264, 319 (1976); P. Patriat and J Achache, ibid. 311, 615 (1984).
6. P. Molnar and P Tapponnier, Science 189, 419 (1975); P. Tapponnier and P. Molnar, Nature 264, 319 (1976); P. Patriat and J Achache, ibid. 311, 615 (1984).
7. C. R. Scotese, L. M. Gahagan, R. L. Larson, Tectonophysics 155, 27 (1988). Solutions of these computer reconstructions are, however, not unique.
8. S. A. P L. Cloetingh and M. J. R Wortel, ibid. 132, 49 (1986)
9. B. Isacks and P. Molnar, Rev. Geophys Space Phys. 9, 103 (1971).
10. K R Newcomb and W R McCann, J. Geophys. Res. 92, 421 (1987).
11. M. J. R. Wortel and N. J Vlaar, Pure Appl. Geophys. 128, 625 (1988).
12. S. Kirby, S Stein, E. Okal, D Rubie, in preparation.
13. Y. Fukao, M. Obayashi, H. Inoue, M. Nenbai. J. Geophys Res 97, 4809 (1992)
14. N. T Pusprto, Y. Yamanaka, T. Miyatake, K. Shimazaki, K. Hirahara, Tectonophysics 220, 175 (1992).
15. R D. van der Hilst, E. R. Engdahl, W Spakman, G. Nolet, Nature 353, 37 (1991); R D. van der Hilst and T. Seno, Earth Planet. Sci. Lett. 120, 395 (1993).
16. R D. van der Hilst, E. R. Engdahl, W Spakman, G. Nolet, Nature 353, 37 (1991); R D. van der Hilst and T. Seno, Earth Planet. Sci. Lett. 120, 395 (1993).
17. R. D van der Hilst, Nature 374, 154 (1995).
18. The data were inverted for variations in seismic velocity and their effects on hypocenter location [W. Spakman and G. Nolet, in Mathematical Geophysics, N.J. Vlaar et al., Eds. (Reidel, Dordrecht. 1988), pp. 155-188; R. D van der Hilst and E R Engdahl, Phys. Earth Planet Inter. 75, 39 (1992), R. D van der Hilst, E. R Engdahl, W. Spakman, Geophys J. Int 115, 264 (1993)]. We used the LSQR algorithm [C. C. Paige and M. A Saunders, ACM Trans. Math. Software 8, 43 (1982); ibid, p. 195; G Nolet, J Comput. Phys. 61, 463 (1985)] to solve the large system of equations. The inversion was linearized around the radially stratified iesp91 reference velocity model (16). Smooth models of slab structure were produced by forcing the solution to the reference values in areas of poor resolution (norm damping) and by minimizing differences in velocity values in neighboring blocks (gradient damping) [G Nolet, in Seismic Tomography, G Nolet, Ed. (Reidel, Dordrecht, 1987), pp. 1-23]. As a result of the damping, the amplitudes of the velocity variations were underestimated in most parts of the model space, and short wavelength variations were suppressed.
19. The data were inverted for variations in seismic velocity and their effects on hypocenter location [W. Spakman and G. Nolet, in Mathematical Geophysics, N.J. Vlaar et al., Eds. (Reidel, Dordrecht. 1988), pp. 155-188; R. D van der Hilst and E R Engdahl, Phys. Earth Planet Inter. 75, 39 (1992), R. D van der Hilst, E. R Engdahl, W. Spakman, Geophys J. Int 115, 264 (1993)]. We used the LSQR algorithm [C. C. Paige and M. A Saunders, ACM Trans. Math. Software 8, 43 (1982); ibid, p. 195; G Nolet, J Comput. Phys. 61, 463 (1985)] to solve the large system of equations. The inversion was linearized around the radially stratified iesp91 reference velocity model (16). Smooth models of slab structure were produced by forcing the solution to the reference values in areas of poor resolution (norm damping) and by minimizing differences in velocity values in neighboring blocks (gradient damping) [G Nolet, in Seismic Tomography, G Nolet, Ed. (Reidel, Dordrecht, 1987), pp. 1-23]. As a result of the damping, the amplitudes of the velocity variations were underestimated in most parts of the model space, and short wavelength variations were suppressed.
20. The data were inverted for variations in seismic velocity and their effects on hypocenter location [W. Spakman and G. Nolet, in Mathematical Geophysics, N.J. Vlaar et al., Eds. (Reidel, Dordrecht. 1988), pp. 155-188; R. D van der Hilst and E R Engdahl, Phys. Earth Planet Inter. 75, 39 (1992), R. D van der Hilst, E. R Engdahl, W. Spakman, Geophys J. Int 115, 264 (1993)]. We used the LSQR algorithm [C. C. Paige and M. A Saunders, ACM Trans. Math. Software 8, 43 (1982); ibid, p. 195; G Nolet, J Comput. Phys. 61, 463 (1985)] to solve the large system of equations. The inversion was linearized around the radially stratified iesp91 reference velocity model (16). Smooth models of slab structure were produced by forcing the solution to the reference values in areas of poor resolution (norm damping) and by minimizing differences in velocity values in neighboring blocks (gradient damping) [G Nolet, in Seismic Tomography, G Nolet, Ed. (Reidel, Dordrecht, 1987), pp. 1-23]. As a result of the damping, the amplitudes of the velocity variations were underestimated in most parts of the model space, and short wavelength variations were suppressed.
21. The data were inverted for variations in seismic velocity and their effects on hypocenter location [W. Spakman and G. Nolet, in Mathematical Geophysics, N.J. Vlaar et al., Eds. (Reidel, Dordrecht. 1988), pp. 155-188; R. D van der Hilst and E R Engdahl, Phys. Earth Planet Inter. 75, 39 (1992), R. D van der Hilst, E. R Engdahl, W. Spakman, Geophys J. Int 115, 264 (1993)]. We used the LSQR algorithm [C. C. Paige and M. A Saunders, ACM Trans. Math. Software 8, 43 (1982); ibid, p. 195; G Nolet, J Comput. Phys. 61, 463 (1985)] to solve the large system of equations. The inversion was linearized around the radially stratified iesp91 reference velocity model (16). Smooth models of slab structure were produced by forcing the solution to the reference values in areas of poor resolution (norm damping) and by minimizing differences in velocity values in neighboring blocks (gradient damping) [G Nolet, in Seismic Tomography, G Nolet, Ed. (Reidel, Dordrecht, 1987), pp. 1-23]. As a result of the damping, the amplitudes of the velocity variations were underestimated in most parts of the model space, and short wavelength variations were suppressed.
22. The data were inverted for variations in seismic velocity and their effects on hypocenter location [W. Spakman and G. Nolet, in Mathematical Geophysics, N.J. Vlaar et al., Eds. (Reidel, Dordrecht. 1988), pp. 155-188; R. D van der Hilst and E R Engdahl, Phys. Earth Planet Inter. 75, 39 (1992), R. D van der Hilst, E. R Engdahl, W. Spakman, Geophys J. Int 115, 264 (1993)]. We used the LSQR algorithm [C. C. Paige and M. A Saunders, ACM Trans. Math. Software 8, 43 (1982); ibid, p. 195; G Nolet, J Comput. Phys. 61, 463 (1985)] to solve the large system of equations. The inversion was linearized around the radially stratified iesp91 reference velocity model (16). Smooth models of slab structure were produced by forcing the solution to the reference values in areas of poor resolution (norm damping) and by minimizing differences in velocity values in neighboring blocks (gradient damping) [G Nolet, in Seismic Tomography, G Nolet, Ed. (Reidel, Dordrecht, 1987), pp. 1-23]. As a result of the damping, the amplitudes of the velocity variations were underestimated in most parts of the model space, and short wavelength variations were suppressed.
23. The data were inverted for variations in seismic velocity and their effects on hypocenter location [W. Spakman and G. Nolet, in Mathematical Geophysics, N.J. Vlaar et al., Eds. (Reidel, Dordrecht. 1988), pp. 155-188; R. D van der Hilst and E R Engdahl, Phys. Earth Planet Inter. 75, 39 (1992), R. D van der Hilst, E. R Engdahl, W. Spakman, Geophys J. Int 115, 264 (1993)]. We used the LSQR algorithm [C. C. Paige and M. A Saunders, ACM Trans. Math. Software 8, 43 (1982); ibid, p. 195; G Nolet, J Comput. Phys. 61, 463 (1985)] to solve the large system of equations. The inversion was linearized around the radially stratified iesp91 reference velocity model (16). Smooth models of slab structure were produced by forcing the solution to the reference values in areas of poor resolution (norm damping) and by minimizing differences in velocity values in neighboring blocks (gradient damping) [G Nolet, in Seismic Tomography, G Nolet, Ed. (Reidel, Dordrecht, 1987), pp. 1-23]. As a result of the damping, the amplitudes of the velocity variations were underestimated in most parts of the model space, and short wavelength variations were suppressed.
24. The data were inverted for variations in seismic velocity and their effects on hypocenter location [W. Spakman and G. Nolet, in Mathematical Geophysics, N.J. Vlaar et al., Eds. (Reidel, Dordrecht. 1988), pp. 155-188; R. D van der Hilst and E R Engdahl, Phys. Earth Planet Inter. 75, 39 (1992), R. D van der Hilst, E. R Engdahl, W. Spakman, Geophys J. Int 115, 264 (1993)]. We used the LSQR algorithm [C. C. Paige and M. A Saunders, ACM Trans. Math. Software 8, 43 (1982); ibid, p. 195; G Nolet, J Comput. Phys. 61, 463 (1985)] to solve the large system of equations. The inversion was linearized around the radially stratified iesp91 reference velocity model (16). Smooth models of slab structure were produced by forcing the solution to the reference values in areas of poor resolution (norm damping) and by minimizing differences in velocity values in neighboring blocks (gradient damping) [G Nolet, in Seismic Tomography, G Nolet, Ed. (Reidel, Dordrecht, 1987), pp. 1-23]. As a result of the damping, the amplitudes of the velocity variations were underestimated in most parts of the model space, and short wavelength variations were suppressed.
25. The inversion results for mantle structure beneath the entire Indonesian region, as well as the results of the global imaging, will be published elsewhere.
26. B. L N. Kennett and E. R. Engdahl, Geophys. J. Int. 105, 848 (1991).
27. A procedure that uses an improved global traveltime model, ak135 [B L N Kennett, E. R. Engdahl, R. Buland, Geophys. J. Int 122, 108 (1995)], which for mantle P waves is virtually identical to iasp91 (16), and the arrival times of first-arriving regional and teleseismic P phases, regional S phases, depth phases (pP, pwP, and sP), and the PKPdf branch were used to relocate all teleseismically well-constrained earthquakes that occurred between 1964 and June 1994. For the data processing, corrections are applied for Earth ellipticity, topography, bathymetry (at bounce points), and average upper mantle station effects (within 5○ by 5○ patches). Quality filters for each event include at least 10 reports, azimuthal coverage of more than 180○ at teleseismic stations, and standard errors of less than 35 km in location and 15 km in depth. This procedure (E. R. Engdahl, R. D. van der Hilst, R. Buland, in preparation) ensures that depth errors and the mapping of source heterogeneity into mislocation are minimized, thereby creating a powerful uncontaminated database of P, pP, and pwP residuals for use in tomographic imaging.
28. R. D. van der Hilst, B. L. N. Kennett, D. Christie, J. Grant, Trans Am. Geophys Union 75, 177 (1994).
29. For stations outside the study area, we combined all ray paths from earthquakes in a particular source region to closely spaced stations elsewhere into a single ray, commonly referred to as the summary ray, to reduce the uneven sampling of mantle structure by ray paths, the dimension of the matrices involved in the inversions (and thus computer memory requirements), and the computer time for ray tracing The datum (residual time) assigned to the summary ray was the median of all data considered for that summary ray. The number of rays that contributed to the summary ray was not restricted For stations inside the study area, however, we used Individual rays to optimize the sampling.
30. W. Spakman, S. van der Lee, R van der Hilst, Phys. Earth Planet. Mer. 79, 3 (1993)
31. We inverted data from different time periods and from earthquakes in different depth intervals to test the robustness of structural information in the images In addition, test inversions were conducted with synthetic data (with and without simulated data errors) computed from the ray distribution as in the real data inversion from isolated target anomalies, checkerboard patterns, and specifically designed, semirealistic slab models (15). One has to be cautious with the interpretation of the results of such tests because they do not simulate accurately the effect of data errors and source misfocation and because they are conducted with the same theoretical assumptions (for example, linearization and neglect of anisotropy) and approximations (for exampte, parameterization) as in the inversion of reported phase data. Interpretation is further complicated by the dependence of inferences pertinent to image quality on the spatial characteristics of the input model from which synthetic data were computed [J.-J. Léveque, L. Rivera, G. Wittlinger, Geophys. J. Int. 115, 313 (1993)]. The neglect of the effects of the fast slab on wave velocity and ray path position can result in the overestimation of the thickness of the slab [E R Engdahl and D. Gubbins, J. Geophys. Res. 92, 13855 (1987)] but probably does not produce an effect on the large-scale structures discussed here that is significant enough to invalidate our conclusions.
32. We inverted data from different time periods and from earthquakes in different depth intervals to test the robustness of structural information in the images In addition, test inversions were conducted with synthetic data (with and without simulated data errors) computed from the ray distribution as in the real data inversion from isolated target anomalies, checkerboard patterns, and specifically designed, semirealistic slab models (15). One has to be cautious with the interpretation of the results of such tests because they do not simulate accurately the effect of data errors and source misfocation and because they are conducted with the same theoretical assumptions (for example, linearization and neglect of anisotropy) and approximations (for exampte, parameterization) as in the inversion of reported phase data. Interpretation is further complicated by the dependence of inferences pertinent to image quality on the spatial characteristics of the input model from which synthetic data were computed [J.-J. Léveque, L. Rivera, G. Wittlinger, Geophys. J. Int. 115, 313 (1993)]. The neglect of the effects of the fast slab on wave velocity and ray path position can result in the overestimation of the thickness of the slab [E R Engdahl and D. Gubbins, J. Geophys. Res. 92, 13855 (1987)] but probably does not produce an effect on the large-scale structures discussed here that is significant enough to invalidate our conclusions.
33. R. K. Cardwell and B. L Isacks, J. Geophys. Res. 83, 2825 (1978), P. Lundgren and D Giardini, ibid 99, 15833 (1994).
34. R. K. Cardwell and B. L Isacks, J. Geophys. Res. 83, 2825 (1978), P. Lundgren and D Giardini, ibid 99, 15833 (1994).
35. The absence of earthquakes between 350 and 500 km in depth could be caused by the low deviatoric stress in the transition from tensional to compressional stress (6, 8) or to the weakening of the plate by the exothermic phase change from olivine to β spinel at ∼ 410 km (D. C. Rubie, Nature 308, 505 (1984), M. J. R Wortel, Geophys. Res. Lett. 13, 34 (1986)].
36. The absence of earthquakes between 350 and 500 km in depth could be caused by the low deviatoric stress in the transition from tensional to compressional stress (6, 8) or to the weakening of the plate by the exothermic phase change from olivine to β spinel at ∼ 410 km (D. C. Rubie, Nature 308, 505 (1984), M. J. R Wortel, Geophys. Res. Lett. 13, 34 (1986)].
37. M. G. Audley-Charles, P. D. Ballantyne, R Hall, Tectonophysics 155, 317 (1988), and references therein.
38. E J. Cobbing, D. I. J. Mallick, P. E J Pitfield, L H. Teoh, J. Geol. Soc. London 143, 537 (1986).
39. C S. Hutchison, Geol. Soc. Am. Bull 86, 797 (1975).
40. M. Gurnis and B. H. Hager, Nature 335, 317 (1988); C. Kincaid and P. Olson, J Geophys. Res. 92, 13832 (1987); R. W Griffiths, R. Hackney, R. D. van der Hilst, Earth Planet Sci. Lett. 133, 1 (1995).
41. M. Gurnis and B. H. Hager, Nature 335, 317 (1988); C. Kincaid and P. Olson, J Geophys. Res. 92, 13832 (1987); R. W Griffiths, R. Hackney, R. D. van der Hilst, Earth Planet Sci. Lett. 133, 1 (1995).
42. M. Gurnis and B. H. Hager, Nature 335, 317 (1988); C. Kincaid and P. Olson, J Geophys. Res. 92, 13832 (1987); R. W Griffiths, R. Hackney, R. D. van der Hilst, Earth Planet Sci. Lett. 133, 1 (1995).
43. S Zhong and M Gumis, Science 267, 838 (1995); G. F. Davies. Earth Planet. Sci. Lett. 133, 507 (1995); U. R. Christensen, in preparation.
44. S Zhong and M Gumis, Science 267, 838 (1995); G. F. Davies. Earth Planet. Sci. Lett. 133, 507 (1995); U. R. Christensen, in preparation.
45. U R. Christensen and D. Yuen, J. Geophys. Res. 89, 4389 (1984)
46. J. B. Minster and T H. Jordan, ibid. 83, 5331 (1978).
47. C DeMets, R. G Gordon, D F. Argus, S. Stein, Geophys. J. Int 101, 425 (1990).
48. C-S. Liu, J R. Curray, J M. McDonald, Earth Planet. Sci Lett. 65, 331 (1983); T. W. C Hilde, S. Uyeda, L Kroenke, Tectonophyiscs 38, 145 (1977).
49. C-S. Liu, J R. Curray, J M. McDonald, Earth Planet. Sci Lett. 65, 331 (1983); T. W. C Hilde, S. Uyeda, L Kroenke, Tectonophyiscs 38, 145 (1977).
50. Oblique subduction below Sumatra is comparable to the oblique subduction of the Pacific Plate at the western Aleutian trench, for which K C. Creager and T. M. Boyd [J. Geophys. Res 96, 2293 (1991)] showed from membrane strain calculations that the particle motion in the shallow slab is essentially subhorizontal.
51. M. J. R. Wortel and W. Spakman, Proc. K. Ned. Akad Wet 95, 325 (1992).
52. After initiation of slab detachment, the gravitational force exerted on the detached slab results in an increased slab pull and tensional stress in the hinge area of the adjacent segment of continuous slab, which can lead to shear failure and the lateral propagation of the detachment [also see S. Yoshioka and M. J. R. Wortel, J. Geophys. Res 100, 20223 (1995); S. Yoshioka, D. A. Yuen, T B. Larsen, Island Arc 4, 89 (1995)]. As a result, the slab pull is small for the region of detachment, increases abruptly near the hinge region, and assumes normal values where the deep slab is still attached to the surface plate. The laterally migrating concentration of the tensional forces can also result in more pronounced slab roll back and oceanward trench migration, which has been invoked to explain block rotations in the Mediterranean region. In our case, the hinge region is near the transition from Sumatra to Java, close to 105○E
53. After initiation of slab detachment, the gravitational force exerted on the detached slab results in an increased slab pull and tensional stress in the hinge area of the adjacent segment of continuous slab, which can lead to shear failure and the lateral propagation of the detachment [also see S. Yoshioka and M. J. R. Wortel, J. Geophys. Res 100, 20223 (1995); S. Yoshioka, D. A. Yuen, T B. Larsen, Island Arc 4, 89 (1995)]. As a result, the slab pull is small for the region of detachment, increases abruptly near the hinge region, and assumes normal values where the deep slab is still attached to the surface plate. The laterally migrating concentration of the tensional forces can also result in more pronounced slab roll back and oceanward trench migration, which has been invoked to explain block rotations in the Mediterranean region. In our case, the hinge region is near the transition from Sumatra to Java, close to 105○E
54. We gratefully acknowledge J. Weekes for the excellent picking of phase arrivals in the Skippy records that were used in this study; B. L. N. Kennett, K. C Creager, O Gudmundsson, and anonymous referees for reviews, G Nolet for providing us with his LSQR inversion code, and E. R. Engdahl for early versions of our hypocenter and phase data set. We thank C. Krayshek for producing the final draft of Fig. 4.