See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/227706758

Seismic moment release during slab rupture

beneath the Banda Sea
ARTICLE in GEOPHYSICAL JOURNAL INTERNATIONAL JULY 2008
Impact Factor: 2.72 DOI: 10.1111/j.1365-246X.2008.03838.x

CITATIONS

DOWNLOADS

VIEWS

17

84

87

1 AUTHOR:
Mike Sandiford
University of Melbourne
133 PUBLICATIONS 2,835 CITATIONS
SEE PROFILE

Available from: Mike Sandiford

Retrieved on: 27 June 2015

Geophys. J. Int. (2008) 174, 659671

doi: 10.1111/j.1365-246X.2008.03838.x

Seismic moment release during slab rupture beneath the Banda Sea
Mike Sandiford
School of Earth Sciences, University of Melbourne, Victoria 3010, Australia. E-mail: [email protected]

Accepted 2008 April 30. Received 2008 April 30; in original form 2007 August 15

SUMMARY
The highest intermediate depth moment release rates in Indonesia occur in the slab beneath
the largely submerged segment of the Banda arc in the Banda Sea to the east of Roma, termed
the Damar Zone. The most active, western-part of this zone is characterized by downdip
extension, with moment release rates (1018 Nm yr1 per 50 km strike length) implying
the slab is stretching at 1014 s1 consistent with near complete slab decoupling across the
100200 km depth range. Differential vertical stretching along the length of the Damar Zone is
consistent with a slab rupture front at 100200 km depth beneath Roma propagating eastwards
at 100 km Myr1 . Complexities in the slab deformation field are revealed by a narrow
zone of anomalous in-plane P-axis trends beneath Damar, where subhorizontal constriction
suggests extreme stress concentrations 100 km ahead of the slab rupture front. Such stress
concentrations may explain the anomalously deep ocean gateways in this region, in which case
ongoing slab rupture may have played a key role in modulating the Indonesian throughflow in
the Banda Sea over the last few million years.

1 I N T RO D U C T I O N
The rupture of subducting slabs from the trailing surface plate that
must inevitably accompany continent-arc collision is a crucial process in the plate tectonic cycle (Davies & von Blackenburg 1995).
Consequently, numerous studies have attempted to elucidate the
mechanisms of slab rupture using constraints derived from seismic tomography (Spakman 1990), forward modelling (Yoshioka &
Wortel 1995; Wong et al. 1997; Gerya et al. 2004) and geological
observations (Meulen et al. 1998). Along with many others, these
studies have progressed our understanding of slab rupture processes.
However, they have yet to be adequately tested using constraints
derived from the deformation field in actively rupturing slabs. A
crucial set of data relevant to the instantaneous deformation field
within deforming slabs is provided by focal mechanism solutions
and moment tensor data. With a few notable exceptions (Nothard
et al. 1996), past attempts to resolve the deformation field within
slabs have been hampered by data limitations. However, the growth
of the Harvard CMT catalogue now make it timely to reassess constraints on the deformation field in actively rupturing slab systems
using moment release rates.
The Banda Sea in eastern Indonesia in one of the best locations
to study active slab rupture (McCaffrey et al. 1985; McCaffrey
1989a, b; McCaffrey & Abers 1991). For well over 20 yr, it has been
known that the 500 km long, now extinct, and actively uplifting
(Merritts et al. 1998), section of the volcanic arc around Wetar and
Alor, to the north of Timor, is underlain by a region of anomalously
low intermediate depth seismicity compared to the volcanically active regions to either side (Figs 1 and 2, McCaffrey et al. 1985;

C

2008 The Author

C 2008 RAS
Journal compilation 

McCaffrey 1989b; Das 2004). This study focuses on the nature of

the deformation field at intermediate depths between Wetar and the
zone of much more intense activity beneath the island of Damar,
further east. It builds on the previous studies (Cardwell & Isacks
1978; McCaffrey et al. 1985; Das 2004; Milsom 2005) in utilizing
the now greatly expanded set of data now available through the CMT
catalogue. Seismic activity in this region has been previously investigated in several studies (Das 2004; Milsom 2005). Most notably,
Das (2004) used a set of relocated events assembled in 1999 to describe the geometry of the slab system and reported relevant CMT
solutions for the relocated events. As of 2007 October, the CMT
catalogue comprised 226 intermediate depth (70300 km) events
between 127 and 131 E along the southern limb (south of 6 S) of
the Banda slab in the vicinity of the island of Damar, providing a
density data that allow much better resolution of deformation within
the Banda slab, thereby leading to new insights into the mechanisms
of slab rupture.
2 GEOLOGICAL AND GEOPHYSICAL
S I G NA L S O F S L A B D E TA C H M E N T
B E N E AT H T H E B A N D A S E A
Convergence between the Indo-Australian and Sunda plates is accommodated along both the Sunda and Banda arcs (Fig. 1). The
Sunda arc comprises the active volcanic segment of the arc to the
west of 122 E, where oceanic lithosphere of the Indo-Australian
plate is being subducted. In the eastern part of the Sunda arc, from
eastern Java to Sumbawa, earthquakes define a more or less continuous, steeply dipping, planar slab extending to transition zone

659

GJI Seismology

Key words: Seismicity and tectonics; Subduction zone processes; Asia.

660

M. Sandiford

Figure 1. (a) Map showing composite relocated Indonesian (CIR) catalogue derived from Engdahl & Villasenor (2002) and Schoffel & Das (1999). Triangles
represent events greater than 300 km deep, circles events between 70 and 300 km and squares events less than 70 km. Symbol sizes are proportional to reported
magnitudes (with a range from M 3.57.4). (b) Depth slice show event distribution for those events shown in panel (a) between 12 S and 6 S.

depths at 600 km (Fig. 2a, Schoffel & Das 1999; Das 2004). Tomographic imaging (Replumaz et al. 2004) shows the slab connects
at subtransition zone depths to a large, seismically fast anomaly
beneath the Sunda Plate. The Banda arc is distinguished because
it juxtaposes Australian continental lithosphere against the forearc,
representing one the best examples of active arc-continent collision
on the modern earth. The geometry of the slab is well constrained by
relocated earthquake data (e.g. Schoffel & Das 1999; Milsom 2001;
Das 2004) and tomographic imaging (Widiyantoro & van der Hilst
1997). It comprises either (1) one continuous, but tightly folded, slab
or (2) two or more distinct slabs of opposing vergence that meet at
the surface near 132 E, 6 S (Milsom 2001). In the former case, the
small radius of curvature (200 km) make it one of the most tightly
folded slabs known, attributable to the unique geodynamic setting
involving a combination of backarc spreading that opened the Banda
Sea to the east at around 5 Ma and Pacific Plate motion that drives

a westward motion of crustal blocks to the northeast of the Banda

Sea, relative to Australia (McCaffrey 1989b).
In terms of volcanism, the southern limb of the Banda arc comprises three distinct segments: a 400 km long, volcanically active, segment to the west of Alor (here termed the Flores Zone); a
500 km long volcanically inactive segment including the islands
of Alor, Atauro and Wetar to the north of Timor (termed the Wetar
Zone); and sa 400 km long, volcanically active segment extending
east and north from Damar around the Banda Sea to Ambon (termed
the Damar Zone). Volcanism in the Wetar Zone ceased at about
3 Ma (Abbott & Chamalaun 1981), following collision between the
Australian continental margin as marked by Pliocene tectonism on
Timor (Audley-Charles 2004). Helium and strontium isotopes show
systematic variations along the arc, and evidence significant contamination of magma source zones by continental crust, suggesting the
leading edge of the Australian continental crust has been subducted


C

2008 The Author, GJI, 174, 659671

C 2008 RAS
Journal compilation 

Seismic moment release during slab rupture beneath the Banda Sea

661

Figure 2. Panels (ac): NS sections using 3 -wide bins across segments of the Indonesian slabs using the CIR catalogue. (a) Section through the eastern
Sunda slab. (b) Section through the islands of Timor, Wetar and Buru highlighting the intermediate depth seismic gap, and the distinct band of seismicity at
depths greater than 400 km that connects the north dipping Benioff zone beneath Wetar with a south dipping Benioff zone beneath Buru on the northern side
of the Banda Sea. (c) Section through the island of Damar and Seram in the eastern Banda sea, showing a diffuse band of seismicity connecting the southern
and northern limbs of the Banda slab(s), highlighting the intense seismicity on the southern limb at depths 100200 km. The boxed region is illustrated in
detail in Panels (di). Panels (df) details of Damar Zone seismicity (CIR catalogue) in 1 wide boxes (110 km) projected onto a plane striking N022 W,
perpendicular to the slab trace. The coordinates of the centres of the southern boundaries of the boxed regions are 128.5, 129.5 and 130.5 E and 8.75, 8.25
and 7.75 S, respectively. For scale the shaded region is 30 km wide, highlighting the widening of the zone of seismic activity to the east of Damar panel (f),
near Leyeni. In the western part of the Damar Zone well-located earthquakes show a relatively narrow distribution consistent with seismogenic slab zone of
<30 km wide. Panels (fi) as for panels (df) but using the CMT solutions with focal mechanisms plotted as lower hemisphere projections.


C 2008 The Author, GJI, 174, 659671
C 2008 RAS
Journal compilation 

662

M. Sandiford

Figure 3. Topographic profile of highest points along the Banda Arc between 122 E and 130 E encompassing the mostly exposed Flores and Wetar Zones
and largely submerged Damar Zone. The western limit of intermediate depth seismicity in the Damar Zone occurs at 127.4 E beneath Roma (see Figs 5 and 6).

to depths of at least 100 km during collision with the Banda arc

(Hilton et al. 1992, 1993; Gasparon et al. 1994).
The boundary between the Wetar and Damar Zones, near the island of Roma at 127 30 E, is marked by a significant change in

topography (Fig. 3). To the west the arc is mainly emergent with
only narrow seaways separating the volcanic islands such as Wetar,
Alor and Flores. To the east the arc is largely submerged with significant gateways as deep as 2 km. The Wetar Zone shows abundant

Figure 4. Summed scalar moments for the intermediate depth seismicity in the Indonesian region (contours in Nm yr1 ). (a) depth range 70300 km summed
for cylindrical volumes of surface radius = 115 km, (b) depth range 100200 km computed for cylindrical volumes of surface radius of 50 km. Note that the
intermediate depth seismic moment release rate of the Damar Zone isone to two orders of magnitude higher than elsewhere in the Sunda and Banda arcs,
while the seismic moment release rate of the Wetar Zone is the lowest in SundaBanda arc system.

C

2008 The Author, GJI, 174, 659671

C 2008 RAS
Journal compilation 

Seismic moment release during slab rupture beneath the Banda Sea
evidence for emergence in the form of elevated Quaternary terraces,
now hundreds of metres above sea level over a broad zone extending
from Sumba in the west through to the eastern tip of Timor Leste
(Merritts et al. 1998). Uplift rates are estimated at 0.5 mm yr1 on
Sumba (Pirazzoli et al. 1991) and Atauro (Chappell & Veeh 1978)
and 11.2 mm yr1 on Alor (Hantoro et al. 1994). Similar uplift
rates are indicated for large uplifted, but essentially undeformed,
coral terrace platforms in the eastern part of Timor Leste. Variations in uplift rate of almost an order of magnitude, from 0.2 to
1.5 mm yr1 , have been reported from the islands of Serau and Rote,
respectively (Merritts et al. 1998) and indicate that lithospheric deformation must play an important role in promoting uplift. However,
the long wavelength uplift of about 0.5 mm yr1 extending some
400 km from Alor through Atauro to the eastern end of Timor Leste
is arguably a response to sublithospheric processes that may have
direct bearing on the nature of processes operating in and around
the slab beneath (see later discussion). Little is known of the bathymetric evolution of the largely submerged Damar Zone, which plays
a crucial role in providing the principal gateway for the Indonesian
throughflow, connecting the Pacific and Indian Oceans.
The notion that the Banda Slab has been partially ruptured from
the surface plate in the vicinity of Timor is further informed by
GPS and geological studies. Genrich et al. (1996) showed that both
Timor and parts of the extinct section of the volcanic arc to its north
are now effectively welded to the Australian plate. The implication is that convergence between the Australian and Sunda Plates is

663

now largely accommodated on the northern (or backarc) side of the

extinct arc segment along the Wetar thrust. Shallow-depth seismic
activity to the north of Flores and Wetar is significantly higher than
along the Timor trench, confirming the notion that plate convergence is no longer kinematically connected to the subducted slab in
a simple way. Slab rupture is likely to have commenced at 3 Ma
when volcanism in the now extinct part of the arc ceased (Abbott &
Chamalaun 1981; Audley-Charles 2004; Elburg et al. 2004). Published tomographic imaging (e.g. Widiyantoro & van der Hilst 1997)
has not yet clearly resolved upper mantle structure beneath Wetar
at the level required to elucidate details of slab structure associated
with variations in seismic activity.
3 SEISMIC CHARACTER OF THE
S O U T H E R N B A N DA S E A
In this paper several earthquake catalogues are used characterize
the seismicity in the southern part of Banda Sea (south of 6 S).
First, relocated events from the Centennial catalogue (Engdahl &
Villasenor 2002) have been combined with a catalogue of relocated
Indonesian events assembled by (Schoffel & Das 1999). This composite catalogue of Indonesian relocated events (CIR) comprises a
total of 1019 well-located events for the region between 12 6 S
and 113137 E and provides for the best available constraint on
the spatial distribution of Banda slab events (Figs 1 and 2). Second,
some 747 focal mechanism solutions from the CMT catalogue, up to

Figure 5. Map view for selected CMT focal mechanisms for the Damar Zone. Solutions in the depth range 95205 km are colour coded by mechanism (red
normal fault, green strike-slip and blue reverse). Solutions with depths >205 km are shown in black. P-axis projections are shown for events with depths
shallower than 95 km. The box shows the region analysed in Fig. 8, corresponding to the seismogenic slab in the depth range 100200 km. The dashed segments
within the box demarcate the individual selections used in the analysis shown in Fig. 8. (Colours in the online version only.)


C 2008 The Author, GJI, 174, 659671
C 2008 RAS
Journal compilation 

664

M. Sandiford

and including 2007 October, for the same region are used to explore
the deformation field within the Banda slab.
In accord with the findings of several previous studies
(McCaffrey 1989a; Schoffel & Das 1999; Das 2004; Milsom 2005),
both the CIR and CMT catalogues show that the Wetar Zone is
characterized by a distinct gap in intermediate depth seismicity compared to the volcanically active regions on either side (Figs 1 and 2),
and especially the Damar Zone to the east. At depths greater than
300 km, more or less continuous seismicity extends across the
three zones with CMT solutions dominated by down dip compression typical of deep subduction earthquakes. The Damar Zone is
by far the most seismically active portion of the Banda slab at intermediate depths (Fig. 4). Despite comprising only 15 per cent
by arc-length, it accounts for 60 per cent of the eastern Sunda
and Banda arc intermediate level events in the CMT catalogue. Intermediate depth volumetric scalar moment release rates obtained
by summation of CMT events are more than an order of magnitude greater than in any other portion of the Banda or Sunda arc
(Fig. 3). The Damar Zone includes the M w 8.3 1963 event (depth
100 km deep, epicentre 6.8 E, 129.58 E; Osada & Abe 1981),
one of the largest intermediate known depth events recorded, as
well as the M w 8.5 great Banda Sea earthquake of 1938 (depth
60 km, epicentre 5.06 S, 131.60 E), the ninth largest event of the

20th Century (Genrich et al. 1996; Kanamori 1997; Okal & Reymond 2003). In the last 50 yr, the Damar Zone has experienced six
M w 7+ events, making it one of the most active intermediate depth
earthquake zones on the planet. The CMT catalogue includes 305
events in the Damar Zone including five M w 7+ events, but does
not include either of the M w 8+ events referred to above (Figs 5
and 6). To the west of 129 30 E, intermediate depth seismicity is
dominated by reverse fault mechanisms, reflecting downdip tension,
with both catalogues showing the seismic activity is confined to a
narrow zone 30 km in across-strike width, trending 068 E (Fig.
2). East of 129 30 E, intermediate depth activity is characterized
by a mix of reverse, strike-slip and normal mechanisms distributed
over a zone at least 80 km in across-strike width (Fig. 2f).
Within the Damar Zone the azimuth of principal axes of
the moment tensors (P, T) show spatial variation allowing subdivision into distinct domains (Figs 68). At shallow depths
(<70 km) P-axes tend to be aligned at about 1020 E broadly paralleling the Australian plate velocity (see Das 2004), consistent with
stress regimes being dictated by plate boundary interactions. At intermediate depths, the P-axis alignment relative to the slab trend
changes along the slab. As noted originally by Cardwell & Isacks
(1978), east of 130.5 E, P-axes tend to be aligned in the plane
of the slab. Between 130.5 E and 128.9 E, P-axes trend NNW, at a

Figure 6. Depth slice showing CMT focal mechanism solutions for the southern limb of the Banda slab seismicity in the depth range 35250 km. Note that only
events from within the boxed region in Fig. 5 are used, and that the focal mechanisms are shown as lower hemisphere projections colour coded by mechanism
(red normal fault, green strike-slip and blue reverse). Stereographic nets show P-, T- and B-axes orientations for reverse mechanisms from each of the
three boxed regions. And highlight the anomalous nature of the intermediate depth seismicity beneath the island of Damar at 128.5 E, where the P-axes are
aligned more-or-less parallel to the trend of the slab (068 E). This contrasts the more typical slab character of P-axes aligned at a high angle to the slab
trend. (Colours in the online version only.)

C

2008 The Author, GJI, 174, 659671

C 2008 RAS
Journal compilation 

Seismic moment release during slab rupture beneath the Banda Sea

665

Figure 7. P-axis (a) and T-axis (b) projections for the southern limb of the Banda slab in the vicinity of Damar for the depth range 95205 km. The size of
the indicators are scaled for magnitude, as reflected by variations in the projected width of indicators (most of the variation in the projected length relates to
variations in the plunge of the axes, as indicated by the mainly steeply plunging T-axes and shallow, slab-parallel P-axes near Damar).

high angle to the slab. In a narrow zone centred just south of Damar,
P-axes are aligned within the plane of the slab. West of 128.2 E,
the P-axes are aligned at a moderate angle to the slab trend more
or less parallel to the shallower seismicity. The zone of slab parallel
P-axes near Damar is characterized by a general steepening of Taxes that define a tight cluster with a median plunge of about 80 ,
implying that horizontal, in-plane slab shortening is accompanied
by downdip stretching.

4 T H E D A M A R Z O N E D E F O R M AT I O N
FIELD
The intense intermediate depth seismic activity in the Damar Zone
demands that the slab is failing internally, suggesting that it may be
the site of active slab rupture ahead of a slab window opening from
the west. The plausibility of this model can be tested with reference
to the seismic deformation field within the slab. This section uses
moment tensor data from the CMT catalogue to quantify the deformation field using approaches adapted from previous studies (e.g.
Nothard et al. 1996).
The principal components of the summed seismic moment release rate tensor for Damar Zone intermediate depth seismicity are
summarized in Table 1 and shown in Figs 9 and 10 in terms of the six

C 2008 The Author, GJI, 174, 659671
C 2008 RAS
Journal compilation 

tensor components M rr , M tt , M pp , M rt , M rp and M tp . The summation

M i j =

N


Mikj ,

k=1

where Mijk is the ijth moment component of the kth earthquake,

is taken over vertical cylinders of radius 25 km between 95 and
205 km depth. r, t and p refer to vertical (up), north and east directions, respectively: directions that are well aligned with respect to
the principal moment axes as revealed by orientation of P and Taxes as shown in Figs 6(ac). The mean moment release rates in this
depth range, normalized to a slab volume equivalent to 50 km long
by 100 km deep, with a seismogenic width of 30 km, are shown in
Table 1.
Catalogue moment release rates are invariably biased by the location of the rare, large events, with frequency comparable to catalogue
duration (see Figs 9ac). The long-term spatial pattern of moment
release is therefore better informed by using a reduced catalogue
including only events that have recurrence intervals that are significantly shorter than the catalogue record. Figs 9(di) and 10 show
the reduced-moment release obtained by excluding the three largest
CMT events (with M w > 6.65 or scalar moments greater than about
1019 Mm). Positive M rr show that the Damar Zone slab is extending vertically along its entire length (Fig. 9d) while negative M tt

666

M. Sandiford

Figure 8. Characteristics of intermediate depth (95205 km), CMT focal mechanism principal axes (P, T, B) for the Damar Zone. Panel (a): deviation of
median P-axis trend from the 068 E (i.e. mean trend of the slab in the western Damar Zone). Panel (b): plunge of median T-axis. Panel (c): deviation of median
B-axis trends from 068 E. The shaded zone shows 1 SD. The number of mechanisms in each selected region are prefaced with a #.
Table 1. Summary of intermediate depth (100200 km) seismic moment release rate and strain rate for the Damar Zone. Moment release rates
are normalized to a slab segment 50 km long with a seismogenic width of 30 km. r, t and p refer to vertical (up), north and east directions,
respectively. Positive values imply extension.
rr
M i j (mm yr1 )
i j (s 1 )

tt

9.3
2.8 1015
1017

pp

9.6
2.9 1015
1017

rt

3.5
1.1 1016

show that it is thinning across the extent of the slab (Figs 9e). The
reduced M rr values are about typically three to five times higher
in the western sector (west of 129 40 E) than in the eastern sector,
implying differential vertical stretching along the slab (Fig. 10a).
The sign of M pp varies along the zone (Fig. 9f) with positive values
showing that it is mostly extending along its length except in of the
region near Damar, where negative M pp reflect the in-plane short-

1016

rp

2.9
8.6 1017
1016

tp

7.1
2.1 1015
1017

9.4 1016
2.8 1016

ening, as is indicated by the characteristic alignment of P-axes in

the plane of the slab. Here, negative M pp and M tt imply the slab is
undergoing horizontal constriction, with the associated maxima in
M rr implying vertical stretching at faster rates than any other part
of the slab. In the segment to the west of the Damar constriction,
absolute M pp imply deformation approximates plane strain, while
further east deformation involves mild flattening.

C

2008 The Author, GJI, 174, 659671

C 2008 RAS
Journal compilation 

Seismic moment release during slab rupture beneath the Banda Sea

667

Figure 9. Summed components of the seismic moment tensor for intermediate depth seismicity in the Damar Zone (contours in units of log10 Nm yr1 ). r, t
and p refer to vertical (up), north and east directions, respectively. Positive values are shown by shaded regions. Moments are summed over vertical cylindrical
volumes with radius 25 km extending from 95 to 205 km depth. Figs (ac) show summation over all events, while panels (di) show reduced moment summation
for events with scalar moments less than 1023 Nm. See text for discussion.

Estimates of the seismogenic deformation rates can be derived

from seismic moment release rates following Kostrov (1974):
i j =

1
Mi j ,
2vt

where is the shear modulus (3.5 1010 Nm2 ), v is the relevant

volume over which seismic moments of the individual earthquakes
are summed and t is the time interval of relevance. The mean value
of M rr 1018 Nm yr1 in the depth range 100200 km, when normalized to a 50 km long by 30 km wide slab segment, corresponds
to mean vertical stretching rate of 2.8 1015 s1 . The reduced
moment release profile in Fig. 10(a) suggests that maximum stretching rates are several times the mean for the fastest deforming zone
between 127 and 128.7 E near Damar. Moreover, several factors
suggest that the total seismogenic strain rates are likely to be both
higher, and show more along strike variation, than indicated by the
profiles of reduced moment release shown in Fig. 10.
First, because the very large events that dominate long-term moment release have recurrence intervals much longer than the duration
of seismic records, they are typically not adequately represented in
existing catalogues. Missing large events will result in cataloguebased moment release estimates to underestimate long-term moment
release rates. That much higher moment events than recorded in the
CMT catalogue have occurred is evidenced by the historical record
of two M w 8+ events in the 20th century, with the 1963 M w 8.3
event occurring in the depth range 100200 km relevant to the discussion here. However, the notional increase in moment up to some
arbitrary magnitude M max can be estimated assuming these events
follow some statistical recurrence relation. For GuttenbergRichter

C 2008 The Author, GJI, 174, 659671
C 2008 RAS
Journal compilation 

statistics the up-scaling factor is (e.g. Johnston et al. 1994):



Mo 10(cb)(Mmax Mobs ) ,
where M obs is the largest catalogue event used in determining the
moment, M max is the maximum expected magnitude for earthquakes
in the region of interest, b is the recurrence parameter (typically 1
for GuttenbergRichter statistics) and c is a conversion factor of
1.5 relating moment magnitude scale to seismic moment (Hanks
& Kanamori 1979). For a b-value of 1, increasing M max from 7.4,
the largest event in the CMT database, to 8.3, in order to incorporate the largest known historical event, equates to an increase of
M o 2.8. Because the extent to which large magnitude events
(M w > 7.5) obey the GuttenbergRichter scaling relations is
unknown, the moment release rate estimate using this approach
is clearly subject to significant uncertainty. Nevertheless, it does
highlight the point that summation of CMT catalogue moments is
likely to underestimate the long-term moment release rate by several
factors.
Second, summed moment release rates are sensitive to the assumed width of the seismically active zone. The seismogenic slab
has been assumed to be 30 km wide, consistent with the observed
distribution of both CIR- and CMT-catalogue events in the western
end of the Damar slab (Fig. 2) as well as in other subduction zones
such as Tonga (Nothard et al. 1996). However, other workers have
suggested that slab seismicity may be confined to zones as little as
15 km wide (Houston 1993). Given uncertainties in earthquake locations this is certainly possible for the western part of the Damar
Zone, implying that volumes associated with moment release may
have been overestimated and associated strain rates underestimated.
However, it is also clear that within the Damar slab the seismogenic

668

M. Sandiford

Figure 10. Longitudinal profiles of intermediate-depth, reduced moment release (log10 Nm yr1 ) along the Damar Zone (see Fig. 9 caption for details). Solid
lines with grey shades for +ve summed moments and dashed line for ve summed moments. Reduced moment release rates are used to understand the spatial
pattern of moment release. As discussed in the text moments can be up-scaled to obtained estimates for any assumed maximum magnitude event. For example,
upscaling the reduced moments to M max = 7.4 (the largest Damar Zone event in the CMT catalogue), is obtained with a factor of 2.4, leading to a maximum
M rr value of 1.2 1018 Nm yr1 and to M max = 8.3 (the largest intermediate depth Damar Zone event) with a factor of 6.7 with M rr of 3.4 1018 Nm yr1 .

zone broadens considerably to the west, and must be significantly

wider than 30 km east of 130 E (Fig. 2f). One implication of this
is that the gradient in the volumetric vertical moment release rate
along the Damar slab must be significantly more marked than shown
in Fig. 10(a).
Finally, we currently have no meaningful way of assessing the
relative contributions of seismic and aseismic deformation within
a deforming slab. However, by analogy with continental realms,
where satellite based methods show that aseismic strain accumulation tends to exceed the seismically released strain, it is probable
that a considerable portion of the strain in the deforming slab is
accruing aseismically.
These factors caution that seismogenically determined momentrelease and strain rates are highly uncertain. However, they mostly
suggest that the estimates shown in Fig. 10 are conservative, implying that the mean vertical stretching rate due to seismic activity
in the slab can be assumed to be at least 3 1015 s1 . In the region of greatest volumetric moment release beneath Damar, vertical
stretching due to seismic release is occurring at several times this at
rates approaching 1014 s1 .
In summary, seismic moment release rates suggest the Damar
Zone slab is deforming at rates of 1015 to 1014 s1 in the 100

200 km depth range. The deformation involves vertical stretching

accompanied by across slab thinning and, generally strike parallel
stretching. A localized zone beneath the island of Damar is characterized by horizontal constriction and the highest rates of vertical
stretching. Vertical stretching rates drop rapidly east of 129 E.
5 I M P L I C AT I O N S F O R B A N D A
S L A B DY NA M I C S
Variation in both intensity and character of intermediate depth seismicity points to significant complexity in the seismogenic deformation field in the Banda slab. As noted in a number of previous
studies, the most striking aspect is the variability in intermediate
depth activity between the Wetar and Damar Zones. The seismic
moment data summarized here for the first time indicates vertical
stretching in the Damar Zone at rates of up to 1014 s1 , greater
in the west than the east, lending strong credence to the idea that
the slab is rupturing in this region, propagating eastwards in the
depth range 100200 km and widening a slab-window beneath Wetar (Fig. 11). In this context, it is relevant to note that a mean vertical
stretching rate of 2 1014 s1 over a depth range of 100 km, produces the differential displacement of 60 km Myr1 , more or less

C

2008 The Author, GJI, 174, 659671

C 2008 RAS
Journal compilation 

Seismic moment release during slab rupture beneath the Banda Sea

669

Figure 11. Schematic synthesis of inferred slab dynamics in the Banda Sea based on the analysis of the CMT moment distribution presented in this paper.
Variations in the moment tensor release allow subdivision of the slab and overlying lithosphere in the region 124131 E into a number of distinct domains: (I) a
domain characterized by an intermediate depth seismic gap in the Wetar Zone, (II) the western segment of the Damar Zone, characterized by down dip T-axes,
P-axes at a high angle to the slab trend and a mild flattening strain field (Fig. 6a), (III) the centralwest segment of the Damar Zone characterized by down dip
T-axes, P-axes in the plane of the slab and a constrictional strain field (Fig. 6b), (IV) the centraleast segment of the Damar Zone characterized by slab-normal
P-axes trends and oblique T-axes (Fig. 6c), (V) an eastern segment in which P-axes lie in the plane of the slab as noted by Cardwell & Isacks (1978), (VI) the
lithospheric domain characterized by shallow P-axes aligned with plate convergence (see also Fig. 5) and (VII) the domain of deep seismicity characterized
by near vertical P-axes indicative of downdip compression. The constrictional strain field that characterizes domain III (white arrows) is postulated to reflect
stress concentration imparted from the deeper, now semi-detached segment slab beneath domain I, which may also cause the anomalously deep bathymetry of
the overlying arc. In domain II the slab maybe more or less completely decoupled, with the slab rupture front propagating rapidly eastwards (as suggested by
the narrow dashdotted line). Black arrows show relative inferred velocities of the boundaries of domain I, with the lower boundary assumed to be sinking at
6070 km Myr1 , consistent with complete decoupling from the Australian plate, and the opening of a 200 km deep slab window in domain I since 3 Ma
coincident with the intermediate depth seismic gap.

matching the absolute speed of the Australian plate (Genrich et al.

1996). Assuming that the deeper parts of the slab system are sinking at long term rates governed by Australian plate motion, then
the Banda slab must be close to complete decoupling in the western
segment of the Damar Zone across the depth interval 100200 km.
Indeed, with the possibility that seismic deformation accounts for
no more than 50 per cent of the total strain then this segment of
the slab could already be effectively fully decoupled. Such a notion
is consistent with the absence of volcanism west of Damar as well
as with the occurrence of great earthquakes such as the 1963 M w
8.3 event, the moment of which requires that it ruptured the entire
seismogenic core of slab over several hundred kilometres (Osada &
Abe 1981). As the seismic core of the slab thins deformation should
accelerate, and thus the strain rates deduced form moment release
suggest the characteristic timescale for slab rupture in the Damar
Zone is no more than 106 yr.
As first noted by Cardwell & Isacks (1978) and subsequently
by McCaffrey et al. (1985) the abundance of slab parallel P-axes

C 2008 The Author, GJI, 174, 659671
C 2008 RAS
Journal compilation 

trends in the Banda slab is unusual. Beyond about 130.5 E (domain V in Fig. 11), the slab parallel P-axes have been related to the
unusually tight folding of the eastern most part of the Banda slab
(Cardwell & Isacks 1978). The analysis presented here shows there
is also a narrower zone of slab-parallel P-axes beneath the island of
Damar in the region of most rapid vertical stretching (domain III in
Fig. 11). An apparent small discontinuity in the trend of intermediate depth seismicity allows the possibility of minor folding of the
slab in the vicinity of Damar (Fig. 7) that might help explain the inplane P-axis trends (see also Das 2004). However, probably more
important is the fact that the zone of anomalous P-axis trends occurs in a relative sharp gradient in the vertical moment release rate
as reflected most prominently in the reduced moment release estimates presented here. As such, the zone of anomalous P-axis trends
may simply reflect in-plane horizontal shortening induced by the
differential vertical stretching rates to either side. However, the moment release rates imply that this region is also unusual in that it is
experiencing horizontal constriction. An important consequence of

670

M. Sandiford

partial slab rupture is that negative buoyancy from the ruptured segments must lead to stress concentration in the region ahead of the
rupture front (Fig. 11). The hypothesis advanced here is that the zone
of constriction near Damar reflects stress concentration ahead of an
eastward propagating slab rupture front. The notion that the rupturing slab segment beneath Damar is more highly stressed than other
parts of the system is consistent with the unusually low bathymetry
of this segment arc, evident in the >2 km deep ocean gateways
(Fig. 3). This bathymetric anomaly points to a significant negative
dynamic topography that would be expected to accompany increased
stress transfer through the slab (see further discussion below).
Assuming that the Wetar slab-window has grown to its present
width since the cessation of volcanism at about 3 Ma, then the timeaveraged rate of lateral propagation rate must be 130 km Myr1 .
Presumably, some of the widening of the slab window is occurring at
the western end of the Wetar Zone, beneath Pantar (McCaffrey et al.
1985). However, the intermediate depth seismic moment release
rate beneath the Flores Zone is much lower than beneath the Damar
Zone (Fig. 3), and the westward propagation rate into the Flores
Zone is likely much less than its eastward propagation rate into
the Damar Zone (Fig. 11). It seems likely therefore that the slab
window front is propagating eastward into the Damar Zone at rates of
100 km Myr1 , consistent with the seismogenic stretching rates
in the 100-km long western part of the Damar Zone.
As noted earlier the differences in topographic character of the
arc between the Wetar and Damar Zones are striking and arguably
relate to dynamic topographic effects mediated by the changing
pattern of stress transmission through the rupturing slab. Two consequences of this rupturing should be evident in the evolving surface
topographic field. First, as stress transmission from the deeper, negatively buoyant parts of the slab is reduced, the surface above ruptured segments should be experience uplift (Chatelain et al. 1992).
Second, increased stress transmission through the slab ahead of the
rupture front should induce subsidence above the intact segment of
the slab, as well as enhancing seismicity. While the Wetar Zone is
demonstrably uplifting at a regional average of about 0.5 mm yr1 ,
there is no evidence pertaining to a corresponding active subsidence
above the Damar Zone apart from the indirect evidence implied by
the remarkably low elevation of the arc here. Clearly such evidence
would provide a profound test of a propagating slab rupture, and
would hold important clues as to the dynamics of stress transmission through the slab, as well as the evolution of the Indonesian
seaways that have been postulated as so important to late Neogene
climate change in the continents bounding the Indian Ocean (Cane
& Molnar 2001). For example, if subsidence of the Damar Zone was
antithetic to the uplift of the Wetar Zone at around 0.5 mm yr1 , then
the ocean gateways to the east of Wetar will have deepened on average by more than 50 per cent over the last 1 Myr. Since these are
by far the largest gateways through the Indonesian arc system, such
a scenario may be expected to have significant oceanographic and
climatic ramifications (Cane & Molnar 2001).

AC K N OW L E D G M E N T S
Shamita Das kindly provided her catalogue of relocated Indonesian
earthquakes. Reviews by Meghan Miller and an anonymous reviewer
helped sharpen the focus of this work. This work was funded by ARC
Discovery grant DP0556133.

REFERENCES
Abbott, M.J. & Chamalaun, F.H., 1981. Geochronology of some Banda Arc
volcanics, Special Publication Geological Research and Development
Centre, Republic of Indonesia, 2, 253271.
Audley-Charles, M.G., 2004. Ocean trench blocked and obliterated by Banda
forearc collision with Australian proximal continental slope, Tectonophysics, 389, 6579.
Cane, M.A. & Molnar, P., 2001. Closing of the Indonesian seaway as a precursor to east African aridification around 34 million years ago, Nature,
411, 57162.
Cardwell, R. & Isacks, B., 1978. Geometry of subducted lithosphere beneath the Banda Sea in eastern Indonesia from seismicity and fault-plane
solutions, J. geophys. Res., 28252838.
Chappell, J. & Veeh, H.H., 1978. Late quaternary tectonic movements and
sea-level changes at Timor and Atauro Island, Geol. Soc. Am. Bull., 89,
356368.
Chatelain, J.-L., Molnar, P., Prevot, R. & Isaaks, B., 1992. Detachment of
part of the down going slab and uplift of the New Hebrides (Vanuatau)
Islands, Geophys. Res. Lett., 19, 15071510.
Das, S., 2004. Seismicity gaps and the shape of the seismic zone in the
Banda Sea Region from relocated hypocenters, J. geophys. Res., 109,
doi:10.1029/2004JB003192.
Davies, J.H. & von Blackenburg, F., 1995. Slab breakoff: a model of lithosphere detachment and its test in the magmatism and deformation of
collisional orogens, Earth planet. Sci. Lett., 129, 85102.
Elburg, M.A., van Bergen, M.J. & Foden, J.D., 2004. Subducted upper and
lower continental crust contributes to magmatism in the collision sector
of the Sunda-Banda arc, Indonesia, Geology, 32, 4144.
Engdahl, E.R. & Villasenor, A., 2002. Global seismicity: 19001999, in
International Handbook of Earthquake and Engineering Seismology,
Part A, Chapter 41, pp. 665690, eds W.H.K. Lee, H. Kanamori, P.C.
Jennings & C. Kisslinger, Academic Press, Amsterdam.
Gasparon, M., Hilton, D.R. & Varne, R., 1994. Crustal contamination processes traced by helium isotopes: examples from the Sunda arc, Indonesia,
Earth planet. Sci. Lett., 126, 1522.
Genrich, J.F.Y.B., McCaffrey, R., Calais, E., Stevens, C.W. & Subayra, C.,
1996. Accretion of the southern Banda arc to the Australian plate margin
determined by Global Positioning System measurements, Tectonics, 15,
288295.
Gerya, T., Yuen, D.A. & Maresch, W.V., 2004. Thermo-mechanical modelling of slab detachment, Earth planet. Sci. Lett., 226, 101116.
Hanks, T.C. & Kanamori, H., 1979. A moment-magnitude scale, J. geophys.
Res., 84, 23482350.
Hantoro, W.S. et al., 1994. Quaternary uplifted coral reef terraces on Alor
Island, East Indonesia, Coral reefs, 13, 215223.
Hilton, D.R., Hoogewerff, J.A., van Bergen, M.J. & Hammerschmidt, K.,
1992. Mapping magma sources in the east Sunda-Banda arcs, Indonesia:
constraints from helium isotopes, Geochim. Cosmochim. Acta, 56, 851
859.
Hilton, D.R., Hammerschmidt, K., Loock, G. & Friedrichsen, H., 1993.
Helium and argon isotope systematics of the central Lau Basin and Valu Fa
Ridge: evidence of crust/mantle interactions in a back-arc basin, Geochim.
Cosmochim. Acta, 57, 28192841.
Houston, H., 1993. The non-double-couple component of deep earthquakes
and the width of the seimogenic zone, Geophys. Res. Lett., 20, 16871690.
Johnston, A.C., Coppersmith, K.J., Kanter, L.R. & Cornell, C.A., 1994.
The earthquakes of stable continental regions, Electric Power Research
Institute Publication TR102261V1, Palo Alto, California.
Kanamori, H., 1997. The energy release of great earthquakes, J. geophys.
Res., 82, 29812987.
Kostrov, B.V., 1974. Seismic moment and energy of earthquakes, and seismic
flow of rocks, Izvestia Acad. Sci. USSR Phys. Solid Earth 1, 2340.
McCaffrey, R., 1989a. Seismological constraints and speculations on Banda
arc tectonics, Netherlands J. Sea Res., 24, 141152.


C

2008 The Author, GJI, 174, 659671

C 2008 RAS
Journal compilation 

Seismic moment release during slab rupture beneath the Banda Sea
McCaffrey, R., 1989b. Seismological constraints and speculations on Banda
Arc tectonics, Netherlands J. Sea Res., 24, 141152.
McCaffrey, R. & Abers, G.A., 1991. Orogeny in arccontinent collision: the
Banda Arc and western New Guinea, Geology, 19, 563566.
McCaffrey, R., Molnar, P., Roecker, S.W. & Joyodirwiryo, Y.S., 1985. Microearthquake seismicity and fault plane solutions related to arc continent
collision in the eastern Sunda Arc, Indonesia, J. geophys. Res., 90, 4511
4545.
Merritts, D., Eby, R., Harris, R., Edwards, R.L. & Chang, H., 1998. Variable
rates of Late Quaternary surface uplift along the Banda Arc-Australian
plate collision zone, eastern Indonesia, Geol. Soc., Lond., Special Publications, 146, 213224.
Meulen, M.J.v.d., Meulenkamp, J.E. & Wortel, M.J.R., 1998. Earth planet.
Sci. Lett., 154, 203219.
Milsom, J., 2001. Subduction in eastern Indonesia: how many slabs? Tectonophysics, 338, 167178.
Milsom, J., 2005. The Vrancea seismic zone and its analogue in the Banda
Arc, eastern Indonesia, Tectonophysics, 410, 325336.
Nothard, S., Haines, J., Jackson, J. & Holt, B., 1996. Distributed deformation
in the subducting lithopshere at Tonga, Geophysical. J. Int., 127, 328338.
Okal, E.A. & Reymond, D., 2003. The mechanism of great Banda Sea earthquake of 1 February 1938: applying the method of preliminary determination of focal mechanism to a historical event, Earth planet. Sci. Lett.,
216, 115.


C 2008 The Author, GJI, 174, 659671
C 2008 RAS
Journal compilation 

671

Osada, M. & Abe, K., 1981. Mechanism and tectonic implications of the
great Banda Sea earthquake of November 4, 1963, Phys. Earth planet.
Inter., 25, 129139.
Pirazzoli, P.A., Radtke, U., Hantoro, W.S., Jouannic, C., Hoang, C.T., Causse,
C. & Best, M.B., 1991. Quaternary raised coral-reef terraces on Sumba
Island, Indonesia, Science, 252, 18341836.
Replumaz, A., Karason, H., Van Der Hilst, R.D., Besse, J. & Tapponnier, P., 2004. 4-D evolution of SE Asias mantle from geological reconstructions and seismic tomography, Earth planet. Sci. Lett., 221, 103
115.
Schoffel, H.-J. & Das, S., 1999. Fine details of the Wadati-Benioff zone
under Indonesia and its geodynamic implications, J. geophys. Res., 104,
B6.
Spakman, W., 1990. Tomographic images of the upper mantle below central
Europe and the Mediterranean, Terra Nova, 2, 542553.
Widiyantoro, S. & Van Der Hilst, R.D., 1997. Mantle structure beneath
Indonesia inferred from high-resolution tomographic imaging, Geophys.
J. Int., 130, 167182.
Wong, A., Ton, S.Y.M. & Wortel, M.J.R., 1997. Slab detachment in continental collision zones: an analysis of controlling parameters, Geophys.
Res. Lett., 24, 20952098.
Yoshioka, S. & Wortel, M.J.R., 1995. Three-dimensional modeling of
detachment of subducted lithosphere, J. geophys. Res., 100, 20 223
20 244.