You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/353831834

Source Characteristics of the 2019 Mw 6.5 Ambon, Eastern Indonesia,


Earthquake Inferred from Seismic and Geodetic Data

Article  in  Seismological Research Letters · August 2021


DOI: 10.1785/0220210021

CITATIONS READS

9 389

8 authors, including:

Irwan Meilano Rino Salman


Bandung Institute of Technology Nanyang Technological University
165 PUBLICATIONS   1,842 CITATIONS    8 PUBLICATIONS   79 CITATIONS   

SEE PROFILE SEE PROFILE

Susilo Susilo Pepen Supendi


Badan Riset dan Inovasi Nasional University of Cambridge
61 PUBLICATIONS   371 CITATIONS    81 PUBLICATIONS   556 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Studi Pemantauan Kandungan Uap Air Menggunakan Metode Inversi GPS View project

Updating Indonesian Seismic Hazard Maps 2017 View project

All content following this page was uploaded by Rino Salman on 06 December 2021.

The user has requested enhancement of the downloaded file.


Source Characteristics of the 2019 M w 6.5
Ambon, Eastern Indonesia, Earthquake
Inferred from Seismic and Geodetic Data
Irwan Meilano*1 , Rino Salman2 , Suchi Rahmadani1,3 , Qibin Shi4 , Susilo Susilo5, Eric Lindsey6,
Pepen Supendi7, and Daryono Daryono7

Abstract
The 26 September 2019 M w 6.5 Ambon earthquake has been the largest instrumen-
tally recorded event to occur in Ambon, the capital city of Maluku Islands, eastern
Indonesia, and ruptured a previously unmapped active fault. In this study, we use
seismic and geodetic data to investigate the source characteristics of the event.
Our results show that the rupture process was complex in both the rupture initiation
and slip directions. In addition, the rupture was mostly strike-slip motion with normal
component and pure reverse slip in the north of the inverted fault. Our analysis
of campaign and continuous Global Positioning System (GPS) velocity fields estimates
that the fault has a 4.9 [4.0, 5.5] mm/yr slip rate with an earthquake recurrence interval Cite this article as Meilano, I.,
R. Salman, S. Rahmadani, Q. Shi, S. Susilo,
of 115 [102, 141] yr. In addition, a comparison of the horizontal strain-rate tensor
E. Lindsey, P. Supendi, and D. Daryono
derived from GPS velocity fields with historical earthquake data shows that (2021). Source Characteristics of the 2019
Ambon Island and the nearby regions have a high strain accumulation rate correlated Mw 6.5 Ambon, Eastern Indonesia,
Earthquake Inferred from Seismic and
with the distribution of M w ≥ 6 earthquakes, indicating that the regions are Geodetic Data, Seismol. Res. Lett. 92,
seismically active and possibly will experience more Ambon-type earthquakes in the 3339–3348, doi: 10.1785/0220210021.

future. Supplemental Material

Introduction addition, the moment-tensor solution of the event by the Global


The eastern Indonesia region has a complex tectonic setting Centroid Moment Tensor has a non–double-couple component
caused by the interaction of Pacific, Sunda, and Australian (Fig. 1), indicative of source complexity consisting of two or
plates. A prominent tectonic feature is the Banda arc, which more subevents (Sipkin, 1986; Frohlich and Apperson, 1992).
comprises troughs, an inner volcanic arc, and outer arc islands Determining the underlying fault geometry that produces the
(Hamilton, 1979; Spakman and Hall, 2010). A few islands that event is important for future studies such as seismic hazard
belong to the outer Banda arc are Seram, Saparua, Haruku, and assessments and regional stress field inversions.
Ambon Islands, which are tectonically bounded by the Seram In this study, we use seismic observations as well as Global
trough to the north and mapped active faults (South Buru, Positioning System (GPS) and Interferometric Synthetic
North Buru, and Manifa) (Irsyam et al., 2020) to the east Aperture Radar (InSAR) data to investigate the source charac-
and south regions (Fig. 1). Because of these tectonic environ- teristics of the event. We find that the rupture process was
ments, all the islands have experienced small to moderate complex in both the rupture initiation and slip directions.
earthquakes (3 ≥ M w < 7:0), with M w < 6 events distributed In addition, the rupture was mostly strike-slip motion with
almost evenly, whereas M w > 6 events are mostly concentrated
near the mapped active faults, except the latest 26 September 1. Faculty of Earth Science and Technology, Institute of Technology Bandung (ITB),
2019 M w 6.5 Ambon earthquake (Sahara et al., 2021) discussed Bandung, Indonesia, https://orcid.org/0000-0003-3488-9154 (IM); https://
orcid.org/0000-0001-6643-0597 (SR); 2. Earth Observatory of Singapore, Nanyang
in this article.
Technological University, Singapore, https://orcid.org/0000-0001-6999-8516 (RS);
The Ambon earthquake was the first instrumentally recorded 3. Center for Earthquake Science and Technology (CEST), Institute of Technology
M w > 6 event to occur near the city of Ambon. According to Bandung (ITB), Bandung, Indonesia; 4. Asian School of the Environment, Nanyang
Technological University, Singapore, https://orcid.org/0000-0002-4211-9187 (QS);
the National Disaster Management Authority of Indonesia, 5. Geospatial Information Agency (BIG), Bogor, Indonesia; 6. Department of Earth and
the event killed 41 people and damaged 6355 houses and 489 Planetary Sciences, University of New Mexico, Albuquerque, New Mexico, U.S.A.;
7. Agency for Meteorological, Climatology, and Geophysics (BMKG), Jakarta,
public facilities. The epicentral location of the event has been Indonesia
estimated between Ambon Island and Haruku Island, indicating *Corresponding author: irwanm@itb.ac.id
that the event ruptured a previously unmapped active fault. In © Seismological Society of America

Volume 92 • Number 6 • November 2021 • www.srl-online.org Seismological Research Letters 3339

Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/92/6/3339/5452232/srl-2021021.1.pdf


by Nanyang Technological University user
Mw Data Analysis and
Topography (km)
Previous events Mapped fault t6 7 Coseismic Slip
Modeling
0 1 2 3 Mw t3 t5 6
Multiple-point-source
inversions
Seram tro To understand the earthquake
ugh
rupture process and inferring
the fault geometry of the main-
shock event, we performed
multiple-point-source inver-
sions following the method
of Shi et al. (2018). We used
16 regional broadband
−3° North Buru seismic network of Indonesia’s
Seram Island Kaw Meteorological, Climatological,
a and Geophysical Agency
(BMKG) and 38 teleseismic
Mw 5.5 network downloaded from
Mw 6.5
Incorporated Research
Institutions for Seismology
Saparua Island
Ambon Island (Fig. S1, available in the supple-
Haruku Island mental material to this article).
Before running the inversion,
Manifa

we removed the instrument


−4° response from the three compo-
nents of the waveforms data,
downsampled the waveforms
South Buru to 10 Hz, and integrated the
waveforms from velocity to dis-
128° 129°
placement. To extract as much
information as possible from
Figure 1. Focal mechanisms with labels are for the mainshock of the Ambon event and one Mw 5.5
the waveforms, we carefully
aftershock event in the following day. Focal mechanisms without labels are past 5 < Mw < 7:0
earthquakes during the period 1976–2019. All focal mechanisms are based on the Global Centroid tested the frequency range to
Moment Tensor (CMT) catalog (Dziewonski et al., 1981; Ekström et al., 2012). Orange circles filter the waveforms of each sta-
represent earthquakes during the period 1973–2009 (U.S. Geological Survey [USGS] catalog) and tion and discarded the wave-
2009–2018 (Supendi et al., 2020). Black lines are nearby mapped faults from Irsyam et al. (2020). form components that were
Inset figure shows that Ambon Island is surrounded by the interaction of Australian, Pacific, and too noisy or of low quality.
Sunda plates. The color version of this figure is available only in the electronic edition.
Overall, the station-dependent
filtering frequency for regional
Pnl wave and surface wave
normal component and pure reverse slip in the north of the are 0.01–0.12 Hz and 0.01–0.05 Hz, respectively. In addition,
inverted fault. Our analysis of campaign (1994–2015) and the filtering frequencies for teleseismic P wave and S wave
continuous (2010–2018) GPS velocity fields estimates that are 0.01–0.09 Hz and 0.01–0.04 Hz, respectively. The multi-
the approximately north–south fault has a 4.9 [4.0, 5.5] ple-point-source inversion results are shown in Figure 2.
mm/yr slip rate with an earthquake recurrence interval of
115 [102, 141] yr. In addition, the comparison of horizontal Aftershock relocation
strain-rate tensor derived from the GPS velocity fields with We used aftershock events recorded by 17 regional seismic sta-
earthquakes at depths <30 km during the period 1973–2018 tions of Indonesia’s BMKG (Fig. S2) between 26 September and
shows that Ambon Island and the nearby regions have a 31 October 2019. During this period, the stations recorded 402
high strain accumulation rate that matches the distribution aftershock events of local magnitude (M L ) 2–5 at depths of
of M w ≥ 6 earthquakes, indicating that the region is seismically <30 km. The initial hypocenter locations of the events were gen-
active and possibly will experience more Ambon-type earth- erated by the GFZ SeisComP3 program, which uses the IASP91
quakes in the future. velocity model (Kennett and Engdahl, 1991).

3340 Seismological Research Letters www.srl-online.org • Volume 92 • Number 6 • November 2021

Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/92/6/3339/5452232/srl-2021021.1.pdf


by Nanyang Technological University user
(a) (b)

Mw 6.45
Mw 6.47
Mw 6.38
10 km 10 km
3.4˚S 0 10 20 0 10 20
Time (s) Time (s)

3.6˚S

128.2˚E 128.4˚E 128.6˚E 128.2˚E 128.4˚E 128.6˚E

(c) (d)
P vertical S tangential P vertical S tangential
4886 km
IU.WAKE
57.55° 65% 92% 93% 91%
3553 km
IU.HNR
102.2° 93% 77% 98% 80%
4584 km
G.NOUC
120.0° 94% 24% 98% 59%
8699 km
G.CRZF
223.3° 94% 54% 98% 57%
8960 km
II.ABPO
251.3° 95% 93% 99% 95%
5426 km
II.PALK
282.7° 88% 90% 96% 90%
9112 km
II.ARTI
328.1° 93% 68% 98% 85%

0 50 0 50 100 150 0 50 0 50 100 150

Time (s) Time (s)

Figure 2. Teleseismic waveform fitting for multiple-point-source representative stations are selected to show the waveform fitting
solutions. (a–b) Map view of the point-source solutions, for the (c) single-point-source and (d) two-point-source solution.
where (a) is the single-point-source solution and (b) is the two- The corresponding station code, epicentral distance, and azimuth
point-source solution. The focal mechanisms indicating the are indicated on the left of the waveforms. Data and synthetics are
centroid location of the subevents. Star denotes the relocated shown in black and red, respectively, with their cross-correlation
epicenter using the Meteorological, Climatological, and coefficients shown below the waveforms. The color version of this
Geophysical Agency (BMKG) first-arrival picks. The top right insets figure is available only in the electronic edition.
show the source time functions of the subevents. (c–d) Seven

Volume 92 • Number 6 • November 2021 • www.srl-online.org Seismological Research Letters 3341

Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/92/6/3339/5452232/srl-2021021.1.pdf


by Nanyang Technological University user
2017). The final estimated coseismic offset is shown in
Aftershocks Figure 4.
(ML)

−3.375° t4 5 InSAR data processing


t3 4 In addition to GPS coseismic offset, we also used Advanced
t2 3
Land Observing Satellite-2 InSAR scenes (Table 1). We proc-
essed the InSAR observations using modules of the InSAR
Depth (km) Scientific Computing Environment (Rosen et al., 2012; Liang
and Fielding, 2017) to get the line-of-sight (LoS) displace-
−3.5° 0 10 20
ments. After the processing, we manually fixed unwrapping
errors by shifting the displacements with multiples of 2π (see
Fig. S3 and Text S1). Then we applied Generic Atmospheric
Correction Online Service (GACOS) for InSAR data (Yu et al.,
2018) to correct tropospheric phase delay. However, we found
−3.625° that the GACOS correction worsened the LoS displacements:
the corrected LoS displacements are inconsistent with the GPS
10 km offset projected into the satellite look direction and have higher
128.25° 128.5°
standard deviation than the LoS displacements before GACOS
correction (Fig. S4). The reason is possibly because of the lack
of meteorological data in this region, resulting in inaccurate
Figure 3. Relocated aftershock distribution from this study. The
solid red line is the inverted fault model projected to the surface, estimation of the tropospheric phase delay. Thus, we used
and the dashed line is the inverted fault model in the subsurface. the LoS displacements without the GACOS tropospheric cor-
The color version of this figure is available only in the electronic rection for our inversion modeling (Fig. 4). Finally, we down-
edition. sampled the LoS displacements using the quadtree algorithm
(Jónsson et al., 2002) to produce a smaller set of data points for
the inversion modeling.

To relocate the aftershock events, we used the HypoDD Coseismic slip modeling
program that applies the double-difference algorithm of We computed the coseismic slip distribution by jointly
Waldhauser and Ellsworth (2000) and minimizes residuals inverting the coseismic offsets from GPS and the LoS displace-
between the observed and calculated travel-time differences ments. Our inversion modeling assumes fault dislocations in a
in an iterative procedure (Waldhauser, 2001). The method homogeneous elastic half-space (Okada, 1992). We fixed the
assumes that ray paths from two or more hypocenter events strike and dip angle of the fault to our multiple-point-source
to seismic stations are considered similar and propagate inversion result (strike = 344°, dip = 84°), divided the faults
through the same medium if the distances between the hypo- into small rectangles of ~2 by 2 km, determined a smoothing
center events are smaller than the distances between the hypo- factor following an L-curve criterion (Aster et al., 2005;
centers to seismic stations. We used the 1D seismic-velocity Fig. S5), used a Laplacian smoothing, set the initial rake based
model from ak135 (Kennett et al., 1995). The relocated after- on our multiple-point-source inversion result of 160° and
shock events are shown in Figure 3. allowed it to vary within ±45°, and used LoS uncertainties esti-
mated from an experimental semivariogram over nondeform-
GPS data processing ing regions in the LoS displacements (Bagnardi and Hooper,
The coseismic event was recorded by one continuous GPS 2018). To assess the modeling misfit, we compared the coseis-
station ~28 km southwest of the epicenter. To estimate the mic offsets observations and predictions using percentage of
coseismic offset, we processed the GPS data using GPS at MIT variance reduction, defined as
(GAMIT)/Global Kalman filter (GLOBK) software (Herring
et al., 2018) by incorporating nearby International Global  
Gm − d2
Navigation Satellite System Service stations to obtain the GPS p%  1 − × 100; 1
d2
EQ-TARGET;temp:intralink-;df1;320;171

time series in International Terrestrial Reference Frame


(ITRF) 2014. Then we used the time series to estimate a linear
velocity by removing outliers over 2σ rejection levels and by in which G is the Green’s functions calculated using Okada
eliminating the seasonal variations. Finally, we referenced the equations, m is our model, and d is our coseismic observations.
linear velocity to the Sunda frame using the following Euler To find an optimal relative weight between each data type, we
pole parameters: 44.16° N, −87.22° E, 0.349 °/Ma (Yong et al., ran a suite of inversion models, examined the effect to the

3342 Seismological Research Letters www.srl-online.org • Volume 92 • Number 6 • November 2021

Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/92/6/3339/5452232/srl-2021021.1.pdf


by Nanyang Technological University user
Results
Fault model ALOS Our multiple-point-source
−2.75° 2 - D0
20 inversions results suggest that
two-point-source solution fit
the amplitude and shape of
teleseismic P and S waves in
various azimuths and distances
−3°
20
20 kkm
m with higher cross-correlation
coefficient than a single-
point-source solution (Fig. 2).
Adding the third point source
−3.25° to the inversion does not
improve the overall data fit-
ting. Hence, we use the two-
point-source solution to
represent the earthquake rup-
−3.5° ture process, as shown in
Figure 2b. The initiation time
of the two subevents is similar,
1 cm
with only a 2 s difference
InSAR time span
(Fig. 2b inset). However, their
−3.75° Coseismic only
CAMB centroid times differ about
cm 8.5 s, indicating different fault
LOS
properties on the north and
20 km −10 0 10
south fault segments. In addi-
128° 128.5° 129° tion, a major difference of the
two segments is in their focal
Figure 4. Coseismic offsets from Advanced Land Observing Satellite-2 (ALOS-2) Interferometric mechanisms, which show
Synthetic Aperture Radar (InSAR) observations along the descending track and from Global oblique-slip faulting on the
Positioning System (GPS) data. Positive values represent line-of-sight (LoS) displacements toward north fault segment and
the satellite. The black vector is the offset based on GPS displacements of 15 days before the event
strike-slip faulting on the south
and one day after the event. The purple vector is the total offset covering the time span as the
InSAR data (includes surface displacements four days after the event). Uncertainty ellipses fault segment. The moment
represent 95% confidence levels. The colored square represents the purple GPS vector projected magnitude of the south sube-
onto the satellite look direction. The comparison of black and purple vectors suggests that our LoS vent (M w 6.45) is relatively
displacements also captured both the coseismic and ~1 cm postseismic displacements. The color larger than that of the north
version of this figure is available only in the electronic edition. subevent (M w 6.38). The spa-
tial linear alignment of the
two subevents as well as the
epicenter (Fig. 2b) are consis-
misfit of each data type, and then made adjustments in the rel- tent with the relocated aftershock events (Fig. 3). We
ative weight until each data type fit well. Through this take a note that the single-point-source solution (Fig. 2a) is
trial-and-error procedure, the relative weight of the InSAR very similar to the south subevent in the two-point-source
to GPS is 20. solution (Fig. 2b), but the single-point-source inversion did
not capture the north subevent, maybe because the rupture
to the north was relatively slow and weak. Therefore, our
TABLE 1 two-point-source solution shows that the rupture process of
Interferometric Synthetic Aperture Radar (InSAR) the M w 6.5 earthquake was complex in both the rupture ini-
Observations along Descending Track Used in This tiation and slip directions on the north and south fault
Study
segments.
Satellite Mode Path Frame Acquisition Date (2019) Our relocated aftershock events indicate that the earthquake
ruptured two faults that are trending almost orthogonal to each
ALOS-2 ScanSAR 020 3700 18 August 29 September
other: the first fault trending approximately north–south,
ALOS-2, Advanced Land Observing Satellite-2. whereas the second fault trending approximately east–west

Volume 92 • Number 6 • November 2021 • www.srl-online.org Seismological Research Letters 3343

Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/92/6/3339/5452232/srl-2021021.1.pdf


by Nanyang Technological University user
S N observation indicates that our LoS displacements also captured
0
both the coseismic and at least ~1 cm postseismic dis-
placements.
5
Our modeling result (Fig. 5) shows that the mainshock
50
rupture was mostly strike-slip motion with some normal com-
Depth (km)

10 10 50
0 ponent and pure reverse slip in the north. The model also sug-

50

50
gests that most of the slips on the fault are in the ocean, where
15 Global This the surface deformation cannot be captured by our InSAR data
CMT study cm (Figs. 4 and 6). Therefore, our InSAR inversion cannot resolve
20 50 the fault geometry and rupture complexity to greater details.
0 50 100 150
Our checkerboard test (produced using the same smoothing
−10 0 10 20 factor as that used in the inversion using real data) shows that
Distance along strike (km) the data have the resolution to resolve slip on the upper 10 km
depths (Fig. S7). The comparison of the estimated slip and
Figure 5. Our preferred coseismic slip model shows that the relocated aftershock events show that the depths of maximum
earthquake rupture was mostly strike-slip motion with normal slip (~100 cm) coincide with the maximum depths of the after-
component and pure reverse slip in the north. The color version shock events at 11 km, indicating the seismogenic depth of
of this figure is available only in the electronic edition.
the fault.

Discussion
The effect of stress changes on nearby faults
(Fig. 3). However, the relocated aftershock events by Sahara To evaluate the effect of the Ambon earthquake on the nearby
et al. (2021) who deployed a temporary seismic network mapped faults, we use our slip distribution to calculate
nearby the mainshock event suggest that the approximately Coulomb stress changes using Coulomb 3.3 software (Lin and
east–west fault does not exist (Fig. S6). Nevertheless, the Stein, 2004; Toda et al., 2005). Our Coulomb stress changes at a
two relocated events share similar characteristic for the exten- depth of 5 km, based on a friction coefficient of 0.4, shows that
sion of the approximately north–south fault. the Ambon event imparted at least 10 KPa Coulomb stress
Our GPS data show that the earthquake produced an ~4 cm changes, a typical earthquake triggering threshold (King et al.,
coseismic offset at GPS station CAMB ~28 km to the southwest 1994; Hardebeck et al., 1998), on Kawa and South Buru faults
of the epicenter location (black vector in Fig. 4). In addition, (Fig. 7). Our calculations at different depths (Fig. S8) also show
our InSAR data that contain LoS displacements in a large area similar features to Figure 7.
show that the event produced surface deformations in all
islands close to the epicenter location. Particularly, the event Seismic hazard assessments
produced ~10 cm displacements along the western coastline Regional areas. To evaluate the regional seismic hazard,
of Haruku Island (Fig. 4). The comparison of GPS coseismic we use campaign GPS velocity fields (1994–2015) derived
offset (black vector) and cumulative GPS offset (purple vector) by Susilo et al. (2016), and we process continuous GPS data
covering the same time span as the InSAR data (Fig. 4) suggests (2010–2018) by following the same processing procedures to
that GPS station CAMB recorded ~1 cm cumulative postseis- prepare the GPS coseismic offset. The GPS velocities in
mic displacement during four days after the event. This ITRF2008 and Sunda reference frame are available in
Tables S1 and S2, respectively.
Then we interpolate the east
and north GPS velocity com-
Observations Predictions Residuals
−3.25° cm ponents using spline interpola-
10 tion (Wessel and Bercovici,
1998) to a grid size of 1°.
−3.5° 0 Finally, we use the interpolated
GPS velocity to compute the
Model −10
−3.75° 20 km p = 79% Data (1 cm) second invariant of the hori-
128.25° 128.625° 128.25° 128.625° 128.25° 128.625° zontal strain-rate tensor fol-
lowing the approach of Hackl
Figure 6. Observations, predictions, and residuals LoS displacements. In the last panel, black and red et al. (2009). We compare
vectors are observed and synthetic GPS offsets, respectively. The color version of this figure is the second invariant with
available only in the electronic edition. M w ≥ 3 earthquakes of depths

3344 Seismological Research Letters www.srl-online.org • Volume 92 • Number 6 • November 2021

Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/92/6/3339/5452232/srl-2021021.1.pdf


by Nanyang Technological University user
11 km. The new model results
KPa in 4.9 [4.0, 5.5] mm/yr slip rate
50
5
500 km
kkm
m (Fig. 9). Based on the fault
−100 0 100 length of ~34 km (Fig. 5),

10
the seismogenic depth of
North Buru
11 km (Fig. 9), the estimated
10 slip rate, and its
−3°
uncertainty, we estimate the
10
Kaw recurrence interval of the fault
a
is 115 [102, 141] yr.

National regions. After


Manifa

10 the 2004 M w 9.2 Aceh earth-


quake that claimed about a
quarter million of people, sev-
eral megathrust earthquakes
10

10
have hit the country: the 2005
M w 8.7 Nias earthquake, the
2006 M w 7.7 Java earthquake,
−4°
the 2008 M w 8.4 Bengkulu
earthquake, and the 2010
South Buru M w 7.8 Mentawai earthquake.
South Buru In addition to megathrust
128° 129° earthquakes, numerous crustal
fault earthquakes with magni-
tude sizes of 6–7 also struck
Figure 7. Coulomb stress changes at a depth of 5 km due to the earthquake shows that the event
imparted at least ~10 KPa, an earthquake-triggering threshold (King et al., 1994; Hardebeck et al., the national regions during the
1998), positive stress changes on Kawa and South Buru faults. The color version of this figure is period 2006–2019. Despite
available only in the electronic edition. being smaller in size than the
megathrust earthquakes, the
threat was worrying because
these events were generated
<30 km during the period 1973–2009 (U.S. Geological Survey by the ruptures of shallow crustal faults that lie beneath dense
catalog) and 2009–2018 (Supendi et al., 2020). The comparison population centers. Since 2006, including the Ambon event,
shows that the city of Ambon and the nearby regions have seven shallow crustal fault earthquakes have claimed signifi-
higher strain rates than the rest of the regions and match cant casualties in Indonesia (Table 2). Of these seven events,
the distribution of M w ≥ 6 earthquakes (Fig. 8), indicating that two earthquakes (~30%) ruptured mapped faults, whereas the
the areas are seismically active and possibly will experience remaining five events (~70%) ruptured faults that were previ-
more Ambon-type earthquakes in the future. ously unmapped. These facts highlight that the seismic hazards
To better understand the earthquake cycle of the Ambon during 2006–2019 have also been posed by unmapped active
earthquake, we calculate slip rate and locking depth of the faults. Therefore, we strongly recommend escalating efforts on
fault using GPS velocity fields within 100 km from the fault identifying unknown active faults to better improve the
(Fig. 8). In the calculation, we fit a half-space dislocation model national seismic hazard assessments.
(Savage and Burford, 1973) by sampling model parameters
using a slice sampling algorithm (Neal, 2003) (see Text S2). Conclusion
From the maximum a posteriori (MAP) of probability distri- We study the 26 September 2019 M w 6.5 Ambon earthquake
bution function of model parameters, we estimate that the fault that was the largest instrumentally recorded event to strike
has a 4.3 [3.6, 5.2] mm/yr slip rate and a 3.6 [0.6, 12.5] km Ambon, the capital city of Maluku Islands, eastern Indonesia.
locking depth (Fig. S9). We consider the MAP of the locking Our results based on seismic and geodetic data show that the
depth that is <4 km less accurate because it is not consistent rupture process was complex in both the rupture initiation and
with the seismogenic depth of 11 km inferred from the com- slip directions. In addition, the rupture was mostly strike-slip
parison of our slip model and the relocated aftershock events. motion with normal component and pure reverse slip in the
Therefore, we run another model by fixing the locking depth to north of the inverted fault. Our analysis of campaign and

Volume 92 • Number 6 • November 2021 • www.srl-online.org Seismological Research Letters 3345

Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/92/6/3339/5452232/srl-2021021.1.pdf


by Nanyang Technological University user
Data and Resources
Focal mechanisms We are grateful to The Japanese

(Mw ) Aerospace Exploration Agency
t5 6 (JAXA) for giving us Advanced
t6 7 Land Observing Satellite-2
t7 8 (ALOS-2) data access through a
M w t3 earthquakes proposal granted to Eric Lindsey
0° under the Second Research
nanostrain/yr
Announcement for Earth
Observations PI No. 017; owner-
0 50 100
km ship and copyright of raw data
0 100 200 maintained by JAXA. We created
30 mm/yr all figures using the Generic
−2° 6.5/2019 Mapping Tools (GMT; Wessel
et al., 2013) and a color map from
Crameri (2018). Supplemental
material for this article includes
a text describing slip rate and lock-
B
A ing depth calculation and addi-
−4°
tional figures referenced in the
article.

Declaration of
−6° Competing
Interests
The authors acknowledge there
are no conflicts of interest
recorded.
−8°

12 4° 128° 132° Acknowledgments


The authors thank Meteorological,
Climatological, and Geophysical
Figure 8. Background color is the second invariant of the horizontal strain-rate tensor derived from
Agency (BMKG) for providing
our regional GPS velocities (red vectors) with respect to the Sunda reference frame. We masked out
the resulting second invariant located outside the coverage of our GPS velocities to avoid over- regional seismic data and BIG
interpreting artifacts due to the interpolation. Yellow circles represent earthquakes with depths for providing continuous Global
<30 km during the period 1973–2009 (USGS catalog) and 2009–2018 (Supendi et al., 2020). The Positioning System (GPS) data. I.
focal mechanisms are from the Global CMT catalog (Dziewonski et al., 1981; Ekström et al., 2012). M. is partially supported by
Black lines are the nearby mapped faults from Irsyam et al. (2020). The cyan line represents an east– “Hibah Penelitian Dasar
west GPS profile used to compute slip rate and locking depth of the fault. The color version of this Unggulan Perguruan Tinggi,
figure is available only in the electronic edition. Kemenristek (No. 2/E1/
KP.PTNBH/2020)”. S. R. is sup-
ported by the Indonesian
Ministry of Research,
Technology, and Higher Education through a scholarship
continuous GPS velocity fields estimates that the fault has a 4.9
program of “Pendidikan Magister Menuju Doktor Untuk Sarjana
[4.0, 5.5] mm/yr slip rate with an earthquake recurrence
Unggul (PMDSU) Batch IV (2018)”. R. S. was supported by the
interval of 115 [102, 141] yr. In addition, the comparison of Earth Observatory of Singapore (EOS) via its funding from the
the horizontal strain-rate tensor derived from the same GPS National Research Foundation Singapore and the Singapore
velocity fields with historical earthquake data shows that Ministry of Education under the Research Centres of Excellence
Ambon Island and the nearby regions have a high strain accu- Initiative and by the National Research Foundation
mulation rate that matches the distribution of M w ≥ 6 earth- Investigatorship Scheme (Award Number NRF-NRFI05-2019-0009
quakes, indicating that the regions are seismically active and to Emma Hill). R. S. thanks Shengji Wei for his comments on the
possibly will experience more Ambon-type earthquakes in article. This is Earth Observatory of Singapore Contribution
the future. Number 362.

3346 Seismological Research Letters www.srl-online.org • Volume 92 • Number 6 • November 2021

Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/92/6/3339/5452232/srl-2021021.1.pdf


by Nanyang Technological University user
(a) (b) situation-update-no-10-aceh-pidie-jaya-earthquake-wednesday-
1 21-december-2016 (retrieved January 2021).
Fault-parallel velocity (mm/yr)

Slip rate:
2
4.9 mm/yr [4.0, 5.5]
ASEAN Humanitarian Assistance (AHA) Centre (2018). Situation
update No. 15—M 7.4 earthquake and tsunami, Sulawesi,
Locking depth (fixed):

PDF
0
11 km
Indonesia, 1–19, available at https://ahacentre.org/situation-update/
situation-update-no-15-sulawesi-earthquake-26-october-2018/
–2 (retrieved January 2021).
0 Aster, R. C., B. Borchers, and C. H. Thurber (2005). Parameter
–40 0 40 80 3 4 5 6
Distance from fault (km) Slip rate (mm/yr) Estimation and Inverse Problems, Elsevier Academic, San Diego,
California.
Badan Nasional Penanggulan Bencana (BNPB) (2013). Rencana aksi
Figure 9. (a) Fault-parallel GPS velocity fields (black dots) and the
synthetic model (cyan line) based on the estimated slip rate and rehabilitasi dan rekonstruksi wilayah pasca bencana gempa bumi
locking depth of the fault. (b) Posterior probability distribution Kabupaten Aceh Tengah dan Bener Meriah tahun 2013–2014,
function (PDF) of the slip rate based on a locking depth fixed to Retrieved January 2021 (in Indonesian).
11 km. The solution for the slip rate is based on the maximum a Badan Nasional Penanggulan Bencana (BNPB) (2019). BNPB: 41 tewas
posteriori (MAP) of the PDF, which is drawn as a vertical solid line, and 103 ribu orang masih mengungsi akibat gempa Ambon, avail-
with 95% confidence intervals drawn as dash lines. The color able at https://bpbd.bengkaliskab.go.id/web/detailberita/18/bnpb-
version of this figure is available only in the electronic edition. 41-tewas-dan-103-ribu-orang-masih-mengungsi-akibat-gempa-
ambon (last accessed January 2021) (in Indonesian).
Bagnardi, M., and A. J. G. Hooper (2018). Inversion of surface defor-
mation data for rapid estimates of source parameters and uncer-
tainties: A Bayesian approach, Geochem. Geophys. Geosys. 19,
TABLE 2 no. 7, 2194–2211.
Deadliest Shallow Crustal Fault Earthquakes in BBC Indonesia (2007). Gempa kuat guncang Sumbar, available at http://
Indonesia since 2006 www.bbc.co.uk/indonesian/news/story/2007/03/070306_indonquake
.shtml (last accessed January 2021) (in Indonesian).
Fault Consultative Group on Indonesia (2006). Preliminary damage and
Number Earthquake Casualty Source
loss assessment, Yogyakarta and central Java natural disaster: A
1 2006 Mw 5700 Unmapped
joint report of BAPPENAS, the provincial and local governments
6.3 Jogjakarta (CGI, 2006) (Tsuji et al., 2009)
of D.I. Yogyakarta, the provincial and local governments of central
2 2018 Mw 3390 Mapped Java, and international partners, 15th Meeting of the Consultative
7.5 Sulawesi (AHA Centre, (Hamilton, 1979) Group on Indonesia (CGI) Jakarta, Jakarta, Indonesia, 14 June
2018) 2006.
Crameri, F. (2018). Geodynamic diagnostics, scientific visualisation
3 2016 Mw 104 Unmapped
6.5 Aceh and StagLab 3.0, Geosci. Model Dev. 11, no. 6, 2541–2562.
(AHA Centre, (Muzli et al., 2018)
Daryono, M. R., and A. Tohari (2016). Surface rupture and geotech-
2016)
nical features of the July 2, 2013 Tanah Gayo earthquake, Indones.
4 2007 Mw 69 Mapped J. Geosci. 3, no. 2, 95–105.
6.4 Padang (BBC Indonesia, (Sieh and Natawidjaja, Dziewonski, A. M., T. A. Chou, and J. H. Woodhouse (1981).
2007) 2000) Determination of earthquake source parameters from waveform
5 2013 Mw 42 Unmapped data for studies of global and regional seismicity, J. Geophys.
6.1 Aceh (BNPB, 2013) (Daryono and Tohari, Res. 86, no. B4, 2825–2852.
2016) Ekström, G., M. Nettles, and A. M. Dziewoński (2012). The global
CMT project 2004–2010: Centroid-moment tensors for 13,017
6 2019 Mw 41 Unmapped
earthquakes, Phys. Earth Planet. In. 200, 1–9.
6.5 Ambon (BNPB, 2019) (This study)
Frohlich, C., and K. D. Apperson (1992). Earthquake focal mecha-
7 2019 Mw 7.2 13 Unmapped nisms, moment tensors, and the consistency of seismic activity
Halmahera (KOMPAS, 2019) (Unstudied) near plate boundaries, Tectonics 11, no. 2, 279–296.
Hackl, M., R. Malservisi, and S. Wdowinski (2009). Strain rate pat-
BBC Indonesia and KOMPAS are national newspapers. AHA, ASEAN Humanitarian
terns from dense GPS networks, Nat. Hazards Earth Syst. Sci.
Assistance; BNPB, National Disaster Management Authority of Indonesia; CGI, The
Consultative Group on Indonesia.
9, 1177–1187.
Hamilton, W. B. (1979). Tectonics of the Indonesian region
(No. 1078), U.S. Government Printing Office, i–ix, 1–344, available
References at https://pubs.er.usgs.gov/publication/pp1078 (last accessed
ASEAN Humanitarian Assistance (AHA) Centre (2016). Situation January 2021).
update No. 10—Aceh Pidie Jaya earthquake, Indonesia, 1–3, Hardebeck, J. L., J. J. Nazareth, and E. Hauksson (1998). The static
available at https://reliefweb.int/report/indonesia/form-2- stress change triggering model: Constraints from two southern

Volume 92 • Number 6 • November 2021 • www.srl-online.org Seismological Research Letters 3347

Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/92/6/3339/5452232/srl-2021021.1.pdf


by Nanyang Technological University user
California aftershock sequences, J. Geophys. Res. 103, no. B10, Savage, J. C., and R. O. Burford (1973). Geodetic determination of
24,427–24,437. relative plate motion in central California, J. Geophys. Res. 78,
Herring, T. A., R. W. King, M. A. Floyd, and S. C. McClusky (2018). no. 5, 832–845.
Introduction to GAMIT/GLOBK, Release 10.7, Massachusetts Shi, Q., S. Wei, and M. Chen (2018). An MCMC multiple point sources
Institute of Technology, Cambridge, Massachusetts, available at inversion scheme and its application to the 2016 Kumamoto M w 6.2
http://geoweb.mit.edu/gg/Intro_GG.pdf (last accessed December earthquake, Geophys. J. Int. 215, no. 2, 737–752.
2020). Sieh, K., and D. H. Natawidjaja (2000). Neotectonics of the Sumatran
Irsyam, M., P. R. Cummins, M. Asrurifak, L. Faizal, D. H. Natawidjaja, fault, Indonesia, J. Geophys. Res. 105, no. B12, 28,295–28,326.
S. Widiyantoro, I. Meilano, W. Triyoso, A. Rudiyanto, S. Hidayati, Sipkin, S. A. (1986). Interpretation of non-double-couple earthquake
et al. (2020). Development of the 2017 national seismic hazard mechanisms derived from moment tensor inversion, J. Geophys.
maps of Indonesia, Earthq. Spectra 36, no. 1_suppl, 112–136, Res. 91, no. B1, 531–547.
doi: 10.1177/8755293020951206. Spakman, W., and R. Hall (2010). Surface deformation and slab–man-
Jónsson, S., H. Zebker, P. Segall, and F. Amelung (2002). Fault slip tle interaction during Banda arc subduction rollback, Nature
distribution of the 1999 M w 7.1 Hector Mine, California, earth- Geosci. 3, no. 8, 562–566.
quake, estimated from satellite radar and GPS measurements, Supendi, P., A. D. Nugraha, S. Widiyantoro, J. D. Pesicek, C. H.
Bull. Seismol. Soc. Am. 92, no. 4, 1377–1389. Thurber, C. I. Abdullah, D. Daryono, S. H. Wiyono, H. A.
Kennett, B. L. N., and E. R. Engdahl (1991). Travel times for global Shiddiqi, and S. Rosalia (2020). Relocated aftershocks and back-
earthquake location and phase association, Geophys. J. Int. 105, ground seismicity in eastern Indonesia shed light on the 2018
429–465. Lombok and Palu earthquake sequences, Geophys. J. Int. 221,
Kennett, B. L. N., E. R. Engdahl, and R. Buland (1995). Constraints on no. 3, 1845–1855.
seismic velocities in the Earth from traveltimes, Geophys. J. Int. Susilo, S., I. Meilano, H. Z. Abidin, B. Sapiie, J. Efendi, and A. B.
122, 108–124. Wijanarto (2016). Velocity field from twenty-two years of com-
King, G. C., R. S. Stein, and J. Lin (1994). Static stress changes and the bined GPS daily coordinate time series analysis, AIP Conf. Proc.
triggering of earthquakes, Bull. Seismol. Soc. Am. 84, no. 3, 935–953. 1730, no. 1, 040003, doi: 10.1063/1.4947393.
KOMPAS (2019). Pasca-gempa magnitudo 7.2 di Halmahera Selatan, 13 Toda, S., R. S. Stein, K. Richards-Dinger, and S. B. Bozkurt (2005).
meninggal hingga 26 sekolah rusak, available at https://regional Forecasting the evolution of seismicity in southern California:
.kompas.com/read/2019/07/26/16410061/pasca-gempa-magnitudo- Animations built on earthquake stress transfer, J. Geophys. Res.
7-2-di-halmahera-selatan-13-meninggal-hingga-26?page=all (last 110, no. B5, doi: 10.1029/2004JB003415.
accessed January 2021) (in Indonesian). Tsuji, T., K. Yamamoto, T. Matsuoka, Y. Yamada, K. Onishi, A. Bahar,
Liang, C., and E. J. Fielding (2017). Interferometry with ALOS-2 full- I. Meilano, and H. Z. Abidin (2009). Earthquake fault of the 26
aperture ScanSAR Data, IEEE Trans. Geosci. Rem. Sens. 55, no. 5, May 2006 Yogyakarta earthquake observed by SAR interferometry,
2739–2750, doi: 10.1109/TGRS.2017.2653190. Earth Planets Space 61, no. 7, e29–e32.
Lin, J., and R. S. Stein (2004). Stress triggering in thrust and subduc- Waldhauser, F. (2001). HypoDD—A program to compute double-dif-
tion earthquakes and stress interaction between the southern San ference hypocenter locations, U.S. Geol. Surv. Open-File Rept.
Andreas and nearby thrust and strike-slip faults, J. Geophys. Res. 2001-113.
109, no. B2, doi: 10.1029/2003JB002607. Waldhauser, F., and W. L. Ellsworth (2000). A double-difference
Muzli, M., M. Umar, A. D. Nugraha, K. E. Bradley, S. Widiyantoro, K. earthquake location algorithm: Method and application to the
Erbas, P. Jousset, S. Rohadi, I. Nurdin, and S. Wei (2018). The 2016 Northern Hayward fault, California, Bull. Seismol. Soc. Am. 90,
M w 6.5 Pidie Jaya, Aceh, North Sumatra, earthquake: Reactivation 1353–1368, doi: 10.1785/0120000006.
of an unidentified sinistral fault in a region of distributed defor- Wessel, P., and D. Bercovici (1998). Interpolation with splines in ten-
mation, Seismol. Res. Lett. 89, no. 5, 1761–1772. sion: A Green’s function approach, Math. Geol. 30, 77–93.
Neal, R. M. (2003). Slice sampling, Ann. Stat. 31, 705––741. Wessel, P., W. H. Smith, R. Scharroo, J. Luis, and F. Wobbe (2013).
Okada, Y. (1992). Internal deformation due to shear and tensile Generic mapping tools: Improved version released, Eos Trans.
faults in a half-space, Bull. Seismol. Soc. Am. 82, no. 2, AGU 94, no. 45, 409–410.
1018–1040. Yong, C. Z., P. H. Denys, and C. F. Pearson (2017). Present-day kin-
Rosen, P. A., E. Gurrola, G. F. Sacco, and H. Zebker (2012). The ematics of the Sundaland plate, J. Appl. Geod. 11, no. 3, 169–177.
InSAR scientific computing environment, EUSAR 2012, 9th Yu, C., Z. Li, N. T. Penna, and P. Crippa (2018). Generic atmospheric
European Conf. on Synthetic Aperture Radar, 730–733. correction model for Interferometric Synthetic Aperture Radar
Sahara, D. P., A. D. Nugraha, A. Muhari, A. A. Rusdin, S. Rosalia, A. observations, J. Geophys. Res. 123, no. 10, 9202–9222.
Priyono, Z. Zulfakriza, S. Widiyantoro, N. T. Puspito, A. Rietbrock,
et al. (2021). Source mechanism and triggered large aftershocks of
the M w 6.5 Ambon, Indonesia earthquake, Tectonophysics 799, Manuscript received 21 January 2021
228709, doi: 10.1016/j.tecto.2020.228709. Published online 11 August 2021

3348 Seismological Research Letters www.srl-online.org • Volume 92 • Number 6 • November 2021

Downloaded fromView
http://pubs.geoscienceworld.org/ssa/srl/article-pdf/92/6/3339/5452232/srl-2021021.1.pdf
publication stats
by Nanyang Technological University user

You might also like