Introduction

On September 28, 2018, an Mw 7.5 devastating earthquake and tsunami affected the city of Palu in Central Sulawesi, Indonesia. As of November 22, 2018, the Indonesian Agency for Meteorology, Climatology, and Geophysics (BMKG) has recorded 554 aftershocks in this area with a significant number of events having a magnitude ≥ 3 with the use of a dense regional seismic network. According to the BMKG catalog, the mainshock occurred at 10:02:44 UTC and the epicenter was located at 0.18° S; 119.85° E with a depth of 10 km depth (Fig. 1). About 12 min after the mainshock, a sequence of aftershocks continued occurring until the final date that these data were downloaded, which was November 23, 2018, but even after that aftershocks still occurred. The shake map from BMKG and the community reports indicate that the earthquake was rated VII to VIII on the Modified Mercalli Intensity (MMI) scale in Palu and the surrounding area.

Fig. 1
figure 1

Map of the study area. The red star depicts the Mw 7.5 mainshock; the green inverted triangles are the BMKG seismic station used in this study; blue traces represent the Palu-Koro Fault; and red traces correspond to the other major crustal faults in the region extracted from Irsyam et al. (2017). The black boxes show the map regions of Figs. 2, 3, 4, 5, 6, 8 and 9

Full size image

According to the National Disaster Management Authority (NDMA) report (http://bnpb.go.id/en), the tsunami and liquefaction caused more than 4000 fatalities. The geological map made by Watkinson (2011) shows that the city of Palu consists of Holocene sedimentary rock. Pramono et al. (2017) used the multichannel analysis of surface waves (MASW) method to conclude that the city of Palu and its surrounding area consist of alluvium and soft soil. Thus, the damage caused by the earthquake and the liquefaction was massive in the Palu area. Gusman et al. (2019) showed that the tsunami was caused by a combination of sudden ground and seafloor changes due to the earthquake, along with landslides, and a high tide at the time of the event.

The earthquake was generated by the strike-slip faulting of the Palu-Koro Fault (Bao et al. 2019; Socquet et al. 2019). The Palu-Koro Fault has a slip rate of about 42 mm/year, which was estimated by Global Positioning System (GPS) and slip rate modeling (Socquet et al. 2006). Daryono (2016) suggested that this fault includes active faults with a slip rate of around 30–40 mm yearly and can potentially generate a co-seismic slip. As a result, seismic hazard along Palu-Koro Fault segment in the vicinity of a highly populated area is also increasing.

Interestingly, a month after the mainshock, a swarm earthquake occurred in Mamasa, which is ~ 230 km to the south of Palu (Fig. 1). As of November 22, 2018, the BMKG has recorded 556 events with a magnitude of M > 2. These were located in the area at an average depth of 10 km. Unfortunately, some of the earthquakes caused damage to several houses and economic loss. However, the source of these swarm earthquakes is still unclear: whether they occurred due to known activity at currently dormant volcanoes or a static triggering as a result of the devastating Palu earthquake. Therefore, this study aims to relocate the aftershocks of the Palu earthquake and the swarm earthquakes to obtain more precise hypocenter locations, as well as to conduct a focal mechanism analysis to estimate the fault type in the Mamasa area.

Data and method

The arrival time data used in this study were obtained from September 28 to November 22, 2018, at BMKG seismic stations in Sulawesi and Borneo (Fig. 1). During this period, there were 554 aftershocks from the Palu earthquake and 556 swarm earthquakes from Mamasa, constituting 5608 and 2649 P- and S-wave arrival times, respectively. The velocity model from IASPEI91 (Kennett and Engdahl 1991) was used for the initial hypocenter determination of the BMKG catalog, using the SeisComP3 program (GFZ).

We used the HypoDD program (Waldhauser 2001) to perform the double-difference method (Waldhauser and Ellsworth 2000) for relocating the aftershock hypocenters. The method assumes that if there are two earthquakes with a hypocentral distance smaller than the distance from the hypocenters to the station, then the ray paths of these two earthquakes to the station can be assumed to be the same and therefore, propagate through the same medium. This method has been successful in relocating earthquakes in Indonesia using the BMKG network data with some prominent tectonic interpretations: for example, in Sumatra (Nugraha et al. 2018a), West Java (Supendi et al. 2018a), Sulawesi (Ismullah et al. 2017; Supendi et al. 2018b) and Molucca (Utama et al. 2015; Nugraha et al. 2018b).

We applied a statistical resampling approach “bootstrap” method (Efron 1982; Billings 1994; Shearer 1997) to assess the reliability of the error estimates. For the final hypocenters, we replaced the final residuals with samples drawn with replacements from the observed residual distribution and relocated all events with these bootstrap sample data and unit weights to determine the shift in location with the resampled data vector. We applied Gaussian noise to the data with a standard deviation 0.1 s. The process was then repeated 1000 times.

For selected events in the Mamasa earthquakes, we used the ISOLA package (Sokos and Zahradnik 2008) to perform moment tensor inversions from at least four BMKG seismic stations (see inverted green triangles in Fig. 1). The observed waveforms were preprocessed using a high-pass filter with a corner frequency of 0.075 Hz to 0.15 Hz. For hypocenter relocation and focal mechanism determination, we used the 1-D seismic velocity model AK135 (Kennett et al. 1995).

Results and discussion

We have relocated 386 aftershocks from the Palu earthquake (Fig. 2). We first compared the relocated aftershocks with the initial locations (Fig. 3). The relocated hypocenters were then plotted in the vertical cross section and show a northwest–southeast trending (Fig. 3). The relocated hypocenters exhibit an improvement in clustering both horizontally and vertically, as shown in Fig. 3. Relative location errors for the 386 aftershocks along the Palu-Koro Fault are shown in Fig. 4. Relative horizontal and vertical error ellipses are shown to be at the 95% confidence level. Ellipses are computed from the major axes of the horizontal and vertical projection of the 95% confidence ellipsoids obtained from a bootstrap analysis of the final double-difference vector. The distribution of the major and minor axes of the horizontal and vertical projections of the ellipsoids for the Palu aftershocks is shown in Fig. 7a. Average mislocations horizontally and vertically are generally less than 2 km, and the maximum dislocation is less than 13 km (Table 1).

Fig. 2
figure 2

Map view for relocated events of the Palu aftershocks; red-to-blue circles represent the epicenters of earthquakes as a function of the focal depths. Red star illustrates the epicenter of the mainshock, red beach ball diagram denotes the global centroid moment tensor (gCMT) solution, blue traces correspond to the Palu-Koro Fault and red traces represent other major crustal faults in the region extracted from Irsyam et al. (2017)

Full size image
Fig. 3
figure 3

Vertical cross section of the aftershocks parallel to the fault before relocation (left panel); after relocation (right panel), 386 events, respectively. The red star represents the Mw 7.5 mainshock, and the blue line depicts the Palu-Koro Fault

Full size image
Fig. 4
figure 4

a Map view of relative location errors for the 386 aftershocks along the Palu-Koro Fault Zone; b depth view along latitude; and c depth view along longitude. Relative horizontal and vertical error ellipses are shown at the 95% confidence level. Ellipses are computed from the major axes of the horizontal and vertical projection of the 95% confidence ellipsoids obtained from a bootstrap analysis of the final double-difference vector

Full size image
Table 1 Horizontal (DX, DY) and vertical (DZ) deviation shift with Gaussian noise (0.1 s) for the Palu aftershocks
Full size table

The distribution of aftershocks extended from the north to the south of the mainshock (Fig. 2). The location of the aftershocks is consistent with the crustal deformation data in the area. The Geospatial Information Authority of Japan (GSI) applied interferometric analysis using ALOS-2/PALSAR-2 data to show that crustal deformation occurred in the part of the island (https://www.gsi.go.jp). Based on the vertical cross section in parallel to the fault (cross section A), the aftershocks were mostly located less than a depth of 20 km, which stay within the seismogenic zone, whereas the trend shown by the hypocenter in the northern part (close to the Mw 7.5 mainshock) is shallower than in the southern part. Based on the distribution of relocated aftershocks, it can be seen that the events have a NW–SE trending about ~ 200 km in length and ~ 50 km in width.

We have relocated 535 of the 556 swarm earthquakes in Mamasa with a magnitude of M 2 to M 5.4 (Fig. 5). The events that had previously been held fixed at 10 km could now be relocated/resolved (Fig. 5). Our results show that the earthquake swarms probably correspond to the activity of the local fault in the area, indicated by the fact that the seismicity pattern has a dip that becomes shallower toward the west (dipping at a ~ 45° angle) and extends from north to south with a length of ~ 50 km (Fig. 8b). Relative location errors for the 535 swarm earthquakes in Mamasa are shown in Fig. 6. The distribution of the major and minor axes of the horizontal and vertical projections of these ellipsoids for the events is shown in Fig. 7b. The spatial distribution of relative error agrees with the relocated seismicity pattern. This confirms that the seismic swarm sequence has a dip with a 45° angle. Average mislocations horizontally and vertically are less than 1.1 km, and the maximum dislocation is less than 9 km (Table 2).

Fig. 5
figure 5

Map of 535 relocated swarm earthquakes in Mamasa, West Sulawesi; for (left panel) the initial location of the BMKG catalogue; (right panel) after relocation using the double-difference method used in this study

Full size image
Fig. 6
figure 6

a Map view of relative location errors for the 535 earthquakes in Mamasa; b depth view along latitude; c depth view along longitude. Relative horizontal and vertical error ellipses are shown at the 95% confidence level. Ellipses are computed from the major axes of the horizontal and vertical projection of the 95% confidence ellipsoids obtained from a bootstrap analysis of the final double-difference vector

Full size image
Fig. 7
figure 7

Histograms of lateral and vertical relative location errors of double-difference solutions for a the Palu aftershocks; b the Mamasa swarm earthquakes. Errors are computed from the major axes of the horizontal and vertical projection of the 95% confidence ellipsoids obtained from a bootstrap analysis of the final double-difference vector based on 1000 samples with replacement

Full size image
Table 2 Horizontal (DX, DY) and vertical (DZ) deviation shift with Gaussian noise (0.1 s) for the Mamasa sequence
Full size table

We also conducted a focal mechanism analysis to estimate the type of fault slip for selected events with a magnitude of M > 4.5 (Fig. 8a). Most of the focal mechanism solutions show the normal fault type. We plotted a spatiotemporal distribution of the aftershocks right after the Mw 7.5 mainshock and the swarm earthquakes in Mamasa (Fig. 9).

Fig. 8
figure 8

a Focal mechanism solution for selected events (M > 4.5) Mamasa swarm earthquakes; b cross section A after relocation; dashed blue line is the first-order interpretation of dipping fault, cross section location in Fig. 4

Full size image
Fig. 9
figure 9

Map view of spatiotemporal distribution of relocated Palu aftershocks and the Mamasa swarm earthquakes. Colored dots depict the sequence number of events (days) relative to the Mw 7.5 Palu earthquake (September 28, 2018, to November 22, 2018)

Full size image

As noted from the spatiotemporal distribution of the relocated seismicity, the swarm earthquakes in the Mamasa area are not related to the devastating Palu earthquake. Figure 9 indicates that the Mamasa swarm earthquakes occurred approximately 30 days after the larger magnitude (M > 3) aftershocks had stopped. Furthermore, it seems very unlikely that a direct dynamic triggering would respond from such a large distance (~ 230 km) (O’Malley et al. 2018) and the timing is beyond the timescale of the dynamic stress transfer. The evidence of a large earthquake triggering other earthquake sequences only occurs at a magnitude of M > 8 and is very rare (Johnson et al. 2015). However, the static stress triggering may have contributed to the stress accumulation at the Mamasa earthquake sequence. Therefore, a more rigorous study of static stress change, incorporating the area of the Mamasa earthquake sequence, needs to be performed.

Conclusions

We have conducted hypocenter relocations of the aftershocks of the Mw 7.5 earthquake in Palu since the September 28, 2018 event. Our results show that the aftershocks were located to the east of the Palu-Koro Fault Line, and these results are consistent with the deformation data of the area. The relocated swarm earthquakes in Mamasa most likely correspond to the activity of the local fault (dipping at a ~ 45° angle) and extend from north to south for a length of ~ 50 km.