GNSS VELOCITY AND STRAIN FIELD IN THE NORTHERN SUMATRA 15 YEARS AFTER THE 2004 M9.2 SUMATRA ANDAMAN EARTHQUAKE

1. INTRODUCTION

Background. The study of past earthquakes is important for earthquake potential study in the future, and considering the postseismic deformation was the big issue. The postseismic deformations of the 2004 M9.2 Sumatra-Andaman or other earthquakes would overweigh the earthquake potential calculation. The earthquake, which occurred on 26 December 2004 00:58:53 UTC at 3.295° N, 95.982° E at a depth of 30.0 km according to the United States Geological Survey (USGS), produced coseismic slip of 16–25 m off the northern Sumatra [Fujii et al., 2021]. The coseismic slip model and the rheology (viscosity of asthenosphere) are the key estimate for the postseismic deformation of the earthquake [Lubis et al., 2013]. The question asises, however, what earthquake mechanism contributes most to the postseismic effects [Gunawan et al., 2014]. The studies using the data for more than a 10-year period after the earthquake could be a tool for the existing model validation and studying whether the postseismic deformation still takes place in the affected area [Mendoza et al., 2022].

Geological Settings. The northern Sumatra is one of the most tectonically active regions in the world, especially after the 2004 M9.2 Sumatra-Andaman earthquake. It is due to the oblique subduction of the Indo-Australia plate beneath the Sundaland plate [McCaffrey, 2009] of the western part of the Sumatra Island with the rate of plate convergence 44–48 mm/yr [DeMets et al., 2010]. The Sumatran fault zone that extends from southern to northern Sumatra [Natawidjaja, 2018] and exhibits a slip rate of 20 mm/yr [Ito et al., 2012] in the region makes up to 40 % of the trench-parallel component of the subduction (Fig. 1, a). The Northern Sumatra, which covers the Aceh Province and the North Sumatra Province in Indonesia, is the most affected region of the disastrous 2004 M9.2 Sumatra-Andaman earthquake and has been the subject of numerous studies, especially coseismic and postseismic modeling [Chlieh et al., 2007; Panet et al., 2010]. Unlike Alaska and Chile that also had produced M>9 earthquakes, the northern Sumatra was in a thermal state which led to a higher potential for other large earthquakes [Simoes et al., 2004].

The 2004 M9.2 Sumatra-Andaman earthquake in the northern Sumatra was followed by a large number of earthquakes in the Sumatra region, the most nearby and the largest of which is the 2005 M8.6 Nias earthquake that occurred in Nias, also in the northern Sumatra. The slip distribution of this earthquake is separated from that of the 2004 earthquake by a persistent barrier beneath the Simeulue Island [Morgan et al., 2017]. Other earthquakes (M>7) in Sumatra after the main 2004 events are the 2007 M8.5 Bengkulu earthquake, the 2008 M7.4 Simeulue earthquake, the 2009 M7.6 Padang earthquake, the 2010 M7.8 Mentawai earthquake, and the 2010 M7.8 Simeulue earthquake. The 2012 M8.6 earthquake which occurred west of the Sunda Trench, the boundary between the Indo-Australian plate and the Sundaland plate, was also affecting the northern and even the southern Sumatra [Alif et al., 2016]. These earthquakes were studied somewhat better than the 2004 earthquake, since the need emerged for more Global Navigation Satellite System (GNSS) monitoring sites after the 2004 earthquake.

The GNSS-based velocity monitoring tools become indispensable in the study of postseismic [Anugrah et al., 2015] or crustal deformations [Alif et al., 2021, 2023] and volcanoes [Anggara et al., 2023; Natadikara et al., 2023]. The study which only utilizes the GNSS velocities could be a preliminary indicator whether the postseismic deformation still takes place. The GNSS velocities relative to the plate (Sundaland plate in case of the northern Sumatra), should be directed toward the epicenter or toward the trench if the postseismic deformation still takes place. No study on GNSS velocities in northern Sumatra has been conducted to clarify the existence of the postseismic deformation. A similar approach has been applied to the 2007 M8.5 Bengkulu earthquake [Lubis, 2021] which showed no postseismic deformations during the period 2018–2020. This study will utilize the GNSS velocities and strain field 15 years after the 2004 M9.2 Sumatra-Andaman earthquake to obtain a preliminary indicator whether the postseismic deformation still takes place in the northern Sumatra.

2. METHODS AND MATERIALS

Materials. The data used in this research are the GNSS data from 2017 to 2021 and the GNSS-velocities before 2004. These velocities are calculated from the campaign GNSS data collected at 34 sites scattered throughout Sumatra [Prawirodirdjo et al., 2010] before the 2004 M9.2 Sumatra Andaman earthquake. Meanwhile, the GNSS data recently collected at 28 permanent sites (Table 1) were provided by the Geospatial Agency of Indonesia (BIG) and the Earth Observatory of Singapore (EOS), named as Indonesia Continuous Operating Reference Stations (InaCORS) and Sumatran GPS Array (SuGAr) [McLoughlin et al., 2011], respectively. 7 SuGAr sites and 21 InaCORS sites, used in this research, are located in Aceh Province and North Sumatra Province, in mainland of Sumatra and surrounding small islands (Fig. 1, b). These sites were installed a relatively long time ago as compared to other InaCORS sites in Indonesia and SuGAr sites in Sumatra, and have been used for other crustal deformation studies [Maulida et al., 2016].

Methods. The new GNSS data were processed to obtain recent velocities in North Sumatra. The data which had the sampling interval of 30 seconds were processed using Bernese 5.2 [Dach et al., 2015] in a consistent reference frame. The daily solution of coordinates was obtained based on the double-difference positioning with the International GNSS Service (IGS) sites used as constrained sites (ALIC, DARW, DGAR, IISC, KARR, PIMO, YAR2) [Johnston et al., 2017]. The data from all sites were processed in ITRF2014 [Altamimi et al., 2016]. The IGS final ephemeris, the Earth rotation parameters, the IERS Conventions 2010, and the DRY-GMF model were utilized in the GNSS data processing to obtain more precise coordinates.

The recent velocities were calculated using linear regression of the processed coordinates. This linear regression can be conducted since the logarithmic trend or exponential trendlines were not found in the coordinate time series. Nevertheless, considering coordinate jumps induced by antenna replacement or some unknown factors, the stepwise linear regression [Feng et al., 2015] was applied to each clear offset. The least-square linear regression was performed after removing outliers with larger than 95 % confidence in the coordinate time series. In order to obtain recent velocities and velocities before 2004 in the same reference frame, the recent velocities were transformed from ITRF2014 position to the Sundaland plate reference frame. Prior to this transformation, all velocities were first transformed to ITRF2000 [Altamimi et al., 2016], since the velocities before 2004 were measured with respect to the ITRF2000-based Sundaland plate reference frame. The transformation to the Sundaland plate reference frame was calculated using Euler’s formula [Stein, Wysession, 2009] and the Euler pole parameter from the ITRF2000-based Sundaland plate model [Simons et al., 2007].

The postseismic deformations from the 2004 earthquake were analysed by doing a comparison between recent velocities and velocities before 2004 in the similar Sundaland Plate reference frame. The comparison was made in three ways. The first way is to perform a direct analysis of the velocity vector in these two time periods. The second way is to calculate strain in the area for these two time periods. The strain is calculated in each grid cell at a spacing of 25 km. The strain analysis was conducted to understand extension and shortening behaviour in this area before and 15 years after the 2004 earthquake. The third way is to perform the velocity subtraction between these two periods. The two steps necessarily conducted prior to the subtraction were as follows: interpolating inner-region velocities in each time period on a grid with spacing of 0.25° and finding all points between two given points on a grid. In short, the final comparison has been drawn between the interpolated velocities for common grid points.

3. RESULTS AND ANALYSIS

The recent velocities range from 7.6 to 45.6 mm/yr within the ITRF2014 (Table 2). These velocities are horizontal velocities calculated from linear regression for horizontal coordinate time series of 28 GNSS sites (Fig. 2). There are three types of spatial patterns of these velocities (Fig. 3). The first type is the northeastward velocity pattern obtained from a relatively high velocity in small islands. This pattern follows the direction of subduction of the Indo-Australian plate. The second type refers to the southeastward velocities of the southern GNSS sites of the Sumatra Island. This pattern follows the direction of motion of the Sundaland plate with the velocity comparable to that of 15.3–31.4 mm/yr, reported by [Hanifa et al., 2014]. The third type follows the southward direction of motion of the GNSS sites in the northernmost area. This pattern appears to be highly related to the effect of postseismic deformation since it does not follow the direction of either subducting plate or plate motion.

There is also a distinguishable pattern of velocity with reference to the Sundaland plate and its derived grid velocities. This pattern direction is changed for the velocities which refer to the Sundaland plate (Fig. 4). The first-type pattern for the small islands changes its direction to northward, the second-type pattern changes its direction to northwestward, and the third-type pattern changes its direction to westward. The velocities of 5.5 to 40.5 mm/yr refer to the Sundaland plate. The northernmost area is dominated by high velocities. The recent pattern for this area is also much different from the pattern of velocity prior to 2004 in the same area (Fig. 5). In general, the velocities prior to 2004 had smaller values and were more uniform as compared to the recent velocities.

The strains calculated from the velocities show a different shortening-extension pattern for the recent velocities and the velocities prior to 2004. These strains are calculated from the interpolated grid points. The statistics of the velocities on the grid points shows that the recent velocities are lower than the velocities prior to 2004 (Table 3). Therefore, the recent strain is lower than the strain prior to 2004. However, the strain field pattern for these two periods is different. In general, the dominant pattern is shortening due to convergence of the Indo-Australian Plate and the Sundaland plate [Alif, Pratama, 2019]. High strain rates prior to 2004 are concentrated on the southern area between small islands and the Sumatra Island (Fig. 6). Meanwhile, the amount of shortening in that area is lower for the recent strain (up to 47.2 μstrain/yr as compared to up to 33.8 μstrain/yr). Furthermore, the recent strain rates are associated with the extension in the northern Sumatara Island (Fig. 7). This is possibly caused by the postseismic deformation therein.

4. DISCUSSION

The postseismic deformation in the northern Sumatra is associated with velocities and strain field. The first, direct analysis approach for the velocity vectors shows that the vector in the northern area has been recently directed towards the west and was directed towards the north prior to 2004. The second, strain comparison approach shows the extension in the northern area while the dominant shortening direction indicates no postseismic deformation. The third approach which consists in subtracting velocities between these two periods shows the trenchward orientation of the vector (Fig. 8). This trenchward motion at a rate of up to 26.3 mm/yr indicates postseismic deformation [Wang et al., 2012].

5. CONCLUSIONS

The postseismic deformation of the 2004 M9.2 Sumatra-Andaman earthquake still occurs in the northern Sumatra 15 years thereafter, as shown by the study on the GNSS-based velocities and strain field determinations. The trenchward velocity vector direction and the extensional strain field were the herein-determined indicators of the postseismic deformation. The difference between GNSS velocities from 2017 to 2021 and the velocities prior to 2004 shows different deformation pattern and behavior of the crust in the northern Sumatra. These postseismic deformations should be considered when studying earthquake potential in this area. These results can help in evaluating the validity of the existing postseismic model, or even in generating a new postseismic model.

6. ACKNOWLEDGMENTS

Figures were drawn using the Generic Mapping Tools (GMT) software [Wessel et al., 2013]. Thanks are given to the Geospatial Information Agency of Indonesia (BIG) for providing us with continuous GPS data.

7. CONTRIBUTION OF THE AUTHORS

All authors made an equivalent contribution to this article, read and approved the final manuscript.

8. DISCLOSURE

The authors declare that they have no conflicts of interest relevant to this manuscript.