Relocation of Earthquake Hypocenter Using Double-Difference Method in Western Part of Sumatera

The western part of Sumatra has a fairly complex tectonic setting where there is a convergent meeting between the Indo-Australian plate and the Eurasian plate. These plates form a subduction zone and also the Sumatran Fault. The subduction zone and the Sumatran Fault cause many earthquakes in western Sumatra. This study will identify the location of the earthquake hypocenter, which is quite accurately obtained by relocating the hypocenter and making a cross-section in order to analyze the subduction zone and the Sumatran Fault. Hypocenter relocation can be done using the double difference (DD) method using the hypoDD program. This study used data on the arrival time of P and S waves from BMKG in the range of January 2016 to December 2020. The results of the relocation indicate that contains several earthquakes in groups with an average depth of fewer than 60 km that is close to the main source of the earthquake, namely the subduction zone and the Sumatran Fault, so it can be said that this area has a high level of seismicity.


Introduction
The western part of Sumatra is one of the areas in Indonesia that is prone to earthquakes due to its tectonic location close to the subduction zone.The Indo-Australian Plate subduction into the Eurasian Plate continuously can cause the formation of a shallow subduction zone or megathrust.West Sumatra also has one of the most active faults in the world, namely the Sumatran Fault, which extends from the tip of Aceh to the Sunda Strait, so it can cause land earthquakes.In addition, Natawidjadja and Sieh (2000) stated that the Sumatran Fault is divided into 19 fault segments.The mainland area of western Sumatra is traversed by the Sumatran fault system that consists of four active fault segments, i.e. the Sumpur fault segment, the Sianok fault segment, the Sumani fault segment, and the Difficult fault segment, which stretches from South Solok to the Rao area, Pasaman Regency [1].The following is a history of earthquakes in the western part of Sumatra that was destructive near the subduction zone: in 1779 (M 8.4), 1883 (M 9.2), 1861 (M 8.3), 2004 (M 9.2), 2007 (M 7.9 and M 8.4), and 2009 (M 7.6).The Sumatran Fault also has a history of seismicity in the western part of Sumatra, including the Sumani Segment in 1943 (M 7.4) and 2004 (M 5.6), in the Sump Segment in 1977 (M 5.5), and in the Sianok Segment in 2007 (M 6.4) [2].
The earthquake hypocenter is one of the earthquake parameters that can be used as a consideration in disaster mitigation efforts.The accuracy of the earthquake hypocenter position is one of the most important things for accurate seismic investigations, which can reconstruct seismogenic structures [3].If the fault plane or subsurface geological structure can be described by the appearance of the earthquake hypocenter, it can help identify earthquake-prone areas.Earthquake hypocenter relocation is a method used to correct or recalculate the hypocenter position in order to provide a more accurate position description [4].
The determination of the common hypocenter is performed using the single hypocenter determination method or SED, wherein the use of this method, there are several velocity structures that are not modelled so that the resulting hypocenter still contains errors.The quality of the velocity model also affects the accuracy of the earthquake hypocenter.The more suitable an earthquake wave velocity model in an area, the accuracy of the hypocenter will be better [5].The velocity model which is generally used in determining the hypocenter is the global velocity model, the model eliminates the differences in the structure of the subsurface layer.Therefore, it is prominent to relocate the earthquake hypocenter.In this study, the method used in the relocation of the earthquake hypocenter was the DD method.The DD method has been widely used to relocate earthquake hypocenters, as has been done by Waldhauser and Ellsworth in 2000 to relocate the distribution of hypocenters in the northern Heyward fault, California, United States.The DD method was also used [6] to relocate the distribution of the hypocenter on the 30 September 2009 earthquake in Padang.In this study, the DD method was used to relocate earthquakes in the West Sumatra region to obtain an overview of the faults below the surface in disaster mitigation efforts.

Data
Earthquake catalogue data used are earthquake events in the western part of Sumatra with an area limit of 2°00'00" North Latitude to 4°00'00" South Latitude and 93°00'00" East Longitude to 103°00'00" East Longitude for 4 years.The data used as initial input for relocation consisted of 3497 earthquake events during the time period from January 1, 2016, to January 1, 2020.The method used by BMKG in determining the hypocenter was the SED (single event determination) method using a global velocity model, namely the IASP91 velocity model [7].In this study, we modified velocity models as shown in Table 1.The distribution of the earthquake epicentre data used is shown in Fig. 1.Table 1.Reference table for the 1D wave velocity model for the western part of Sumatra (modification of [8] and [9]).

Method.
The DD method relies on the idea that if the distance between two paired earthquakes is significantly smaller than the distance between each earthquake and a station, their ray path and waveform can be seen as nearly identical.The method assumes that the difference in travel time between these two earthquakes, as recorded at the same station, can be attributed to the distance between their hypocenters without any need for station correction [10].Based on the following assumptions, the model error can be minimized [10] with the following equation: residual travel time    from two earthquakes i and earthquake j at observation station k which is calculated based on the travel time difference of observations and calculations for the two earthquakes.   is the travel time for earthquake i to the observation station k, while    is the travel time for earthquake j to the observation station k.

Result and Discussion
The relocation of the hypocenter using the DD method was carried out using the BMKG earthquake catalogue data for the period 2016-2020, which consisted of 3497 events recorded at 103 stations, as shown in Figure 1.The dark blue inverted triangle represents the recording station point, and the black line represents the Sumatran Fault.The data that were successfully relocated using the hypoDD program were 3368 events, with 129 events not being relocated or, in other words, around 96% of the data that could be used for the relocation process.The existence of events that were not relocated was due to a pair of earthquake events that did not meet the criteria in accordance with the input parameters given to the ph2dt and hypoDD programs.In addition, other causes of events that were not relocated were the presence of several earthquake pairs that had weak attachments between one event and another, and during the iteration process, the position of the earthquake hypocenter shifted to above the surface (air quake).Quantitatively, the results of the relocation using the DD method show good results, where validation can be done by making a histogram of the Root Mean Square (RMS) residual travel time from the data before and after relocation (Figure 2).The value of the RMS histogram of the residual travel time after the relocation is tight to zero when compared with before the relocation.When the RMS residual time value is close to zero, then the calculation results using the DD method are precise.The residual value of travel time that is close to zero represents that the projected travel time value is close to the value of the observed travel time, so it can indicate that the model or structure of the inverted seismic wave velocity used to compute the calculated travel time is close to the Earth's actual conditions [11].Thus, it can be said that the relocation using the DD method with its grouping technique is able to reduce the effects due to speed variations that are not modelled, so it can be seen that the velocity model used is good enough and can be used in determining the position of the hypocenter to be more accurate.But basically, the residual value of the travel time cannot be used as the main reference in the accuracy of the hypocenter position.
Based on Figure 3, it can be shown that the results of the relocation using the DD method provide a better shift in the position of the hypocenter based on pairs of earthquake events grouped by the ph2dt program.The epicentre position before the relocation has not shown the seismicity of the study area better because it has not formed a trend line that follows the main fault structure.The results after the relocation show that the earthquakes form clusters on a trend line, so it can be assumed that these earthquakes are associated with the main fault structure around the study area.The large shift of earthquakes that do not form clusters is thought to be due to an error when reading the arrival time of the waves.Figure 4 shows the change in the angle vector direction of the earthquake hypocenter.In this diagram, the 0-360 scale shows the shift angle interval after relocation, while the 0-14 scale shows the earthquake number.The change in position of the hypocentre shown by the rose diagram to the southwest and northeast occurs because the earthquake in the research area is in the northern and southern parts of Sumatra, while the recording station is on the mainland of Sumatra.
Based on the data from the distribution of the epicentre distribution before and after the relocation, a vertical section analysis will be carried out in the direction of the subduction zone to see the process of the Indo-Australian plate subduction against the Eurasian plate.A-A' cross-section in the north and a B-B'-section in the south for analysis of the vertical cross-section (Figure 5).Based on the distribution of the epicentre before and after the relocation, it can be seen that there is a change in the position of the epicentre.Changes in the position of the epicentre are seen where the position of the epicentre after relocation tends to gather to form clusters or seismic groups compared to the results before relocation.[12].The red circle shows the earthquake epicentre before the relocation, while the yellow circle shows the earthquake epicentre after the relocation.The dotted line shows the width of the cross-section line.
The western part of the Sumatra subduction zone is the result of the Indo-Australian plate subduction into the Eurasian plate with a subduction angle of about 31.33o with an average relative subduction speed of 6.32 cm/year.This value provides information that the subduction that occurs in western Sumatra is classified as shallow, so it is prone to tectonic earthquakes (Akmam, 2011).In the vertical section, A-A' it is known that the earthquake that occurred had a shallow depth (less than 60 km) to medium (between 60 km to 300 km) with an earthquake depth range of 10-300 km.The depth value of 10 km has been pre-arranged by the BMKG as a fixed depth.The seismic pattern shown by the A-A' cross-section after the relocation shows a more clustered pattern and forms earthquake clusters closer to the subduction line when compared to the results of the vertical cross-section before the relocation.The earthquake was dispersed within the subduction zone, primarily in the slab situated between the Indo-Australian plate contact area and the Eurasian plate.Earthquakes that accumulate tend to be at a depth of 10 km to 60 km and a depth of 100 km to 200 km.Groups of earthquakes at a distance of 200 km to 400 km with a depth of fewer than 60 km are thought to be caused by subduction activity.Then at a distance of 500 km to 600 km with a depth of fewer than 60 km, it is thought to be due to the Sumatran fault and earthquakes at a depth of 100 km to 200 km are thought to be due to slab subduction.
The position of the hypocenter that has been relocated 50 tends to gather to form a seismic group when compared to the position of the hypocenter before relocation.Based on the results of the vertical cross-section of B-B' before and after the relocation shown in Figure 6, the results of the vertical crosssection before the relocation show that the distribution of earthquake is still spreading and does not form earthquake clusters when compared to the results after the relocation.In this section, there are several earthquakes that form clusters at a depth of fewer than 40 km, these earthquakes are shallow earthquakes.Earthquakes are heavily influenced by plate subduction, so earthquakes located in subduction zones will have shallow depths.There were also earthquakes that formed clusters at a depth of 100 km to 210 km, the earthquake was classified as a deep earthquake.This earthquake is in the Sumatra Fault zone, so the earthquake caused by tectonic activity is in the form of a sudden plate shift.[12].The red circle shows the earthquake epicentre before the relocation, while the yellow circle shows the earthquake epicentre after the relocation.The dotted line shows the width of the cross-section line.

Conclusion
The hypocenter relocation of the earthquake using the DD method shows a fairly accurate change in the hypocenter position.This is supported by the histogram results of the residual travel time from the relocation results, which have a value that is close to zero after the relocation.Thus, it can be said that the relocation using the DD method shows that the speed model used is closer to the actual speed model.2. The results show that there are several earthquake clusters that follow the existing geological structure pattern.The distribution of hypocenters after relocation tends to converge in the subduction zone and on the Sumatra Fault, which is dominated by shallow earthquakes (0-60 km).

Figure 1 .
Figure 1.Map of the distribution of the epicentre of the western Sumatra earthquake for the 2016-2020 period before the relocation.The red to blue colours indicate the depth range of the earthquake's epicentre, while the size of the circle indicates the magnitude of the earthquake.

Figure 2 .Figure 3 .
Figure 2. Histogram of the residual value distribution of travel time a) before relocation, b) after relocation

Figure 4 .
Figure 4. Result of rose diagram after relocation.

Figure 5 .
Figure 5.The results of the A-A' cross-section, a) after relocation, b) before relocation, and the red line indicating the subduction slab line are based on the research of[12].The red circle shows the earthquake epicentre before the relocation, while the yellow circle shows the earthquake epicentre after the relocation.The dotted line shows the width of the cross-section line.

Figure 6 .
Figure 6.Results of the B-B' cross-section, a) After relocation, b) Before relocation, the red line indicating the subduction slab line is based on the research of[12].The red circle shows the earthquake epicentre before the relocation, while the yellow circle shows the earthquake epicentre after the relocation.The dotted line shows the width of the cross-section line.