Assessing the threat to Western Australia from tsunami generated by earthquakes along the Sunda Arc
- First Online:
- Cite this article as:
- Burbidge, D. & Cummins, P. Nat Hazards (2007) 43: 319. doi:10.1007/s11069-007-9116-3
- 167 Views
A suite of tsunami spaced evenly along the subduction zone to the south of Indonesia (the Sunda Arc) were numerically modelled in order to make a preliminary estimate of the level of threat faced by Western Australia from tsunami generated along the Arc. Offshore wave heights from these tsunami were predicted to be significantly higher along the northern part of the west Australian coast than for the rest of the coast south of the town of Exmouth. In particular, the area around Exmouth may face a higher tsunami hazard than other areas of the West Australian coast nearby. Large earthquakes offshore of Java and Sumbawa are likely to be a greater hazard to WA than those offshore of Sumatra. Our numerical models indicate that a magnitude 9 or above earthquake along the eastern part of the Sunda Arc has the potential to significantly impact a large part of the West Australian coastline.
KeywordsTsunami deterministic hazard assessmentWestern AustraliaSunda Arc subduction zoneTsunami numerical propagation modelSubduction zone earthquakes
The occurrence of the Indian Ocean Tsunami on 26 December, 2004 has raised concern amongst Australian emergency management authorities about the lack of information on the tsunami threat to Australia, which makes it difficult for them to determine appropriate mitigation measures. This paper attempts to redress this by considering the possible effect on the western coastline of Australia from a tsunami generated by a large subduction thrust earthquake along the Sunda Arc (south of Indonesia). This is the most likely source of large tsunami that could affect West Australia (WA). Volcanoes, landslides and asteroid impacts are also possible sources of tsunami that could affect WA, but historically they have been less frequent than the tsunami generated by subduction zone earthquakes and are not considered further in this paper. The focus of this paper is on determining the maximum “credible” offshore tsunami wave height along the WA coast from tsunami generated along the Sunda Arc. This is assessed by finding the maximum positive tsunami height along the WA coast from any tsunami generated by a magnitude 8.5 or 9.0 earthquake along the Sunda Arc. A probabilistic study considering all ranges of earthquakes (such as described by Geist and Parsons 2006) is not covered by this paper. The maximum positive tsunami height at the coast produced by this study will show which locations face the greatest threat from a large tsunami, and therefore which regions should be of most concern to emergency managers and providers of tsunami warnings.
Estimate the maximum magnitude of a “credible” tsunamigenic earthquake over the period of consideration in all the source areas that can effect WA;
Estimate the physical properties of the maximum earthquake (e.g. length, width, rake and mean slip);
Calculate the sea-floor deformation from all the earthquakes to be considered in the study; and
Calculate the maximum positive wave height of any of the resulting tsunami off the coast.
Each of these steps will now be covered in turn.
2.1 Maximum magnitude of earthquakes along the Sunda Arc
The geological or palaeo-tsunami record,
The historical record predating the widespread use of instrumentation and
The modern (∼40–50 year) record of seismicity when quantitative data is available.
Further to the east along the Sunda Arc earthquakes as large as about magnitude 8.5 have occurred in the outer rise seaward of the subduction zone (e.g. the 1977 Sumbawa earthquake). While the faulting mechanism of these earthquakes is different from that of the interplate thrust earthquakes usually associated with tsunami, it still involves vertical movement of the sea floor and can be an efficient generator of tsunami (e.g. Satake et al. 1992). In 1994 a MW 7.8 thrust earthquake occurred off Java. The resulting tsunami killed hundreds of people in Java, and also created significant waves across the northwest coast of Australia. The largest waves washed away structures up to 6 m above the high water mark and stranded fish hundreds of metres inland (Gregson and van Reeken 1998). Fortunately no-one was hurt, due partly to the fact that the tsunami arrived in the early hours of the morning. Therefore, we believe that earthquakes up to magnitude 8.5 are entirely plausible along the whole Sunda Arc.
While there is now no question that earthquakes of magnitude 9 and greater occur in the western Sunda Arc off Sumatra, whether the maximum magnitude of earthquakes occurring in the eastern Sunda Arc is as high as 9.0 is unknown. On the one hand, there are arguments that, because the age of the Australian plate being pushed beneath Java is relatively old, the cooler temperature of the interplate contact will lead to a narrow seismogenic zone, hence smaller earthquakes and a smaller maximum magnitude (Ruff and Kanamori 1980; Hyndman and Wang 1993; Oleskevich et al. 1999). This argument is supported by the fact that very few large thrust earthquakes have occurred in the subduction zone off Java since records were kept during the Dutch colonial period (Newcomb and McCann 1987), and by the apparent lack of a wide zone of interplate coupling suggested by geodetic measurements (Bock et al. 2003). On the other hand, other studies suggest that there is little dependence of subduction zone seismicity on plate age (Bird and Kagan 2004; Nishenko 1991; Pacheco et al. 1993), in which case there is no basis for inferring that magnitude 9 earthquakes cannot occur off Java.
In summary, we believe that the maximum magnitude of earthquakes that can occur anywhere off the Sunda Arc is at least 8.5, but whether it might be as high as 9.0 for the eastern end of the Arc is less clear. While the difference between 8.5 and 9.0 may seem small, we will show that this an important effect on the level of tsunami hazard for Western Australia. It should be noted that we would also expect magnitude 7–8 earthquakes to also occur with a greater frequency than the large events covered in this study. However, judging from the historic events, they are likely to pose a comparatively low and highly localised hazard to Australia (i.e. isolated areas may experience run-ups of a few metres and there may be unusually strong currents in specific areas similar to the effects of the 1994 and 2006 Java tsunami). However any earthquake greater than approximate magnitude 7.5 is likely to be very damaging to the southern coasts of the islands of Indonesia.
The parameters used in the numerical model
Depth to top edge of rupture
Seismogenic width of rupture
Time step size
Total number of time steps
Total length of rupture
∼200 km (Suite I)–1000 km (Suite II)
Parallel to plate motion from Bird (2003) or pure thrust
Strike of sub-faults
Variable along fault—values taken from the digitisation of the Sunda Arc by Bird (2003)
Length of sub-faults
Variable along fault—values taken from the digitisation of the Sunda Arc by Bird (2003)
Variable—scaled to a particular MW according to area
2.2 Fault geometry
In order to determine the height of the tsunami we need to know the amount of vertical movement the earthquake imparts to the sea floor. This depends on the amount of fault slip during the event and the dip and geometry of the fault. The dip of a subduction zone typically varies from about 8° at the trench to about 30° below the arc (Bird and Kagan 2004). However, not all of this interface will be seismogenic. In this study, we assume that the entire subduction zone has a down-dip width of 150 km (from Pacheco et al. 1993) and the rupture starts 5 km below the surface. The mean lower depth for transition from unstable to stable sliding along the Sumatra section of the subduction zone is probably 50 km (based on the distribution of earthquakes), while near Java it could be as high as 70 km (Pacheco et al. 1993). In this study we assume that the dip of the seismogenic part of the subduction zone is 19.5° (i.e. the arcsin of 50/150). Note that there is considerable uncertainty in many of these parameters. A good case could be made for any dip between 10° and 25°, and similarly the down-dip width could be anywhere between 50 km and over 200 km.
Each earthquake rupture is made up from several sub-faults of varying strike and length. The strike and length of each sub-fault was taken from the digitised plate boundary model PB2002 (Bird 2003). The total length of each earthquake varied slightly from event to event because not all the sub-faults had the same length. The number of sub-faults was chosen to sum to approximately 200 km in total length for the magnitude 8.5 events and 1,000 km in total length for the 9.0 events.
2.3 Fault slip
The scalar moment for an earthquake, MO, is given by
Therefore, for an earthquake of a given area and magnitude, we need an estimate of the shear modulus in order to calculate the mean slip, u. For crustal rocks, the shear modulus is about 30 GPa; however it is much higher for mantle rocks. Subduction zone faults like the one along the Sunda Arc probably consist of both types of rocks. In this study we assume that the shear modulus for all the Sunda Arc faults is 49 GPa (an approximate mean of crustal and mantle rocks used by Bird and Kagan 2004).
For both Suites, the area and shear modulus chosen above implies that the total slip along each segment was in the range of 4–6 m for any earthquake, depending on the exact area of that particular earthquake. This slip is close to the mean slip for the magnitude 9.0 2004 Aceh-Andaman earthquake (5.0 m, see Lay et al. 2005). However, one could argue that mean slip may, in general, be greater than 5.0 m for the Suite II events and that the Aceh-Andaman earthquake had exceptionally low mean slip for such a large event. If one uses the higher magnitude for the 2004 event based on normal modes (MW = 9.3), then the mean slip for the 2004 earthquake would be around 11 m (Stein and Okal 2005). However it is not clear if this slip was all “fast” and thus tsunamigenic. For the models shown in this paper, the parameters described above were chosen so that mean slip was close to 5 m. Higher mean slip events would have shorter length ruptures, which generate more uplift for the same magnitude than the earthquakes considered here.
Given the rate of convergence across the subduction zone, 4–6 m of slip implies the recurrence time for each individual MW 8.5 earthquake on a given section of fault is about 100–200 years. Since a MW 9.0 event would require a large part of the subduction zone to be close to failure simultaneously, it probably would have a recurrence time greater than 200 years on a given section of fault. Exact recurrence times for these events are difficult to estimate due to the short seismic record and lack of palaeotsunami data (particularly for the eastern section of the Sunda Arc).
The direction of slip (i.e. the rake) was assumed to be parallel to the relative plate motion for all the events in Suites I and the three events in Suite II shown in Fig. 3. The relative motion is virtually perpendicular to the plate boundary for the Java events and becomes more oblique for events off Sumatra. The further north the earthquake is along the Sumatran section of the subduction zone, the greater the component of fault parallel motion. However it is believed that a large component of this strike-slip motion along the Sumatran section of the subduction zone is taken up by the Great Sumatran fault along the centre of the island of Sumatra (Beck 1983) since the angle of convergence for much of this section of Sunda Arc is favourable for strain partitioning (Burbidge and Braun 1998). Since the amount of strike-slip motion has a large effect on the amount of vertical uplift (and consequently the size of the tsunami), an additional MW 9.0 event off Sumatra was included in Suite II which did not have a strike-slip component to the slip direction (i.e. it was a pure thrust event). This event otherwise had the same source geometry as the other 9.0 event off the coast of Sumatra.
2.4 Modelling the tsunami
Once the amount of slip for each segment is determined, the amount of vertical deformation of the sea floor was calculated using a programme developed as part of the Method of Splitting Tsunami (MOST) tsunami modelling software (Titov and Gonzalez 1997). This programme calculates the vertical component of the sea-floor deformation using linear elastic dislocation theory (Gusiakov 1972). This sea-floor deformation is assumed to instantaneously cause the sea surface to change shape to match the shape of the sea-floor deformation. This wave is then propagated across the ocean using the MOST numerical model, which uses the finite difference method to calculate the water height and velocity up to the 10 m water depth contour. We do not model the inundation of the coast above the 10 m contour since this requires a more complicated model which includes properties such as bottom friction and it requires a detailed knowledge of the local bathymetry and topography. At the end of this process we have a suite of tsunami generated by MW 8.5 and MW 9.0 events spaced evenly along the Sunda Arc.
3.1 Suite I: magnitude 8.5 earthquakes along the Sunda Arc
Figure 2b shows a magnitude 8.5 event off Java (close to the location of the 1994 event but much larger). This has a much larger effect on Australia than the tsunami shown in Fig. 2a and the effect is more localised to the region between Exmouth and Port Hedland. Smaller waves are predicted along the rest of the north-west shelf coastal areas. This result is very similar to the observed distribution of the impact for the much smaller 1994 Java tsunami (Fig. 1).
Figure 2c shows an MW 8.5 event near the eastern end of the subduction zone (close to the location of the 1977 Sumbawa event see Fig. 1). It has a much larger effect to the northeast of Broome than the tsunami shown in Fig. 2a and b. Most of the northwest coast could experience some tsunami similar to the one in 1977 (Gregson et al. 1979). Neither of the modelled tsunami shown in Fig. 2b and c are likely to pose a significant threat to the Fremantle area.
3.2 Suite II: magnitude 9.0 earthquakes along the Sunda Arc
The maximum positive offshore heights along the coast northeast of Exmouth for Suite II (illustrated in Fig. 3) reflect the tsunami generated by the event to the east of Java (Fig. 3c). While these offshore heights are typically only 1–2 m, they may run up to much larger onshore heights. The maximum wave heights around Exmouth and further south are controlled by the event offshore Java (Fig. 3b). We note that the heights along the northwest coast for Suite I are closer to what has been experienced historically (e.g. the 2–4 m onshore heights observed during the 1977 event, Gregson et al. 1979). The offshore heights in Fig. 4b are 2–3 times those of Fig. 4a, which suggests that the onshore heights of the Suite II events may be higher than any of the known tsunami that has affected WA. From the tsunami waveforms that were examined, the minimum wave height (i.e. the maximum drawdown) was typically about the same as the maximum wave height shown in the figures.
The event illustrated in Fig. 3a is very similar to what might be expected during a repeat of the Great 1833 Sumatra Earthquake. The probability that this segment may rupture within the next few decades is high, since we know that it has been building up strain energy since 1833, and as discussed, the recurrence interval is estimated to be close to 200 years. While recent work on coral microatolls (Zachariasen et al. 1999) suggests a somewhat larger recurrence interval (250 years), the accumulation of strain energy building up to the next 1833-like rupture may have been increased due to the occurrence of the 28 March 2005 earthquake in the subduction zone immediately to the northwest of the 1833 rupture area (McCloskey et al. 2005). In any case, it seems likely that the next major Indian Ocean tsunami will have a character similar to what is illustrated in Fig. 3a. Our calculations suggest that the height of such a tsunami along the WA coast will be very similar to that of the 26 December 2004 tsunami (which produced no deaths or significant damage to Western Australia).
The magnitude 9.0 events off Java and Sumbawa shown in Fig. 3b and c would have the highest impact on Australia of any events simulated in this study. As stated earlier, it is very difficult to assess the likelihood of such events occurring—none has occurred historically. But if they did occur, it is likely to be of major concern to the whole west coast of WA. In particular, the magnitude 9.0 earthquake off Java illustrated in Fig. 3b would have a considerable impact along the western coast from Exmouth south. When these computed offshore wave heights are compared with those for the Aceh-Andaman 2004 tsunami (from Dominey-Howes et al. 2006) they are 1–4 times larger; suggesting onshore run-up heights for this event would be higher than that for the Aceh-Andaman 2004 tsunami for points along the WA coast.
While the tsunami generated by a magnitude 9.0 earthquake off Sumbawa (Fig. 3c) does not have a pronounced effect on the coast south of Exmouth, it would have a substantial effect on the whole of the north-west shelf area to the northeast of Exmouth. The situation would be analogous to that of Thailand or Sri Lanka during the 26 December 2004 tsunami. However, unlike those areas, this area of Australia is much more sparsely populated and the resulting impacts on infrastructure and human life would therefore be less.
The first wave for the tsunami from the magnitude 9.0 earthquake off Sumbawa reached Port Hedland about 3 and half hours after the earthquake and the maximum wave for Port Hedland arrived about an hour later. The tsunami was slowed down significantly by the extended continental shelf off Port Hedland. At Exmouth the first wave and the largest wave arrived about an hour earlier than they did at Port Hedland. The tsunami arrived half an hour later at Geraldton than it did at Exmouth and was significantly smaller. In all cases the first wave would have caused a rise in sea level, with no draw down before the first wave. The second wave was usually much larger than the first. The tsunami generated by a similar sized event off Java were similar (i.e. the first wave arrived in Exmouth before Port Hedland, and the later waves were higher than the first wave). Therefore emergency services have about 2 h to detect and measure the event and promulgate the warning to the communities along the northwest coast before the tsunami starts to reach parts of the WA coast.
The maximum positive wave heights are at their highest near the Exmouth region, but are also moderately high all along the north-west coast of WA. The WA coast south of Carnavon faces a much lower threat from tsunami generated by earthquakes off the coast of Indonesia.
The modelling suggests that the tsunami from a magnitude 9 earthquake along the Java section of the Sunda Arc reaches the area offshore of Exmouth before it arrives offshore of the communities further to the north. This is due to the broad continental shelf slowing any wave down which comes from Java. Any event from this location of the arc is therefore likely to affect the Exmouth region before the other communities along the coast.
In the event of a ‘repeat’ of the 1833 event (i.e., a magnitude 9 earthquake off central Sumatra), which has a high likelihood of occurrence within the next 20–50 years, the tsunami impact on the west Australian coast is likely to be very similar to that of the 26 December, 2004 event.
Magnitude 8.5 events off Java could produce tsunami with onshore heights of several metres at specific restricted locations of the Australian coast, similar to what has been observed historically from the 1994 Java and 1977 Sumbawa tsunami. More detailed inundation modelling will be needed to quantify this effect at particularly vulnerable sites (i.e., those with relatively high offshore wave heights and significant population and/or infrastructure).
Magnitude 9.0 events from the eastern section of the Sunda Arc, which have never occurred historically and whose probability of occurrence is unknown, could have considerable impact all along the western coast of Australia. Maximum positive offshore heights for these events are 2–3 times the offshore heights calculated for historical events and therefore onshore heights may be higher than anything that has happened historically. Earthquakes of this size also have the potential to affect a large part of the coast within a few hours of an earthquake.
One general point to note from this study is the greatly increased far-field effect for the magnitude 9.0 events when compared to the 8.5 events. This observation is very likely to be true for tsunami generated along other subduction zones in other parts of the world. This implies that constraining the likelihood of the very large (above magnitude 8.5) events is going to be crucial for constraining the tsunami probabilistic hazard for far-field locations. Unfortunately, the length of historic and instrumental record is often too short and/or too incomplete to confidently constrain the recurrence of these very rare, but very important, events. The current best hope of constraining the recurrence of these events is from detailed palaeo-tsunami studies in order to constrain whether these events occur for a particular subduction zone and how often they might occur. In the case of Western Australia, the Java to Sumbawa section of the Sunda Arc is clearly one such area which requires such a study.
It should also be noted that this study only models the deep-water (greater than 50 m) wave height. The actual level of inundation at the coast would be strongly influenced by the local bathymetry and topography. For example, the presence of bays could act to shield areas from a tsunami or could act to resonate it, depending on the direction of the tsunami and its period. The other point to note is that this study does not include other possible tsunami sources such as those created by large intra-plate earthquakes, volcanoes, landslide or asteroids, all of which would be less likely to occur than earthquakes along the Sunda Arc, but they could still potentially have a large effect on Australia depending on their location and size.
The authors would like to thank the Fire and Emergency Services Authority (FESA) of Western Australia for its financial support of this project. We would also like to thank Vasily Titov for providing the source code for the tsunami numerical model. The authors would also like to acknowledge the contribution of the two anonymous reviewers, Mark Leonard, Clive Collins and the editor for improving this manuscript.