Tsunami Source of the 2010 Mentawai, Indonesia Earthquake Inferred from Tsunami Field Survey and Waveform Modeling

3.1 Asahan and Purorougat

On the Asahan coast, facing the mouth of a ~2-km-wide bay on the west coast of South Pagai Island, a large coral boulder (~1 m diameter) was transported ~70 m beyond a beach ridge of ~3 m altitude. Traces on trees indicated tsunami heights of 6.4–9.3 m, with an average of 7.4 m (Fig. 6a). These were the largest tsunami heights measured by our survey. On the contrary, in Asahan village, located near the head of the bay, no tsunami damage occurred. At Purorougat, facing a smaller (~0.5 km wide), U-shaped bay, the tsunami washed away most houses, and 75 out of 235 residents lost their lives. A trace on a tree indicated a flow depth of ~5.5 m.

3.2 Lopon Lakau

Located in a bay ~2 km wide on the southwest coast of South Pagai Island, Lopon Lakau (a logging port) was damaged by the tsunami. Numerous coral boulders were found on the beach, and a trace on a tree indicated a 2.5-m tsunami height. At Lakau village, an eyewitness recounted that the ground shaking was strong enough for him to be awakened. Having heard an airplane-like loud noise, villagers escaped inland, and all survived.

3.3 Maonai

Located at the end of a U-shaped bay (~0.5 km wide) on the southwest coast of South Pagai Island, this village also suffered significantly from the tsunami. On both sides of the bay, the tsunami deposited numerous large (~2 m diameter) coral boulders. The inundation distance was >470 m. Five measurements of traces on trees indicated that the tsunami heights were 6.7–7.3 m, with an average of 6.9 m (Fig. 5a). The flow depth ranged from 3.5 to 3.9 m. The entire village was washed away (Fig. 6b), and the casualties were 38 out of the original population of 139.

Open image in new window

Open image in new window

3.4 Sabeu Gunggung

Located on the end of a U-shaped bay (~3 km wide) on the southwestern coast of North Pagai Island, this village suffered significantly from the tsunami. We surveyed a profile up to ~300 m from the coast, but could not reach the inundation limit because of the river in the back of the village. The tsunami height was 4.3–7.0 m (average of four measurements was 5.5 m) measured on a surface ~2 m above the sea level (Fig. 5b). The flow depth ranged from 3.9 to 4.8 m. There had been 65 houses, but none remained. Of the original population of 260, 120 people lost their lives (Fig. 6c). According to a survivor, ground shaking was not strong, and he escaped after hearing a loud sound ~5 min after the shaking. Some people were watching TV and saw running text of a tsunami warning, but the tsunami was arriving at that time.

3.5 Muntei Barubaru

Located ~5 km SE of Sabeu Gunggung, this village faces into a small V-shaped bay approximately 1 km wide. The inundation distance was ~420 m, with a run-up height of 4.8 m: the tsunami heights ranged from 4.6 to 5.7 m with one exceptional measurement of 7.8 m (Fig. 5c). The average of the eight measurements was 5.2 m. Many large (~2 m diameter) coral boulders were also transported onto land, and the entire village was washed away (Fig. 6d). Of the original population of 310, 149 lost their lives. On the opposite (western) side of the bay, traces on trees indicate tsunami heights ranging from 3.9 to 8.8 m, with an average of 6.0 m of the four measurements.

3.6 Macaronis Resort

Located on a ~2-km-long peninsula, extending in the SW direction, at the western tip of North Pagai Island, this resort consisted of a main hotel tower (Fig. 6e), several single-story buildings, and beach cottages. The main tower was damaged on the ground floor; however, it remained standing while the second and higher floors provided shelter for approximately 20 guests, all of whom survived. The cottages on the beach were all washed away, and the buildings were also damaged. We surveyed across the peninsula (Fig. 5d) and found that the tsunami arrived from both directions. Clear watermarks on the building indicated that the flow depth was ~1.2 m on the peninsula (~2 m above sea level). Higher tsunami traces ranging from 2.9 to 5.4 m were found on trees, with an average tsunami height of 4.0 m based on eight measurements.

3.7 Tumalei

Located on the northwest coast of North Pagai Island, in a small bay facing north, the main village, with a population of approximately 200, was on a flat area ~100 m wide flanked by a steep hill. The tsunami height ranged from 4.0 to 6.1 m, and the average of six measurements was 5.0 m. The inundation distance was ~140 m to the hill slope (Fig. 5e). Most houses were washed to the base of the hill, but some were not completely broken (Fig. 6f). Given the proximity to the hill, the residents spontaneously escaped to high ground. In addition, tsunami education and training had been conducted by national and foreign Non-Governmental Organizations (NGOs), such as TimPB and SurfAID, with four drills since 2008. These factors contributed to the small number of fatalities.

4.1 Observed Tsunami Waveforms

The tsunami was recorded at more than ten tide gauge stations around the Indian Ocean, including four tide gauge stations in Indonesia (Padang, Enggano, Tanahbala, and Telukdalam; Fig. 7). In addition, the tsunami was recorded on a surface GPS buoy (GITEWS SUMATRA-03) located off Mentawai just west of the source area and a DART station 56001, ~1,600 km southeast of the epicenter. The tsunami waveforms were collected from the websites of the West Coast/Alaska Tsunami Warning Center (WCATWC http://wcatwc.arh.noaa.gov/), Flanders Marine Institute (VLIZ http://www.ioc-sealevelmonitoring.org/) of UNESCO/IOC, and the National Oceanic and Atmospheric Administration (NOAA, http://nctr.pmel.noaa.gov/Dart/), and are shown in Fig. 8.

Open image in new window

Open image in new window

Within 10 min of the earthquake, the Mentawai buoy recorded two pulses of tsunami waves with amplitudes of ~15 cm and a period of a few minutes. At Padang, the nearest tide gauge station, the tsunami started with a small negative pulse at approximately 1 h after the earthquake, followed by a crest of ~0.3 m. The tsunami amplitudes were mostly <0.1 m at other stations in Indonesia, Sri Lanka, and Maldives. Notable exceptions were at Rodrigues and Port Louis (Mauritius) where a second wave with ~0.2 m amplitude arrived ~6 and ~7 h after the earthquake, respectively. The tsunami was also recorded at La Reunion Island with amplitude ~0.2 m (Sahal and Morin 2012). At the DART station, the tsunami arrived at ~2 h after the earthquake with an amplitude of just 0.01 m.

4.2 Tsunami Computation and Bathymetry Data

We carried out three types of tsunami numerical computations. The first one was in the entire Indian Ocean for the inversion of the observed tsunami waveforms to estimate the earthquake slip distribution (Sect. 3.3). The second one was a more detailed computation around the Pagai Islands with tsunami inundation to compare with the measured tsunami heights (Sect. 3.4). The third type involved a one-dimensional computation based on the measured topographic profiles at selected sites (Sect. 3.5).

For the tsunami propagation in the Indian Ocean, we computed tsunami waveforms by using linear computations on a spherical coordinate system (Satake 1995). The computational region is as shown in Fig. 7 with grid interval of 2′ (~3.7 km). Near the tide gauge stations, finer (24″, ~0.74 km) grids were adopted. The bathymetric data was resampled from GEBCO_08 30″ gridded data. Around the Indonesian tide gauge stations (Padang, Enggano, Tanahbalah, and Telukdalam), we made 12″ (~0.37 km) grid data from GEBCO_08 and Indonesian Navy charts.

For the computations around the Pagai Islands, we used a system of nested grids. The largest domain that included the source region had a 27″ (~833 m) grid, while the finest domain had a 1″ (~30 m) grid. For the coarser domains (27″ and 9″), the linear shallow-water equations were used, while at the finer domains (3″ and 1″), both linear computation and nonlinear computation with inundation were made. For these detailed computations, we merged ETOPO1 data with Indonesian Navy chart data at the 200-m contour depth, and modified the coastline and SRTM data, as described below.

For the one-dimensional tsunami inundation simulations along the surveyed transect lines, the nonlinear shallow-water equations were solved using a finite difference scheme in the Cartesian coordinate system (Gusman et al., 2012). For this computation, tsunami waveforms computed in the detailed simulation at points several hundreds meters from the shoreline at a water depth of ~5 m were used as input waveforms. Tsunami inundation along a surveyed topographic profile was computed with a 1-m grid size.

We found that publicly available bathymetric data, such as GEBCO_08 or ETOPO1, as well as SRTM topographic data, were not very accurate around the Pagai Islands. Comparison of GEBCO_08 and Indonesian Navy chart data around the Pagai Islands (Fig. 9) indicated that the GEBCO_08 data showed abnormally shallow areas off the west coast of the Pagai Islands. In addition, the coastlines are different between the GEBCO_08 and Navy Chart data. Various satellite imagery maps were made and provided by the Center for Satellite Based Crisis Information (ZKI) of the German Aerospace Center (DLR) and the Centre for Remote Imaging, Sensing, and Processing (CRISP) of the National University of Singapore (NUS). Comparisons of these images indicated that the coastlines in the Navy charts were more accurate. We therefore manually digitized the coastlines and water depth points from the Indonesian Navy Charts (241, 242, and 277) and made nearshore bathymetry extending to the 200 m depth contour. We then merged this data with the GEBCO or ETOPO1 data. For the finer domains (3″ and 1″), we directly digitized the coastlines from the satellite images.

Open image in new window

For the topography data of the Pagai Islands, we compared the Shuttle Radar Topography Mission (SRTM) data (1″ grid) provided by the US National Aeronautics and Space Administration (NASA) with topographic profile data obtained through a field survey, and found that the SRTM data were consistently higher by several meters, likely because of the vegetation effects. Therefore, we reduced 7 m in elevation from the SRTM data to form the topography data and then merged this with the bathymetry data.

4.3 Tsunami Inversion and Slip Distribution

The result of the waveform inversion is shown in Fig. 10a and Table 2. This shows that most slip occurred on the shallower subfaults (60 km width). Except at a few subfaults, the slips on deeper subfaults were mostly zero. Late first arrival of the tsunami at Padang prohibited large slip on the deep fault off North Pagai Island. This was in contrast to the 2007 Bengkulu earthquake, which had the most slip at the deeper part of the plate interface (Fujii and Satake 2008; Fig. 10a). The largest slip, 6.1 m, was estimated on the shallowest subfault near the epicenter. The average slip on shallower subfaults was approximately 2 m. The resultant seafloor deformation (Fig. 10b) showed that a small amount (up to a few tens of cm) of subsidence was expected only on a part of North Pagai Island. The maximum subsidence recorded at GPS stations was approximately 4 cm on the west coat of South Pagai island (Hill et al., 2012). The computed tsunami waveforms from this seafloor deformation reproduced the observed waveforms very well (Fig. 8). In particular, the two pulses recorded at the Mentawai buoy seemed to be from two large slip patches (subfaults 4 and 6) and lack of slip in between (subfault 5).

Open image in new window

Slip on the shallow subfault ranged from 1 to 6 m, while slip on the deep subfaults was smaller. The maximum computed slip of 6.1 m was somewhat smaller than that estimated from seismic wave analysis (4.5 to 9.6 m; Lay et al., 2011; Bilek et al., 2011; Newman et al., 2011) or from the GPS and tsunami modeling (9.7 m; Hill et al., 2012), although the size of slip patches of the above analyses was smaller. The seismic moment was computed as 1.0 × 10²¹ Nm, assuming the rigitidy of 3 × 10¹⁰ N/m². The corresponding moment magnitude was M_w = 7.9.

4.4 Detailed Simulations Around Pagai Islands

Nearshore tsunami heights on 9-arcsec (9″) grids were calcualted using the linear shallow-water wave equations and the source model described in the previous section. On the east coast of the Pagai Islands, the computed nearshore tsunami heights were less than 2 m. The computed heights on the west coasts varied from place to place, ranging from 2 to 14 m (Fig. 11). In general, the computed heights were larger on the northern coast of North Pagai Island and the southern coast of South Pagai Island than on the coasts between, as can be seen in the color of coasts in the figure. Our computation showed a maxmimum nearshore tsunami height of 7.8 m around Sibigau Island, where the extreme run-up height of >16 m was reported by Hill et al., (2012).

Open image in new window

Comparison of the measured tsunami heights with the computed nearshore heights within 500 m of the measurement points showed that they are similar at most locations. At Maonai, the computed heights ranged from 4.9–8.1 m, while the measured heights ranged from 6.7–7.3 m. On North Pagai Island, the computed heights ranged from 2.8 to 6.5 m at Tumalei (measured heights are 4.0–6.1 m), from 2.2 to 3.8 m at Macaronis resort (measured heights are 2.9–5.4 m), from 2.0–4.0 m at Muntei Barubaru (measured heights are 4.6–5.7 m in the village and 3.9–8.8 m on the western side), and from 3.0–6.8 m in Sabeu Gunggung (measured heights are 4.3–7.0 m). An exception is at Asahan where the computed heights ranging from 1.8 to 3.1 m were much smaller than the measured heights ranging from 6.4 to 9.3 m.

Nonlinear computations, including inundation on land, were carried out on finer (3″ and 1″) grids. The coastal and inundation heights on the finest grids (1″) were shown in Fig. 12 with enlarged maps at Muntei Barubaru, Sabeu Gunggung, and Macaroni Resort. Comparison of the computed inundation distance and heights with the measurements showed that the computed results are smaller than the measurements. A comparison of coastal heights from linear to nonlinear computations indicated that the nonlinear computations produced smaller coastal heights than the linear computations by a factor ranging from 1.3 to 2.2. Nonlinear effects include advection terms and bottom friction, and both depend on bathymetry. As described in Sect. 3.2, our bathymetry data were not accurate enough for 3″ or 1″ grids, particularly the arbitrary correction applied to topography data (7 m elevation reduced from the SRTM data). In order to examine the effects of topographic data, we carried out one-dimensional inundation computation using topographic profiles measured during the field survey.

Open image in new window

4.5 One-dimensional Inundation Computations

The tsunami run-up heights along the measured topography profiles were computed and compared with the measured tsunami heights (Fig. 5). The waveforms calculated at offshore points of Sabeu Gunggung, Muntei, and Macaroni Resort on a 1″ grid with linear and nonlinear computations, at Tumalei on a 3″ grid with linear and nonlinear computations, and at Maonai on a 9″ grid with linear computations were used as input for the 1D computation. When the linear input waveforms were used, the computed tsunami inundation heights were mostly similar or larger than the measured heights projected on the profile, except at Sabeu Gunggung. Considering that not all the measured points were located on the profile, the agreement was rather satisfactory. When the nonlinear input waveforms were used, the computed inundation heights were somewhat smaller than the measured heights. The discrepancy came from the different amplitudes of input waveforms. The amplitudes on the 1″ grid were smaller than those on the 9″ grid by a factor of 1–1.7; the nonlinear computation produced smaller amplitudes than the linear computation.