Hydroacoustic contributions to understanding the December 26th

0 downloads 0 Views 401KB Size Report
Dec 26, 2004 - earth. By exploiting these characteristics, regional hydroacoustic arrays can provide ... is the best indicator of tsunomigenic potential, arguing that for very large events the duration of the ... minimizes travel time residuals (e.g., Fox et al. ... rupture was ∼8 min, with an average rupture velocity of ∼2.5 km/s.
Surv Geophys DOI 10.1007/s10712-006-9003-6 REVIEW

Hydroacoustic contributions to understanding the December 26th 2004 great Sumatra–Andaman Earthquake Maya Tolstoy Æ DelWayne R. Bohnenstiehl

Received: 20 November 2005 / Accepted: 31 May 2006  Springer Science+Business Media B.V. 2006

Abstract Within the era of modern digital-recording, the December 26th 2004 Sumatra–Andaman Earthquake represents an event of unprecedented scale. Hydroacoustic observations have made significant contributions toward our understanding of this great rupture and serve to reiterate the potential use of tertiary (T) waves as a tool in tsunami warning. Small-aperture arrays of hydrophones operated by the International Monitoring System (IMS) recorded the seismically generated, water-borne T-wave within the Indian Ocean. Due to the velocity structure of the oceanic water column, T-wave propagation is both slower and more efficient than radiation within the solid earth. This results in a relatively large amplitude signal that arrives within a time window distinct from the more complex and overlapping pattern of solid earth seismic phases. Hydroacoustic analysis has constrained the rupture length of the fault to be ~1,200 km and the duration on the order of 8 min, with 2–3 phases exhibiting progressively decreasing rupture velocity. These data also indicate that aftershock rates in the first hours following the mainshock correlate with spatial variability in the sourced T-wave amplitude, with far fewer events along the northern section of the main rupture. Although IMS stations telemeter data in near real time, data access for scientists was restricted due to the provisions of the Comprehensive Test Ban Treaty. The swift dissemination of data will be critical in using hydroacoustic methods to assess the magnitude and tsunamigenic potential of future events. Keywords T-wave Æ Hydrophone Æ Hydroacoustic Æ Sumatra-Andaman earthquake Æ Rupture speed

M. Tolstoy (&) Æ D. R. Bohnenstiehl Lamont-Doherty Earth Observatory, Columbia University, 61 Route 9W, Palisades, NY 10964-8000, USA e-mail: [email protected]

123

Surv Geophys

Introduction to hydroacoustic methods Acoustic energy from sub-sea earthquakes may become trapped and propagate laterally within the ocean’s low-velocity waveguide, known as the SOund Fixing And Ranging or SOFAR channel. Such signals were first described on coastal seismic stations, where the water-borne phase is converted back to a seismic phase upon incidence with the submerged shelf (e.g., Tolstoy and Ewing 1950). These waterborne signals have been termed tertiary (T) waves (Linehan 1940) due to their arrival after the primary (P) and secondary (S) seismic phases, which propagate at higher velocities within the solid earth. T-waves also may be recorded directly within the sound channel using either bottom-mounted or moored hydrophone systems. This technology was pioneered for acoustic surveillance during the cold war, with subsequent civilian use (e.g., Fox et al. 1994) and the further academic development of autonomous instruments for long-term monitoring of geological, biological and environmental ocean noise (e.g., Fox et al. 2001). The use of hydroacoustic data for earthquake detection and location has two principle advantages relative to traditional seismic methods: (1) energy loss is cylindrical (~ r–1) within the sound channel, as opposed to the spherical spreading (~ r –2) that dominates propagation within the solid earth; (2) the sound velocity structure of the ocean is reasonably well defined and less variable than the solid earth. By exploiting these characteristics, regional hydroacoustic arrays can provide significant improvements in detection threshold and location accuracy, relative to land-based seismic monitoring (Fox et al. 1994; Bohnenstiehl and Tolstoy 2003). Importantly, T-wave studies have done much to define mid-ocean ridge eruptive processes (e.g., Fox et al. 1995; Dziak and Fox 1999), illuminate the complex structure of subduction zones and slabs (e.g., Fox and Dziak 1999; Okal and Talandier 1997), provide insights into transform fault dynamics (e.g., Dziak et al. 1997; McGuire et al. 2005) and shown great promise as a tool in addressing a variety of other tectonic problems (e.g., Smith et al. 2002; Bohnenstiehl et al. 2002, 2003; Escartı´n et al. 2003). T-wave arrivals are emergent in nature, with characteristics that may reflect their environment of formation. Those sourced from the abyssal setting, along mid-ocean ridge crests or from abyssal plains lying below the sound channel axis, are commonly more symmetric with shorter tails than slope-generated T-waves (Johnson et al. 1963; Keenan and Merriam 1991). In the subduction zone setting, where longer land paths may exist, slope-generated T-waves may exhibit a lower frequency content with multiple peaks in the arrival packet (Johnson and Norris 1968; de Groot-Hedlin and Orcutt 1999; Graeber and Piserchia 2004). These multiple peaks are thought to correspond to specific bathymetric radiators. Several studies, going back to Ewing et al. (1950), have cited the potential applicability of T-wave data to tsunami warming and detection programs (cf. Talandier 1972). Okal and Talandier (1986) showed that the length of the T-wave train is the best indicator of tsunomigenic potential, arguing that for very large events the duration of the source is the major agent controlling this parameter. In these cases, maximum signal amplitude does not appear to correlate well with the event magnitude. Many tsunami generating events in fact exhibit T-wave energy deficiencies that are interpreted by Okal et al (2003) to reflect the slow rupture velocities for these earthquakes.

123

Surv Geophys

Despite the wide-spread use of T-wave signals over the past several decades, the exact mechanism(s) through which seismically generated energy becomes coupled into the oceanic water column remain poorly understood. Seismic rays intersecting the seafloor are predicted to be bent sharply toward vertical due to the velocity contrast across the interface; however, only acoustic energy with low grazing angles ( < ~12) will become trapped directly with the SOFAR channel (Ewing and Worzel 1948). Where seismic waves insonify a sloping seafloor, energy may be converted to low grazing angles through a series of reflections between the sea surface and bottom—a mechanism termed down-slope propagation (Johnson et al. 1963; Talandier and Okal 1998). For abyssal T-waves, scattering from the rough seafloor is most commonly evoked as an excitation mechanism, and in fact such a process also can be used to synthesize slope derived T-waves (Fox et al. 1994; de Groot-Hedlin and Orcutt 1999, 2001; Park et al. 2001; Yang and Forsyth 2003). Other proposed mechanisms that may contribute to the process of T-wave generation include sea surface scattering (Johnson et al. 1968) and scattering and mode coupling associated with internal waves and density neutral thermohaline variations within the water column (Butler 2004; Colosi 2004).

Hydroacoustic instrumentation and processing The International Monitoring System (IMS) of the Comprehensive Test Ban Treaty Organization (CTBO) is deploying a sparse global network of hydrophone stations designed to detect the sounds generated by explosions that may be carried out at or below the ocean surface (Okal 2001). Installations within the Indian Ocean Basin include two stations H08N and H08S that are located ~180 km to the northwest and ~25 km to the south of the Diego Garcia island atoll (6.3 S, 71.0 E and 7.6 S, 72.5 E), a station H01W that is positioned ~100 km to the southwest of the Australian coast near Cape Leeuwin (34.9 S, 114.1 E) and stations H04N and H04S that are ~30 km to the north and south of Crozet Island (46.1 S, 51.8 E and 46.8 S, 51.9 E) (Fig. 1). The stations consist of three hydrophones deployed in a triangular configuration (triads) with ~2 km spacing between instruments. IMS hydrophone sensors are moored to the seafloor and floated near the sound channel axis, where they record continuously at a sample rate of 250 Hz and 24-bit A/D resolution. The system’s pressure response is flat to within 3 dB over the frequency range ~3–100 Hz, with continued sensitivity within the mHz range (Hanson and Bowman 2005a). The small-aperture triangular geometry of the IMS stations can be utilized to estimate the directionality of the arriving signal. For a given station, travel time differences between each pair of hydrophones are derived from the cross correlation of the signal arrivals. These delay times can be inverted to determine the horizontal slowness components and estimate the back-azimuth to the source region (e.g., Hanson and Givens 1998; Del Pezzo and Giudicepietro 2002). Typically, this analysis is done using sliding windows of 10–30 s duration, providing an azimuthal record throughout the T-wave packet. Calculations may be performed on a band passed signal, typically in the 4–6 Hz range where the largest signal to noise ratio is typically observed (e.g., de Groot-Hedlin 2005; Tolstoy and Bohnenstiehl 2005). Alternatively cross-correlations can be derived within multiple frequency bands, using a

123

Surv Geophys Fig. 1 IMS hydroacoustic stations within the Indian Ocean Basin. Great Sumatra– Andaman Earthquake epicenter illustrated with a yellow star. Area of rupture shown with solid red line. Inset shows the geometry of the H08S hydrophone triad located to the south of the Diego Garcia atoll

Progressive Multi-Channel Correlation (PMCC) procedure (e.g., Cansi 1995; Okal and Cansi 1998). As the sound velocity in the hydrophone arrays is well constrained, the slowness value returned by the inversion, along with the correlation coefficient or closure value of the lag times, can be used to assess the quality of the azimuthal estimates (Cansi 1995). Both empirical results and theoretical arguments suggest azimuthal accuracies that are on the order of ~1 or less (Graeber and Piserchia 2004; Hanson et al. 2005a, b; Chapp et al. 2005). Uncertainties arise largely due to the temporal movement of the tethered hydrophones about their watch-circles. When events are detected at multiple stations, absolute arrival time information can be used for location purposes within either a grid search routine or inversion that minimizes travel time residuals (e.g., Fox et al. 2001). Azimuthal and travel time data also may be used jointly to constrain event location (e.g., de Groot-Hedlin 2005; Hanson and Bowman 2005a). When combining data from multiple stations, however, care must be taken to account for azimuthally dependant multi-path effects, which appear to be common in the subduction zone setting (e.g., Graeber and Piserchia 2004).

The Great Sumatra–Andaman Earthquake The Great Sumatra–Andaman earthquake occurred at 00:58:53.5 UTC on December 26th 2004, triggering a devastating tsunami estimated to have killed as many as 300,000 people. Initially classified as a Mw 9.0 event based on body and surface waves (Havard CMT 2004; Ji 2005; Park et al. 2005), analysis of the normal modes excited by this event indicated a moment equivalent to Mw 9.3 (Stien and Okal 2005). Subsequent work showing the extent of the rupture to be on the order of 1,200 km supports this larger moment estimate (Ishii et al. 2005; Kruger and Ohrnberger 2005;

123

Surv Geophys

Tolstoy and Bohnenstiehl 2005; Guilbert et al. 2005). The width of the rupture zone is estimated to have been 150–200 km (Ammon et al. 2005), with the fault rupturing from a depth of ~30 km (Lay et al. 2005) to the seafloor trench. The duration of the rupture was ~8 min, with an average rupture velocity of ~2.5 km/s. Propagation was to the north-northwest along the Sunda Trench, a convergent boundary where the Indian plate is subducting beneath the Burma plate with increasingly oblique motion to the north. T-waves were observed at all of the Indian Ocean IMS hydroacoustic stations, at distances ranging from 2,866 km (H08S) to 7,025 km (H04S) from the epicenter. At the time of the 2004 Sumatra–Andaman event, however, only two channels were functioning at each of the Crozet triads, so azimuth estimates from this station were not as accurate. The strongest arrival was observed at Diego Garcia South (H08S), where the duration of the T-wave was on the order of 800 s. Although the Diego Garcia North triad (H08N) lies at a similar distance to the rupture, T-wave propagation was partially blocked by near-station bathymetric features (de Groot-Hedlin 2005). Figure 2 shows

9

Amplitude (µPa)

x 10 1

A

0.5 0 –0.5

1600

1800

2000

2200

2400

2600

2800

3000

2800

3000

2800

3000

Frequency (Hz)

Time after Event Origin Time (seconds)

100

B

50

0 1600

1800

2000

2200

2400

2600

Azimuth from H08S (deg)

Time after Event Origin Time (seconds) 70

C 60 50 40 1600

1800

2000

2200

2400

2600

Time after Event Origin Time (seconds)

Fig. 2 Broad-band waveform (top) and spectrogram (middle) of T-wave arrival at a single Diego Garcia South hydrophone. The azimuth to the source (bottom) is calculated from cross-correlation of all three hydrophones elements, using a 4–6 Hz band-pass signal with 10 s windows and 50% overlap. After Tolstoy and Bohnenstiehl (2005)

123

Surv Geophys

the spectrogram of the T-wave arrival from one channel of the Diego Garcia South triad. In this review, we summarize three peer-review manuscripts dealing with hydroacoustic recordings of the Great Sumatra–Andaman earthquake of December 26th, 2004. Two of these three studies used exclusively Diego Garcia South (HO8S) data (Tolstoy and Bohnenstiehl 2005; Guilbert et al. 2005) and the other (de GrootHedlin 2005) used data from three Indian Ocean stations (H08S, H01W and H04S). Each study used different approaches for their analysis, as outlined below. In addition, Hanson and Bowman (2005b) and Graeber et al. (2005) were able to observe the actual tsunami arriving at Indian Ocean hydroacoustic and seismic stations through low pass filtering. However, we will focus on the analysis of T-wave signals and the earthquake rupture itself.

Hydroacoustic studies de Groot-Hedlin—Multi-station approach The de Groot-Hedlin (2005) paper used a multi-station approach, whereby azimuthal estimates derived from a plane wave inversion were used jointly with arrival time information. For each station, arrival and predicted hydroacoustic travel times were used to compute T-wave source times for a set of potential radiator points within a swath defined by the estimated back-azimuth ( ± the predicted error). Radiator positions were then determined by minimizing differences between the source times computed separately for each station. This method can be used to locate sources with as few as two triads. Although only two hydrophones in each of the Crozet triads were operational at the time of the event, de Groot-Hedlin (2005) was able to make an additional azimuth estimate using the station H04S, where the two operational hydrophones were oriented normal to the propagation direction. Of the two possible solutions for this two-sensor geometry, the northeast one was assumed. Final source locations were then derived based on the arrival times and estimated azimuths at stations H08S, H0W1 and H04S. This approach assumes that the T-wave energy arriving at each station shares a common radiator point and source time. However, using well-constrained epicenters from moderate magnitude (4.4–5.7 mb) earthquakes in the Sumatra region, Graeber and Piserchia (2004) have shown that estimates of T-wave radiator location depend on the station recording the event. This is a logical result when the mechanism for generating T-waves is considered. Since the T-waves in this setting are primarily Pwave energy converted at a crust-water interface (Graeber and Piserchia 2004), the area of seafloor insonified by seismically radiated energy is dependent on the depth of the earthquake, and the strongest radiation points will be controlled by the local topography within this zone of generation (de Groot-Hedlin and Orcutt 1999; Grauber and Piserchia 2004). Thus, stations at different azimuths may have different T-wave radiation points, depending on the morphology and orientation of the shelf. This observation, combined with differences in land-path travel times may introduce errors in source location and timing (Fig. 3). The de Groot-Hedlin (2005) analysis concludes that the rupture had two phases with the first phase having a faster rupture velocity (2.4 km/s) than the latter phase

123

Surv Geophys Fig. 3 Diagram illustrating how different radiator locations might lead to errors in timing and location of the apparent source location if a single source is assumed. The black line between the brown (solid earth propagation) and blue (ocean propagation) represents radiator location which is likely to be steep portion of the shelf edge within the peak area of the SOFAR channel (1,000–2,000 m) (Graeber and Piserchia 2004). Note that arrivals from the same earthquake at different stations have different T-wave radiator locations, and different length crustal and water paths

(1.5 km/s). Other hydroacoustic and seismic studies also report a faster rupture initially and slower rupture toward the north (Kru¨ger and Ohrnberger 2005; Tolstoy and Bohnenstiehl 2005; Guilbert et al. 2005), with seismic observations suggesting an average rupture velocity 2.5 km/s (e.g., Ishii et al. 2005; Kru¨ger and Ohrnberger 2005). Throughout the ruputure, velocity estimates by de Groot-Hedlin (2005) are somewhat slower (~0.3–0.5 km/s) than those cited in other studies (Kru¨ger and Ohrnberger 2005; Tolstoy and Bohnenstiehl 2005; Guilbert et al. 2005). This discrepancy may be explained through bias associated with the multi-station method, or associated with the fact that de Groot-Hedlin measured distance from the epicenter, rather than along a fault line. Nevertheless, the general pattern of faster rupture to the south and slower rupture to the north is consistent with other hydroacoustic studies. The multi-station location analysis of the de Groot-Hedlin tracks the rupture for ~800 km, extending from the epicentral region near ~3 N to ~930¢ N latitude along the subduction zone. She notes, however, the coherent T-wave arrival at both H08S and H01W suggest that the rupture may continue to the north. This suggestion is supported by other hydroacoustic analysis (Tolstoy and Bohnenstiehl 2005; Guilbert et al. 2005). Tolstoy and Bohnenstiehl—single station approach A single station method employed by Tolstoy and Bohnenstiehl (2005) used only data from the closest hydroacoustic station, Diego Garcia South. This approach avoids the assumption of a single radiator point at any given stage in the rupture regardless of station azimuth. The trade-off is that given only a series azimuthal constraints an assumption must be made regarding the depth from which the dominant T-wave radiation occurs. Graeber and Piserchia (2004) found that for moderate magnitude events in this subduction setting radiation points cluster around steeply sloping features such as

123

Surv Geophys

the edge of the shelf and trench walls, at or near the depth of the SOFAR channel in this region (~1,000–2,000 m). Characteristic sound velocity profiles for the month of December are shown in Fig. 4(https://www.idbms.navo.navy.mil/gdemv/ gdemv.html). Tolstoy and Bohnenstiehl (2005) selected the 2,000 m contour in their analysis, projecting azimuthal information onto this shelf depth and estimating the source time by subtracting a predicted acoustic travel time from the observed arrival time. The 2,000 m contour was selected because it approximates a natural break in the slope that lies near the mid-depth position of the over-thrust shelf. For an event this large, where a broad area of the shelf likely radiates energy, the actual choice of a depth contour is less important than the fact that it be representative of the trend of the shelf. For example, the choice of a 1,000 m contour in this region would results in a more circuitous path along the flatter upper part of the shelf. Although the geometry of the steep shelf edge is relatively well defined by the 2,000 m contour, Tolstoy and Bohnenstiehl (2005) applied an additional correction to account for variations in travel time that might be due to long-wavelength meandering of the shelf radiator zone. T-wave source times were corrected for variable land-paths about a line that mimicked the geometry of the trench (i.e., the strike of the subduction interface), but lay to the east of the 2,000 m contour. This was done assuming a seismic velocity of 6 km/s, consistent with P-wave energy being the dominant contributor to T-wave amplitude (Graeber and Piserchia 2004). The results of their analysis indicate a source duration of ~480 s and rupture length of ~1,200 km, consistent with other seismic (e.g., Ishii et al. 2005; Kru¨ger and Ohrnberger 2005) and hydroacoustic (Guilbert et al. 2005) studies. The single station approach provides considerable detail regarding variations in rupture velocity with time. As in the de Groot-Hedlin (2005) paper, Tolstoy and Bohnenstiehl characterize the rupture as having two phases. They used a least squares linear regression to determine the velocity based on the corrected source time versus distance along the fault trace. The timing of the velocity transition between these two phases was determined by minimizing the distance misfit using a series of two-slope velocity models. The best-fitting two-slope models were associated with break points occurring at source times within the 150–190 s range, with a broader minimum between 150 and 290 s. In their preferred model, Tolstoy and Bohnenstiehl find an average rupture velocity of 2.8 km/s ± 0.1 km/s for the first Water Column Velocity Profiles near T-wave radiator sites. 0 Depth below seasurface (meters)

Fig. 4 Example sound velocity profiles for the area. Data are the December averages from the U.S. Navy’s Generalized Digital Environmental Model (GDEM) for a variety of latitudes off the edge of the shelf along the rupture length of the earthquake

3°N 93°E 7°N 91°E 12°N 91°E

500 1000 1500 2000 2500 3000 3500 4000 4500 5000 1490

1500

1510

1520

1530

1540

Sound Speed (meters per second)

123

1550

Surv Geophys

~180 s, and an average rupture velocity of 2.1 km/s ± 0.2 km/s for the remaining ~300 s. These phases also are marked by two pulses in T-wave amplitude with a transition around 7 N, and with the second pulse showing particularly low amplitudes north of ~10 N (Fig. 5). Guilbert et al.—combining hydroacoustic and seismic stations Guilbert et al. (2005) combined hydroacoustic data from Diego Garcia South with data from the CMAR-seismic array in Thailand and utilized azimuthal measurements from both arrays to constrain the rupture length, duration and speed. The hydroacoustic data were analyzed using the PMCC method, using 10 sub-bands within the 1–50 Hz range. Their detection results found coherent arrivals continuing 15° N

91° E

92° E

93° E

94° E

9 5° E

96° E

97° E

14° N

15° N

14° N

2.0 km/s

13° N

12° N

13° N

12° N

11° N

10° N

10° N

2.5 k

11° N

9° N

/s

m/s

2.1 km

Fig. 5 Comparison of hydroacoustically determined rupture velocity. Yellow star indicates earthquake epicenter at 3.3 N 95.8 E. The blue line shows the location of three different rupture velocity phases interpreted by Guilbert et al. (2005). The red line shows the two different rupture velocity phases inferred by Tolstoy and Bohnenstiehl (2005). The yellow circle indicates the position at which de Groot-Hedlin (2005) interprets a change in velocity from 2.4 km/s to 1.5 km/s, and the white dot indicates the limit of the area where rupture could be constrained using the three station method. The black circles indicate the relative amplitude of the broad-band T-wave recorded at H08S (from Tolstoy and Bohnenstiehls 2005) projected along the source region

9° N

8° N

8° N

7° N

7° N

km/s

5° N

m/s 2.7 k

2.8

6° N

6° N

5° N

4° N

4° N

3° N

3° N

2° N

2° N 91° E

92° E

93° E

94° E

95° E

–5250 –4500 –3750 –3000 –2250 –1500 –750

96° E

0

97° E

3000

Elevation (m)

123

Surv Geophys

for ~800 s, with azimuths to the source starting at ~66 (the azimuth to the epicenter) and progressing to ~43. Guilbert et al. (2005) constrain the T-wave radiator points to lie along three linear segments that approximate the intersection of the shelf with the axis of the SOFAR channel. They invert for the rupture velocity within these segments, with results indicating decreasing rupture speeds of 2.7 km/s, 2.5 km/s and 2.0 km/s toward the north (Fig. 5). The length of the rupture zone was determined to be 1,235 km, with duration of 515 s, only slightly larger than the values proposed by Tolstoy and Bohnenstiehl (2005). To test the results of their inversion, the solution derived from the H08S data was used to predict the azimuthal pattern of arrivals at the CMAR-seismic array. This simulation was successful; however, it was necessary to apply an empirical correction to the CMAR observations to account for complexities in the seismic velocity structure. These corrections were derived from a series of well-located earthquakes distributed along the length of the mainshock rupture zone. This approach, combining hydroacoustic and seismic data, is innovative and demonstrates the possibilities for using these technologies jointly to associate and locate events recorded on both networks—a procedure not yet implemented in routine processing at the International Data Center.

Aftershocks Guilbert et al. note that aftershocks from the initial part of the fault rupture begin before the rupture is complete in the north. This can be clearly seen in the azimuth data (Fig. 6) as arrivals coming in at azimuths above the main line of rupture. The hydroacoustic data provides a useful tool to observe these early aftershocks, whereas they might otherwise be difficult to resolve using traditional seismic analysis. Early seismic reports suggested that there were no aftershocks in the northern most section of the rupture in the first 60+ min, supporting the conclusion that the latter portion of the fault must have slipped in a later slow-slip event (Bilham 2005). Investigation of the hydroacoustic detection results indicates that there where fewer events in the north, where lower T-wave amplitudes were sourced and seismic inversions show lower mean slips (e.g., Ammon et al. 2005). As aftershocks represent a process of stress relaxation within the lithosphere, the correlation between aftershock rates and local moment release along the fault is intuitive and consistent with a productivity law whereby the triggering rate scales with the mainshock magnitude (e.g., Ogata 1988). Using these observations, an argument also could be made for a dearth of aftershock events in the north in the first ~50 min; however, there is no evidence for an abrupt increase in event rate that might represent the lithosphere’s response to a slow earthquake occurring in the hours immediately following the mainshock (Fig. 6).

Conclusions Hydroacoustic analysis has proven to be a valuable tool in understanding the rupture of this very large submarine earthquake. All studies indicate a unilateral propagation of the rupture to the north along the Sunda trench. Rupture length estimates range from >800 km using a multi-station approach (de Groot-Hedlin 2005) to

123

Surv Geophys

Azimuth (N-S)

A

Start of Mainshock

70 60 50 40

2000

4000

End of Mainshock

6000

8000

10000

12000

14000

Time (Seconds after start of main shock)

B

Number of detections

100

80

60

40

20

0 40

45

50

55

60

65

70

Azimuth from Diego Garcia South (NORTH TO SOUTH) Fig. 6 (A) Azimuths (in degrees) of Diego Garcia South (H08S) detections during the 4 h following the mainshock (black x’s). Note that it takes ~1,900 s for the first arrival to reach this station, with a slight increase in T-wave travel time as the rupture propagates to the north. Continuous band of detections between 1,900 s and 2,700 s is the mainshock T-wave, consistent with the predicted arrival time for a water-borne path based on the seismically determined origin time and epicenter. Detections at later times are aftershocks, which occur within the southern section (higher azimuth) before the rupture has stopped in the north. B: Histogram of number of detections versus azimuth (in degrees) for the first hour (grey) and first ~3.5 h (dark grey) following the arrival of the mainshock. The number of detections does not correspond directly to the number of aftershock, as a single event may produce coherent arrivals spanning multiple 10 s detection windows. Rather, the histogram is broadly representative of the total T-wave duration (energy) resulting from aftershock activity. The black x’s indicate the normalized waveform amplitude of the mainshock T-wave arrival at a given azimuth. The two major amplitude peaks coincide with the two largest peaks in aftershock detection. The grey dashed line indicates the area of maximum slip ( > ~6 m) inferred by Ammon et al. (2005) based on their Method II. Note that aftershocks peak at the end of zone of maximum slip consistent with an increase in stress at the edge of the slip patch

~1,200–1,235 km using a single station analysis (Tolstoy and Bohnenstiehl 2005; Guilbert et al. 2005). The latter estimates show that the rupture terminated near the latitude of the pole of rotation between the Indian and Burma plates (Bird 2003),

123

Surv Geophys

and are consistent with a variety of seismic results (e.g., Ni et al. 2005; Ishii et al. 2005; Ammon et al. 2005). Unlike seismic data, however, which has been equivocal on the change in rupture speed, the hydroacoustic papers have all clearly shown that the average rupture speed decreased as the rupture progressed. It appears that the first ~450 km ruptured with an average rupture speed of 2.7–2.8 km/s, with the latter ~750 km rupturing with an average speed between 2.4 km/s and 2.0 km/s (Tolstoy and Bohnenstiehl 2005; Guilbert et al. 2005). T-wave data allows a detailed assessment of the aftershocks associated with the earthquake, in particular smaller aftershocks not easily located by global arrays, as well as aftershocks that were occurring while the primary rupture was still taking place or immediately following it. The rate of aftershock activity correlates with along strike changes in the radiated T-wave amplitude, with fewer events along northern section of the rupture in the hours immediately following the propagation of the mainshock. Finally, the results of these studies reiterate the usefulness of hydroacoustic data as a tool for rapid assessment of the scale and character of large, submarine earthquakes. In this regard, T-wave duration remains the simplest tool at our disposal. While the T-wave travels more slowly than solid earth waves it still arrives with minutes to 10’s of minutes of the solid earth waves, depending on the distance from the source. With the total duration of coherent T-wave signal being >600 s at all the T-wave monitoring stations in the Indian Ocean basin, a comparison with existing duration-moment data (e.g., Okal and Talandier 1986) would have identified the great size and tsunamigenic potential of this event within 45 min of the start of the start of rupture. Acknowledgements We thank an anonymous reviewer for constuctive comments. LDEO contribution number 6942

References Ammon CJ, Ji C, Thio H-K, Robinson D, Ni S, Hjorleifsdottir V, Kanamori H, Lay T, Das S, Helmberger D, Ichinose G, Polet J, Wald D (2005) Rupture process of the 2004 Sumatra– Andaman Earthquake. Science 308:1133–1139 Bilham R (2005) A flying start, then a slow slip. Science 308:1126–1127 Bird P (2003) An updated digital model of plate boundaries. Geochem Geophys Geosyst 4, doi:10.1029/2001GC000252 Bohnenstiehl DR, Tolstoy M, Smith DK, Fox CG, Dziak R (2002) Aftershock sequences in the midocean ridge environment: an analysis using hydroacoustic data. Tectonophysics 354:49–70 Bohnenstiehl DR, Tolstoy M, Smith DK, Fox CG, Dziak R (2003) Time-clustering behavior of spreading-center seismicity between 15 and 35 N on the Mid-Atlantic Ridge: Observations from hydroacoustic monitoring. Phys Earth Planet Inter 138:147–161 Bohnenstiehl DR, Tolstoy M (2003) Comparison of teleseismically and hydroacoustically derived earthquake locations along the north-central Mid-Atlantic Ridge and Equatorial East Pacific Rise. Seismol Res Lett 74:790–801 Butler R (2004) Observation of Seismo-Acoustic T-waves at and Beneath the Seafloor. In: Proceedings of Seismo-Acoustic Applications in Marine Geology and Geophysics Workshop, Woods Hole Oceanographic Institution, APL-UW Tech. Rep. Number 0406 Cansi Y (1995) An automatic seismic event processing for detection and location: the PMCC method. Geophys Res Lett 22:1021–1024 Chapp E, Bohnenstiehl DR, Tolstoy M (2005) Sound-Channel Observations of Ice-Generated Tremor in the Indian Ocean. Geochem Geophys Geosyst 6, Q06003, doi:10.1029/2004GC000889

123

Surv Geophys Colosi J (2004) Theory of long-range acoustic propagation. In: Proceedings of Seismo-Acoustic Applications in Marine Geology and Geophysics Workshop, Woods Hole Oceanographic Institution, APL-UW Tech. Rep. Number 0406 de Groot-Hedlin CD (2005) Estimation of the rupture length and velocity of the great Sumatra earthquake of Dec 26, 2004 using hydroacoustic singals. Geophys Res Lett 32, doi: 10.1029/ 2005GL022695 de Groot-Hedlin CD, Orcutt JA (1999) Synthesis of earthquake-generated T-waves. Geophys Res Lett 26:1227–1230 de Groot-Hedlin CD, Orcutt JA (2001) Excitation of T-phases by seafloor scattering. J Acoust Soc Amer 109:1944–1954 Del Pezzo E, Giudicepietro F (2002) Plane wave fitting method for a plane, small aperture, short period seismic array: a MATHCAD program. Computational Geoscience 28:59–64 Dziak RP, Fox CG (1999) The January 1998 earthquake swarm at axial volcano, Juan de Fuca Ridge: hydroacoustic evidence of a seafloor volcanic activity. Geophys Res Lett 26:3429–3432 Dziak RP, Fox CG, Matsumoto H, Schreiner AE (1997) The 1992 cape mendocino earthquake sequence: Seismo-Acoustic analysis using fixed hydrophone arrays. Mar Geophys Res 19:137–162 Escartı´n J, Smith DK, Cannat M (2003) Parallel bands of seismicity at the mid-atlantic ridge, 12–14 N, Geophys Res Lett 30, doi:10.1029/2003GL017226 Ewing M, Worzel JL (1948), Long-range sound transmission. In: Propagation of sound in the ocean, GSA Memoir 27 Fox CG, Dziak RP, Matsumoto H, Schreiner AE (1994) Potential for Monitoring Low-level Seismicity on the Juan de Fuca Ridge Using Fixed Hydrophone Arrays. Mar Tech Soc 27:22–30 Fox CG, Dziak RP (1999) Internal deformation of the gorda plate using hydroacoustic monitoring methods. J Geophys Res 104:17603–17615 Fox CG, Radford WE, Dziak RP, Lau T-K, Matsumoto H, Schreiner AE (1995) Acoustic detection of a seafloor spreading episode on the Juan de Fuca Ridge using military hydrophone arrays. Geophys Res Lett 22:131–134 Fox CG, Matsumoto H, Lau TKA (2001) Monitoring Pacific Ocean seismicity from an autonomous hydrophone array. J Geophys Res 106:4183–4206 Graeber FM, Piserchia PF (2004) Zones of T-wave excitation in the NE Indian Ocean mapped using variations in back azimuth over time obtained from multi-channel correlation of IMS hydrophone triplet data. Geophys J Int 158, doi: 10.1111/j.1365–246X.2004.02301.x Graeber FM, Grenard P, Koch K (2005) Observations at IMS Hydrophone Stations from the December 2004 Indian Ocean Tsunami Event. Abstract EGU05-A-08773 Guilbert J, Vergoz J, Schissele E, Roueff A, Cansi Y (2005) Use of hydroacoustic and seismic arrays to observe rupture propagation and source extent of the Mw—9.0 Sumatra earthquake. Geophys Res Letts 32, L15310, doi: 10.1029/2005GL022966 Hanson JA, Bowman JR (2005a) Indian Ocean ridge seismicity observed with a permanent hydroacoustic network. Geophys Res Letts 32, L06301, doi: 10.1029/2004GL021931 Hanson JA, Bowman JR (2005b) Dispersive and reflected tsunami signals from the 2004 Indian Ocean tsunami observed on hydrophones and seismic stations. Geophys Res Letts 32, L17606, doi: 10.1029/2005GL023783 Hanson JA, Givens HK (1998) Accurate azimuthal estimates from large aperture hydrophone array using t-phase waveforms. Geophys Res Lett 25:365–368 Harvard CMT, http://www.neic.usgs.gov/neis/eq_depot/2004/eq_041226/neic_slav_hrv.html, 2004 Ishii M, Shearer PM, Houston H, Vidale JE (2005) Extent, duration and speed of the 2004 Sumatra– Andaman earthquake imaged by the Hi-Net array. Nature 435:933–936 Ji C (2005) http://www.neic.usgs.gov/neis/eq_depot/2004/eq_041226/neic_slav_ff.html Johnson RH, Northrop J, Eppley R (1963) Sources of Pacific T-phases. J Geophys Res 68:4251–4260 Johnson RH, Norris RA (1968) T-phase radiators in the western aleutians. Bull Seism Soc Am 58:1–10 Keenan RE, Merriam LRL (1991) Arctic abyssal T phases: coupling seismic energy into the ocean sound channel via under-ice scattering. J Acoust Soc Am 89:1128–1133 Kru¨ger F, Ohrnberger M (2005) Tracking the rupture of the Mw 5 9.3 Sumatra earthquake over 1,150 km at teleseismic distance. Nature 435:937–939 Lay T, Kanamori H, Ammon CJ, Nettles M, Ward SN, Aster RC, Beck SL, Bilek SL, Brudzinski MR, Butler R, DeShon HR, Ekstro¨m G, Satake K, Sipkin S (2005) The great Sumatra–Andaman Earthquake of 26 December 2004. Science 308:1127–1133 Linehan D (1940) Earthquakes in the West Indian Region. Trans AGU 21:229–232

123

Surv Geophys McGuire JJ, Boettcher MS, Jordan TH (2005) Foreshock sequences and short-term earthquake predictability on east pacific rise transform faults. Nature 434:457–461 Ni S, Kanamori H, Helmberger D (2005) Energy radiation from the Sumatra earthquake. Nature 484:582 Okal EA (2001) T-phase T-phase stations for the international monitoring system of the comprehensive nuclear-test ban treaty: a global perspective. Seism Res Lett 72:186–196 Okal EA, Cansi Y (1998) Detection of PKJKP at intermediate periods by progressive mult-channel correlation. Earth Planet Sci Lett 164:23–30 Okal EA, Talandier J (1997) T waves from the great 1994 Bolivian Deep earthquake in relation to channeling of S wave energy up the slab. J Geophys Res 102:27421–27437 Okal EA, Talandier J (1986) T-wave duration, magnitudes and seismic moment of an earthquake—application of tsunami warning. J Phys Earth 34:19–42 Okal EA, Alasset P-J, Hyvernaud O, Schindele´ F (2003) The deficient T waves of tsunami earthquakes. Geophys J Int 152:416–432 Ogata Y (1988) Statistical models for earthquake occurrences and residual analysis for point processes. J Amer Statist Assoc 83:9–27 Park M, Odom RI, Soukup DJ (2001) Modal scattering: a key to understanding oceanic T-waves. Geophys Res Lett 28:3401–3404 Park J, Anderson K, Aster R, Butler R, Lay T, Simpson D (2005) Global seismographic network records the great Sumatra–Andaman earthquake. EOS Trans AGU 86:57–64 Smith DK, Tolstoy M, Fox CG, Bohnenstiehl DR, Matsumoto H, Fowler MJ (2002) Hydroacoustic Monitoring of Seismicity at the Slow-Spreading Mid-Atlantic Ridge. Geophys Res Lett 10.1029/ 2001GL013912 Stein S, Okal EA (2005) Sumatra earthquake: gigantic and slow. Nature 484:581–582 Talandier J (1972) Estude et precision des tsunamis. These d’ University Paris, p 128 Talandier J, Okal EA (1998) On the mechanism of conversion of seismic waves to and from T-waves in the Vicinity of Island Shores. Bull Seism Soc Am 88:621–632 Tolstoy I, Ewing WM (1950) The T phase of shallow-focus earthquakes. Bull Seism Soc Am 40:25–51 Tolstoy M, Bohnenstiehl DR (2005) Hydroacoustic constraints on the rupture duration, length and speed of the great Sumatra–Andaman earthquake. Seis Res Lett 76:419–425 Yang Y, Forsyth DW (2003) Improving epicentral and magnitude estimation of earthquakes from T-phases by considering the excitation function. Bull Seism Soc Am 93:2106–2122

123