Hydroacoustic events located at the intersection of the Atlantis (30°N) and Kane (23°40′N) Transform Faults with the Mid-Atlantic Ridge

Authors


Abstract

We investigate the characteristics of T-phase events located at the ends of two segments of the Mid-Atlantic Ridge. Our motivations for the study were to understand whether T-phase locations represent earthquake epicenters (and thus whether accurate geological inferences can be made from their spatial patterns) and to understand further the relationship between T-phase event characteristics and earthquake properties. We examine the characteristics of 158 T-phase events with respect to both event location water depth and source to receiver distance. The propagation paths of the T-phases are also modeled to study the effects of encountering seafloor topography. We find that existing models for T-phase excitation and propagation cannot explain adequately all of our observations. The amplitudes (Received Levels) of T-phases at the hydrophones show no dependence on event water depths, in contrast to current excitation models which predict a decrease in event magnitude with increasing water depth. The Received Levels are observed to decrease with increasing source to receiver distance, and events from the two study areas exhibit different trends in relative Received Levels between hydrophones, once attenuation is taken into account. Our acoustic ray trace model is able to reproduce similar trends in relative amplitudes at the hydrophones based on 1-D topography between the event and each hydrophone, but the variances in both the observed data and model are high. We observe a pattern of short T-phase onset times for shallow water events and long onset times for deep water events, where onset time is defined as the time interval between the appearance of the T-phase envelope above the ambient noise and its first peak. This suggests that the onset time may be a function of several variables, including efficiency of energy conversion based on local topography, efficiency of propagation based on event water depth, and hypocentral depth in the crust. The results of this study underscore the complexity of T-phase excitation and propagation and argue that current models of T-phase excitation and propagation need to be improved to explain the observed characteristics of T-phase data.

1. Introduction

The T-phase (T-wave or Tertiary Wave) from an earthquake in the oceanic crust and upper mantle is a hydroacoustic wave that travels at the speed of sound in water (∼1.5 km/s) and consequently arrives at an oceanic or coastal sensor after the faster traveling P- (Primary) and S- (Secondary) body waves, which propagate through the earth [Tolstoy and Ewing, 1950; Johnson et al., 1963]. The origin time and seafloor location of T-phase generating events can be obtained from the data recorded by sparse arrays of single hydrophones moored in the ocean sound channel. Catalogs of T-phase events have become increasingly comprehensive and accessible as the networks of hydrophone arrays deployed in the world's oceans expand [e.g., Fox et al., 1994, 2001; Smith et al., 2002]. Long term monitoring of seismicity in remote areas of the oceans using hydrophone arrays has proven to be very successful over the last decade. This is especially true at mid-ocean ridges where T-phase data are providing information on the overall seismicity of the ridge axis and are helping to constrain models of mid-ocean ridge crustal structure [e.g., Dziak and Fox, 1999a, 1999b; Dziak et al., 1995, 2004; Schreiner et al., 1995; Fox et al., 1994, 1995, 2001; Bohnenstiehl and Tolstoy, 2003; Bohnenstiehl et al., 2002, 2003, 2004; Smith et al., 2002, 2003]. T-phase data are also an important source of information for studying hydroacoustic wave propagation in the oceans [Johnson et al., 1968; Pulli et al., 1999; Pulli and Upton, 2002; Harben et al., 2003].

When a submarine earthquake occurs within the crust or upper mantle, seismic energy insonifies an area of the seafloor. The seismic energy at the seafloor couples with the water column to form acoustic energy which then propagates as a T-phase. However, there is much still to understand about the fundamental processes of T-phase excitation and propagation, a situation which limits the extent to which geological inferences can be made from T-phase characteristics. For example, because the area of seafloor that is significantly insonified by the seismic energy may be large, with a diameter up to a few tens of kilometers, it is not known whether the location estimated for the T-phase event is the same as the earthquake epicenter. We do not understand the coupling mechanism which converts seismic energy at the seafloor to acoustic energy in the sound channel. Consequently we do not understand the relationship between T-phase event characteristics and earthquake properties such as magnitude, event depth and focal mechanism. Also we do not understand how seafloor bathymetry effects the propagation of the T-phase.

Motivated by the deficit in our understanding of T-phases we address some of these fundamental issues by investigating the characteristics of T-phase events recorded from two study areas on the Mid-Atlantic Ridge (MAR) by the North Atlantic Hydrophone Array (NAHA) [Smith et al., 2002] (Figure 1). The study areas are located at the eastern end of the Atlantis Transform Fault (TF) (30°N) and at the eastern end of the Kane TF (23°40′N). The two areas were chosen on the basis of their comparable geological setting, their geological relevance to ocean ridge processes, their dramatic bathymetric relief, and their similar number and spatial distribution of T-phase events.

Figure 1.

Location map of the North Atlantic Hydrophone Array (NAHA) and the two study areas at the eastern intersection of the Atlantis (30°N) and Kane (23°40′N) Transform Faults (red boxes) on the northern Mid-Atlantic Ridge (MAR). The red dots represent the locations of all the T-phase events recorded between February 1999 and February 2001 (∼3500 events). Each of the six hydrophones is identified by its location in the array (white stars labeled NE, NW, CE, CW, SE, and SW). The hydrophones are positioned ∼1000 km apart.

Within these study areas, we examine the spatial distribution of T-phase event locations as a function of water depth and geologic setting. We assess whether the amplitudes of the T-phase events are dependent on water depth at the event location and on source to receiver distance. We also quantify and study the significance of the onset time of the T-phase events, which is defined as the time interval between the appearance of the T-phase and the first peak in the T-phase envelope (Figure 2a). Finally, we model bathymetric blockage along the inferred propagation path on the basis of known topography and ray trace theory and compare the model results to the observed data.

Figure 2.

(a) An example of a T-phase time series. The T-phase is composed of a lens or envelope of energy less than 90 seconds in duration. The T-phase has a short onset time, measured from its appearance above the ambient noise to its maximum amplitude, and then gradually decays back to ambient noise levels. (b) Example of a sound velocity profile for the North Atlantic. The SOFAR channel extends from the sea surface to the critical depth (red line), and the sound axis minimum is shown by the darker blue rectangle. (c) Ray trace modeling for events that occur by excitation scattering at the seafloor at varying event water depths. Events in the SOFAR channel will propagate great distances (green, black, blue, gray, and cyan rays). However, those events below the critical depth will be severely attenuated by scattering at the seafloor (red dashed ray) and will not propagate far. The rays plotted represent only the ray path of a horizontally traveling ray from each event. All rays with emergent angles that lie within the colored cones will propagate through the SOFAR channel. As water depth increases, the width of the cone decreases to zero at and below the critical depth. We define the process of decreasing energy with increasing depth as event location water depth dependence.

This work is a first step in comparing hydroacoustic modeling results with data. Quantifying the differences between predictions and observations will guide the future development of acoustic propagation models which, in turn, will extend the applicability of T-phase data to the study of oceanic crustal processes.

2. Acoustics Background

T-phase energy appears in the frequency band of 1–100Hz. The time series of a T-phase usually resembles a lens or envelope of energy where the wave amplitude increases to a maximum and then slowly decays to ambient noise levels (Figure 2a). The T-phases in this study are typically less than 90 seconds in duration. The maximum amplitude of the T-phase lens is assumed to correspond to the point on the seafloor with the maximum conversion of seismic to acoustic energy in the T-phase excitation region [Slack et al., 1999]. Models of T-phases generally consider two regions: a relatively short excitation region where the T-phase characteristics are established and a propagation region where the T-phase energy is totally trapped in the ocean sound channel. Since scattering of some kind is usually invoked in one or both regions we distinguish them by region as excitation scattering and propagation scattering.

The waveguide that traps the T-phase propagation is defined by the sound velocity structure of the water column. A typical sound velocity profile for the North Atlantic is shown in Figure 2b. The depth at which the sound velocity is equal to the sea surface sound velocity is called the critical depth and the water column from the surface to the critical depth is defined as the Sound Fixing and Ranging (SOFAR) channel. The axis of the channel is at the sound velocity minimum, normally located around 1 km water depth in the tropics and midlatitudes (Figure 2b). Steeply emergent rays (i.e., rays with a low incidence angle, which is measured from the vertical) and any energy originating below the critical depth will reflect and scatter from the sea surface and the seafloor and will not travel far horizontally before being attenuated by propagation scattering processes (red dashed ray, Figure 2c). In contrast, low grazing angle (high incidence angle) energy will become totally trapped in the SOFAR channel. This energy will then propagate efficiently because the anelastic attenuation in the water column is practically negligible [Talandier and Okal, 1998] and energy divergence in the channel is cylindrical, compared with spherical divergence for body waves (r−1 compared with r−2 respectively). The colored cones in Figure 2c represent the effective energy or range of ray angles which will be totally trapped in the water column from omnidirectional scattering at various seafloor depths. The effective energy decreases with increasing water depth to zero at the critical depth and below. Earthquakes that occur in regions where the water depths are below the critical depth should have strongly attenuated T-phases (Figure 3c).

Figure 3.

Illustrations of coupling mechanisms that may generate T-phases. (a) Downslope propagation and (b) rough seafloor scattering. (c) Events that occur below the critical depth should not be able to propagate in the SOFAR channel based on ray trace theory (Figure 2c). (d) A local bathymetric high located in the SOFAR channel axis may act as a radiator, where acoustic energy from a deep water event enters the SOFAR channel efficiently and results in T-phase event locations that are not epicenters.

The “T-phase problem” is that models of T-phase excitation produce acoustic energy with low incidence angles, due to the sound velocity contrast between the seafloor and water column. Several mechanisms have been proposed for the conversion from low incidence to high incidence energy. For earthquakes which insonify sloping seafloor, the T-phase energy is converted into high incidence angles through multiple reflections and downslope propagation [Johnson et al., 1963; Chiu, 1994; Talandier and Okal, 1998] (Figure 3a). In deep water, scattering from the rough seafloor [Walker et al., 1992; de Groot-Hedlin and Orcutt, 1999, 2001; Park et al., 2001; Yang and Forsyth, 2003] (Figure 3b), sea surface [Johnson et al., 1968], or the underside of sea ice [Keenan and Merriam, 1991] may excite high incidence angle energy. It has also been suggested that scattering (or mode coupling) at internal waves and SPICE (density neutral thermohaline variations) [Butler, 2004; Colosi, 2004], can excite high incidence angle propagation [Park et al., 2001] and these processes may cause propagation scattering as well.

The complexity of T-phase excitation makes it difficult to infer hypocenter locations, magnitudes and focal mechanisms from the characteristics of the T-phase. For example, the peak of a bathymetric high may act as a radiator of acoustic energy for local seismic events (Figure 3d). As a result, the T-phase location does not coincide with a point on the seafloor directly above the earthquake hypocenter. There have been several efforts to derive a relationship between seismic magnitude and acoustic magnitude [Dziak et al., 1997; Dziak, 2001; Fox et al., 2001; Bohnenstiehl, 2001; Pulli and Upton, 2002; Yang and Forsyth, 2003; Pan and Dziewonski, 2005]. There are several problems, however, because in the past large magnitude earthquakes were clipped by the hydrophone sensors (mb > 4.7 [Fox et al., 2001]) making the estimate of T-phase magnitude unreliable (this problem is being eliminated with the larger dynamic range in more recent hydrophone systems). Another problem is that small magnitude events (mb < ∼4) recorded by the hydrophones are not recorded by land-based seismometers and therefore do not have a calculated seismic magnitude. Understanding the relationship between seismic magnitude and acoustic magnitude will require a concurrent hydrophone and ocean bottom seismometer (OBS) study.

In addition to quantifying acoustic magnitude, variations in risetimes have been used for geological interpretation of T-phase events. Risetime is defined as the time between the appearance of the T-phase and its maximum peak and it has been related to the focal depth of the earthquake [Norris and Johnson, 1969]. For example a shorter risetime may reflect a shallower earthquake and longer risetime may mean a deeper earthquake. The combined observations of migration in seismic activity, a decrease in risetime and acoustic magnitude of T-phase events were interpreted as a shallowing magma dike intrusion at Axial [Dziak and Fox, 1999b] and CoAxial [Schreiner et al., 1995] segments of the Juan de Fuca Ridge and the Lucky Strike segment of the MAR [Dziak et al., 2004].

Finally, T-phases are commonly assumed to be blocked by bathymetry along the propagation path [e.g., Fox et al., 2001], but little work has been done to quantify and model the effect. Pulli and Upton [2002] investigated a T-phase event recorded by the Diego Garcia hydroacoustic array which was associated with a continental earthquake and its aftershocks in India. T-phase acoustic magnitudes in the blocked (or shadow) region of the islands were reduced by ∼40 dB confirming a transmission loss along the propagation path but not total blockage. Modeling and accurate assessment of bathymetric blockage is hindered by a lack of high-resolution bathymetry over large areas of seafloor [Harben et al., 2003].

3. Study Areas

The MAR between 35°N and 15°N is a slow spreading (∼12 mm/yr, average half spreading rate) mid-ocean ridge (MOR) separated into multiple segments by large offset transform faults and nontransform offsets (Figure 1). Both of our study areas are located at a ridge-transform intersection (RTI): the eastern intersection of the MAR with the Atlantis TF (Atlantis study area, Figure 4a) and the eastern intersection of the MAR with the Kane TF (Kane study area, Figure 4b).

Figure 4.

Locations of the maps in Figures 4a and 4b are shown in Figure 1. (a) Bathymetric map of the Atlantis study area with T-phase events represented by the red circles and the error in their locations (provided in the T-phase catalog) shown by the red bars. The ICH massif may have formed by movement on a low angle detachment fault that first nucleated along the break-away. The termination is the present-day contact between the footwall and the hanging wall of the fault. The cross section through the ICH massif illustrates the inferred low angle detachment fault and the unroofing of lower crustal (gabbros) and upper mantle (peridotites) rocks to form the summit. Volcanic seafloor (basalts) makes up the hanging wall and covers the axial zone. (b) Bathymetric map of the Kane study area with T-phase event locations and errors as in Figure 4a. The locations of 2 teleseismic events recorded by land-based networks that occurred during the period of this study are indicated by blue triangles. A third teleseismic event is located to the north, but its T-phase is located within our study area. Teleseismic locations and corresponding T-phase event locations are connected by dashed blue lines. The three T-phases associated with these teleseismic events are labeled with their event number and shown by the yellow circles. The black open circles surround two clusters of three T-phase events which are shown in Figure 7. The bathymetry data for both study areas were obtained from the Ridge Multibeam Synthesis Project Web site (http://ocean-ridge.ldeo.columbia.edu/general/html/home.html). The bathymetry contour interval is 200 m in both Figures 4a and 4b.

The typical features of a RTI include a deep valley at the intersection of the ridge and transform (known as a nodal basin), a high massif located at the corner of the RTI toward the offset ridge axis (the inside corner high (ICH)), and relatively lower topography on the other side of the ridge segment (the outside corner high (OCH)) [Severinghaus and Macdonald, 1988]. It is thought that the ICH is composed of lower crust and upper mantle rocks that have been unroofed along a low-angle detachment fault [Mutter and Karson, 1992; Tucholke and Lin, 1994; Escartin and Lin, 1995; Tucholke et al., 1996, 1998; Cann et al., 1997; Blackman et al., 1998, 2002; Canales et al., 2004]. Both study regions have approximately the same spatial area (∼56 by 50 km) and are centered on the ICH massif.

3.1. Atlantis Study Area

The Atlantis TF is a 75 km long, left lateral offset fault (Figure 4a). The ICH massif, located at the eastern end of the RTI, has a smooth domed shape and a corrugated surface and rises to ∼0.7 km water depth at its southern edge [Cann et al., 1997; Blackman et al., 1998, 2002]. The nodal basin has water depths of ∼4.6 km, on average. This results in a change in water depth of ∼4 km over ∼20 km lateral distance from the summit of the ICH to the valley of the nodal basin. Serpentinized upper mantle rocks (serpentinized harzburgites) have been sampled from the summit of the Atlantis ICH massif and along its eastern slope [Blackman et al., 1998]. Extensive mass wasting has shaped the south slope of the massif leaving large scarps and talus slopes [Blackman et al., 1998]. The southern slope is associated with hydrothermal activity. A serpentinite-hosted, carbonate vent field, “Lost City”, is located near the top of the steep south slope [Kelley et al., 2001].

3.2. Kane Study Area

The Kane TF is a 150 km long, right lateral offset fault (Figure 4b). The nodal basin has water depths that exceed ∼6 km and the top of the ICH rises to a water depth of ∼1.2 km (∼4.8 km difference in water depth over ∼20 km lateral distance). Gabbroic rocks have been sampled from the top of the Kane ICH and along the eastern slopes [Karson, 1998]. Serpentinite has not been sampled at the Kane study area and there are no corrugations on the summit of the Kane ICH. In addition, no hydrothermal activity has been found, to date, associated with the Kane ICH. The closest known hydrothermal activity is the “Snake Pit” sulphide vent field located on the neovolcanic ridge within the valley floor (<5 kyr old crust) [Brown and Karson, 1989; Lalou et al., 1993].

3.3. Seismicity

Both study areas have a large number of T-phase events located on the summit and flanks of the ICH compared with the nearby ridge axis, transform fault, and OCH (Figures 4a and 4b). Previous seismicity studies of mid-ocean ridge transform and nontransform offsets have observed clustering of earthquakes on the ICH (nontransform offset at 29°N [Wolfe et al., 1995], Oceanographer TF (35°N [Rowlett, 1981]), St Paul's TF (0°40′N, [Francis et al., 1978]) and Vema TF (10°45′N [Rowlett and Forsyth, 1984]) on the MAR and Rivera TF (19°N [Reid, 1976]) on the East Pacific Rise (EPR)). The source of the seismicity may be associated with movement on the inferred detachment fault, along which the lower crustal and upper mantle rocks have been exposed (cross section shown in Figure 4a). Oblique faults with normal [Rowlett, 1981] and reverse focal mechanisms [Engeln et al., 1986; Behn et al., 2002], and faults which form in response to continuous rotation of the footwall of the detachment fault [Tilmann et al., 2004] might also play a role in generating seismicity. At the Atlantis ICH seismicity may be associated with the increase in volume and heat generation during the process of serpentinization, which may cause flexing and brittle failure as well as fluid expulsion events [Kelley et al., 2001].

4. Hydrophone-Recorded and Teleseismic Data

The NAHA consists of six autonomous hydrophone moorings that straddle the MAR from 35°N–15°N, as shown in Figure 1 [Smith et al., 2002, 2003]. We refer to the hydrophones as Northeast (NE), Northwest (NW), Central East (CE), Central West (CW), Southeast (SE), and Southwest (SW), reflecting their relative positions within the array. The moorings are anchored to the seafloor and the hydrophones float in the SOFAR channel. The hydrophones continuously and autonomously record low frequency energy propagating through the SOFAR channel and the data are retrieved annually when the hydrophones are recovered and redeployed. The National Oceanic and Atmospheric Administration (NOAA) Pacific Marine Environmental Laboratory (PMEL) processes the data to produce a T-phase catalog containing event time, number and identity of the hydrophones that recorded the event, location in latitude and longitude (and errors) and acoustic magnitude (and error). The data are available at http://www.pmel.noaa.gov/vents/data/index.html, and detailed descriptions of the hydrophones and PMEL processing methods are given by Fox et al. [2001], Dziak [2001], and Smith et al. [2003].

The NAHA was first deployed in February 1999 and was recovered in 2005. The data used in this report were recorded between 25 February 1999 and 9 March 2001. During this period all six instruments in the array functioned properly. The NW hydrophone results were removed from our analyses, however, because this hydrophone experienced high ambient noise levels due to strumming of the mooring line. As a result, many of the T-phase events cannot be distinguished in the time series or the spectrogram.

In total 95 events were located in the Atlantis study area and 63 in the Kane study area during the two year period. Tables 1 and 2 list the T-phase catalog information for all the events at Atlantis and Kane, respectively. We make the following assumptions about the T-phase catalog event locations for this study: (1) the T-phase event locations are accurate, and (2) the locations represent earthquake epicenters.

Table 1. Atlantis Study Area T-Phase Events
Event NumberYearDayHourMinSecs x10No. of HydaHyd IDbLon, °ELat, °NSource Level, dB re: 1μPa@1mError TimeError Lon, °EError Lat, °NError Source Level
  • a

    No. of Hyd: total number of hydrophones that recorded the event.

  • b

    Hyd ID: identity of the hydrophones (1-NW, 2-CW, 3-SW, 4-SE, 5-CE, 6-NE).

119996385555556243−42.130.12202.060.3750.010.012.6
219997021102376156243−42.1430227.20.6240.010.023
319997310471806156243−42.2729.93211.690.4320.010.019.2
419997321285156156243−42.229.91210.990.460.010.015.7
519998375445441564−42.0930.12209.110.049004.5
61999851853146556243−42.0630.14210.480.340.010.014.2
719991002255197556243−42.0930.12204.670.3460.010.012.4
81999115744308515243−42.0330.31205.850004.1
919991171620696156243−4230.33203.30.50.010.012.9
10199911819321376156243−41.9630.34211.611.1020.020.033.1
1119991191559318515624−42.3130.05201.372.0870.030.062.2
121999190735106515643−41.9230.33205.190.4520.010.011.6
13199920018415546156243−42.0529.96211.320.330.010.014.2
1419992052203513564−42.1130.12202.245.8990.070.120.8
151999209172930515624−42.0129.97206.150.4810.010.012.6
161999209198215515624−42.1729.95202.550.5250.010.012.4
171999218176381515624−42.229.91215.081.350.020.042
1819992307201606156243−42.1129.94208.270.7960.010.024.3
19199923011434756156243−42.0830.11223.360.18500.0112.6
2019992301215876156243−42.0330.14208.610.19100.016.8
2119992311543116156243−42.0730.11219.390.4060.010.014.3
2219992312236355515624−42.130.14208.470.8010.010.022.8
2319992312246355515624−41.9830208.290.9150.020.036.4
24199923317322236156243−41.9729.91232.750.7310.010.022.4
25199923320173276156243−42.1830212.830.4170.010.013.8
2619992332136766156243−42.1830.01212.420.3050.010.014.1
271999234928145515643−42.2230207.130.7480.010.0210.3
2819992349282646156243−42.1830.022050.9710.020.039.7
2919992390113686156243−42.2530.03211.740.6420.010.026.3
30199924018452276156243−42.0929.93221.510.7670.010.027.7
31199924019453076156243−42.0230.14224.050.2200.015.2
321999259114246741564−42.0930.09211.321.2140.020.045.5
33199926112451245243−42.1530.03202.620.9150.010.014.7
34199926353026541564−42.2530.08208.180.7640.010.029.5
351999266333226156243−41.9330.13209.871.0140.020.033.9
3619992791531259556243−42.1829.91200.880.50.010.011.6
371999314175119515624−42.1130.17208.150.9110.020.033.7
381999314175119515624−42.230208.150.9110.020.033.7
3919993191954516156243−42.230204.540.4240.010.012.8
40199932215134976156243−42.1130.12215.390.3720.010.015.3
4119993231562756156243−42.130.13209.730.004003.6
42199933718455836156243−42.0730.12203.070.6710.010.021.8
43199934122304106156243−4229.95214.720.710.010.024.2
441999347824606156243−42.1630.04210.630.920.020.034.6
45199934917314946156243−42.1429.95210.080.9940.020.037.3
46199935023251206156243−41.9630.3219.390.5380.010.025.4
47199935912404426156243−42.1930.03212.521.2880.020.043
482000101112186156243−42.3330.1206.111.0150.020.032.7
4920001817391186156243−42.1530219.431.0810.020.035.9
50200035232369556243−41.9729.92217.011.3360.020.025.4
51200045955936156243−42.1130.14225.610.6490.010.026.2
5220004743837341524−42.2830.04220.410.7050.010.0312.8
53200063859507556243−42.0930.09212.595.3840.10.092.7
542000905114726156243−42.2630.23224.221.2770.020.043.3
5520009111412866156243−42.2629.98227.391.2620.020.035.4
5620009115264016156243−42.1929.95231.283.9210.060.114.8
572000926756515624−42.1530.01211.480.3830.010.011.8
582000105192350741524−42.2129.94206.412.7350.040.13.5
59200012139308515243−42.1829.96209.113.7260.050.131.1
6020001448238345624−42.130.28212.761.1530.020.023.4
612000157656166156243−42.1330.1221.040.7660.010.025.8
622000173114122041562−42.1330.12209.690.5010.010.014
6320001731157544515624−4230.33206.421.0670.020.032.4
64200018834916156243−41.9330.28218.10.22800.013.4
6520001731157495515624−41.9930.28207.690.9050.020.032.3
6620001731157516515624−41.9830.35207.330.004002.7
672000234193853241524−41.8830.13199.420.510.010.023.5
6820002376531756156243−42.130.15221.490.530.010.013.2
6920002551128400515624−42.1329.96210.980.9710.020.032.5
702000256328041564−42.1430.01206.860.18200.012.3
71200025616341041564−42.0429.94203.931.5880.020.052.1
722000256163517441564−42.1229.95204.591.1550.020.041.8
7320002853113736156243−41.9729.96210.430.7250.010.023.4
74200028744526041524−41.9729.93202.440.026000.5
752000288212038941564−41.9930.28209.540.6470.010.022.7
76200028822238041564−41.9330.25207.830.6140.010.022.1
77200028822349841564−41.8930.25206.310.8980.010.031.5
78200028917323896156243−41.9730.29216.480.3530.010.014.5
792000289203817041564−41.8830.29202.120.7330.010.022.8
802000296020196515624−41.9330.28209.290.4330.010.011.4
81200029761016641543−42.2530.04214.650.3190.010.024.3
82200029761039841543−42.330.09215.811.1630.020.071.7
8320002971259251515624−42.3130.04205.880.6410.010.021.6
8420002985383241524−42.2530.03202.711.0190.010.041.1
8520002992116333515624−42.0730.1205.910.8750.010.022.8
862000320358259515624−41.9130.27202.60.7690.010.022.3
872000320924368515624−42.129.99205.210.5670.010.021.5
8820003231947377515624−41.9629.95202.470.002001.8
89200033718451456156243−42.2929.99207.250.4670.010.011.8
9020003483115426156243−42.2730.05203.790.2700.011.8
9120011411113516156243−41.9630.04205.650.3440.010.011.4
9220011919235406156243−41.9929.9202.960.129001.3
932001392028141524−42.0930.12204.490.7240.010.032.3
942001397151646156243−42.1330.01204.690.1900.011.5
952001397151646156243−42.1630.16204.690.1900.011.5
Table 2. Kane Study Area T-Phase Events
Event NumberYearDayHourMinSecs x10No. of HydaHyd IDbLon, °ELat, °NSource Level, dB re: 1μPa@1mError TimeError Lon, °EError Lat, °NError Source Level
  • a

    No. of Hyd: total number of hydrophones that recorded the event.

  • b

    Hyd ID: identity of the hydrophones (1-NW, 2-CW, 3-SW, 4-SE, 5-CE, 6-NE).

119995895580552346−45.16923.586204.10.0060.0060.36.6
219996612250942346−45.12423.523205.540.0120.0150.724.2
319996618192552346−44.99723.349203.550.0240.0231.155.1
4199968128521552346−45.12123.54206.670.0060.0050.269.5
51999741243545552346−45.12823.538209.150.0030.0030.1413.1
61999876501242341−44.8223.45203.570.020.0331.422.3
719991091835456552346−45.00723.377219.580.0130.0120.66.1
819991091837383552346−44.98923.33215.960.0110.010.518.1
91999111434214552346−45.1223.539207.420.0050.0050.234.3
1019991252115330552346−44.87123.577212.630.0220.0190.995.1
11199914320554542346−45.12423.517207.730.0080.010.512.5
12199915143397552346−45.1723.582211.310.0090.0080.48.9
1319991526418645234−45.15623.533205.060.0140.0120.654.7
141999159312554552346−45.07923.472206.30.0160.0140.742.1
15199916602846545346−45.17923.683209.710.0480.0582.578
16199919702045745346−44.95823.417209.280.0010.0020.071.5
1719992123749845346−45.05323.676203.640.0250.031.324
1819992153630445346−45.14623.556208.240.0240.031.39.7
191999247355252552346−45.09423.513207.440.0170.0160.823.9
201999265031504552346−45.0423.434204.130.0120.0110.560.9
2119992802328045346−45.15123.565207.910.030.0371.575.2
22199928023324552346−45.11223.506201.890.0010.0010.053.3
2319993036210552346−44.95423.588215.750.0160.0140.742.3
2419993051341339552346−45.18223.582202.240.0070.0060.332.5
25199932223242376523416−45.08323.539210.930.0090.0090.468.1
26199933533216552346−45.24323.579210.590.0210.0190.984.3
2719993352027411552346−45.14823.566211.70.0090.0080.428.6
281999340238441552346−45.10323.559203.330.0060.0060.32.1
29199935218338345234−45.09123.578205.680.0030.0030.144.1
302000317401616523416−44.81623.506217.310.0240.0241.194.4
3120001022334346523416−45.11923.573218.090.0050.0050.236.7
322000131426106523416−45.11323.61234.120.0170.0170.846.8
332000311026598552346−45.20923.604207.80.0140.0130.675
34200033327106552346−44.92923.373205.670.0210.021.013.3
3520005214353946523416−45.12423.673222.930.0030.0030.164.7
36200070164857145234−45.0223.689209.070.0350.031.63.5
37200088233049545234−44.90923.608203.010.1090.0955.043.8
3820008920559745234−45.0823.678204.650.0530.0452.422.9
39200090233180552346−45.23723.646205.030.0350.0311.63.7
40200096166125552346−45.03823.488219.880.0380.0351.81.7
412000961615350552346−44.98423.456221.890.0280.0251.33.3
42200096172438445234−45.03923.396211.770.0220.021.062.3
432000984265466523416−45.0923.424222.130.0410.0422.066.2
4420009851228645234−45.09123.406208.980.0240.0221.143.3
45200010274646345234−44.91223.564206.260.0390.0341.792.4
4620001280268945234−44.95823.666209.810.0040.0040.24.6
47200015412175543526−45.05923.31199.660.1050.0123.921.9
4820001641950502552346−45.14323.626210.540.0150.0140.722.4
49200018832543245234−45.14423.573206.720.0070.0060.331.9
50200022719714552346−45.05723.45206.360.0310.0281.452.2
51200023355510552346−45.01523.325213.080.0310.0291.463.7
522000258135539245346−45.19823.639209.80.0410.0512.232.2
53200028773445552346−45.13523.577205.120.0130.0120.63.5
5420002877131533526−44.94823.74204.970.0850.0112.742.1
5520002877522726523416−45.11823.553209.880005.5
562000310654253552346−45.14323.577205.280.0130.0110.591.9
57200111824217552346−45.09923.535209.310.0030.0030.138
582001131126260552346−44.99623.355202.780.0020.0020.112.3
592001161136367552346−45.02123.608204.990.0060.0060.292.5
6020012624129745234−44.95223.557203.010.0120.0110.582.4
61200137934542552346−45.00223.351207.470.0020.0010.072.6
622001631254401552346−45.11223.673204.880.0120.0110.553
6320016873025845234−44.80423.676202.226E-045E-040.031.8

Searches for all teleseismic events at the two locations during the two year time period were made on the Harvard Centroid Moment Tensor (CMT) Catalog (http://www.seismology.harvard.edu/CMTsearch.html), the National Earthquake Information Center (NEIC) Database (http://neic.usgs.gov/neis/epic/epic_rect.html) and the International Seismology Center (ISC) which includes data from the International Database Center (IDC)/European International Data Center (EIDC) (http://www.isc.ac.uk/Bulletin/arrivals.htm). Three events were large enough to be recorded both as teleseisms and T-phases. All three teleseimic events were located in the Kane study area (Figure 4b) but no focal mechanism solutions are available for these events from the Harvard CMT catalog.

5. Results and Interpretation

In the following section, the T-phase data from the Atlantis and Kane study regions are compared by geological setting (e.g., top of massif, transform valley, ridge axis). The data for each study area are also combined to allow for more robust comparison with the ray trace modeling results and to assess trends in T-phase character with varying event water depths, source to receiver distances and onset time. Details of the data processing, picking of events, modeling and curve fitting are provided in Appendices ATransmission Loss CurvesCurve FittingD.

5.1. Spatial Distribution of the T-Phases

The average number of events per unit area has been calculated for each of the study areas and normalized to 1000 km2. At the Atlantis study area there are ∼35 events/1000 km2 and at Kane there are ∼22 events/1000 km2. By comparison, the MAR segments adjacent to our study areas, segment 34 for Atlantis and 17 for Kane (as defined by Smith et al. [2003]) both fall within the “medium” T-phase activity group (∼10 events/1000 km2) over the same two year time period [Smith et al., 2003]. This difference in the density of the T-phase events implies that there are significantly more seismic events associated with the RTIs in our two study areas compared with the ridge axis.

The spatial density of T-phase events in the study areas was also examined as a function of the event location water depth. Water depth of an event was obtained by sampling a multibeam bathymetry grid (data spacing 150 m) at the latitude and longitude of the event location. Event water depths for those events with T-phase location errors <0.03° were sorted into 1 km sized bins and the seafloor surface area calculated within each bin (81 events from Atlantis and 48 events from Kane). The normalized number of T-phase events per 100 km2 for each bin shows that a greater number of events are located in water depths shallower than 2 km (i.e., the shallow summit and flanks of the ICH) compared with deep water events along the transform and ridge axis in both survey areas (Figure 5). We suggest three possible reasons that could explain the large number of events located on the ICH:

Figure 5.

Histograms of the T-phase spatial density with increasing water depth for (a) Atlantis and (b) Kane study areas. Note the greater number of events located on the summit and flanks of the ICH massif (between 1 and 2 km water depth) compared with all other water depths at both study areas.

1. The ICH acts as a radiator of seismic energy from events located beneath the ridge axis or transform fault (Figure 3d). Seismic energy from the events propagate through the ICH crust and couples with the water column at the ICH summit, which intercepts the axis of the SOFAR channel in both study areas.

2. The same number of events occur at all water depths, but a larger number of events are recorded from the summit and the shallow flanks of the ICH because the energy couples more easily into the SOFAR channel, as in (1) above.

3. A greater number of seismic events occur at shallow water depths compared with deep water depths and are associated with the building and maintenance of the ICH massif.

5.2. Amplitude of the T-Phases

The amplitude of a T-phase event is the first characteristic we quantify from the T-phase time series. Our definition of the amplitude of a T-phase is different from the T-phase catalog. The value of acoustic magnitude (called Source Level) provided in the T-phase catalog measures the amplitude of the broadest band portion of the T-phase spectrogram. This amplitude, recorded by the hydrophone, is corrected for instrument gain and transmission loss over the source to receiver distance to give a magnitude ∼1 m above the seafloor at the event location. The Source Levels for all the hydrophones that record the T-phase are averaged to give a mean Source Level for each event. The range in Source Level for the Atlantis study area is 199.42–232.75 dB re: 1μPa @ 1m and 199.66–234.12 dB re: 1μPa @ 1m for the Kane study area.

In this study we calculate the Received Level, which is the Root Mean Square (RMS) of a 90 second window of the time series containing the T-phase event. Ambient noise levels are quantified from the RMS of the first 10 seconds of each 90 second window of the time series, with the assumption that this is “quiet time” before the arrival of the T-phase. We then select T-phase events with locations errors <0.03° and a RMS Signal >1 dB re: counts (72 events from Atlantis and 39 events from Kane, see Appendix A). One value of Received Level is calculated for each event at each hydrophone. The Received Level calculation was done in counts and the results converted to decibels (decibel conversion, dB re: counts = 20 log 10 (Amplitude)). The variability in Received Level between hydrophones for the same event is expressed by the horizontal range in values for each event in Figure 6. To summarize, the difference between the Received Level (used here) and Source Level (given in the T-phase catalog) is that Source Level is the magnitude of the T-phase ∼1 m above the seafloor at the event location while the Received Level is the magnitude at the hydrophone (with no range correction).

Figure 6.

Plot of T-phase Received Level versus event location water depth for events at (a) Atlantis and (b) Kane. The different colored circles represent the 5 different hydrophones. The range in Received Level is ∼5–30 dB re: counts for all water depths at both study areas. This result is contrary to ray trace theory which predicts decreasing Received Level with increasing water depth.

A comparison of hydrophone catalog Source Level, mean Received Level (a geometric mean calculated in dB re: counts), and seismic event magnitude (mb), for the three teleseismic events recorded at the Kane study area is shown in Table 3 and the event locations are labeled in Figure 4b. The results highlight a difference between seismic body wave magnitudes and T-phase amplitudes: Event 32 has a significantly higher mean Received Level (and Source Level) than events 41 and 43 but the lowest mean body wave magnitude of the three events. The discrepancy between the magnitude values may be due to different water depths and topography at the event locations. Event 43 is the shallowest of the three (water depth of 2.96 km) and is located on the broad southern side of the ICH. Event 32 is at a water depth of 3.36 km and located on the steep northern side of the ICH. Event 41 is the deepest with a water depth of 4.18 km and located within the inner valley floor. The comparison between the seismic magnitude and Received Level for the three teleseismic events suggests that the relationship between seismic and Received Level is more complex than we currently understand. However, the small number of events makes it difficult to draw conclusions from our comparison. It is surprising that there were no teleseismic events located in the Atlantis study area because seven events have a mean Received Level greater than Event 41 in the Kane area. This suggests that the relationship between seismic and acoustic magnitude may be location specific.

Table 3. Teleseismic Events From the Kane Study Area
Event NumberaISC, mbIDC/EIDC, mbNEIC, mbMean mbMean Received Level, dB re: countsSource Level. dB re: 1μPa @ 1mEvent Location Water Depth, km
324.14.04.04.0327.78234.153.358
414.23.94.44.1717.66221.894.177
434,13.94.54.1718.3222.132.966

5.3. Received Level of T-Phases Versus Event Location Water Depth

Ray trace theory predicts that the range of ray angles that can propagate through the SOFAR channel decreases with increasing water depth, approaching zero at the critical depth (Figure 2c). If we assume that the range of ray angles can be related to the T-phase Received Level we would expect the Received Level to decrease with increasing water depth. However, this trend is not observed at either study region (Figures 6a and 6b). Instead, the spread in Received Levels recorded by the hydrophones for one event appears to remain fairly constant, ∼5–30 dB re: counts for both study areas, at all water depths (a similar trend is observed for Source Level as well). We assume that all hydrophone-recorded events have magnitudes <4.0 mb, (except the 3 teleseismic events at Kane) because they were not listed in the teleseismic earthquake catalogs. We also assume that most of the T-phase events have magnitudes >3 mb because 3 mb is the level of completeness for the NAHA [Bohnenstiehl et al., 2002]. Therefore the range of earthquake magnitudes for all our events has relatively small scatter (between ∼3–4 mb) and variable earthquake magnitudes are unlikely to be the source of the range of Received Levels between T-phase events. We confirm this assumption by comparing groups of events with similar Received Level and Source Level but no trends with changing water depth were observed. To highlight the variability in Received Level for T-phase events located at the same water depth we compare the spectrograms from two clusters of 3 events located at ∼1.2 km (Figures 7a–7c) and ∼4 km water depth at Kane (Figures 7d–7f). The locations of these two clusters are outlined by the open black circles in Figure 4b. Within each of the two clusters the T-phase event time series and spectrograms look similar but the hydrophones record a range of Received Levels (dB re: counts values shown in the large black text in Figure 7).

Figure 7.

Time series and spectrograms of the two clusters of events identified by the open black circles in Figure 4b. All six columns show data for the CW-SW-SE-CE-NE hydrophones from top to bottom. (a–c) Time series and spectrograms for events 4, 5, and 9 located at the top of the ICH massif at ∼1.2 km water depth. The T-phase Received Level at the hydrophones (in dB re: counts, labeled by black text) varies for the three events, but the appearances of their T-phase time series and spectrogram are similar. All three have short onset times (defined as the time interval between the appearance of the T-phase envelope above the ambient noise and its first peak). (d–f) Time series and spectrograms for events 3, 58, and 61 located near the neovolcanic zone at ∼4 km water depth. These events also have varying Received Levels, but the T-phases have longer onset times than events 4, 5, and 9 above.

Ray trace theory also predicts that energy from events which occur at water depths below the critical depth will not propagate to the hydrophones in the SOFAR channel (Figure 2c), however several T-phase events are located below the average critical depth in both our study areas (∼4.5 km water depth). One explanation for the hydrophones recording these deep water events may be insonification of large areas of seafloor by earthquakes. Ray trace modeling of an earthquake with a hypocenter at 4 km depth in the crust shows that energy insonifies a seafloor area of ∼20 km radius [Stephen et al., 2002]. The source of the T-phase therefore may be a large area of the seafloor that includes water depths above and below the critical depth (a broad area of insonification is also discussed by de Groot-Hedlin and Orcutt [1999, 2001] and Yang and Forsyth [2003]). Until the mechanisms of seismic energy coupling into the water column and conversion to acoustic energy are understood (whether it can be explained by ray or wave theory) this problem will remain.

5.4. Received Level of T-Phases Versus Distance From Event to Hydrophone

Received Levels decrease with increasing source to receiver distance, as shown by the mean Received Level (black solid line) in Figures 8b and 8d. This result is expected because T-phase energy is reduced by transmission loss, attenuation and scattering along the propagation path. Transmission loss curves are calculated for each study region by summing attenuation due to cylindrical spreading and intrinsic attenuation in the water column (Figures 8a and 8c; see Appendix B). The curves are then used to correct for transmission loss at each hydrophone relative to the hydrophone closest to the event (i.e., the transmission loss at the furthest hydrophone is maximum and zero at the closest hydrophone). The mean Received Level (black line in Figures 8b and 8d) is then corrected for transmission loss so as to reduce the number of processes that affect the T-phase as a function of distance from the event to the hydrophones. The correction flattens the trend of mean Received Level versus distance (red solid line in Figures 8b and 8d) and highlights the relative differences in mean Received Level recorded by each hydrophone.

Figure 8.

(a and c) Plots show transmission loss curves due to cylindrical spreading and intrinsic attenuation from each study area to the hydrophones. We define the transmission loss to be zero at the hydrophone closest to the event and a maximum at the furthest hydrophone. (b and d) Received Level versus distance from the event to the hydrophones for the Atlantis and Kane study areas respectively. The solid black line represents the mean Received Level (in dB re: counts) for each hydrophone. The mean Received Level is corrected for transmission loss and shown by the red line. The vertical colored bars at each hydrophone represent one standard deviation from the mean Received Level. Events from the Atlantis region have the highest Received Level at the CE and SE hydrophones and the lowest at the CW and NE. Events from the Kane region have the highest Received Level at the CE and NE hydrophones and the lowest at the CW hydrophone.

We believe the corrected mean Received Level (red solid line in Figures 8b and 8d) may reflect bathymetric blockage along the propagation path for different source and receiver pairs. The trend of the corrected mean Received Level is different for each study area. Events from the Atlantis study area have a corrected mean Received Level which varies by ∼5 dB re: counts between hydrophones (Figure 8b). The CW hydrophone records events that have the lowest magnitudes, closely followed by the NE hydrophone. The SE hydrophone records events with the highest magnitudes and the CE and SW are intermediate. The greatest difference in magnitude is between the two southern hydrophones, which have only 50 km difference in the length of the propagation path but have an average difference of ∼4 dB re: counts. The corrected mean Received Level from the Kane study area varies by ∼8 dB re: counts between hydrophones (Figure 8d). As with the Atlantis events, the CW hydrophone records the lowest magnitudes for the Kane events but the NE hydrophone records the highest (however, there are few data points for the NE hydrophone). Unlike the Atlantis events, the Kane events have approximately equal magnitudes at the two southern hydrophones. These results and their significance for bathymetric blockage are discussed in more detail in section 5.6.

5.5. Onset Time

The onset time is defined as the time interval from when the T-phase envelope or lens emerges above the ambient noise to the first peak in the T-phase (Figure 2a). We quantify the onset time following the curve fitting method of Yang and Forsyth [2003] (details can be found in Appendix C). The onset time was only calculated for those events with a signal-to-noise ratio (SNR) >2 and with latitude and longitude errors <0.03° (27 events at Atlantis and 17 events at Kane) so as to exclude those events with anomalously high ambient noise levels.

The first peak in many of the events in our study area is not the maximum peak of the T-phase time series, and thus the onset time differs from the “risetime” [Schreiner et al., 1995; Dziak and Fox, 1999b] which is defined as the time from the emergence of the T-phase to its maximum peak. T-phase locations are traditionally based on either the arrival time of the largest amplitude of the smoothed envelope of the time series or on the arrival time of the broadest spectral content in a spectrogram. For simple lens-shaped arrivals these are usually at the same time (for example the two events shown in Figure 9). For more complicated arrivals with multiple peaks (for example, events 4, 5 and 9 shown in Figures 7a–7c) the broadest spectrum may not correspond to the peak of the envelope or multiple peaks may have indistinguishable amplitude and spectral content. As we do not understand the mechanism for multiple peaks we focus only on the first peak in the T-phase.

Figure 9.

Examples of the curve fitting process used to calculate the onset time of the T-phases (the time interval between the appearance of the T-phase envelope above the ambient noise and its first peak). A short onset time event in shown in the left column, and a long onset time event is shown in the right column. (a) Time series of an event recorded at one hydrophone. The time series is band pass filtered in an octave centered on 16 Hz. (b) A complex envelope of the time series is calculated and smoothed with a one second running average (black line). (c) The log of the smoothed complex envelope with a fitted curve (red line). The value of the onset time, tb, is shown in seconds. (d) Spectrogram of the event.

We quantify the onset time by fitting a curve to the log of the complex envelope of the event (Figure 9, Appendix C). We then select and empirically define a critical onset time, tc, based on our visual inspection of the T-phase coda. A T-phase event that has onset times <tc, for all the hydrophones which record the event are termed “short onset” and onset times >tc “long onset” events. The tc is 4 seconds for the Atlantis and 6 seconds for the Kane study areas. Those events with hydrophone onset times that bridge the critical onset time are termed “mixed onset”.

Our results show that all of the short onset events in both study areas are located on the top and slopes of the ICH in water depths <2 km (Figure 10). All of the long onset events are located on the seafloor in water depths >2 km. The mixed onset events occur in both shallow and deep water. This is strictly an empirical observation but the correlation between onset time and event water depth is remarkable. At this time we do not have a physical explanation for this correlation. It is interesting to note that all of the events with mixed onset at the Kane were due to a long onset time recorded by the CE hydrophone alone. Only two of the mixed onset events at Atlantis, located at the top of the massif, were due to a long onset time for the CE hydrophone alone. The CE hydrophone is the closest hydrophone to the Kane study area (∼460 km) and the second closest to the Atlantis study area (∼560 km). We therefore conclude from this that length of the propagation path does not influence the onset time of the event.

Figure 10.

Map showing the spatial distribution of onset times for (a) Atlantis and (b) Kane. The critical onset times at Atlantis and Kane are 4 seconds and 6 seconds, respectively. The white circles represent events for which arrivals at all the hydrophones were shorter than the critical onset time, and the black circles represent all arrivals longer than the critical time. Those events that have arrivals which bridge the critical onset time are shown in red. The general trend in the two study areas is short onset events at water depths <2 km and long onset events at water depths >2 km. Only those events with SNR >2 are shown in this figure. The bathymetry contour interval is 200 m in both Figures 10a and 10b.

Variations in risetime of T-phases have been noted prior to this study [Johnson et al., 1968, Schreiner et al., 1995; Dziak and Fox, 1999b; de Groot-Hedlin and Orcutt, 2001; Yang and Forsyth, 2003]. Schreiner et al. [1995] interpret the change in risetime to reflect the crustal depth of the seismic source: a deep crustal event will insonify a larger surface area of the seafloor and have a long risetime. The opposite is true for a shallow crustal event which may have a tighter wave packet and shorter risetime. However, our onset time could also be a function of other variables apart from hypocenter depth, such as efficiency of energy conversion from shallow topography and efficiency of propagation based on water depth of the event (Figure 2c).

Onset times can vary depending on the T-phase excitation process. Downslope propagation on a steep slope requires few reflections (Figure 3a) and may be considered a more efficient coupling mechanism than rough seafloor and sea surface scattering, where only a small percentage of the acoustic energy can propagate in the SOFAR channel [Johnson et al., 1968; Talandier and Okal, 1998]. Johnson et al. [1968] suggested that downslope propagation and seafloor scattering excite different types of T-phases: (1) Slope T-phase, generated by downslope propagation, and (2) Abyssal T-phase, generated by excitation scattering processes. The Slope T-phase has a short onset time and low frequencies compared with the Abyssal T-phase which has a long onset and high frequencies. de Groot-Hedlin and Orcutt [2001] also observed differences in the onset times of T-phases, excited by scattering alone, at different water depths. Their synthesized T-phases show deep water events have higher frequencies, shorter durations and are near-symmetrical compared with shallow T-phase events generated on slopes. If we compare shallow and deep water events in our study areas (Figure 7 and Figure 10) the deep events have long onset times and lower frequencies compared with the shallower events (Figure 7). Therefore our results do not reflect the same frequency or duration characteristics as seen by Johnson et al. [1968] and de Groot-Hedlin and Orcutt [2001]. This suggests that the T-phases generated at RTI locations (average slopes >∼10°) are different from Abyssal and Slope type T-phases.

Efficiency of propagation based on water depth over the event is the same hypothesis discussed in section 5.1. An event at the depth of the SOFAR channel axis will propagate more efficiently because the majority of its energy will enter the waveguide and propagate with little attenuation [Pulli et al., 1999].

5.6. Ray Trace Model

A ray trace model was used to investigate bathymetric blocking of T-phase energy assuming a coupling mechanism of rough seafloor scattering. The model assumes Snell's Law and takes into account 1-D topography along the propagation path (for details see Appendix D). The model was run five times for each event, each time with a different sound velocity profile. The five profiles were used to encompass the seasonal variability observed in the water column (Figure 11). They were obtained from the Generalised Digital Environmental Model (GDEM) [Davis et al., 1986; Teague et al., 1990]. The range in critical depth for the five profiles is 3.35–4.85 km (average ∼4.1 km, Figure 11). In the ray trace model, ninety rays are emitted from each T-phase event at 1° increments from horizontal to vertical (Figure 12). When a ray intersects the seafloor or sea surface it is terminated. The number of rays that reach the range of each hydrophone for an event is averaged over the five runs of the model and this value is used for the rest of our analyses. We find that only 1–9 rays reach each of the hydrophones from events in the Atlantis study region (Figure 13a) and 1–12 rays from events in the Kane study region (Figure 13b). These numbers represent a maximum of 10% of the original energy of the T-phase (assuming omnidirectional scattering). Note that the model results show that no rays should have reached the hydrophones for ∼65% (61 events) of the events that were recorded by the hydrophones from the Atlantis area and ∼33% (21 events) of the events from the Kane area because they encounter the high relief of the transform and median valley walls (e.g., MAR location shown by gray box in Figure 12). The ray trace model also predicts that rays should have reached one or more hydrophones for 9 events from Atlantis and 6 events from Kane that were not included in the hydrophone catalog. Of the 9 events from the Atlantis area, the data from 5 of them have noise from air guns and other unidentified noise sources which mask the T-phase to some extent. The data from 2 events from the Kane area also have noise from air guns. The large percentage of events for which no rays should have reached the hydrophones implies that this model is not a good predictor of acoustic energy levels. To identify where and why the ray trace model does not work we consider sound velocity profile variations, source location water depth, and distance from the event to the hydrophones.

Figure 11.

Two hundred and forty sound velocity profiles between each of the six hydrophones and the Kane study area are shown by the black dots. The data were acquired from the Generalized Digital Environmental Model (GDEM) for four times of year (March, June, September, and December). The five colored lines represent the five profiles used in the ray trace modeling to represent all the variability in the water column sound velocity. We assume that the sound velocity profile is constant from the event to the hydrophone. The critical depth of the five profiles varies greatly (3.35–4.85 km, shown by black dashed line) due to the large variation in sound velocity in the thermocline.

Figure 12.

Example of the ray trace model output for one event located on the top of the Kane massif. The blue line is bathymetry, and the red curves are the ray traces. Ninety rays emerge from the event on the left-hand side of the plot, and any ray that intersects the bathymetry is terminated. We assume the number of rays that reach the hydrophone on the right-hand side of the plot are a proxy for the amount of energy that propagates to the range of each hydrophone. Each source to receiver path encounters very different topography; some cross the MAR axis (marked by gray column on CW hydrophone), while others travel only along the deeper water of the ridge flanks (NE).

We find that the model results for the number of rays that reach a hydrophone are strongly dependent on the sound velocity profiles used. As an example, for one T-phase event at the Atlantis study region, the number of rays that reached the range of the NE hydrophone varies between 2–12 rays depending on the sound velocity profile used. The standard deviation from the average number of rays that reach the range of the hydrophones is largest for the shallow events and generally decreases with increasing event location water depth. Our results may be influenced by the large variation in the SOFAR channel sound velocities observed around the same depths as these shallow events (∼10 m/s difference over 0.5–1.5 km water depth, Figure 11). This variation in the sound velocity profile may be caused by changes in Antarctic Intermediate Water [Roemmich and Wunch, 1985] or possibly high salinity Mediterranean Eddies (“Meddies”), both of which lie around this depth in the water column [Richardson et al., 1988, 1991]. Whatever the source, future T-phase propagation models should take sound velocity variations into account.

The ray trace model results combine two processes: the event location depth dependence from ray trace theory (see Figure 2c) and bathymetric blockage along the direct line propagation path from the source to the receiver. We have included all T-phase events at all hydrophones, apart from those with large location errors (> 0.03°), for the ray trace model so as to take into account events that may not have been recorded due to bathymetric blockage (79 events at Atlantis and 43 events at Kane). Therefore there are more T-phase events shown in Figures 13 and 14 than in Figures 6 and 8.

Figure 13.

Ray trace model results versus water depth for events from (a) Atlantis and (b) Kane. The results are a combination of an event location depth dependence (from Figure 2) and bathymetric blockage. Those events which appear to have only one black circle represent events where no rays made it to any of the 5 hydrophones and the five colored dots are stacked at zero, with the NE hydrophone (black circle) on the top. The overall trend is decreasing number of rays with increasing event water depth at both study locations. The CE and SE hydrophones receive the highest number of rays relative to all the other hydrophones from the Atlantis region, and the CE and NE hydrophones receive the highest from the Kane study region.

Figure 14.

(a and c) Received Level versus range from the event to each hydrophone for the Atlantis and Kane study areas, respectively. These plots differ from the data used for Figure 8 because all of the data are used (we incorporate all of those hydrophones not used in the T-phase catalog to locate the T-phase events and all of the T-phase events with RMS Signal <1 dB re: counts). The mean Received Level (black line) is corrected for transmission loss to give the corrected mean Received Level (red line) in the same way as Figure 8. The vertical boxes represent one standard deviation from the mean. (b and d) The ray trace modeling results. The red line represents the mean number of rays for each hydrophone, for direct comparison with the red lines in Figures 14a and 14c, and the colored vertical bars are one standard deviation from the mean. The mean appears to be low for the model results, but this is due to the large number of events for which no rays made it to the hydrophone because of bathymetric blockage (i.e., there are multiple values of zero). In each of the study areas the ray trace model results and the observed Received Level reflect similar trends in the relative amount of energy that reaches the range of the hydrophones.

In this model we use the number of rays that travel unblocked to the hydrophone range as a proxy for the energy arrival at each hydrophone. Admittedly this is a simple model but the results have shown it is applicable. We take the same approach as the observed data and compare the number of rays that reach the range of the hydrophones with the event location water depth and source to receiver distance. When we compare the number of rays with event water depth we observe a decrease in the number of rays with increasing water depth, approaching zero rays as the average critical depth (4.1 km) is reached for the five sound velocity profiles (Figure 13). This is the relationship we expected to observe in the Received Levels. However, Received Levels remain fairly constant at all water depths (Figure 6) suggesting that omnidirectional scattering combined with the event location water depth dependence does not describe the process of coupling at the seafloor. The process of T-phase generation is more complex and does not appear to be strictly bounded by the sound velocity profile.

The ray trace model results (Figures 14b and 14d) show two different trends in the mean number of rays versus distance from the event for the two study areas and there is some correlation between the model trend and the observed data trend (Figures 14a and 14). For events from the Atlantis region, the model has a minimum mean number of rays at the CW hydrophone and maximum mean number of rays at the CE and SE hydrophones. The maximum values at the CE and SE hydrophones may be due to the rays entering deep water more quickly along the first 50 km of the propagation path, compared with propagation paths to the NE and SW hydrophones. The minimum at the CW hydrophone is most likely due to the shallow MAR axis topography between the source and the hydrophone.

For Kane events the model shows the CE and NE hydrophones have the maximum mean number of rays and the CW hydrophone the minimum. Again, the rays enter deeper water more quickly along propagation paths to the CE and NE hydrophones which may explain why more rays, in general, make it to these hydrophones (Figure 12). Even though the CW hydrophone is located close to the Kane region (∼500 km), the T-phases must propagate out of the transform fault and over the MAR axis, resulting in much interaction with the topography and disruption of the T-phases before they reach the CW hydrophone.

The high standard deviation of the observed data and the model results makes it difficult to interpret the results extensively. We conclude that the ray trace model under predicts the amount of energy that propagates from deep water events, and the model condition that the rays terminate when they intersect the bathymetry is likely too rigorous. Instead some percentage of the energy may make it past topographic features, either by propagation scattering or wave front healing [Claerbout, 1985; Pulli and Upton, 2002]. Alternatively the theory of a critical depth is incorrect for T-phase propagation and some portion of the T-phase energy from an event below the critical depth is able to enter the SOFAR channel. In summary, the results of our model do not explain the Received Level of the T-phases based on the water depth of the events and highlight the complexity of the coupling process. However, the model does show that bathymetric blockage along the propagation path affects the amplitude of a T-phase. The model predicts two trends in relative energy levels between hydrophones as a function of distance from source to receiver for the two study areas and these trends are similar to the trends observed in the Received Levels.

6. Discussion

A T-phase excitation and propagation model that explains all of the observed characteristics of T-phase arrivals (lens shape, risetime, multiple peaks, spectrogram behavior, etc.) does not exist. In general, there are two regions to consider: a relatively short excitation region where the T-phase characteristics are established, and a propagation region where the T-phase energy is totally trapped in the ocean sound channel (there have been occasional observations of basin scale multipathing, with secondary scattering from continental margins [e.g., Shurbet and Ewing, 1957]). Some common hypotheses for excitation within the short region where the T-phase characteristics are established are summarized in Figure 3. Scattering from roughness and heterogeneities at and near the seafloor has been invoked to explain the excitation of abyssal (water depths greater than 3 km) T-phases (Figure 3b). Even this mechanism, however, is inadequate when the seafloor is well below the critical depth (Figure 3c).

All T-phases by definition involve long range propagation in the ocean. There is growing evidence however that T-phase propagation involves the coupling of energy between the ocean sound channel and the shallow oceanic crust including the almost ubiquitous sediment layers [Butler and Lomnitz, 2002]. Recent observations have shown that the reciprocal process to excitation, getting energy out of the SOFAR channel into the crust, is commonplace even in deep water [Butler and Lomnitz, 2002; Butler, 2004]. In fact earthquake generated T-phases have been observed on borehole seismometers in the Western Pacific (WP-2) and Philippine Sea (WP-1) that are over 400 m below the seafloor in water depths exceeding 5.5 km [Araki et al., 2004]. Hence scattering may also play a significant role in long-range T-phase propagation as well. The absence of water depth dependence in our T-phase data is consistent with the notion that T-phases are inadequately explained by models with perfect, laterally homogeneous waveguides, even when seafloor scattering is invoked in the excitation region.

Our discussion of the water-depth dependence of T-phase excitation (section 2 and Figure 2) is based on a ray model. Long-range propagation in the ocean can be described by rays in three categories: refracted refracted paths, refracted surface-reflected paths and surface-reflected bottom-reflected paths [Jensen et al., 1994]. For simple velocity depth profiles (e.g., Figures 2 and 11) the critical depth defines the waveguide for the refracted refracted paths. These totally trapped paths are the most efficient for transmitting energy because they are only subject to cylindrical spreading and the intrinsic attenuation of water. The other paths suffer scattering losses at each bounce with either the sea surface or seafloor interface. Although these losses may be small for a single bounce, long range propagation involves many bounces so the losses can be significant. Multiple water bounces may play a role in the near-field T-phase excitation region [Yang and Forsyth, 2003], but refracted surface-reflected paths and surface-reflected bottom-reflected paths are not traditionally used to explain the long-range propagation.

Alternatively, long-range ocean acoustic propagation can be described by modes. The low order modes correspond to the refracted refracted paths and the smallest grazing angles (high incidence angles). Observations of T-phases on vertical arrays show that the energy is concentrated in the lowest order modes (usually less than mode 7) with grazing angles less than 10° (incidence angles greater than 80°) [D'Spain et al., 2001]. For example, Figure 2 of de Groot-Hedlin and Orcutt [2001] shows the acoustic excitation as a function of seafloor depth for the first three modes at frequencies from 5 to 20 Hz. All of the modes decay substantially by 3 km water depth. This is the modal explanation of what we have described above using rays.

Our knowledge of T-phase dynamics is deficient because we do not know the physical mechanisms responsible for getting abyssal T-phase energy from the earthquake epicenter into the SOFAR channel (in the excitation region) or for getting energy out of the SOFAR channel into the deep seafloor (in the propagation region). Some form of scattering at or near the seafloor has been shown to be necessary in the excitation region to convert the compressional and shear body waves from earthquakes into the high incidence angle paths necessary for propagation in the SOFAR channel [de Groot-Hedlin and Orcutt, 2001; Park et al., 2001] but this process would not work when the seafloor is substantially below the critical depth. Also the ubiquitous observation of T-phases in the propagation region on seismometers at and below the seafloor even in water depths well below the critical depth, is not adequately explained by existing models. An improved understanding of the dynamics of T-phases is required. Until the coupling between the sound channel and the crust is adequately understood, quantitative consideration of more complex phenomena, such as reflection from continental margins, blockage by islands and seamounts, or the inversion of T-phase characteristics for earthquake parameters, is premature.

7. Conclusions

In this study we made two assumptions about the T-phase events located in the Atlantis and Kane study areas: (1) the T-phase event locations are accurate, and (2) the locations represent earthquake epicenters. We examined the detailed characteristics of 158 T-phase events and modeled their propagation to assess the dependence of T-phase amplitude on event water depth, source to receiver distance, the relationship to onset time and the effects of bathymetric blockage along the propagation path. From our analyses we make the following conclusions:

1. A greater number of T-phase events are located at shallow water depths (on the ICH massif and slopes at depths of 1–2 km) than deep water in both our study areas, when the number of events is normalized to the seafloor area. The events may be correctly located and our observations reflect clustering of real seismic events on the massifs due to local tectonic or volcanic processes or more efficient generation or propagation of shallow water T-phases. Alternatively, the events could be mislocated and the massifs may act as radiators for energy from earthquakes located in the adjacent valleys. It is impossible to say from our data which of these processes and mechanisms, if any, controls the spatial distribution of the T-phase events. Our two assumptions that T-phase event locations are accurate and the locations represent earthquake epicenters can only be addressed by comparing concurrent ocean bottom seismometer and hydrophone data. These types of surveys have been deployed recently and should provide answers in the near future.

2. Three events located in the Kane study area were recorded as both teleseisms and T-phases, the smallest of which was 4 mb. The seismic body wave magnitude and mean amplitude (mean Received Level) recorded at the hydrophones for these three events are not correlated, but the small number of events makes it difficult to draw conclusions from this. However, seven events in the Atlantis study area with mean Received Level greater than the smallest of the three teleseismic events were not recorded as teleseisms. Therefore the relationship between seismic and acoustic magnitude may be location specific and more complex than previous studies have suggested.

3. The Received Level of a T-phase is not sensitive to water depth at the T-phase event location. T-phase data show large variations in the Received Level of an individual event at the five hydrophones but little variation in overall Received Level with changing event water depth. In contrast, our ray trace model predicts that the number of rays which reach a hydrophone from a T-phase event decreases with increasing water depth of the event location, and no rays reach hydrophones from events where the seafloor is located below the critical depth. Therefore omnidirectional scattering combined with ray trace theory cannot be used to explain T-phase excitation.

4. The ray trace model results combine two processes: the event location depth dependence from ray trace theory (which includes seasonal variations in the sound velocity profiles) and 1-D bathymetric blockage along the propagation path. On the basis of our model only 35% of the T-phase events included in the hydrophone catalog located in the Atlantis study region and 64% of the events in the Kane study region should have been recorded by the hydrophones. The assumption that a ray is terminated when it intersects the seafloor is likely too rigorous. Some portion of the T-phase energy is able to propagate past shallow features to the hydrophones, possibly by scattering or wave front healing, or the theory of a critical depth (based on a laterally homogeneous waveguide) is not appropriate for T-phase excitation and propagation.

5. The ray trace model did show that there is a correlation between the model mean number of rays and the mean Received Level of events versus distance from the source (earthquake epicenter) to the receiver (hydrophones). Having removed the effects of transmission loss from the signal, it appears that bathymetric blockage is an important process that causes variations in the magnitude of an event as recorded by different hydrophones in an array. The relative change in mean Received Level for increasing source to receiver distance is very different for the two study areas. The model is able to reproduce similar relative changes in the mean number of rays. Ray trace theory can therefore explain to some extent the process of bathymetric blockage but the data have large variances and more analysis with a larger number of events would be needed to confirm the accuracy of the model.

6. The onset time of the T-phase, defined as the time interval from when the T-phase emerges above the ambient noise to the first peak in the T-phase, follows a pattern of short onset times for shallow water events and long onset times for deep water events. This is a remarkable correlation. At this time we do not have a physical explanation for this empirical observation. We suggest that the onset time is a function of several variables including: efficiency of energy conversion based on local topography, efficiency of propagation based on water depth of the event and hypocentral depth of the event. Length of the propagation path does not appear to correlate with onset time. The duration of the onset time and the frequency content of our T-phase events do not agree with earlier work which identify Abyssal (long onset time and high frequencies) and Slope (short onset time and low frequencies) type T-phases. This suggests that the T-phases generated at mid-ocean ridges cannot be classified using the same criteria and may define a new type of T-phase.

Our observations of T-phase characteristics require an improved model for T-phase excitation and propagation that addresses: the process of introducing energy into the SOFAR channel from events located at water depths which exceed the critical depth, the influence of seafloor roughness on seismic to acoustic conversion and the affect of variations in the sound velocity profile between the event and the hydrophone, particularly around the sound channel axis minimum, on T-phase propagation.

Appendix A:: Picking the Events

The hydrophones in the NAHA record 1-byte resolution data at a sample rate of 112 samples per second (C. Fox, personal communication, 2001) with a 1–40 Hz band pass filter [Fox et al., 2001]. The frequency-pressure response function used corresponds with SW = 3 in Figure 2 of Fox et al. [2001].

The first step in processing the T-phase data was to locate the seismic arrivals from the T-phase catalog in the raw hydrophone data. Summary plots of time series and spectrograms for arrivals at all six hydrophones for each event were compiled (examples shown in Figure 7). All 158 events recorded in the Atlantis and Kane study areas between 25 February 1999 and 9 March 2001 are shown in Tables 1 and 2, respectively. The spatial distribution of the T-phases includes all these events (Figure 5). For subsequent analysis of the T-phases those events with location errors greater than 0.03° (∼3.4 km) in either latitude or longitude were removed from the data set to improve the data quality (∼15 events were removed from each data set).

Ambient noise levels vary between hydrophones and high ambient noise levels reduce the detectable level of T-phases. In order to quantify the ambient noise, the RMS of the first 10 seconds of each 90 second window of the time series was computed with the assumption that this was “quiet time” before the arrival of the T-phase. We define this value as the RMS Ambient Noise. The RMS Signal of the T-phase was then calculated by subtracting the RMS Ambient Noise from the RMS Received Level in dB re: counts. Several of the RMS Signal calculations resulted in negative values. The majority of these negative values were traced back to air guns dominating the hydrophone recording as high amplitude, high frequency discrete signals (their removal is discussed below). Clipping of events that exceed the dynamic range of the hydrophones could also affect the calculated RMS Signal. Only one T-phase event in our analysis was clipped by the hydrophone. This was a teleseismic event (Event 32) located at the Kane study area on the CE hydrophone. Event 32 has an average magnitude of 4.0 mb, however Fox et al. [2001] proposed the general relationship that >4.7 mb would exceed the dynamic range of the hydrophones 8-bit systems. Newer generation PMEL instrument have a 16-bit system and others are 24-bit, leading to an increased dynamic range and reducing the occurrence of clipping for large events.

To improve the quality of the data the following steps were carried out before plotting Received Level versus both event water depth and source to receiver distance:

1. The hydrophone catalog used 3 or more hydrophones to locate their events. On occasion we saw T-phase arrivals on hydrophones not used in the hydrophone catalog for locating the event. We only used arrivals that were identified in the catalog, except for the comparisons with ray trace modeling (because of the possibility that bathymetric blocking is responsible for the lack of a T-phase arrival on some hydrophones).

2. The negative RMS Signal values were dealt with in two ways. For those events with a small contribution from the air guns (one negative hydrophone signal per event) the individual negative hydrophone picks were removed. For an event with a large contribution (greater than three negative values per event) the whole event was removed from the data set because the RMS Received Level values represent the noise and not the T-phase amplitude.

3. In order to work only with events of a significant magnitude, all events with an RMS Signal level of <1 dB re: counts were removed (except for comparison with ray trace modeling results because of possible bathymetric blocking as discussed in point (1)).

After these steps, 72 events from Atlantis and 39 events from Kane were used to plot Received Level versus event location water depth and source to receiver distance. T-phase event amplitude versus distance was plotted using the Received Level rather than the RMS Signal because the Received Level defines the magnitude of the actual event as recorded by the hydrophone. The RMS Signal is a modified value and depends on the ambient noise levels at each hydrophone.

Appendix B:: Transmission Loss Curves

The range from source to receiver will affect the magnitude of the event due to transmission loss. One transmission loss curve was calculated for each study area on the basis of two contributions:

1. Cylindrical spreading - 10 log10 r in dB where r is in kilometers [Jensen et al., 1994].

2. Intrinsic attenuation - 0.00333r or 3.3 dB per 1000 km [Fisher and Simmons, 1977; Clay and Medwin, 1977].

Each transmission loss curve was subtracted from the observed data to reduce the number of processes that affect the T-phase as a function of range in each region. Transmission loss was calculated as a relative value between hydrophones by assuming that the nearest hydrophone has zero transmission loss and the furthest hydrophone has the maximum.

Appendix C:: Curve Fitting

We quantify the onset time following the curve fitting method of Yang and Forsyth [2003]. Before computing the complex envelope, the time series is band pass filtered in an octave centered at 16 Hz. The complex envelope is then smoothed with a one second running average and the log of the smoothed envelope is taken.

To define the onset time of the T-phase we fit a curve to the beginning of the log of the smoothed envelope. The amplitude of the envelope as a function of time is given by

equation image

where A0 is the amplitude of the ambient noise, A1 is the amplitude of the first peak in the T-phase, t0 is the arrival time of the peak and tb is the onset time. A0, A1 and t0 are picked from the log of the smoothed envelope. The onset time, tb is then calculated by fitting a curve with an exponential value which gives the minimum RMS fit.

The average ambient noise levels recorded by the hydrophones are ∼10.4 dB re: counts for Atlantis events and ∼9.2 dB re: counts for Kane events. There is no significant difference between the ambient noise level of shallow events (<2 km) and deep events (>2 km) (difference is ∼0.3 dB re: counts for Atlantis and 1.2 dB re: counts for Kane). Variability in ambient noise levels is therefore not a significant signal and does not affect our onset time results.

Appendix D:: Ray Trace Model

equation image

where i is the incidence angle and v is the velocity for a two layer model.

All those events with an RMS signal <1 dB re: counts were included in the data set for the modeling. A total of 79 events from Atlantis and 43 events from Kane were used to compare with the ray trace model results.

Seasonal changes in the sound velocity profile were examined using data from the Generalized Digital Environmental Model (GDEM). Ten different velocity profiles between the Kane Study area and each of the six hydrophones were obtained for the months of December, March, June and September (i.e., 240 profiles in total). Five profiles, shown in Figure 11, were selected to encompass the full extent of this variation and were used in the modeling.

Bathymetric profiles along the propagation path were obtained from Seabeam bathymetry where available (grid spacing ∼150 m) and from Smith and Sandwell [1997] satellite altimetry data (grid spacing ∼3 km) otherwise. The bathymetric profiles were created by sampling the 2-D grids along 1-D tracks.

For the model, ninety rays at equal 1° increments from horizontal to vertical are emitted from the event location. Two main assumptions are made. The first is that the seismic energy couples with the water column through the process of rough seafloor scattering, yielding a range in emergent angles. It is also assumed that when the rays intersect the seafloor or the sea surface, they are terminated and the number of rays which make it to the hydrophone are proportional to the amount of energy propagated (Figure 12). The model was run 5 times for each event with a different sound velocity profile each time. Those rays with low incidence angles (i.e., near vertical) are terminated at the first interaction with the sea surface. The remaining cone of rays that have a high enough incidence angle are able to refract downward, away from the sea surface and continue along their path to the hydrophone, unless they intersect the seafloor.

Acknowledgments

The authors would like to thank C. Fox, R. Dziak, H. Matsumoto, M. Fowler, and J. Cann for their support and encouragement. The data were acquired and made available to us by the Pacific Marine Environmental Laboratory (National Oceanic and Atmospheric Administration). We also thank D. Bohnenstiehl and one anonymous reviewer for their constructive comments. This study was supported by National Science Foundation (grants OCE-0136808 and OCE-0221832) and the Deep Ocean Exploration Institute at Woods Hole Oceanographic Institution.

Ancillary