Dynamic Modeling of Hydroacoustic Reflections, and a First

DYNAMIC MODELING OF HYDROACOUSTIC REFLECTIONS AND A FIRST LOOK AT THE DATA FROM THE NEW HYDROACOUSTIC ARRAYS AT ASCENSION ISLAND

Jay J. Pulli1, Zachary M. Upton1, Jeff Wagoner2, and Phil Harben2

BBN Technologies1 and Lawrence Livermore National Laboratory2

Sponsored by Army Space and Missile Defense Command

Contract No. W9113M-05-C-01361,2

ABSTRACT We report on two of our efforts to better understand hydroacoustic signals for the purpose of event discrimination. The first effort involves the dynamic modeling of hydroacoustic reflections. The goal is to have a frequency dependent model-based predictive tool for assessing where consistent reflections will occur and the frequency dependence of the reflected energy that becomes trapped in the SOFAR channel. This will be a necessary capability if reflected signals are used in event discrimination. The second effort is a preliminary examination of the new hydroacoustic data coming from Ascension Island. A specific bathymetric feature was chosen for the reflection modeling—the steep bathymetry on the eastern side of the Seychelles-Mauritius Plateau in the western Indian Ocean. This site was chosen to coincide with a known zone of reflections that have been observed from both earthquake and explosive sources. Lawrence Livermore National Laboratory (LLNL) set up the base model for dynamic reflection analysis using data from the General Bathymetric Chart of the Oceans (GEBCO) website. These data are in a format of one arc-minute grid spacing and are available as 20X20 degree tiles in netCDF format. The bathymetry data were imported into EarthVision, a 3-D surface modeling code developed by Dynamic Graphics, Inc. The geographic coordinates were then converted to a Universal Transverse Mercator (UTM) projection in the model area. The data were gridded using minimum tension gridding with a bi-harmonic cubic spline function. The basal grid elevation ranges from 18 to -4268 meters. This binary grid is then converted into a format that can be read by the Wave Propagation Project code (WPP). HydroCAM is used to predict propagation along the source-to-reflector path, which is used as input to the WPP code. The WPP code will initially be run in 3-D at low frequencies (less than 10 Hz) and compared to reflected signals from earthquake sources. Dynamic runs will be conducted in July 2006. These calculations will then be handed back to HydroCAM, where the signals will be propagated to longer distances. These modeling calculations will be compared to observed earthquake reflections to help validate the modeling and quantify the effects of parameters such as bathymetry slope at reflection depths. In the second effort, we report on an analysis of the data from the new hydroacoustic arrays around Ascension Island. As part of this effort, the Air Force Technical Applications Center (AFTAC) data mirror at LLNL was modified to allow BBN to access data from the International Monitoring System hydroacoustic stations at Diego Garcia, Cape Leeuwin, Crozet Island, and Ascension Island. Blockage calculations at Ascension Island predict that both arrays will simultaneously record events in most of the southern Atlantic Ocean, with the exception of a small wedge of azimuths to the southeast. The azimuth that is blocked at the north array is covered by the south array. Hence, two back azimuth calculations can be used to locate an event in most areas. The arrays record numerous events along the southern Mid-Atlantic Ridge, and reasonably accurate locations and origin times are available for most of these events above magnitude 4.5 from the United States Geological Survey. These events often produce acoustic energy up to at least the anti-alias frequency of 100 Hz. We attribute this high-frequency energy to the shallow focal depths and hence the short seismic propagation path. Additionally, airgun signals and biological signals are often seen at the arrays.

28th Seismic Research Review: Ground-Based Nuclear Explosion Monitoring Technologies

742

OBJECTIVES

The overall goal of this project is to further our understanding of hydroacoustic blockage and reflections, and assess how these processes affect the discrimination of underwater events. An additional goal is to provide a quick-look assessment of data from the new hydroacoustic arrays as they come online in the Atlantic and Pacific Oceans. Specific objectives during this reporting period include:

• The dynamic modeling of hydroacoustic reflections from the eastern side of the Seychelles-Mauritius Plateau in the western Indian Ocean. Here, we are using the WPP being run at LLNL. Reflections are often observed from this bathymetric feature at the Diego Garcia hydroacoustic array.

• An examination of data from the new hydroacoustic arrays at Ascension Island, which became available to us during 2006. We have examined signals from 15 earthquakes during this time period, as well as numerous airgun and biological signals.

RESEARCH ACCOMPLISHED

Dynamic Modeling of Hydroacoustic Reflections

In earlier studies (Upton et al., 2006) we developed a model that estimates the impulse response of the ocean basins based on high-resolution bathymetry. For a given source-receiver pair, the model uses travel-time ellipses to predict which bathymetric features will contribute reflections to the impulse response. The strength of each reflection was based on the area of the bathymetric feature that intersects the sound channel axis, and its orientation with respect to the source-receiver travel-time ellipse. This model has been successfully applied in a matched-field approach to event localization based on a single station received wave train (Upton and Pulli, 2002). Although the travel times of the reflections match observed data very well, the simple model does not do well at predicting their amplitudes. In this study, we use a dynamic modeling technique at the bathymetric reflector, in combination with long-range predictions using HydroCAM, to more accurately model the reflection process. A specific bathymetric feature was chosen for the reflection modeling: the steep bathymetry on the eastern side of the Seychelles-Mauritius Plateau in the western Indian Ocean (Figures 1 and 2). This site was chosen to coincide with a known zone of reflections that have been observed from both earthquake and explosive sources (Harben and Boro, 2001). LLNL has set up the base model for dynamic reflection analysis using data from the GEBCO website. These data are in a format of one arc-minute grid spacing and are available as 20X20 degree tiles in netCDF format. The bathymetry data were then imported into EarthVision, a 3-D surface modeling code developed by Dynamic Graphics, Inc. The geographic coordinates were then converted to a UTM projection in the model area. The data were gridded using minimum tension gridding with a bi-harmonic cubic spline function. The basal grid elevation ranges from 18 to -4268 meters. Four perspective views of this region are shown in Figure 3. This binary grid is then converted into a format that can be read by the WPP code. HydroCAM is used to predict propagation along the source-to-reflector path, which is used as input to the WPP code. The WPP code will initially be run in 3-D at low frequencies (less than 10 Hz) and compared to reflected signals from earthquake sources. These calculations will then be handed back to HydroCAM, where the signals will be propagated to longer distances. These modeling calculations will be compared to observed earthquake reflections to help validate the modeling and quantify the effects of parameters such as bathymetry slope at reflection depths. As of this writing, the first runs of the dynamic modeling are being conducted at LLNL and will be shown on the accompanying poster in September at the 2006 Seismic Research Review (SRR).


743

Figure 1. Map of the western Indian Ocean downloaded from GEBCO, showing the Seychelles-Mauritius Plateau. The study area is identified by an arrow at the center of the map.

Figure 2. Detailed bathymetry of the area of interest. These data have been converted to UTM coordinates (Zone 40). Note the sharp definition of the northeastern tip of the Seychelles-Mauritius Plateau.


744

Figure 3. Four perspective views of the Seychelles-Mauritius Plateau. The intersection of the 2 orange axes represents the southwest corner of the diagram. Each plot has been rotated slightly to give the viewer a different angle of the plateau. These plots have a vertical exaggeration of 20X.

First Look at Ascension Island Hydrophone Data

Hydrophones were first installed around Ascension Island in the 1960’s. These hydrophones recorded many of the large underwater explosions detonated in the Atlantic Ocean in the 1960’s and 70’s (Pulli et al., 2000). These hydrophone channels were of low dynamic range, so multiple gain channels were used to record the direct arrivals of the large explosions on scale. Many of these hydrophones remained operative through 2001, until equipment failures eventually rendered all of them inoperative. The new installations, which were installed in 2005 and whose data became available to us in 2006, are of the same high dynamic range configurations as those in the Indian Ocean. The geometries of these arrays are shown in Figure 4.


745

Figure 4. Geometric configurations of the North and South Arrays at Ascension Island.

In order to better understand the geographic coverage of these new arrays, blockage calculations were performed using HydroCAM. These calculations are made by shooting fans of rays from the center of each array over 360 degrees at 1-degree increments. Two different criteria are currently used to end the ray path. Figure 5 shows the calculations using the stop criteria of when the sound channel axis depth reaches the surface and becomes nonexistent. Here we see that both arrays provide detection coverage over most of the western North and South Atlantic Oceans. The wedge of blocked signal paths to the southeast for the north array is covered by the south array. Work to better understand the blockage process using an adapted version of the Naval Research Laboratory’s Adiabatic Mode Parabolic Equation (Collins, 1993) model is currently being undertaken on a separate project during this SRR (Upton et al., 2006).

Figure 5. Hydroacoustic blockage calculations for the Ascension North (left) and South (right) arrays.


746

Access to the Ascension data is provided to us via two mechanisms. One is an AutoDRM message to the AFTAC data server. We perform this type of access on an event basis, typically acquiring two hours of data per event starting at the event origin time. Data access is also provided via secure ftp connection to the continuous data archive at LLNL. Most of the events recorded at Ascension have occurred along the Mid Atlantic Ridge. Fifteen events have been examined as of this writing (see Table 1 and Figure 5). These events range in magnitude from 4.6 to 6.0, although aftershocks of these events are frequently observed which are less than magnitude 4.6. In addition, airgun signals are often observed coming from the east of the arrays.

Table 1. Events analyzed at the Ascension hydroacoustic arrays. Epicentral references are from IRIS.

Date O.T. Lat. Lon. Mag. Depth (km) AREA 17-Feb-06 13:24:04 -1.85 -15.07 5.1 10 NO OF ASCENSION ISLAND 18-Feb-06 14:59:04 -6.48 -11.08 4.9 10 MID ATLANTIC RIDGE 4-Mar-06 00:53:32 0.9 -28.04 5.1 10 CENTRAL MID-ATLANTIC RIDGE

27-Mar-06 01:10:33 7.17 -34.26 5.3 10 CENTRAL MID-ATLANTIC RIDGE 30-Mar-06 03:14:43 -1.23 -15.89 5.0 30 NO OF ASCENSION ISLAND 8-Apr-06 22:03:04 -0.12 -18.13 5.0 10 CENTRAL MID-ATLANTIC RIDGE

10-Apr-06 06:26:13 7.45 -37.02 4.8 10 CENTRAL MID-ATLANTIC RIDGE 28-Apr-06 07:22:42 3.83 -31.5 4.9 10 CENTRAL MID-ATLANTIC RIDGE 23-May-06 07:24:23 -30.93 -13.37 4.7 10 SOUTHERN MID-ATLANTIC RIDGE 5-Jun-06 06:34:31 1.02 -28.17 5.6 10 CENTRAL MID-ATLANTIC RIDGE 5-Jun-06 06:27:08 1.18 -28.06 6.0 10 CENTRAL MID-ATLANTIC RIDGE 5-Jun-06 06:18:44 1.05 -28.12 4.9 10 CENTRAL MID-ATLANTIC RIDGE 8-Jun-06 16:29:12 4.55 -51.93 5.2 10 FRENCH GUIANA

18-Jun-06 18:28:02 32.96 -39.62 5.9 10 NO MID-ATLANTIC RIDGE 23-Jun-06 08:02:47 19.1 -46.08 4.6 10 NO MID-ATLANTIC RIDGE

Figure 6. Map of events studied at the Ascension Island arrays (blue dots). Origin information is in Table 1.


747

We now show four examples of events recorded at the Ascension arrays. The first example (Figure 7) is of a magnitude 4.9 event that was located 405 km to the northeast of the north array. The P wave and T wave from this event are seen in the data, as well as some small aftershocks. In addition, airgun signals are seen in the background. Frequencies up to 70 Hz are seen in the P and T waves. The computed back azimuth from the f-k solution for the T waves is 66.45 degrees, whereas the actual back azimuth is 68.62 degrees. The apparent velocity of the T wave is 1.567 km/sec, which is faster than the actual sound speed at this depth and location (1.482 km/sec, see Figure 8). This may mean that the T-wave arrival is bending away from the horizontal as it approaches the array. Although there is adequate signal-to-noise ratio on the south array, the f-k solution is not as sharply defined as it is on the north array. This may be due to the interference of the louder airgun signals on the south array.

Figure 7. Time-frequency and frequency-wavenumber spectra of a magnitude 4.9 event to the northeast of Ascension Island. Origin information is given in Table 1.


748

Figure 8. Sound velocity profiles in the vicinity of the Ascension arrays, from the Generalized Digital Environmental Model database.

Two other examples of events from the Mid-Atlantic Ridge are shown in Figures 9 and 10. Here we see the high-frequency content of T waves from mid-ocean ridges, due to the very shallow depths and hence short distances of seismic propagation (Salzberg, 2006). In both of these cases, the f-k solutions for the T phases point back to the source locations, however on the south array we find that the apparent velocity of the wavefront is again faster than the actual propagation velocity.

Figure 9. Time-frequency and frequency-wavenumber spectra of a magnitude 4.9 event to the northwest of Ascension Island along the Central Mid Atlantic Ridge. Origin information is given in Table 1.


749

Figure 10. Time-frequency and frequency-wavenumber spectra of a magnitude 5.9 event to the northwest of Ascension Island along the Central Mid Atlantic Ridge. Origin information is given in Table 1.

The final example is an event that occurred on the coast (in continental crust) of French Guiana on June 8, 2006, (Figure 11). T waves from this event arrived at Ascension approximately 48 minutes after the event origin. Two groups of T waves were observed, separated by approximately 3 minutes. The second T-wave arrival could be from an aftershock or from a second seismic-to-acoustic conversion point. Because this is a continental earthquake, the bandwidth of the T waves is lower, and is limited to less than 15 Hz by the time the signals reach Ascension. In this case, there are no seismic stations to the east of the event, which may bias the computed location. But this T-wave recording could be used to provide azimuthal constraint on the location.

Figure 11. Time-frequency and frequency-wavenumber spectra of the T-waves from a magnitude 5.2 event on the coast of French Guiana, recorded at the Ascension arrays. Origin information is given in Table 1.


750

CONCLUSIONS AND RECOMMENDATIONS

Long-range hydroacoustic reflections have been reported in the literature for nearly forty years (e.g., Northrop, 1968). With the advent of modern high dynamic range acoustic arrays, we are now beginning to understand how to use these signals to supplement hydroacoustic locations and discrimination analysis (Upton et al., 2006). Our present study seeks a better understanding of the reflection process by modeling one particular bathymetric reflector, the eastern side of the Seychelles-Mauritius Plateau in the western Indian Ocean. We are currently using the WPP at LLNL. Calculations are being performed by LLNL during July and will be reported at the SRR meeting in September.

We have also been examining the data from the new hydroacoustic arrays around Ascension Island, which have replaced the hydrophones that operated in this area since the 1960’s. These combined arrays have a detection coverage area that spans the western North Atlantic Ocean and nearly all of the South Atlantic Ocean, with the exception of a few small areas behind islands. So far, we have examined fifteen earthquakes in the Atlantic Ocean that have been recorded at Ascension Island. High frequency T-wave signals are observed from the shallow events along the Mid-Atlantic Ridge. A coastal event in French Guiana was also recorded by the arrays. In addition, airgun signals are often observed coming from the area east of the arrays.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the cooperation of the Air Force Technical Applications Center in providing access to the Ascension Island hydrophone array data.

REFERENCES

Collins, M.D. (1993). The Adiabatic Mode Parabolic Equation, J. Acoust. Soc. Am. 94: (4), pp. 2269–2248.

Harben, P. and C. Boro (2001). Implosion source development and Diego Garcia reflections, in Proceedings of the 23rd Seismic Research Review: Worldwide Monitoring of Nuclear Explosions, LA-UR-01-4454, Vol. 2, pp. 23–31.

Northrop, J. (1968). Submarine topographic echoes from Chase V, J. Geophys. Res.,73: 3,909–3,916.

Pulli, J. J., Z. Upton, R. Gibson, J. Angell, T. Farrell, and R. Nadel (2000). Characterization of reflected arrivals and implications for hydroacoustic test-ban treaty monitoring, Defense Threat Reduction Agency Report DTRA-TR-00-36, 85 pp.

Salzberg, D. (2006). Hydroacoustic estimates of earthquake source depth: applications to tsunami warning, submitted to Geophys. Res. Letters.

Upton, Z. M. and J. J. Pulli, (2002). Localization of sub-sea earthquakes using hydroacoustic reflections and matched field processing, in Proceedings of the 24th Seismic Research Review—Nuclear Explosion Monitoring: Innovation and Integration, LA-UR-02-5048, Vol. 2, pp. 676–685.

Upton, Z.M., J. J. Pulli, B. Myhre, and D. Blau (2006). A reflected energy prediction model for long-range hydroacoustic reflections in the oceans, J. Acoustical Soc. Amer. Union 119: (1), 153–160.


751

Documents

Dynamic Modeling of Hydroacoustic Reflections, and a First