Volume 113, Issue B8
Geomagnetism and Paleomagnetism/Marine Geology and Geophysics
Free Access

Crustal structure of the Caribbean–northeastern South America arc-continent collision zone

Gail L. Christeson

Gail L. Christeson

University of Texas Institute for Geophysics, Jackson School of Geosciences, Austin, Texas, USA

Search for more papers by this author
Paul Mann

Paul Mann

University of Texas Institute for Geophysics, Jackson School of Geosciences, Austin, Texas, USA

Search for more papers by this author
Alejandro Escalona

Alejandro Escalona

University of Texas Institute for Geophysics, Jackson School of Geosciences, Austin, Texas, USA

Now at Department of Petroleum Engineering, University of Stavanger, Stavanger, Norway.

Search for more papers by this author
Trevor J. Aitken

Trevor J. Aitken

University of Texas Institute for Geophysics, Jackson School of Geosciences, Austin, Texas, USA

Now at Devon Energy, Houston, Texas, USA.

Search for more papers by this author
First published: 30 August 2008
Citations: 56

Abstract

[1] We present the results of a 568-km-long regional wide-angle seismic profile conducted in the southeastern Caribbean that crosses an active island arc, a remnant arc, two basins possibly floored by oceanic crust, an allochthonous terrane of forearc affinity, and the passive margin of northern South America. The velocity structures of the Late Cretaceous Aves Ridge remnant arc and Miocene and younger Lesser Antilles arc are remarkably similar, which implies that magmatic processes have remained moderately steady over time. Crustal thickness is ∼26 km at the Aves Ridge and ∼24 km at the Lesser Antilles arc. In comparison to the Izu-Bonin and Aleutian arcs, the Lesser Antilles arc is thinner and has no evidence for a lower crustal cumulate layer, which is consistent with the estimated low magma production rates of the Lesser Antilles arc. Crustal thickness beneath the Grenada and Tobago basins is 4–10 km, and the velocity structure suggests that these basins could be floored by oceanic crust. A decrease of ∼1 km/s in average seismic velocity of the upper crust is observed from NW to SE across the North Coast fault zone; we argue that this marks the suture between the far-traveled Caribbean arc and the passive margin of the South American continent. Current strike-slip motion between the Caribbean and South American plates is located ∼30 km to the south, and thus material originally deposited on the South American passive margin has now been transferred to the Caribbean plate.

1. Introduction

[2] The Lesser Antilles arc, together with the Greater Antilles arc, the Aves Ridge, and the Leeward Antilles arc (Figure 1a), form segments of a continuous Great Caribbean island arc system that was constructed in the Pacific and migrated into the Atlantic as the Caribbean plate moved eastward with respect to the North America and South America plates [e.g., Burke, 1988; Pindell et al., 1988; Pindell and Barrett, 1990]. Collision between the Caribbean plate and South America began about 55 Ma in northwestern Venezuela, with progressive arc collision proceeding from west to east (Figure 2); this has resulted in eastward progressing cessation of subduction and arc volcanism in the Leeward Antilles arc as these terranes collided with the South American continent [e.g., Burke, 1988; Pindell et al., 1988; Pindell and Barrett, 1990; Pindell and Kennan, 2007]. The Caribbean plate is currently moving at ∼20 mm/a approximately due east relative to South America [Pérez et al., 2001; Weber et al., 2001].

Details are in the caption following the image
(a) Plate boundaries (blue) for the Caribbean region (from the University of Texas PLATES database). Black vectors show the direction of motion with respect to the Caribbean plate for the Pacific plate (85.5 mm/a) and the North America plate (11.9 mm/a) according to the Nuvel-1A model [DeMets et al., 1994]. Red box marks the region plotted in Figures 1b and 1c. (b) Topographic map of the study region [Smith and Sandwell, 1997]. The BOLIVAR active seismic program is indicated by red lines (seismic reflection profiles), white dots (ocean bottom seismometers), blue dots (onshore seismic geophones), and yellow stars (onshore sources). White lines show the seismic refraction profiles presented by Officer et al. [1959] and Edgar et al. [1971]. Yellow dots and maroon triangles show the shots and receivers, respectively, of the Lesser Antilles Seismic Project (LASP) experiment [Boynton et al., 1979]. Black cross outlines the region of the seismic reflection/refraction experiment presented by Bangs et al. [2003] and Christeson et al. [2003], and the white dashed line shows the location of velocity profile S1 plotted in Figure 12a. (c) Satellite gravity map of the study region [Sandwell and Smith, 1997]. Other symbols are the same as those in Figure 1a.
Details are in the caption following the image
Time evolution of the Caribbean plate boundary from 90 to 0 Ma as labeled. Since about 60 Ma the southern edge of the Caribbean plate boundary has been interacting with the northern boundary of the South America plate. The dashed line marks the position of Profile TRIN discussed in the text. Figure is modified from Lugo and Mann [1995].

[3] The ∼850 km long, north-south trending Lesser Antilles island arc forms the eastern boundary of the Caribbean plate beneath which the Atlantic crust is subducted at a rate of ∼20 mm/a [DeMets et al., 2000]. This slow convergence rate might be responsible for the relatively low magma production rates at the Lesser Antilles arc in comparison to other arc systems [Macdonald et al., 2000]. Despite a low production rate, the estimated 30- to35-km crustal thickness of the Lesser Antilles arc [Westbrook, 1975; Boynton et al., 1979; Maury et al., 1990] is comparable to that of other well studied arcs with much higher production rates such as the Izu-Bonin (26–32 km [Kodaira et al., 2007a]) and Aleutian (35–37 km [Shillington et al., 2004]) arcs.

[4] Here we present results from a coincident seismic reflection/refraction profile carried out in the SE Caribbean (Figure 1) as part of the BOLIVAR project [Levander et al., 2006]. Profile TRIN provides new constraints on crustal structure of the Aves Ridge, Grenada Basin, Lesser Antilles island arc, Tobago Basin, and Caribbean–South America plate boundary zone. This paper focuses on the analysis and interpretation of the wide-angle data; a companion paper will present the seismic reflection data (P. Mann et al., manuscript in preparation, 2008). A primary result of this study is that the crustal thickness of the Lesser Antilles arc is ∼24 km, which is 5–10 km thinner than previously estimates. Thus the Lesser Antilles is thinner than other well studied arcs, which is consistent with a low production rate. A secondary result is the location of the possible suture between Caribbean plate material and the South America continental margin.

2. Regional Tectonics

[5] The southeast Caribbean is a complex geologic environment which includes (Figure 1) an active island arc (Lesser Antilles arc), a remnant arc (Aves Ridge), and two basins possibly floored by 12- to 14-km-thick oceanic crust (Grenada and Tobago Basins [Boynton et al., 1979]). The Aves Ridge is a presently extinct island arc that drilling and dredging results suggest was active until the Late Cretaceous to early Cenozoic ∼90–55 Ma [Fox et al., 1971; Edgar et al., 1973; Bouysse et al., 1985; Bouysse, 1988]. The southern Lesser Antilles arc has been active for the past 12–15 Ma, with the submarine volcano, Kick-Em-Jenny, in the Grenadines the most southerly active volcano in the chain [Briden et al., 1979; Maury et al., 1990; Speed et al., 1993]. The southern extent of subduction of the North America and perhaps South America plates beneath the Caribbean plate is poorly defined by seismicity data, but seismic data does suggest that a detached and detaching westward subducted South American slab lies beneath continental South America [VanDecar et al., 2003].

[6] The southern Lesser Antilles arc is separated from the Aves Ridge by the Grenada Basin. Limited seismic refraction data, combined with modeling of gravity data, suggests that the Grenada Basin basement corresponds to oceanic crust with a thickness of ∼14 km [Boynton et al., 1979]. Most investigators have argued that the basin formed through back-arc spreading, although extension orientation varies in these models from east-west [Tomblin, 1975; Bird et al., 1993, 1999], to north-south [Pindell and Barrett, 1990], to northeast-southwest [Bouysse, 1988]. An alternate model for Grenada Basin origin is that it is trapped oceanic crust or forearc that formed during an eastward jump or rollback of the subduction zone [Bunce et al., 1970; Malfait and Dinkleman, 1972; Kearey, 1974]. The Tobago Basin may have formed as a consequence of uplift of the Barbados Ridge to its east [Westbrook, 1975]. Alternately, the Tobago basin may have formed a continuous back arc with the Grenada Basin, with the two basins separating during uplift of the southern Lesser Antilles Arc between 16–28 Ma [Speed and Walker, 1991].

[7] The southern Caribbean plate boundary zone with South America is a broad, active transpressional zone that geologic mapping and seismic data suggests may be 100–250 km wide with dextral slip motion occurring on several major fault systems [e.g., Robertson and Burke, 1989; Ysaccis, 1997; Audemard et al., 2005]. The Hinge Line fault zone is located near the shelf break into the Tobago basin (Figure 3), and may mark the northern edge of Cretaceous arc material [Robertson and Burke, 1989]. It is interpreted as a relatively minor strike-slip fault with offset less than ∼10 km [Robertson and Burke, 1989]. The North Coast fault zone consists of two discrete major strands (Figure 3). Deformation that affects the uppermost reflectors imaged in seismic reflection data suggests that the fault zone has had recent activity, with a dip-slip component of offset of as much as 4 km [Robertson and Burke, 1989] but no significant strike-slip motion since 4 Ma [Pindell et al., 2005]. The North Coast fault zone may mark the boundary between the deformed passive margin sedimentary and metasedimentary rocks of South American and allochthonous Caribbean rocks of arc affinity [Algar and Pindell, 1993]. The El Pilar fault zone marks the southern boundary of the Araya-Paria peninsula in eastern Venezuela [Vierbuchen, 1984]; GPS observations indicate that 80% of the dextral slip between the Caribbean and South American plates is contained within a 80-km-wide shear zone centered on the El Pilar fault system [Pérez et al., 2001]. The fault zone continues into Trinidad where it forms the southern boundary of the Northern Range (Figure 3), but estimates are that only 0–20% of the total Caribbean–South American plate motion is accommodated along the El Pilar fault in Trinidad [Robertson and Burke, 1989; Weber et al., 2001]. Instead, GPS measurements indicate that the Central Range fault is the major active strike-slip fault in Trinidad, and accommodates ∼14 mm/a of the 20 mm/a plate motion [Weber et al., 2001]. Offshore the Central Range and El Pilar fault zones merge into a single, unnamed northeast striking fault (Figure 3) [Robertson and Burke, 1989; Garciacaro, 2006].

Details are in the caption following the image
Expanded (a) topographic map of the study region [Smith and Sandwell, 1997] and (b) satellite gravity map [Smith and Sandwell, 1997] of the study region. Red line shows position of Profile TRIN, with ocean bottom seismometers shown by white dots. Striped patterns outline the Tobago terrane as mapped by Speed and Smith-Horowitz [1998]. Deep Sea Drilling Project (DSDP) [Bader et al., 1970; Edgar et al., 1971] and industry wells [Ysaccis, 1997] discussed in the text are shown by black dots; the ellipse encloses five wells that encountered 87–102 Ma primitive island arc basement rocks [Ysaccis, 1997]. Grey lines show the prominent fault zones of the region [Robertson and Burke, 1989; Ysaccis, 1997; Garciacaro, 2006]. A, Araya Peninsula; CRFZ, Central Range fault zone; EPFZ, El Pilar fault zone; HLFZ, Hinge Line fault zone; LT, Los Testigos; NCFZ, North Coast fault zone; NR, Northern Range; P, Paria Peninsula; TCFZ, Tortuga Coche fault zone; SGBDF, Southern Grenada Basin deformation front.

[8] The extensive allochthonous Tobago terrane is at the leading edge of the Caribbean plate and is currently colliding with and suturing to the South America plate within the plate boundary zone [Speed and Walker, 1991; Russo and Speed, 1992; Speed and Smith-Horowitz, 1998] (Figure 3). This terrane derives it name from the island of Tobago where plutonic and volcanic rocks are exposed that are interpreted as middle to Late Cretaceous oceanic-arc crust [Snoke et al., 2001]; these mainly crystalline assemblages differ from metasedimentary rocks exposed at nearby Trinidad. Speed and Smith-Horowitz [1998] include in the Tobago terrane (Figure 3) outcrops from Margarita island, Los Testigos, and the northern coast of Paria, and wells in the Carupano basin (Figure 3) with 87–102 Ma primitive island arc basement rocks [Ysaccis, 1997]. Robertson and Burke [1989] propose that the Hinge Line fault zone may mark the northern edge of the Tobago terrane, but Speed and Walker [1991] and Speed and Smith-Horowitz [1998] argue that the Tobago terrane forms the basement beneath the southernmost Tobago basin (Figure 3). There is a general consensus that the terrane formed west of its present position and has moved to the east along strike-slip faults as the leading edge of the Caribbean plate [e.g., Robertson and Burke, 1989; Erlich and Barrett, 1990; Pindell and Barrett, 1990].

3. Data Acquisition and Processing

[9] Profile TRIN was acquired in May 2004 during the BOLIVAR [Levander et al., 2006] active source seismic experiment. The source was a 20-element seismic source array with a volume of 113 l (6947 in3) towed by the R/V Maurice Ewing. The profile was shot at 50-m shot spacing (∼20-s shot interval) for seismic reflection recording and again at 150-m shot spacing (∼60-s shot interval) for optimal wide-angle data acquisition. Seismic reflection data were recorded on a 6-km-long, 480-channel Syntron digital streamer. 28 WHOI D2 and 11 SIO LC2000 ocean bottom seismograph (OBS) receivers were positioned at 11- to 16-km spacing along the profile (Figure 3), with a few larger gaps owing to recording or logistical difficulties. The WHOI D2 OBS is built and operated by the Woods Hole Oceanographic Institution; these instruments recorded the vertical, hydrophone, and two horizontal channels with a sample interval of 5 ms. The SIO LC2000 OBSs were provided by the Institute of Geophysics and Planetary Physics at Scripps Institution of Oceanography; these instruments recorded a vertical and hydrophone channel at a sample interval of 4 ms.

[10] Standard processing of the OBS data included applying a correction for clock drift during deployment and inverting the direct water-wave arrivals for instrument location. Maximum instrument location errors are estimated to be 300 m. Further processing consisted of applying a Butterworth filter with a low cut of 3 Hz, a high cut of 15 Hz, and a 48 dB/octave roll-off. For most instruments, the vertical channel recorded higher signal-to-noise ratios than the hydrophone channel; however, signals from the hydrophone channel were sometimes superior especially in water depths <200 m. Representative record sections are displayed in Figure 4. Clear first arrivals are observed to an average offset of 55 km, but signal-to-noise ratios are variable with clear arrivals observed on some instruments to offsets of 100–140 km (e.g., negative distances for OBS 409 in Figure 4b) and on other instruments only to offsets <25 km (e.g., negative distances for OBS 435 in Figure 4g). Traveltime picks of first arrivals and PmP secondary arrivals were made by hand using the unfiltered record sections at the closest offsets (generally 0–10 km) and filtered record sections at all other offsets; both vertical and hydrophone channels were inspected and picks were made on the channel with higher signal-to-noise ratios.

Details are in the caption following the image
(a–d) Representative record sections plotted with a reduction velocity of 8 km/s and with a 1-s automatic gain control applied. First-arrival and PmP traveltime picks are shown with the cyan and red lines, respectively. Locations of ocean bottom seismometers (OBSs) are shown in Figures 1 and 5. (e–h) Representative record sections plotted with a reduction velocity of 8 km/s and with a 1-s automatic gain control applied. First-arrival and PmP traveltime picks are shown with the cyan and red lines, respectively. Locations of OBSs are shown in Figures 1 and 5.
Details are in the caption following the image
(continued)

4. Data Analysis

4.1. Tomographic Inversion

[11] For each OBS, we picked first arrival traveltimes; estimated picking errors are 25 ms for offsets <10 km, 50 ms for offsets 10–20 km, 75 ms for offsets 20–40 km, and 125 ms for offsets >40 km. We then used the two-dimensional FAST tomographic inversion method of Zelt and Barton [1998] to constrain the velocity structure along the profile. This approach uses a finite difference traveltime scheme for the forward step [Vidale, 1988; Hole and Zelt, 1995], and a regularized inversion in which a combination of data misfit and model roughness is minimized for the inverse step [Zelt and Barton, 1998]. A modification to the algorithm allows a perturbation constraint [Zelt et al., 1999]. The grid spacing of the velocity model was chosen as 100 m in both the horizontal and vertical directions for the forward calculation, and we chose to parameterize the inverse grid at 1 km in the horizontal direction and 500 m in the vertical direction. A smaller inverse grid spacing results in fine-scale lateral heterogeneities that cannot be adequately resolved with the experimental geometry and is not required to fit the data within the estimated uncertainties.

[12] The starting model for the first tomographic inversion was a simple layered velocity model that fit the first arrivals at the nearest offsets. We used a tomographic procedure that consisted of three steps:

[13] 1. The tomographic inversion is carried out for seven iterations, and at each iteration three smoothing parameters (the lambda parameter of Zelt and Barton [1998]) are tested. The preferred model at each iteration is chosen as the iteration that produced the smallest chi-square value [Bevington, 1969] that is ≥1.0, and the lambda value is then reduced for the next iteration. The final preferred model is the iteration that produced the smoothest model with a chi-square value closest to 1.0 (i.e., the model fits the observed traveltimes within their independently estimated uncertainties).

[14] 2. This preferred model is smoothed below the seafloor and low-velocity zones are removed; the low-velocity zones limit the raypaths and are often an artifact of the starting model and not representative of the true structure. The resulting velocity model is then used as the starting model for another iterative tomographic inversion.

[15] 3. Calculated traveltimes through the resulting velocity model are compared with the observed traveltimes; this allows us to remove traveltime picks that are clearly outliers and to pick additional traveltimes at the further offsets where signal-to-noise ratios are lower. We repeated these steps several times, and the final tomographic velocity model is displayed in Figure 5a. The final inversion included ∼24,000 traveltime picks from 39 OBS receivers, with a RMS traveltime residual of 74 ms and a chi-square value of 1.0.

Details are in the caption following the image
(a) Velocity model obtained by the tomographic inversion of first-arrival traveltimes; only the region constrained by raypaths is shown. Contours are plotted every 0.5 km/s. Dots at seafloor mark the positions of OBS receivers. Arrows mark the locations of prominent faults that are crossed by the profile. (b) Final velocity model for Profile TRIN; only the region constrained by raypaths is shown. Moho depth is plotted by a grey solid line, and white squares indicate the position of nodes used in inversion for Moho interface depth. Velocity values for the base of the crust and upper mantle are labeled.

4.2. Moho Modeling

[16] Velocities of 7.5–7.9 km/s are observed in the tomographic velocity model (Figure 5a) at depths >19 km and provide an estimate of Moho structure beneath the profile. However, this velocity model is constrained only by first arrival refracted traveltimes; therefore we carried out a joint inversion of both reflected and refracted arrivals to better resolve Moho depth. We used a hybrid forward/inverse traveltime modeling code based on the inversion method of Zelt and Smith [1992]. This technique requires a layered velocity model. Our starting velocity model was computed from the tomographic velocity model, with the velocity grid between the seafloor and a depth near the 6.5 km/s contour converted into 10 layers. The initial Moho interface depth was positioned approximately at the 7.5 km/s contour of the tomographic velocity model, and velocity values at the base of the crust and in the upper mantle were set at 7.0 km/s and 8.0 km/s, respectively. PmP and Pn traveltime picks were made from the OBS record sections; identification of these arrivals was assisted by calculating PmP and Pn traveltimes for the starting velocity model.

[17] The inversion method uses a damped, least squares technique to minimize traveltime residuals [Zelt and Smith, 1992]. We set pick uncertainty for Pn and PmP arrivals at 125 ms and 150 ms, respectively; PmP arrival uncertainties were higher since the secondary arrivals are more difficult to pick and might inadvertently include some diffraction energy. We inverted for velocity at the base of the crust, velocity within the upper mantle, and Moho depth at 14 nodes; Moho depth nodes were spaced 25 km apart except where ray coverage is poor (e.g., model offsets >125 km) or where finer spacing was required owing to rapid lateral changes in structure (e.g., near model offset 100 km). The inversion failed to converge for the closely spaced depth nodes located between 75–125 km and thus these nodes were adjusted through forward modeling to minimize traveltime misfits. The final velocity model and Moho depth nodes are displayed in Figure 5b. We carried out inversions that included additional Moho depth nodes and/or additional velocity nodes at the base of the crust and upper mantle, but the resulting models decreased the RMS traveltime residuals by <5 ms and resulted in greater Moho structure; thus the simpler model is preferred.

[18] Figure 6 compares observed and calculated traveltimes through the final velocity model for a subset of the receivers along the profile, and Figure 7 compares misfits for all PmP and Pn arrivals. For some arrivals, Figures 6 and 7 show that the calculated traveltimes are systematically too early or too late; however, adjusting the model to better fit one set of traveltimes will result in a poorer fit to other travel times. The RMS traveltime residuals for the velocity model displayed in Figure 5b are 138 ms for the PmP arrivals and 114 ms for the Pn arrivals. The overall chi-square value for all PmP and Pn arrivals is 0.84, which indicates that the model fits the observed data slightly better than the assigned uncertainties.

Details are in the caption following the image
(a–j) Comparison of observed and predicted traveltime picks for selected OBS receivers. Observed first-arrival and PmP traveltime picks are shown in cyan and red, respectively, with height equal to the pick uncertainty. Calculated first-arrival traveltime curves are shown with blue dashed lines, and calculated traveltimes for PmP reflections from the Moho interface are shown with maroon dashed lines. All picks are plotted with a reduction velocity of 6 km/s. (k–t) Comparison of observed and predicted traveltime picks for selected OBS receivers. Observed first-arrival and PmP traveltime picks are shown in cyan and red, respectively, with height equal to the pick uncertainty. Calculated first-arrival traveltime curves are shown with blue dashed lines, and calculated traveltimes for PmP reflections from the Moho interface are shown with maroon dashed lines. All picks are plotted with a reduction velocity of 6 km/s.
Details are in the caption following the image
(continued)
Details are in the caption following the image
Comparison of observed and predicted PmP and Pn traveltime picks plotted at a reduction velocity of 6 km/s. All observed picks included in the final modeling for Moho depth are shown.

4.3. Tomographic Model Resolution

[19] An objective measure of model resolution can be obtained by computing a series of checkerboard tests. Using the method detailed by Zelt [1998], we carried out a total of 28 checkerboard tests. Each checkerboard consisted of alternating positive and negative velocity anomalies of 10% superimposed on the final velocity model at depths below the seafloor. Cell sizes in the horizontal direction were 12.5, 25, 50, 75, 100, 125, and 150 km. Cell sizes in the vertical direction were 10% of the horizontal direction (i.e., 1.25–15 km). There were 4 different checkerboard patterns for each cell size in order to reduce the nonlinearity associated with the grid polarity and registration [Zelt, 1998]. Traveltimes were calculated through each checkerboard model using the real shot-receiver geometry. Random noise was added to the calculated traveltimes, which were then inverted using the final velocity model as a starting model. An example of a model checkerboard and a recovered checkerboard are shown in Figures 8a and 8b. The semblance was then computed between the actual and recovered velocity perturbations; a node with a semblance of >0.7 is considered to be well resolved [Zelt, 1998]. The semblances for all 4 checkerboards for each cell size were then averaged; an example semblance result is shown in Figure 8c. The horizontal resolution at each node was obtained by calculating the minimum cell size at that point that could be recovered with an average semblance of 0.7 (Figure 8d). These results show that average horizontal resolution is <20 km to depths ∼5 km beneath the seafloor, and 20–60 km at depths 5–10 km beneath the seafloor. At greater depths the horizontal resolution is more variable along the profile (Figure 8d).

Details are in the caption following the image
(a) Sample checkerboard for an anomaly width of 25 km. (b) Recovered velocity perturbation for an anomaly width of 25 km. (c) Average semblance between actual and recovered velocity perturbation for four checkerboard patterns with an anomaly width of 25 km. A location with a semblance of >0.7 is considered to be well resolved. (d) Estimated horizontal resolution.

4.4. Moho Model Uncertainty

[20] We can obtain an estimate of Moho depth uncertainties by comparing three independent methods to determine Moho depth: the tomographic inversion of first arrivals (Moho depth is estimated from the 7.3–7.7 km/s contours shown as red lines in Figure 9), the preferred joint inversion of PmP and Pn arrivals using upper structure determined from tomographic inversion (black line in Figure 9), and an inversion for interface depth [Zelt et al., 2003] using only PmP arrivals with upper structure determined from tomographic inversion and a lower crustal velocity of 7.3 km/s (cyan dots in Figure 9). The maximum difference between model Moho depths is 4.0 km. Our estimated uncertainty for Moho depth is ±2 km; 95% of the Moho depths from the two alternate models are within this error range. Moho depths between model offsets 110–160 km and 250–265 km are constrained by only a few instruments with no reversed arrivals, and thus could have a larger depth uncertainty than elsewhere in the model.

Details are in the caption following the image
Comparison of Moho depth calculated using three different methods; depths are only plotted where constrained by data. Solid black line shows preferred model with Moho depth calculated from a joint inversion of PmP and Pn arrivals (Figure 5b). Red lines indicate the 7.3, 7.4, 7.5, 7.6, and 7.7 km/s velocity contours from the tomographic inversion of first-arrival traveltimes (Figure 5a). Cyan dots show the Moho depth from an inversion of PmP traveltimes. The dashed lines plot the preferred Moho depth ±2 km, which is the estimated depth uncertainty.

4.5. Comparison With Seismic Reflection Interpretation

[21] Aitken [2005] and P. Mann et al. (manuscript in preparation, 2008) provide an interpretation of the seismic reflection data acquired along Profile TRIN (Figures 10a and 11). We converted this interpretation to depth using our final velocity model, and compare the interpreted horizons and faults with our velocity structure in Figure 10b. In addition, we have computed sediment and crustal thickness variations along the profile (Figure 10c). Sediment thickness was computed using the vertical distance between the seafloor and interpreted basement horizon (from seismic reflection interpretation), and crustal thickness was computed using the vertical distance between the interpreted basement horizon and the modeled Moho depth (from velocity model). Calculated sedimentary and crustal thicknesses vary widely along the profile. There is little to no sediment at the Aves Ridge, Lesser Antilles Arc, and near Tobago; in contrast there is as much as 10–15 km of sediment in the Grenada and Tobago Basins. Crustal thickness is greatest (25–30 km) beneath the Aves Ridge, Lesser Antilles Arc, and the southeastern end of the Tobago Basin; the crust is considerably thinner (4–7 km) beneath the Grenada and Tobago Basins. A basement horizon could not be interpreted southeast of the North Coast fault zone (Figure 11c).

Details are in the caption following the image
(a) Interpretation of seismic reflection data for Profile TRIN: faults (black), seafloor (dark blue), basement (cyan), and sedimentary horizons (various colors). The Moho interface from the model displayed in Figure 5b is shown in purple and is dashed where unconstrained by data. (b) Velocity model overlain with interpreted faults and horizons. The interpreted basement horizon is shown with heavy black lines, and the modeled Moho interface is shown with heavy yellow lines. (c) Calculated thicknesses of sediment (red) and crust (blue) along the profile. Sediment thickness is computed between the seafloor and interpreted basement interface, and crustal thickness is computed between the basement and Moho interfaces. Dashed lines show regions where either the basement horizon or Moho interface depths are unconstrained.
Details are in the caption following the image
Seismic reflection images for the (a) NW portion of Profile TRIN, (b) central portion of Profile TRIN, and (c) SE portion of Profile TRIN. Interpreted faults are shown in yellow, the basement horizon is shown in cyan, and sedimentary horizons are shown in other colors. Overlain in white are the velocity contours from the model displayed in Figure 5b; contours are only plotted where constrained by raypaths. The Moho interface is shown in purple and is dashed where unconstrained by data.

5. Discussion

5.1. Lesser Antilles Arc and Aves Ridge

[22] The Aves Ridge is interpreted as a remnant island arc, and exhibits a comparable crustal structure to the active Lesser Antilles arc as shown by a plot comparing velocity with depth below seafloor for the two features (Figure 12a). Both features show a large increase in velocity from ∼2 km/s at the seafloor to 6.0–6.3 km/s at 6- to 7-km depth, and then a more gradual increase to 7.3 km/s at the base of the crust at 24–26 km. Heat flow measurements indicate that the Lesser Antilles arc is slightly hotter than the Aves Ridge, which is consistent with the interpretation of the Aves Ridge as a remnant island arc [Clark et al., 1978]. Seismic velocities decrease with increased temperatures [e.g., Murase and McBirney, 1973; Kern and Richter, 1981] and thus we would expect the Lesser Antilles arc to have slower velocities than the Aves Ridge at similar depths; instead, our observed velocities in the middle and lower crust are ∼0.2 km/s faster at the Lesser Antilles compared to the Aves Ridge. One explanation for the velocity observations is that the faster velocities of the Lesser Antilles results from a composition that is less silicic (lower SiO2 content) or more mafic (higher MgO content) than the Aves Ridge [e.g., Behn and Kelemen, 2003]; however, drilling of the Aves Ridge did not penetrate basement and thus compositional differences between the two regions are currently unknown. In Figure 12a we also plot the velocity-depth profile of the accretionary wedge backstop measured near 16°N [Christeson et al., 2003] (Figure 1). The backstop has been interpreted as having an island arc composition [Bangs et al., 2003; Christeson et al., 2003], and the similarity in velocity structure between the backstop and the island arc supports this interpretation.

Details are in the caption following the image
(a) Velocity-depth profiles from the Lesser Antilles arc (x = 50 km), Aves Ridge (x = −125 km), near Tobago (x = 200 km), and the island arc backstop near 16°N [Christeson et al., 2003] (location shown in Figure 1). (b) Velocity-depth profiles from the Lesser Antilles (x = 50 km), Aleutian [Shillington et al., 2004; Van Avendonk et al., 2004], and Izu-Bonin [Kodaira et al., 2007a] arcs. The Aleutian profile is the average velocity profile through the along-axis profile of Shillington et al. [2004] and Van Avendonk et al. [2004]. The Izu-Bonin profiles are the average of the slow-average velocity and fast-average velocity profiles from Kodaira et al. [2007a]. Also plotted are the average continental crust values presented by Christensen and Mooney [1995].

[23] Crustal thickness in Figure 5b is ∼26 km at the Aves Ridge and ∼24 km at the Lesser Antilles arc; similar crustal thickness estimates of 26–28 km are observed on the eastern extension of these features near 64°W and 67°W [Clark et al., 2008; M. B. Magnani et al., Crustal structure of the South America–Caribbean plate boundary at 67°W from controlled-source seismic data, submitted to Journal of Geophysical Research, 2008]. These values are less than previous crustal thickness estimates of 30–35 km across the Lesser Antilles arc [Westbrook, 1975; Boynton et al., 1979; Maury et al., 1990] and 30–40 km [Kearey, 1974; Boynton et al., 1979] for the Aves Ridge. The differences in crustal thickness estimates may reflect spatial variability (previous experiments were farther to the north, Figure 1), or could be an artifact of the previous techniques which relied primarily on nonunique gravity modeling with some sparse seismic constraints [Kearey, 1974; Westbrook, 1975; Boynton et al., 1979; Maury et al., 1990]. In order to determine which explanation is more likely, we carried out gravity modeling (Figure 13) over the Aves Ridge and Lesser Antilles arc using a density model based on the Profile TRIN crustal structure. Our density model is composed of water, upper sediment, lower sediment, basement, and mantle layers. The average velocity for each layer was converted to density using the standard relationship of Ludwig et al. [1970]. Gravity was computed, model densities were adjusted of the Lesser Antilles arc basement and the layers northeast of the Aves Ridge, and crustal thickness was decreased at the northeast edge of the model where thinner crust is observed on a nearby seismic profile [Clark et al., 2008]. Gravity modeling is nonunique, but the density model plotted in Figure 13b is one model that fits the crustal structure constrained by seismic data and the primary amplitude variations of the observed gravity signal.

Details are in the caption following the image
(a) Comparison of observed and modeled gravity signal over the Aves Ridge and Lesser Antilles arc. Observed signal is shipboard free-air gravity acquired along Profile TRIN during the seismic experiment; modeled signal is computed for the density model shown in Figure 13b. (b) Density model based on crustal structure observed in the Profile TRIN velocity model. Units are g/cc.

[24] The average crustal velocity of the Aves Ridge and Lesser Antilles arc along Profile TRIN 6.1–6.2 km/s, which is consistent with a density of ∼2.7 g/cc [Ludwig et al., 1970]. The gravity modeling suggests that this density value and measured Moho depth fits the Aves Ridge and Grenada Basin gravity signal reasonably well, but that a decreased density of ∼2.66 g/cc provided a better fit to the Lesser Antilles arc gravity signal. The modeled density difference is consistent with the observed heat flow measurements [Clark et al., 1978], as hotter Lesser Antilles arc rocks would be expected to have a lower density than the Aves Ridge remnant arc [e.g., Murase and McBirney, 1973]. Previous gravity models of the Lesser Antilles arc and the Aves Ridge assumed densities of 2.8–2.9 g/cc [Kearey, 1974; Westbrook, 1975; Boynton et al., 1979; Maury et al., 1990] which are typically associated with velocities of ∼6.5–6.8 km/s [Ludwig et al., 1970]; crustal velocities of 6.5–6.8 km/s are much higher than observed in our velocity model or further to the north near 16°N [Christeson et al., 2003] (Figure 12a). Modeling indicates that increasing the density from 2.7 to 2.85 g/cc results in an increased crustal thickness of 8 km, which would explain the discrepancy between the crustal thicknesses of 24–26 km observed along Profile TRIN and previous crustal thickness estimates of 30–40 km from gravity models. We argue that a density of 2.65–2.7 g/cc is a more realistic value for the Lesser Antilles Arc and Aves Ridge than the density values used in previous models; this implies that the Lesser Antilles arc is ∼25 km thick at the locations of previous gravity studies near St. Vincent, Dominica, and Antiqua [Maury et al., 1990]. The similarity in seismic velocity structure and crustal thickness between the Aves Ridge and Lesser Antilles arc implies that magmatic processes have remained moderately steady over time.

5.2. Comparison of Lesser Antilles Arc to Other Arc Systems

[25] Oceanic island arcs that have been studied with high-resolution seismic experiments include the Izu-Bonin [Suyehiro et al., 1996; Kodaira et al., 2007a; Takahashi et al., 2007] arc in the western Pacific and the Aleutian arc [Shillington et al., 2004; Van Avendonk et al., 2004] in the northern Pacific; in Figure 12b we compare velocity-depth profiles of the Izu-Bonin, Aleutian, and Lesser Antilles arcs. The Lesser Antilles arc is composed of an upper crust with velocities increasing from 2 km/s at the seafloor to 5.9–6.3 km/s at 6.0–7.0 km depth, and a lower crust with velocities reaching ∼7.3 km/s at the base of the crust. The Aleutian island arc has an ∼10-km-thick upper crustal layer with velocities of 6.0–6.5 km/s, a ∼10-km-thick middle crustal layer with velocities of 6.5–7.3 km/s, and a ∼15- to 17-km-thick lower crustal layer with velocities of 7.3–7.7 km/s; total crustal thickness of the Aleutian arc is 35–37 km [Shillington et al., 2004]. The Izu-Bonin arc has a 5- to 7-km-thick upper crust with velocities of 1.8–5.8 km/s, a 3- to 13-km-thick middle crust with velocities of 6.0–6.8 km/s, and a 12- to 18-km-thick lower crust with velocities of 6.8–7.6 km/s; total crustal thickness of the Izu-Bonin arc is 26–32 km [Kodaira et al., 2007a].

[26] The upper crust of the Izu-Bonin arc is interpreted as a mixture of sediments, volcaniclastics, and volcanic rocks with a mainly basaltic composition [Kodaira et al., 2007a]. The Aleutian arc has significantly faster velocities in the upper crust near the seafloor (∼6 km/s) compared to the Lesser Antilles and Izu-Bonin arcs (∼2 km/s); these faster velocities are interpreted as fractured plutonic rocks with some volcanic flows [Shillington et al., 2004]. The upper crust of the Lesser Antilles arc has a velocity structure comparable to the Izu-Bonin arc but significantly slower than the Aleutian arc (Figure 12b); this suggests that the upper crust of the Lesser Antilles and Izu-Bonin arcs are similar. Observations of basaltic rocks on the nearby island of Grenada [Devine, 1995] and elsewhere in the Lesser Antilles island arc [Macdonald et al., 2000] support this interpretation.

[27] The middle crust of the Izu-Bonin arc is a 3- to 13-km-thick layer with velocities of 6.0–6.8 km/s at depths between 5–20 km [Kodaira et al., 2007a]. The Izu-Bonin velocity profiles shown in Figure 12b correspond to regions of thick (slow average crustal velocities) and thin (fast average crustal velocities) middle crust. Kodaira et al. [2007a] argue that the upper part of the Izu-Bonin middle crust consists mainly of tonalite containing high SiO2, and that the lower part consists of plutonic rocks of intermediate compositions. The Izu-Bonin middle crust is interpreted to form through the differentiation of an initial basaltic crust which produces a crust with intermediate to felsic components having a higher SiO2 content [Tatsumi, 2000; Takahashi et al., 2007]. The middle crust of the Aleutian arc is ∼10 km thick with velocities of 6.5–7.3 km/s, and is interpreted as oceanic crust from the relic Kula plate intruded and thickened by plutons of andesitic to basaltic composition [Shillington et al., 2004]. The slower middle crust of the Izu-Bonin arc compared to the Aleutian arc is attributed to the tonalite composition of the Izu-Bonin middle crust [Kodaira et al., 2007a]; seismic velocities decrease with increasing silica content [e.g., Behn and Kelemen, 2003]. The middle and lower crust of the Lesser Antilles arc are associated with velocities that increase smoothly from 5.9–6.3 km/s at 6.0- to 7.0-km depth to 7.3 km/s at the base of the crust. The velocity structure in the 5- to 20-km-depth range is most similar to the Izu-Bonin fast-average crustal velocity profile that is associated with a thin middle crust. The absence of a distinctive middle crust makes it unlikely that a significant (>5 km?) tonalite layer is present at the Lesser Antilles arc.

[28] The 15- to 17-km-thick lower crust of the Aleutian arc has velocities of 7.3–7.7 km/s and is interpreted as a mixture of mafic and ultramafic cumulates [Shillington et al., 2004]. The Izu-Bonin arc has a 12- to 18-km-thick lower crust with velocities of 6.8–7.6 km/s which is interpreted as gabbroic plutons (6.8–7.2 km/s) overlying mafic and ultramafic cumulates (7.2–7.6 km/s) [Kodaira et al., 2007a]. The interpreted cumulates beneath the Aleutian and Izu-Bonin arcs likely formed during intracrustal fractionation. There are no observed lower crustal velocities >7.3 km/s at the Lesser Antilles arc, and thus no evidence for mafic or ultramafic cumulates underlying this arc. The Lesser Antilles arc is associated with low magma production rates [Wadge, 1984] which is consistent with a lack of a lower crustal cumulate layer and the thinner crust (∼24 km) in comparison to the Aleutian (35–37 km [Shillington et al., 2004]) and Izu-Bonin (26–32 km [Kodaira et al., 2007a]) arcs.

[29] Previous studies have proposed that island arcs form a significant source of continental crust [e.g., McLennan and Taylor, 1982; Kelemen, 1995; Holbrook et al., 1999; Kodaira et al., 2007b]. In Figure 12b we plot the mean velocity structure of continental crust [Christensen and Mooney, 1995] which has an overall average velocity of ∼6.55 km/s; in comparison the Lesser Antilles arc has a mean crustal velocity of 6.75 km/s below 5.5 km depth. Similar to other arcs [e.g., Holbrook et al., 1999; Kodaira et al., 2007b], the Lesser Antilles arc would need to be thickened and modified by the addition of silica during the accretion process to generate a velocity structure of typical continental crust.

5.3. Grenada Basin

[30] The Grenada Basin is positioned between the Aves Ridge, a remnant arc, and the currently active Lesser Antilles arc. Interpretations for Grenada Basin formation are that either it formed through back-arc spreading [Tomblin, 1975; Bouysse, 1988; Pindell and Barrett, 1990; Bird et al., 1993, 1999] or that it is trapped oceanic crust or forearc that formed during an eastward shift or rollback of the subduction zone from the Aves Ridge to the present-day location of the Lesser Antilles arc [Bunce et al., 1970; Malfait and Dinkleman, 1972; Kearey, 1974; Aitken, 2005]. Our profile provides new constraints on the structure of the Grenada basin, and indicates that the Moho depth decreases from 22–27 km beneath the Aves Ridge and Lesser Antilles arc to a minimum of 18–19 km beneath the northwestern portion of the Grenada Basin (Figure 5b). Crustal thickness estimates are 4–10 km beneath most of the basin (Figure 10c); previous estimates from lower-resolution seismic data acquired ∼150 km to the north are 7 km [Officer et al., 1959] and 14 km [Boynton et al., 1979]. Our model suggests a crustal thinning from northwest to southeast across the basin, but we note that the crustal thickness values are heavily reliant on the interpreted depth to top of basement, which is poorly resolved at the greater basement depths in the southeast portion of the basin (Figure 10b).

[31] Back-arc extension thins existing arc crust and can ultimately lead to seafloor spreading and generation of oceanic crust. Our crustal thickness estimates of 4–10 km bracket the average value for oceanic crustal thickness of 7 km [White et al., 1992], and our crustal seismic velocities of 5.5–7.3 km/s are consistent with basaltic and gabbroic compositions [Christensen and Salisbury, 1975; Christensen, 1978]. Trapped oceanic crust cannot be distinguished using seismic velocities from oceanic crust formed by back-arc extension. The crust underlying the Tobago Basin, which has velocities of 5.5–7.3 km/s and a crustal thickness of 6–8 km, is similar to Grenada Basin crust and thus consistent with the trapped forearc model for Grenada Basin formation. Thus our measured seismic structure would support either the backarc or trapped oceanic crust model for the Grenada Basin.

5.4. Tobago Basin

[32] Two models for the formation of the Tobago Basin are (1) that it formed as a consequence of uplift of the Barbados Ridge to its east [Westbrook, 1975] and westward thrusting along the contact between the prism and the basin [Torrini and Speed, 1989] or (2) that it formed a continuous back arc [Speed and Walker, 1991] or forearc [Aitken, 2005] with the Grenada Basin, with the two basins separating during uplift of the southern Lesser Antilles Arc between 16–28 Ma [Speed and Walker, 1991]. Boynton et al. [1979] measure an average thickness of 12 km for crust beneath the Tobago Basin west of Barbados, and interpret this as the oceanic crust upon which the Lesser Antilles arc formed. They suggest that the crust may have been thickened tectonically early in the history of the arc (>24 Ma). We measure a thickness of 6–8 km and velocities of 5.5–7.3 km/s for the crust underlying the Tobago Basin along our profile (Figure 10); these values are consistent with normal oceanic crust. The arc may have been built on crust similar to that of the Venezuela Basin, where the crust includes both normal oceanic crust (thickness ∼4–5 km [Diebold et al., 1999]) and thicker plateau crust (thickness ∼10 km [Diebold et al., 1999]). The measured variability in crustal thickness of Venezuela Basin crust [Diebold et al., 1999] is similar to that observed of the Tobago Basin crust, and is consistent with the Lesser Antilles arc being built on normal 6- to 8-km-thick oceanic crust along our profile at the southern end of the arc, and on 12-km-thick plateau crust west of Barbados. Alternatively, the similarity in crustal thickness and seismic velocities between crust underlying the Grenada and Tobago Basins along our profile would support the model where the two regions formed a continuous basin before separation during uplift of the Lesser Antilles arc [Speed and Walker, 1991].

5.5. Southeast Caribbean Plate Boundary Zone

[33] The Profile TRIN velocity model displays a significant lateral change in structure between model offsets 200–225 km, near the intersection of the North Coast fault zone (Figures 1 and 10b) with the seismic profile. Average velocities in the upper 7 km decrease by ∼1 km/s across the fault zone region (e.g., average velocity from 0–7 km is 5.0 at model offset 200 km and 4.0 km/s at model offset 230 km in Figure 10b). The observed lateral change in velocity structure suggests a change in crustal composition, and would be consistent with the North Coast fault zone forming the southern boundary of the Tobago terrane [Speed and Smith-Horowitz, 1998]. Lateral changes in structure are more gradual northwest of Tobago, but there is a change in basement slope near model offset 149 km, which is near the Hinge Line fault zone (Figure 10b). This is consistent with the Robertson and Burke [1989] model which has this fault zone forming the northern boundary of the Tobago terrane. The alternate model of Speed and Smith-Horowitz [1998], with a Tobago terrane extending to the Grenada Basin, would require that crustal thickness of the terrane decrease from 28 km southeast of the Hinge Line fault zone to 6 km beneath the Tobago basin; this thinning might be related to the formation processes of the Tobago basin. The crust in the Tobago terrane between the Hinge Line and North Coast fault zones has velocities of 5.8–6.1 km/s at depths of 3–8 km, and has a velocity structure at these depths similar to that of the Aves Ridge and Lesser Antilles arc (Figure 11a) which is consistent with an interpretation of oceanic-arc origin for the terrane.

[34] Pindell et al. [2005] recognize little or no significant strike-slip motion on the North Coast fault zone since 4 Ma, and argue that the Caribbean plate has become coupled with the basement of South America across this fault. This implies that the North Coast fault zone is the suture between allochthonous Caribbean material and the South American plate. Strike-slip motion between the Caribbean and South American plates has jumped to the south, and is currently accommodated along the Central Range fault in Trinidad [Weber et al., 2001] which intersects Profile TRIN ∼30 km south of the North Coast fault zone. An alternative model is that the Tobago Terrane continues south of the North Coast fault zone to the El Pilar fault zone and underlies the Northern Range of Trinidad [Pindell and Kennan, 2007]; this model is only consistent with our velocity model (which has velocities of 5.5–6.5 km/s for the terrane, Figure 5) if the Tobago Terrane is located at depths greater than 10 km between the North Coast and El Pilar fault zones. Our preferred model is that the Tobago Terrane ends at the North Coast fault zone; this is consistent with the almost vertical velocity contours observed in the upper crust at this location (Figure 5). The material lying south of the North Coast fault zone along the seismic profile is probably equivalent to the rocks exposed in the Northern Range of Trinidad and on the Paria Peninsula of northern Venezuela: metamorphosed passive margin sedimentary rocks of northern South America [e.g., Algar and Pindell, 1993; Avé Lallemant, 1997]. Metamorphism typically increases seismic velocities; this is consistent with the observed higher average upper crustal seismic velocities of this material in comparison to the sediment deposited in the Columbus Basin (Figure 5). An implication of this interpretation is that passive margin sediments originally deposited on the South American plate have been transferred to the Caribbean plate as the strike-slip motion jumped south of the suture to the Central Range fault zone.

5.6. Columbus Basin

[35] The southernmost edge of the subduction zone of South America beneath the eastern Caribbean plate is positioned approximately southeast of Trinidad and Tobago [e.g., Case and Holcombe, 1980], near the location of our seismic profile. Subduction has ended further to the west following collision of the Caribbean plate with the South American margin; even further west a polarity reversal has occurred with the Caribbean plate starting to subduct beneath South America [Van der Hilst and Mann, 1994]. Velocities >6 km/s are observed beneath the Columbus Basin at a depth of ∼18 km; we interpret these velocities as the top of the subducting passive margin of South America. Reflectors parallel to, but shallower than, the top of the subducting plate are observed at 8–10 s TWTT (11–14 km) at the southeastern end of our profile (Figure 10); these are interpreted as Cretaceous age clastic and carbonate rocks overlying the subducting plate [Garciacaro, 2006]. Wide-angle reflections observed on the OBS data place the base of the subducting crust at ∼29 km depth, which implies a crustal thickness of 11 km.

6. Summary of Conclusions

[36] 1. The velocity structure of the Aves Ridge and Lesser Antilles arc are similar, with a large increase in velocity from ∼2 km/s at the seafloor to 6–6.3 km/s at 6- to 7-km depth, and then a more gradual increase to 7.3 km/s at the base of the crust. Crustal thickness is ∼26 km at the Aves Ridge and ∼24 km at the Lesser Antilles arc, and gravity modeling suggests similar crustal thickness elsewhere in the Lesser Antilles arc. The similarity in seismic velocity structure and crustal thickness between the Aves Ridge and Lesser Antilles arc implies that magmatic processes have remained moderately steady over time.

[37] 2. In comparison to the Izu-Bonin and Aleutian arcs, the Lesser Antilles arc is thinner and has no evidence for a lower crustal cumulate layer. This is consistent with the estimated low magma production rates [Wadge, 1984] of the Lesser Antilles arc.

[38] 3. The crustal thickness beneath the Grenada Basin is 4–10 km with seismic velocities of 5.5–7.3 km/s. Tobago Basin crust is similar to Grenada Basin crust, with velocities of 5.5–7.3 km/s and a crustal thickness of 6–8 km. The structure underlying both basins is consistent with oceanic crust, upon which the Lesser Antilles arc was built [Speed and Walker, 1991], but also supports models of trapped forearc crust underlying the Grenada Basin [Bunce et al., 1970; Malfait and Dinkleman, 1972; Kearey, 1974] or thinned Tobago terrane underlying the Tobago Basin [Speed and Smith-Horowitz, 1998].

[39] 4. Lateral variability in seismic structure suggests that the Tobago terrane is bounded on the south by the North Coast fault zone, and on the north by the Hinge Line fault zone. Seismic velocities of the terrane are similar to that of the Aves Ridge and Lesser Antilles arc, and thus consistent with an arc origin for the Tobago terrane.

[40] 5. We interpret the North Coast fault zone as the suture between allocthonous Caribbean arc rocks derived from the west and the passive margin of South American plate. Oblique motion between the Caribbean and South American plates is now accommodated to the south along the Coastal Range fault zone, and thus material originally deposited at depth on the passive margin of South America has now offscraped and uplifted as a narrow strip of the metasedimentary rocks in the Northern and Central Ranges of Trinidad and been transferred to the Caribbean plate.

[41] 6. Although subduction is ending in the Southeast Caribbean as a result of arc-continent collision, a 11-km-thick downgoing passive margin of South America is identified; the top of this crust is at a depth of ∼18 km beneath the Columbus Basin.

Acknowledgments

[42] We thank Alan Levander for his leadership of the BOLIVAR group, Michael Schmitz for his collaboration, and Colin Zelt, Beatrice Magnani, and Harm van Avendonk for helpful discussions. We thank the captain, crew, and science party of the R/V Seward Johnson II SE Caribbean OBS cruise and R/V Maurice Ewing cruise EW0404 for their assistance in data acquisition, with special thanks to Vic Bender, David DuBois, Rob Handy, Peter Lemmond, Jim Ryder, and Steve Swift of the WHOI OBSIP group. We thank John Collins, Bob Detrick, and Tim Askew for their assistance in coordinating this two-ship project. Colin Zelt, Donna Shillington, Associate Editor Doug Toomey, and an anonymous reviewer provided valuable comments on an earlier draft of this manuscript. Our research was supported by NSF Continental Dynamics grants EAR-0003588 and EAR-0607801. UTIG contribution 1992.