1 The University of Sydney
Edgeworth David Building, F05
Department of Geology and Geophysics
NSW 2006, Australia
E-mail: dietmar@es.su.oz.au
Fax: 61 2 9351 2003
2 Géosciences Azur
06235 Villefranche Sur Mer cedex, France
E-mail: royer@ccrv.obs-vlfr.fr
3 Scripps Institution of Oceanography, UCSD
9500 Gilman Drive, La Jolla, CA 92093-0215
E-Mail: cande@gauss.ucsd.edu
4 Geological Survey of Canada, Geophysics Division,
1 Observatory Crescent, Ottawa K1A 0Y3, Canada
E-mail: roest@agg.emr.ca
5 All-Union Res. Inst. for Geology and Min. Res. of the World Ocean
Saint-Petersburg, Russia
Abstract
Introduction
Magnetic Anomaly Dat
Fracture Zone Data
Reconstraction Metho
Results
North America-Africa and South America-Africa finite rotations
North America-South America
finite rotations
North America-South America
plate motions in the Caribbean are
Discussion
Lesser Antilles to Mid-Atlantic
Ridge
Middle
America Trench to Lesser Antilles
Plate motions relative to the mantle
Conclusion
Acknowledgment
Reference
Movie
We review the plate tectonic evolution of the Caribbean area based on a revised model for the opening of the central North Atlantic and the South Atlantic, as well based on an updated model of the motion of the Americas relative to the Atlantic-Indian hotspot reference frame. We focus on post-83 Ma reconstructions, for which we have combined a set of new magnetic anomaly data in the central North Atlantic between the Kane and Atlantis fracture zones with existing magnetic anomaly data in the central North and South Atlantic oceans and fracture zone identifications from a dense gravity grid from satellite altimetry to compute North America-South America plate motions and their uncertainties. Our results suggest that slow sinistral transtension/strike-slip between the two Americas at rates roughly between 3 and 5 mm/y lasted until chron 25 (55.9 Ma). Subsequently our model results in northeast-southwest oriented convergence until chron 18 (38.4 Ma) at rates ranging between 3.7±1.3 to 6.5±1.5 mm/y from 65°W to 85°W, respectively. This first convergent phase correlates with a Paleocene-Lower Eocene calc-alkaline magmatic stage in the West Indies, which is thought to be related to northward subduction of Caribbean crust during this time. Relatively slow convergence until chron 8 at rates from 1.2±0.9 to 3.6±2.1 mm/y from 65°W to 85°W, respectively, is followed by a drastic increase in convergence velocity. After chron 8 (25.8 Ma), probably at the Oligocene-Miocene boundary, this accelerated convergence resulted in 92±22 km convergence from chron 8-6, 127±25 km from chron 6-5, and 72±17 km from chron 5 to the present measured at 85°W near the North Panama Deformed Belt at convergence rates averaging 9.6±3.1 and 9.6±2.1 mm/y from chron 8-6 and chron 6-5, respectively, slowing down to 5.2±1.3 mm/y after chron 5. Neogene convergence measured at the eastern Muertos Trough, at 17.5°N, 65°W, is 41±18 km from chron 8-6, 58±25 km from chron 6-5, and 22±17 km from chron 5-present day, at rates between 4.4±1.7 and 1.6±1.0 mm/y. These well-resolved differential plate motions clearly show an east-west gradient in plate convergence in the Neogene, correlating well with geological observations. We suggest that the early Miocene onset of underthrusting of the Caribbean oceanic crust below the South American borderland in the Columbia and Venezuela basins, the onset of subduction in the Muertos trough, and folding and thrust faulting at the Beata Ridge and the Bahamas, and the breakup of the main part of the Caribbean plate into the Venezuela and Colombia plates, separated by the Beata Ridge acting as a compressional plate boundary (Mauffret and Leroy, this volume) may all be related to the accelerated convergence between the two Americas.
The main differences with previous analyses are that
(1) our model results in substantial variations in convergence rates between
the two Americas after chron 25 (55.9 Ma), (2) we have computed uncertainties
for our North America - South America plate flow lines, and (3) we show Tertiary
Caribbean plate reconstructions in an Atlantic-Indian hotspot reference system.
Our absolute plate motion model suggests that the Caribbean plate has been
nearly stationary since chron 18 (38.4 Ma). The east-west gradient in
convergence between the Americas in the Neogene has not resulted in substantial
eastward motion of the Caribbean plate, but rather contributed to causing
its breakup into the Colombian and Venezuela plates along the Beata Ridge
where east-west oriented compressional stresses are taken up. Our model
also suggests that the eastward escape of the Caribbean plate in a mantle
reference frame ceased when seafloor spreading started in the Cayman Trough,
if the current interpretation of magnetic anomalies in the Cayman Trough is
not grossly in error. Our model suggests that the opening of the Cayman
Trough was accomplished by westward motion of the North American plate relative
to a stationary Caribbean plate in a mantle reference system. This implies
that subsequent North America-Caribbean and South America-Caribbean tectonic
processes were no longer dominated by Cocos-Caribbean and Nazca-Caribbean
plate interactions, as the latter had ceased to drive the Caribbean plate
eastwards. We conclude that the west-northwestward motion of South America
relative to a trapped, stationary Caribbean Plate caused oblique collision
along the passive margin of eastern Venezuela in the Neogene.
Many decades of research on deciphering the tectonic and sedimentary history of the Caribbean area (Figure 1) have resulted in a fairly well understood tectonic framework of its evolution. The plate tectonic history between the two Americas has been reconstructed based on regional geophysical and geological data and its implications for Caribbean geology and have been evaluated in syntheses including Pindell et al. (1988), Ross and Scotese (1988), Pindell and Barrett (1990) and Stéphan et al. (1990). Our analysis builds on the knowledge that has accumulated from these and many other regional studies pertaining to Caribbean tectonic history. The purpose of this paper is not a comprehensive review of Caribbean tectonic evolution. Hence the reader will not find tables of syntheses of all tectonic events that may have occurred during Caribbean tectonic history, or all models that have been put forward to explain them. This information has been thoroughly reviewed by Pindell and Barrett (1990). Here we rather focus on extracting information from recently declassified dense satellite altimetry data that allow us to map the structure of the ocean floor in much more detail than previously possible and to better constrain past plate motions along consuming or transform plate boundaries such as the boundaries between the North America, South America and Caribbean plates (Figure 2). We utilize these data jointly with new and existing magnetic anomaly data to investigate the Late Cretaceous and Tertiary plate kinematic framework of the Caribbean region, including the computation of uncertainties for our plate reconstructions.
Before the equatorial Atlantic between Africa and South America started opening at about chron M-0 time (120 Ma) (Pindell and Dewey, 1982; Mascle et al., 1988), the Caribbean tectonic framework was largely dependent on North America-Africa plate motions, which were identical to North America-South America motion vectors prior to the opening of the South Atlantic. After initial opening of the equatorial Atlantic, plate boundaries between the two Americas were affected by North America-South America relative plate motions resulting from the difference vectors between North America-Africa and South America-Africa sea floor spreading, as first computed by Ladd (1976). Reconstruction of the Caribbean tectonic frame for this period requires closure of the North America-Africa-South America plate circuit by using magnetic anomaly and fracture zone date sets that are standardized with respect to time. Because of the relatively slow velocity of North America-South America plate motions after chron 34 (83 Ma) it is particularly important to obtain estimates of the uncertainties for the relative plate motion vectors between the two Americas in order to evaluate the resolution of plate motions models.
We have combined a set of new magnetic anomaly data in the central North Atlantic between the Kane and Atlantis fracture zones, the "Canary-Bahamas Transect" (Maschenkov and Pogrebitsky, 1992), with existing magnetic anomaly data in the central North and South Atlantic oceans as well as with Seasat and Geosat satellite altimetry data to create a self-consistent data set for the two ocean basins. The finite motion poles and their uncertainties were estimated for 15 times from chron 34 to the present using an inversion method developed by Chang (1987, 1988), Chang et al. (1990), and Royer and Chang (1991), which allows a simple parameterization of the rotation uncertainties along a plate circuit path. The resolution of the estimates for North America-South America plate motions differs through time and is largely dependent on the velocity of their relative motions, i.e. faster plate motions are better resolved than slower motions. Even though the first-order features of our model are similar to Pindell et al.'s (1988) model, our results differ in detail, especially in the Late Tertiary, and we stress the evaluation of uncertainties of plate motion vectors. In particular, our conclusions differ from Pindell et al. (1988) in that we suggest that North America-South America plate motions may have had substantial effects on the structural development of the Caribbean area after chron 34 (83 Ma), especially during a period of rapid plate convergence in the Neogene.
We have created a digital age grid of the ocean floor
with a grid node interval of 6 arc-minutes using a self-consistent set of
global isochrons and associated plate reconstruction poles. The age at each
grid node was determined by linear interpolation between adjacent isochrons
in the direction of spreading. Ages for ocean floor between the oldest identified
magnetic anomalies and continental crust were interpolated by estimating the
ages of passive continental margin segments from geological data and published
plate models. We have constructed an age grid with error estimates for each
grid cell as a function of (1) the error of ocean floor ages identified from
magnetic anomalies along ship tracks and the age of the corresponding grid
cells in our age grid, (2) the distance of a given grid cell to the nearest
magnetic anomaly identification, and (3) the gradient of the age grid, i.e.
larger errors are associated with high age gradients at fracture zones or
other age discontinuities. Future applications of this digital grid include
studies of the thermal and elastic structure of the lithosphere, the heat
loss of the Earth, ridge-push forces through time, asymmetry of spreading,
and providing constraints for seismic tomography and mantle convection models.
Magnetic Anomaly Data
The central North Atlantic is probably the ocean basin
best covered by magnetic anomaly data. The three most comprehensive
individual data sets are the northeast-southwest trending Kronvlaag data (Collette
et al., 1984), the trans-Atlantic Geotraverse (TAG) data set, which comprises
a number of long east-west oriented lines (Rona, 1980), and the Canary-Bahamas
Transect (Maschenkov and Pogrebitsky, 1992), which comprises a dense set of
survey lines between the Atlantis and Kane fracture zones from the Mid-Atlantic
Ridge to about magnetic anomaly 13 (Figure 3).
These data sets are supplemented by a large number of other geophysical surveys
(see Klitgord and Schouten, 1986, for previous compilation), resulting in
dense data coverage.
A comprehensive analysis of magnetic anomaly data in the South Atlantic was carried out by Cande et al. (1988). In order to create a self-consistent set of magnetic anomaly identifications for closing the North America-African-South America circuit, we use Cande et al.?s (1988) magnetic anomaly crossings in the South Atlantic (Figure 3) and identify the same magnetic chrons as Cande et al. (1988) in the central North Atlantic. All magnetic anomaly identifications correspond to the young end of normal polarity intervals, except for anomaly 33o. The phaseshift angles were determined from paleomagnetic poles for North America from Harrison and Lindh (1982) and from the IGRF90 reference field. We use the young end of the following normal polarity intervals according to the Cande and Kent (1995) magnetic reversal time scale: chron 5 (9.74 Ma), 6 (19.05 Ma), 8 (25.82 Ma), 13 (33.06 Ma), 18 (38.43 Ma), 21 (46.26 Ma), 24 (52.36 Ma), 25 (55.90 Ma), 30 (65.58 Ma), and 32 (71.59 Ma), 33o (79.08 Ma), and 34 (83.00 Ma).
Fracture zones represent important information constraining plate motions and can be used in concert with magnetic anomaly data for computing finite rotations. Geosat, Seasat and ERS-1 altimetry data provide a unique data set to uniformly map the height of the sea surface, whose short-wavelength topography reflects uncompensated basement topography, such as that related to fracture zones. Müller et al. (1991) demonstrated that there is an excellent correlation between the geoid anomaly and the basement structure of the Kane Fracture Zone in the central North Atlantic. They used geoid data from Geosat and subsatellite basement topography profiles of the Kane Fracture Zone to show that the average horizontal mismatch between geoid low and the axis of the basement trough, as mapped by Tucholke and Schouten (1988), is 5 km. The results of this comparative study represent "ground truth" for the use of satellite altimetry data for accurately mapping slowly slipping Atlantic-type fracture zones.
Following the Kane fracture zone study, Müller and Roest (1992) identified a number of small- and medium-offset fracture zones from the along-track Geosat and Seasat gravity data by picking the center of the gravity troughs corresponding to the deepest portion of the central fracture valleys. We re-identified these fracture zones from dense satellite-derived gravity data (Sandwell and Smith, in press). For the central North Atlantic reconstructions we use the Atlantis, Northern and Kane fracture zones (Figure 4). When fracture zone offsets change from medium to small, as has happened in the case of the Northern Fracture Zone, whose offset diminished from 80 km at chron 25 to 20 km at chron 13, they may become unstable and commence to migrate along the ridge, producing V-shaped patterns. We used only those portions of fracture zones that appear to follow flow lines, i.e. have not migrated along the ridge axis. The location of the Kane Fracture zone is constrained by Tucholke and Schouten's (1988) compilation of basement structure.
Numerous fracture zones from the equatorial Atlantic to the southern South Atlantic record plate flow lines of seafloor spreading in the South Atlantic. Shaw and Cande (1990) pointed out that the northernmost fracture zones in this spreading system, i.e. the Marathon, Mercurius, Doldrums, and Four-North fracture zones in the equatorial Atlantic (Figure 2), put important constraints on South America-Africa plate motions due to their proximity to the finite rotation poles. However, because no magnetic anomaly data have been identified in this area, we cannot resolve which portion of a fracture zone is relevant for a particular age, and whether these fracture zones reflect South America-Africa spreading for their entire length. We find that by using fracture zones south of 8oS only, we obtain very similar rotations compared with reconstructions in which equatorial fracture zones are included. Accordingly, our South Atlantic reconstructions are constrained by the Bodo Verde, Martin Vaz, Rio Grande, and some unnamed fracture zones (Figure 5). Both in the central North Atlantic and the South Atlantic only those fracture zones were used whose offsets through time are constrained by magnetic anomaly data. This is important, because the use of incorrect fracture zone segments for constraining a given reconstruction would skew our results. We also avoid large-offset fracture zones, as they are not reliable indicators of plate motion changes over short geological time spans.
The finite plate motion poles and their uncertainties were estimated using an inversion method developed by Chang (1987, 1988), Chang et al. (1990), and Royer and Chang (1991), based on Hellinger's (1981) criterion of fit. The uncertainties of a rotation are expressed as a covariance matrix, which is conceptually equivalent to the "partial uncertainty rotations" described by Stock and Molnar (1983). In this method magnetic anomaly and fracture zone data are both regarded as points on two conjugate isochrons, which consist of great circle segments. The best fit reconstruction is computed by minimizing the sum of the misfits of conjugate sets of magnetic anomaly and fracture zone data points with respect to individual great circle segments. Consequently, both the resulting best-fitting rotations, as well as the sum of the misfits, depend critically on correctly identifying conjugate data points that belong to a common isochron segment. In practice, the application of Hellinger's criterion of fit poses no problem, because rotations have been published for most plates describing their late Cretaceous/Tertiary history of motion. Hence, a starting rotation can be used for an initial reconstruction to identify conjugate isochron segments and data points.
A possible disadvantage of applying Hellinger's (1981) criterion of fit to both magnetic and fracture zone data is that fracture zones cannot necessarily be expected to fit as well as magnetic anomaly data. Although all portions of a fracture zone have at some time been the location of a transform fault and part of an isochron, fracture zone morphology becomes overprinted successively at the transform-ridge intersections during changes in spreading direction. The amplitude of this effect is expected to increase as a function of transform length. For this reason we do not use fracture zones with offsets of more than about 150 km. Shaw and Cande (1990) recognized this problem and suggested an inversion method that incorporates fracture zones by minimizing the misfit of fracture zone data with respect to plate flow lines. The benefits of this model were expected to be a utilization of the "integral constraints" of fracture zones, i.e. their continuity, as well as the possibility of allowing for consistent asymmetries of fracture zone limbs on conjugate plate flanks. Shaw and Cande (1990) implemented this method by minimizing the misfits of fracture zone data to symmetric flow lines. However, forcing fracture zones to fit symmetric flow lines may obscure distinct changes in spreading direction recorded in fracture zones bounded by two asymmetrically spreading corridors. Even though seafloor spreading appears to be remarkably symmetric in the long run, accretion of ocean floor through time periods of short "stages" resolvable by magnetic anomaly data can be quite asymmetric.
A critical aspect in our reconstruction method is the correct
assessment of the uncertainties in the location of the data that will propagate
into the uncertainties on the rotation parameters (location of the rotation
pole and rotation angle). Following a detailed analysis of the dispersion
of magnetic anomaly 5 crossings in the Indian ocean by Royer et al. (1997)
we assigned 1-
nominal uncertainties
of 4 km to the magnetic anomaly crossings and of 5 km to fracture zone crossings
following Müller et al.'s (1991) analysis. The uncertainties assigned
to the data ( 
) are related
to their true unknown estimates (
)
by the quality factor 
. Although 
is unknown,
the method developed by Royer and Chang (1991) allows us to estimate 
from the misfit, the geometry of the plate boundary and the number of data:
where N is the number of points, s the number of great circle segments, and
r the total weighted misfit. Note that N-2s-3 corresponds
to the number of degrees of freedom.
Thus the parameter 
indicates whether the assigned uncertainties are correct (
),
underestimated ( <<1) or
overestimated (
>>1). Our
reconstructions indicate that our nominal uncertainties are generally slightly
overestimated. For the central Atlantic, ranges from 1.07
to 3.08 and all degrees of freedom are larger than 50. This means that
the average true uncertainties range from 2.3 (= 4 /
) to 3.9 (= 4 /
) km for the magnetic
crossings, and 2.8 to 4.8 km for the fracture zone crossings (Table
1a). For the South Atlantic,
ranges from 0.67 to 1.74 with degrees of freedom between 28 and 42.
Thus the uncertainties should lie between 3.0 and 4.9 km for the magnetic
data, and 3.8 to 6.1 for the fracture zone data (Table
1b). Several factors may contribute to this discrepancy in the dispersion
of the magnetic crossings: i) there are much less data in the South
Atlantic, hence we are probably combining data from different spreading corridors
into individual segments; ii) there are much less recent (i.e. GPS-navigated)
cruises in the South Atlantic than in the Central Atlantic. Without
further investigating this question, we decided to keep an uncertainty of
4 km for the magnetic crossings. Our inversions also suggest that the
dispersion of the fracture zone data is generally better than 5 km.
However, since fracture zones are not the optimal records for plate reconstructions,
and since we do not know how well gravity troughs relate to the actual location
of (paleo-)transform segments, we choose to remain conservative and use the
Müller et al. (1991) result of an uncertainty of 5 km for fracture zones.
North America-Africa and South America-Africa
finite rotations
Finite rotations for North America-Africa and South America-Africa
plate motions were computed for 12 times from chron 34 to the present. The
resulting finite rotations, statistical parameters, and covariance matrices
are listed in Tables 1 and 2.
The 95% confidence regions are 3-dimensional ellipsoids in latitude, longitude,
and rotation angle space. In order to represent them on a map, the ellipsoids
are projected onto the latitude-longitude sphere (cf. Royer and Chang, 1991).
The projected uncertainty ellipses include uncertainties in both latitude,
longitude, angle, but one must keep in mind that the true size of the rotation
uncertainties (i.e. ellipsoids), which are described by the covariance matrices,
might not be reflected well by their 2-D projection onto the sphere.
For instance, in the case of the South Atlantic, the 2-D uncertainty in the
location of the chron 5 best-fitting pole appears to be at least 10 times
larger than the 2-D uncertainty in the location of the chron 21 rotation pole,
whereas the volume of the 95% confidence region for chron 5 rotation is only
about 3 times larger than for the chron 21 rotation (7838 versus 2773 km3,
respectively). Conversely, the 95% uncertainty volume for chron 8 (19705
km3) is 2.5 larger than for chron 5, whereas its 2-D projection is about 3
times smaller than for chron 5.
In Figure 6 we compare our results with the North America-Africa pole path from Klitgord and Schouten (1986) and the South America-Africa pole path from Shaw and Cande (1990). The main difference for the central North Atlantic is that our path shows a sharp bend at anomaly 8 time (25.8 Ma), resulting in a pole path for chron 5 to chron 8 that is distinctly different from the chron 21-chron 8 path, whereas Klitgord and Schouten's (1986) path is continuous between chrons 21 and 5. The consequences of this difference for Neogene North America-South America plate motions are discussed in subsequent sections.
Our South Atlantic pole path is generally similar to Shaw and Cande's (1990) path, but less smooth. The similarity reflects the fact that both models are based on very similar data sets, whereas the differences reflect properties of the different inversion methods used to compute finite rotations as well as new fracture zone identifications. The "integral constraints" of fracture zone continuity used in Shaw and Cande's (1990) method is probably the main reason for the smoothness of their pole path. However, the use of symmetric flow lines to evaluate the fit of fracture zone data might smooth out real cusps in a finite pole path that is computed from reconstructing data from individual isochrons independently. The main difference between Shaw and Cande's (1990) model and our model for the South Atlantic (Figure 5) is the anomaly 13 reconstruction, which results in a distinct cusp in our pole path, while there is a less accentuated cusp in their path. One way to qualitatively test whether or not such a cusp is supported by the data is to construct plate flow lines and evaluate their fit to fracture zones.
For South Atlantic spreading this is done best in the equatorial
Atlantic, because the fracture zones here are the best recorders of changes
in spreading direction due to their proximity to the finite motion poles.
However, plotting plate flow lines in this area raises the problem of knowing
where the South America-North America plate was located through time.
Plotting flow lines both for North America-Africa and South America-Africa
for the same fracture zone allows us to address this problem. We plot
South Atlantic plate flow lines derived from our model in Figure
7a with the gridded gravity anomaly field computed from Seasat, Geosat,
and ERS-1 data (Sandwell and Smith, in press). All flow lines are constructed
using both ridge-transform intersections as seedpoints. The area encompassed
by the resulting dual flow lines approximates the amount of transpression
or transtension that occurred during changes in spreading direction, assuming
symmetric plate accretion. The Marathon and Mercurius fracture zones
do not fit the computed flow lines well. However, the Four North and
Doldrums fracture zones show a good fit to our flow lines. The latter
observation confirms the validity of our South Atlantic plate motion model,
whereas the former indicates that the North-South America plate boundary
may be located in the vicinity of the Marathon/ Mercurius fracture zones,
resulting in deviations from South America-Africa flow lines. In comparison,
we show central North Atlantic plate flow lines in the equatorial Atlantic
ocean in Figure 7b. The Marathon fracture zone
flow line clearly does not match the gravity anomaly expression of this fracture
zone. The post-chron 6 flow line of the Mercurius fracture zone fits
its eastern limb better than the South America-Africa flow line (compare with
Figure 7a), but not its western limb. Its is
clear that this fracture zone would not be useful to constrain either North
America-Africa nor South America-Africa plate motions, since it appears to
have been subject to recent plate boundary deformation. We can draw
the conclusion that applying Hellinger's criterion of fit jointly to the magnetic
anomaly data and fracture zone data from continuous fracture zones, which
follow plate flow lines, and reconstructing these data independently for each
isochron, results in plate models that produce smooth continuous flow lines,
even though this property is not utilized or imposed by the model inversion,
as it is in Shaw and Cande's (1990) method.
North America-South America finite rotations
Tables 3a and 3b
list the rotation parameters for the North America-South America relative
motions, for each chron resulting from the product of the South America-Africa
rotation with the corresponding Africa-North America rotation. The resulting
pole path (Figure 8) shows: (i) very stable poles
of motion from chron 25 to 21, and from chron 18 to 8; (ii) an important northward
migration of the rotation poles from chron 34 to chron 18; (iii) followed
by a southward migration until chron 6. For this reason we computed
only 8 stage rotations (Table 3) to describe the main
episodes of relative motion between the two America plates. The North
America-South America rotation stage poles always lie outside the Caribbean
plate (Table 3b); for the ages younger than chron 25
(55.9 Ma), they lie within or in the vicinity of the North America-South America
plate boundary, implying a different sense of motion along strike of this
plate boundary.
North America-South America plate motions in the Caribbean area
Ideally we would like to be able to compute North America-Caribbean and South America-Caribbean plate motions and compare the modeled motion vectors with mapped plate boundary deformation. Accurate estimates for North America-Caribbean plate motions are available for times after chron 6 (19.0 Ma), as recorded by seafloor spreading in the Cayman Trough (Rosencrantz et al., 1988, Mauffret and Leroy, this volume). The magnetic anomalies on older portions of the Cayman Trough are less straightforward to identify, and the chronology of both pre- and post-chron 6 seafloor spreading here is still the subject of debate. For these reasons we first analyze the North America-South America plate kinematic framework, as it is independent on knowledge of the spreading history in the Cayman Trough, and compare the results with geological and geophysical data from the northern and southern Caribbean plate boundaries.
Figure 9 shows North-America-South America plate motion through time, illustrated by the successive motion of 3 points attached to the North America plate with respect to the South America plate for 11 stages from chron 34 (83 Ma) to the present. The paths of these points through time as shown on Figure 9 correspond to conventional plate flow lines illustrating relative North America-South America plate motion. The only difference here is that for each stage rotation vector a simultaneous 95% confidence region is plotted about the young ends of the relative motion vectors, i.e. the ellipse about points at chron 33o reflect the uncertainty for the chron 34-33o stage rotation. A simultaneous confidence region represents the area (on the surface of the Earth) in which a point on a given plate may have been located with equal likelihood for a particular reconstruction time, with 95% confidence. The simultaneous 95% confidence regions increase in size with increasing distance from the stage pole of motion; in the Caribbean area they increase in size correspondingly from east to west (Figure 9).
The reader may note that even though the projected confidence
ellipsoids of our finite rotation poles for North America-South America plate
motion for chrons 18-6 show large overlaps (Figure 8),
the simultaneous 95% confidence regions for the three rotated points shown
show only little overlap. This may appear puzzling, but only reflects
the imperfect nature of the 3-dimensional finite rotation pole error ellipsoid
projections onto a spherical surface on Figure 8,
as discussed before. In other words, this result shows that the covariance
matrices and uncertainty ellipsoids for these reconstructions are different
enough to distinguish different phases of convergence in the Late Tertiary
outside of the 95% error bounds.
We cannot resolve very slow relative motions between chrons
25 and 24; some overlap between the confidence regions for chrons 32 and 30
and chrons 18 and 13 is also observed. However, all other stage vectors
are well resolved. Consequently, we compute a new set of 8 stage rotation
poles (Table 3b) and corresponding relative motion
vectors, including only stages for which we can resolve relative North America-South
America plate motions (Figure 10). The
relative motion vectors in this set of stages can be divided into four age
groups: 1) slow sinistral strike-slip/transtension between chrons 34
and 25 (~2.8±0.8 - 4.8±1.1 mm/y at 85°W), 2) northeast-southwest
oriented convergence from chron 25-18 (6.5±1.5 mm/y at 85°W), 3)
slow compressional motion from chron 18-8 (3.6±2.1 mm/y at 85°W),
and 4) fast north-south oriented convergence from chron 8-6 (9.6±3.1
mm/y for chrons 8-6 and 9.6±2.1 mm/y for chrons 6-5 at 85°W), followed
by a deceleration in north-south oriented convergence post chron 5 (5.2±1.3
mm/y at 85°W).
In order to directly compare our model with Pindell et al.'s (1988) results, we compute North-South America relative motions for the same 7 stages as Pindell et al.'s (1988) (Figure 11). The general shape of both models is similar, reflecting the similarity of the magnetic anomaly data sets used to constrain both models. The main difference between the models is that Pindell et al.'s (1988) model implies relatively constant convergence between the Americas of rates between 6 and 4 mm/y chron 21 (46.3 Ma). Our model results in substantial variations in convergence rates from chron 25 (55.9 Ma), as documented in Figure 10 and Table 6. In particular our model resolves the onset of fast convergence after chron 8 at a speed of nearly 10mm/y, compared with less then 4 mm/y from chrons 18-8 measured at 85°W. Without identifying magnetic anomaly 8, two stages of slow (chron 13-8) and fast (chron 8-6) convergence would be averaged.
Lesser Antilles to Mid-Atlantic Ridge
The North America-South America plate boundary east of the Caribbean region is characterized by a number of anomalous ridges and troughs. The Barracuda and Tiburon ridges east of the Lesser Antilles arc (Figure 12) both exhibit unusually large Bouguer gravity anomalies (up to ~135 mGal). The eastern continuation of the plate boundary is expressed in the Researcher Ridge and Royal Trough (Figure 12). The Royal Trough exhibits en échelon shaped tectonic fabric and fresh basalts on a basement characterized by spreading center type faulting identified from GLORIA data, whereas the Researcher Ridge has a large magnetic anomaly. Both features are interpreted as extensional structures (Collette et al., 1984; Roest and Collette, 1986).
For the area at the Tiburon and Barracuda ridges, our plate model results in a total of 151±31 km left-lateral transtension from chron 34 (83 Ma) to 25 (55.9 Ma). A phase of slow transpression is predicted for chron 25 (55.9 Ma) - 18 (46.3 Ma), which is not well resolved, followed by extremely slow motion between chrons 18 and 6, during which time it is within the computed 95% confidence areas for the entire part of the plate boundary east of the Lesser Antilles subduction zone (Figure 12). The plate boundary area east of 56°W was located quite close to the stage poles of motion; the resulting vectors of relative motion are so small that they cannot be resolved. Convergence in the Barracuda/Tiburon ridge area started at chron 6. We model 32±23 km of convergence from chron 6-5, but virtually no relative motion after chron 5, (4±18 km) which is much too small to be resolved, by our model. The post-chron 8 relative motion in the Royal Trough area can only be resolved for post-chron 5 time. Here our model implies 22±10 km of extension for the last 10 million years.
Our plate model supports the suggestion that the present topographic and gravity expression of the Tiburon and Barracuda ridges may have resulted primarily from Neogene plate convergence (c.f. Müller and Smith, 1993). In particular, the strong Moho uplift at the Tiburon Ridge, where the crust is modeled to be 1.5-2 km thick with a Moho uplifted more than 4 km (Müller and Smith, 1993), is extremely unstable. Without compressive forces, oceanic crust as thin as 1.5-2 km would subside and form a depression, rather than a ridge. The plate model used by Müller and Smith (1993) did not allow them to discriminate when in the Tertiary North-South America convergence in this area and the uplift of the two ridges might have been initiated. However, they argued that middle Eocene-late Oligocene turbidites on the slope of the Tiburon Ridge, which is now located 800 m above the abyssal plain, may indicate that most of its uplift occurred in post-Oligocene times, a conclusion strongly supported by the plate model presented here.
This idea contrasts the interpretation of Dolan et al. (1989, 1990), who suggested that the present topography of the Tiburon Ridge had existed prior to deposition of the turbidites, which would have been emplaced by upslope deposition from the abyssal plain onto the rise. Dolan et al. (1989, 1990) argue that the Tiburon Rise has existed as a bathymetrically shallow feature since the late Cretaceous. An extensive discussion of this question can be found in Müller (1993) and Müller and Smith (1993). We find that there is little evidence that would show conclusively that the present bathymetric elevation of the Tiburon Rise has existed since the late Cretaceous. In contrast, the crustal structural modeling carried out by Müller and Smith (1993), together with the plate model presented here, indicates that much of the unusually shallow Moho topography and crustal uplift of the Tiburon rise is a result of the onset of North America -South America convergence at chron 6 (19.0 Ma). A Neogene formation of the present topography of the Tiburon Rise would also alleviate the need for upslope turbidite deposition on the rise in the middle Eocene-late Oligocene, as suggested by Dolan et al. (1989, 1990). Instead, the present elevation of the turbidites on the slope of the rise would reflect post-depositional uplift.
Middle America Trench to Lesser Antilles
Our results suggest that slow sinistral transtension/strike-slip between the two Americas lasted from chron 34 (83 Ma) until chron 25 (55.9 Ma) (~2.8±0.8 - 4.8±1.1 mm/y at 85°W). However, we cannot attribute any particular Caribbean tectonic events to this phase of slow sinistral motion. Our model results in roughly northeast-southwest oriented slow convergence from chron 25 (55.9 Ma) to chron 18 (38.4 Ma) (6.5±1.5 mm/y at 85°W). This time period of convergence includes a calc-alkaline magmatic stage in the West Indies, dated as Paleocene-Lower Eocene (Perfit and Heezen, 1978; Mercier De Lépinay, 1987), which is thought to be related to southward subduction of proto-Caribbean crust during this time. North-South America relative motion was characterized by relatively slow transpression until chron 8 (25.8 Ma) (3.6±2.1 mm/y at 85°W). This phase of slow relative motion between the two Americas correlates with a period of tectonic quiescence along some parts of the Caribbean margins during the Oligocene-Lower Miocene that was characterized by subsiding basins, unconformably overlying the previously deformed belts (Calais et al., 1989). However, at the same time, collision has occurred off Venezuela (Mann et al., 1995).
Our model is different from Pindell et al.'s (1988) plate model for North-South America plate motions, in that it results in a drastic increase in convergence velocity subsequent to chron 8 (25.8 Ma) (9.6±3.1 mm/y for chrons 8-6 and 9.6 ±2.1 mm/y for chrons 6-5 at 85°W, compare also Figures 10 and 11). A slowdown in convergence after chron 5 resulted in a post-chron 5 convergence rate drop to 5.2±1.3 mm/y at 85°W. The modeled Neogene convergence results in 92±30 km convergence from chron 8-6, 127±27 km from chron 6-5, and 72±17 km from chron 5 to the present near the North Panama Deformed Belt at 85°W (Figure 10). The Neogene convergence measured south of the Muertos Trough at 65°W is 40±13 km from chron 8-6, 58±22 km from chron 6-5, and 22±13 km from chron 5-present day.
Geological data from the circum-Caribbean plate boundaries indicate that the tectonic regime changed drastically in the lower to middle Miocene. The eastern portion of the North America Caribbean plate boundary displays an accretionary wedge along the Muertos Trough (Case et al., 1984). Based on seismological evidence Byrne et al. (1985) showed that the Muertos trough is an active structure and suggested it to be the site of subduction. North-south convergence is accommodated by oceanic crust underthrusting the Greater Antilles, or by folding and thrust faulting only, where the crust is thicker, as at the Beata Ridge and the Bahamas (Ladd et al., 1990). Correlation of seismic reflection data from the turbidite fill in the Muertos Trough with the Venezuelan Basin indicates a Neogene, or possibly late Neogene age for the initiation of underthrusting (Ladd et al., 1990). In the early Miocene, the Peralta and Rio Ocoa sediment groups on Hispaniola were deformed in a southwest verging fold-and-thrust belt (Heubeck et al., 1991), supporting evidence for the onset of early Miocene convergence within this part of the Caribbean-North America plate boundary. Tectonic events at this time have also been mapped in the Dominican Republic, in Haiti, and on Cuba (Calais et al., 1992).
The Caribbean-South America plate boundary comprises a wide and complicated plate boundary zone. It starts at the deformation front of the subduction zone, includes various dextral strike-slip faults of northern South America (e.g. Bocono, Oca, El Pilar) and continues to the south as a fold-thrust belt (e.g Ladd et al., 1984). Biju-Duval et al. (1984) analyzed multichannel seismic reflection data in the Venezuela Basin and concluded that the present configuration of the margin, i.e. underthrusting of the Caribbean oceanic crust below the South America borderland, developed in the early or middle Miocene. They also realized that north-south shortening observed at the North Venezuela margin may be the result of regional North-South America convergence. Recently, a comprehensive analysis of new and existing seismic data from the Beata Ridge and adjacent areas by Mauffret and Leroy (this volume) has shown that the Beata Ridge, a Cretaceous plateau, is bounded to the east by compressive structures reactivated by right-lateral strike-slip, and by normal faults to the west. Uplift of the ridge increases from south to north, and is estimated to have started in the early Miocene (23 Ma), resulting in a total shortening between 170 km and 240 km as a function of latitude (Mauffret and Leroy, this vol.). They interpret the Beata Ridge as a compressional plate boundary, resulting from overthrusting of the Colombia microplate onto the Venezuela microplate. The implied clockwise rotation of the Colombia microplate and convergence between the latter and the Venezuela mircoplate are consistent with differential convergence between the North and South America plates, increasing from east to west, thereby "squeezing" the Columbia plate out to the east, as suggested by Burke et al. (1978). Mauffret and Leroy (this vol.) suggest that the observed deformation may also be caused by the buoyancy of the Cocos plate, which is subducting under the Caribbean plate (England and Wortel, 1980; Meijer, 1992). However, Central America experiences extension in the back arc in the north where the subducting Cocos plate is older than in the south, whereas a younger Cocos plate in the south causes shortening. Alternatively, compression in Costa Rica, as expressed by the April 22, 1991 Costa Rica earthquake (Plafker and Ward, 1992), has been attributed to the subduction of an aseismic ridge (Adamek et al., 1987). Mann and Burke (1984) suggested that the Beata Ridge may be the consequence of northward motion of the Maracaibo block, a tectonic block of South America. Recent work has confirmed a north to northeast directed motion of these blocks relative to the Caribbean plate (Ego, 1995, Geology). In summary it is unclear what the role of the Cocos plate may be in terms of contributing to compression at the Beate Ridge.
An east-west gradient in convergence between the Americas is also supported by a recent analysis of present-day relative plate motions between North and South America based on GPS data (Dixon and Mao, 1997). They found an increase of differential north-south convergence from east to west from about 1mm/y at the Barracuda ridge to about 9 mm/y at 85°E. It is interesting to note that their modeled present-day convergence rate of 9 mm/y at 85°E is similar to the rates we calculate for chron 8-5 time, and significantly faster than our post-chron 5 convergence estimate at 85°W of 5.2±1.3 mm/y. It is not clear whether this difference reflects a recent acceleration in North America-South America plate convergence, or whether it is related to problems in our anomaly 5 reconstruction. The latter may well be the case, since the South Atlantic reconstruction for this time is not well constrained; it is based on a small number of data points only, and Table 1b shows that we have slightly underestimated the uncertainties of both magnetic and fracture zone identifications for this data set. In contrast, the anomaly 5 reconstruction in the central North Atlantic is extremely well constrained.
We propose that the post chron-8 convergence between North and South America has also played a substantial role in the Neogene Panama arc collision and subsequent arc-deformation to an S-shaped pattern. In the collision area, the computed north-south convergence between the Americas resulted in a total of 291 ± 75 km of north-south convergence. During the middle to late Miocene the onset of the collision of the Costa Rica-Panama Arc with the western Cordillera of northwestern South America (Wadge and Burke, 1983; Eva et al., 1989; Mann et al., 1990) started forming the North Panama Deformed Belt.
The tectonic events outlined above are all slightly younger (mostly early Miocene) than the onset of North-South America convergence predicted by our model (~26 Ma, late Oligocene). This may reflect an artifact of our model. Our plate model lacks resolution between anomalies 8 (25.8 Ma) and 6 (19.0 Ma), because it is not straightforward to identify the magnetic anomalies between 6 and 8 with confidence in a slowly spreading tectonic regime. It is possible that the early Miocene tectonic events in the Caribbean correspond to a global change in plate motions. Evidence for this idea comes from a detailed survey of the Pitman Fracture Zone in the South Pacific that shows a distinct change in spreading direction at chron 6c (Cande et al., 1995) which represents the Oligocene-Miocene boundary (23.8 Ma). It is virtually impossible to identify this magnetic anomaly in the slowly spreading central North Atlantic and South Atlantic oceans. Therefore, we consider it possible that convergence started at the Oligocene-Miocene boundary, 2 m.y. later than predicted by our plate model. Pindell et al's (1988) model results in an acceleration in convergence at chron 6 (20 Ma in the DNAG timescale (Kent and Gradstein, 1986), 19 Ma in the timescale used here (Cande and Kent, 1995), with extremely slow convergence from chron 13 to chron 6. This demonstrates that a model not constrained by any magnetic anomaly identification between anomaly 6 and 13 (a time interval 14 million years long) results in an apparent acceleration in convergence at 19 Ma, about 5 million years later than the Oligocene-Miocene boundary.
Plate motions relative to the mantle
We put Caribbean plate motions into an absolute hotspot
reference frame based on Atlantic-Indian ocean hotspot tracks (Müller
et al., 1993) for understanding cause and effects of plate motions between
the Americas and the Caribbean plate(s). Ross and Scotese (1988) used
a paleomagnetic reference frame for their model (which cannot resolve longitudinal
motions of plates), and correspondingly do not show a geographic frame on
their reconstructions. Pindell et al. (1988) used Engebretson's (1982)
and Engebretson et al.'s (1985) absolute plate motion model for the Pacific
based on hotspot tracks to calculate Pacific-Caribbean relative motions, which
have likely exerted controls on Caribbean tectonic evolution in the Mesozoic
and Early Tertiary (Pindell et al., 1988), but not necessarily in the Neogene.
In any case, Pacific absolute plate motions are only of limited use to constrain
absolute motions of plates bordering the Atlantic ocean, as our knowledge
on closing plate circuits crossing the bounday between East and West Antarctica
is still inadequate (Molnar and Stock, 1987; Cande et al., 1995).
We use the relative plate motion model for the central North Atlantic and South Atlantic presented here, a revised relative plate motion model for the Caribbean area, largely based on the tectonic elements and plate hierarchy from Ross and Scotese (1988), and the model for motion of plates in the Atlantic and Indian ocean hemisphere relative to major hotspots (Müller et al., 1993). The combined rotation model has been adapted to the Cande and Kent (1995) timescale for post chron 34 (83 Ma) times and the Gradstein et al. (1994) timescale for earlier times.
The absolute plate motion model by Müller et al. (1993) is based on jointly fitting dated hotspot tracks on the Australian, Indian, African, and North and South America plates relative to present-day hotspots assumed fixed in the mantle. Therefore this model is better constrained than a model solely based on the hotpot tracks of one plate. The main weakness of any absolute plate motion model based on hotspot tracks is the assumption of fixity of the hotspots relative to the spin axis. So far there is no evidence for any large-scale relative motion between Atlantic-Indian hotspots relative to the spin axis except for the time before 90 Ma. Van Fossen and Kent (1992) have reported a large (12°) mismatch between paleolatitudes and the present day latitude of the New England hotspot, which appears to have accumulated between 124 and 90 Ma.
Pindell et al. (1988) suggested that most of the total opening by seafloor spreading between the two Americas was accomplished some time between 100 and 90 Ma, when the Caribbean plate started entering from the west. However, the age of the initial contact of a Caribbean plate originating from the Pacific has been revised to late Campanian-Maestrichtian, when syn-orogenic sedimentation and northward verging folding, thrusting and obduction of ophiolites have occurred at the southern margin of the Yucatan Peninsula in Guatemala (Rosenfeld, 1990). The arguments in favor of an allochtonous nature of the Caribbean ocean crust have been reviewed comprehensively by Pindell and Barrett (1986).
In the Paleocene (see Figure 13 a, chron 25, 55.9 Ma) the Yucatan Basin is opening along a left-lateral strike-slip fault (Rosencrantz, 1990). The prograding arc has started colliding with the Bahama platform in western Cuba (Bralower et al., 1993). For the time after the middle Eocene we have an estimate for North America-Caribbean plate motions from the spreading history in the Cayman Trough (Rosencrantz et al., 1988), even though spreading here may not reflect the total North America-Caribbean motion (Rosencrantz and Mann, 1991). Burke et al. (1980) suggested that cumulative offsets of strike-slip faults on Jamaica suggest a minimum rate of offset of 4 mm/y, in addition to an average of 16mm/a of total opening in the Cayman Trough. We implemented this suggestion in our rotation model, similar to Ross and Scotese (1988), by allowing for 4-6mm/a of left-lateral strike-slip between Jamaica and Southern Hispaniola. In Figure 11 we show the resulting path of two points attached to the Caribbean plate relative to the mantle (without considering relative motion between the Colombia and Venezuela microplates through time, which we cannot reconstruct). The two absolute plate motions paths in Figure 11 as well as the plate reconstructions in Figures 13b-d show that the Caribbean plate has been virtually stationary with respect to the mantle at least since the onset of seafloor spreading in the Cayman Trough. The errors from combining the "absolute" and relative plate motions models involved in this calculation are probably larger than the total length of the path shown. The Caribbean plate could have only maintained a substantial eastward component of motion if either the Cayman Trough opened much later and faster than presently assumed, and/or if there has been much faster strike-slip between the Caribbean plate and Jamaica than suggested by Burke et al. (1980).
North America's and South America's plate motions in the mantle reference frame are both characterized by relatively fast westward motion, with a small component of convergence added at chron 13 due to a clockwise change in South America plate motion, and even faster convergence after chron 8 due to a counterclockwise change in North America absolute plate motions. In contrast, the Caribbean plate appears to have been virtually stationary in a mantle reference frame at least since chron 18. This result agrees with an idea put forward by Sykes et al. (1982), who noted that only a small fraction of its perimeter is attached to a subducting slab. Even though the Caribbean slab under the South America plate has been shown to be longer than previously thought, the forces assumed to be most important for driving plates, namely ridge push, slab pull and trench suction (this force acts to draw plates together at a trench (Elsasser, 1971)) must be relatively small. If they were not, then the Caribbean plate would not rest in a mantle reference frame, as found by our plate kinematic analysis.
Our result is also in accordance with Gripp and Gordon's (1990) analysis of present day Caribbean plate motions with respect to the hotspots. Their analysis, based on motion of the Pacific Plate relative to its underlying hotspots, and the NUVEL-1 relative motion model by DeMets et al. (1990) results in roughly west-southwest oriented motion of the Caribbean plate. However, their motion vectors do not differ significantly from zero.
It must be concluded that tectonic plate boundary processes between the Caribbean plate and the Americas are entirely driven by relatively fast, mostly westward motion of North and South America. The resulting differential motion between North and South America affects a stationary Caribbean plate trapped between two larger plates by edge-driven plate tectonic interactions, equivalent to some small plates in the Middle East (e.g. Aarabia/Anatolian plate, McKenzie, 1972).
Sykes et al. (1982) recognized this possibility, but suggested alternatively that the Caribbean plate may be forced to move eastward in response to the gradient in convergence rate between North and South America, increasing from east to west, as also found by our analysis. In contrast, the plate motion paths plotted in Figure 11 suggest that the eastward motion of the Caribbean Plate with respect to the two Americas is entirely due to westward motion of the latter two plates with respect to the mantle, and that the east-west convergence gradient quoted by Sykes et al. (1982), which has been constrained to post chron 8 (25.8 Ma) times by our model (probably post chron 6c as discussed above), may not have resulted in substantial eastward motion of the Caribbean Plate with respect to the mantle. Rather, the east-west gradient in post-chron 6c convergence may have contributed to causing east-west compression at the Beata Ridge, as described in Mauffret and Leroy (this volume).
Our combination of relative and absolute plate motions indicates that throughout the Tertiary tectonic processes at the northern and southern boundary of the Caribbean plate were governed by the relatively fast westward motion of both the North and South America plates with respect to a nearly stationary Caribbean plate. The differences between North America-Africa and South America African plate motions, as described here, resulted in changes in relative motion between the two Americas whose effects are clearly seen in the tectonic development along the northern and southern margin of the Caribbean area. Since the Caribbean plate does not appear to have moved substantially relative to the mantle during the Neogene, there are no major tectonic processes which can be attributed to the eastward "escape" of the Caribbean plate during this time (e.g. Mann et al., 1997). In particular for the time since chron 6 (19 Ma), for which Caribbean-North America relative motion is better constrained than for earlier times, we find that the Caribbean plate was virtually fixed relative to the mantle.
This observation suggests that accelerated convergence post chron 6c (23.8 Ma) at the Oligocene-Miocene boundary reduced the space within the eastward "escaping" arc could operate such that Caribbean plate motion relative to the mantle ceased. It follows that most deformation at the northern and southern Caribbean plate boundaries in the Miocene and younger was entirely governed by changes in absolute plate motion of the North America and South America plates, and the resulting motion relative to the Caribbean plate. While strike-slip along the northern and southern Caribbean margins continued since the east-west-component of absolute plate motion of the Americas was far larger than the north-south components, the magnitude of the latter increased, probably at the Oligocene-Miocene boundary, resulting in convergence between the two Americas. The rate of convergence increased from east to west, resulting in an eastward directed "squeeze" on the Caribbean plate which caused its breakup along the Beata Ridge, where east-west oriented compressional stresses are absorbed (Mauffret and Leroy, this volume).
Mann et al. (1995) show a model for the formation of Caribbean microplates in six stages from the Maestrichtian to present-day in a fixed South American framework. Their figures show that the Caribbean plate has moved eastward by about 800 km since the mid-Oligocene (relative to South America). Mann et al. (1995) reason that collision ceased in the mid-Eocene in central Cuba since the arc could advance no further to the north-northeast above the Bahamas platform, and that this event rotated Caribbean plate motion clockwise in a more easterly direction. They favor the "tectonic escape" mechanism proposed by Burke and Sengör (1986), which results in the motion of a colliding plate towards the remaining "free face", e.g. an island arc. In case of the Caribbean in mid-Eocene times, the remaining free face would have been towards the east, i.e. the Lesser Antilles Arc. This argument is plausible. However, if we put Mann et al.'s (1995) reconstructions in the Atlantic-Indian hotspot framework, it appears that the eastward motion of the Caribbean plate was virtually stopped at mid-Eocene times.
Our arguments show that the apparent continuing apparent "escape" of an arc system as described by Royden (1993) e.g. for the Scotia Arc between South America and the Antarctic plate, may not necessarily involve the absolute motion of a small plate (e.g. the Scotia Sea Plate) relative to the mantle. A retreating subduction boundary may be initiated by the change in polarity of a subduction system, as in the case of the proto-Caribbean, but the Caribbean plate never reached the "open ocean" as in the case of the Scotia Sea (Royden, 1993), since its eastward migration was inhibited by boundary forces to the north and south, due to progressive convergence between the two Americas.
New gravity anomaly data from satellite altimetry and new magnetic data allow us to construct a modified plate model for plate motions between the two Americas, and calculate its uncertainties. For the North-South America plate boundary area east of the Lesser Antilles Arc our results are in good agreement with the observed strong plate deformation of the oceanic crust at the Barracuda and Tiburon ridges. Müller and Smith (1993) inverted Bouguer anomalies for crustal layer structure, and found that the Moho is uplifted 2-4 km over short wavelengths (~70 km) at the Barracuda and Tiburon ridges, implying large anelastic strains and an unstable density distribution. Together with the plate model presented here, these results indicate that much of the unusually shallow Moho topography and crustal uplift of the Tiburon rise is a result of North America-South America convergence after chron 6 (19.0 Ma). This model is in contrast with Dolan et al.'s (1989, 1990) suggestion that the present topography of the Tiburon Rise has existed since the late Cretaceous.
Our results suggest that slow sinistral transtension/strike-slip between the two Americas lasted from chron 34 (83 Ma) until chron 25 (55.9 Ma), followed by roughly northeast-southwest oriented convergence until chron 18 (38.4 Ma). This first convergent phase correlates with a Paleocene-Lower Eocene calc-alkaline magmatic stage in the Greater Antilles, which is thought to be related to southward subduction of proto-Caribbean crust during this time. Relatively slow transpression until chron 8 is followed by a drastic increase in convergence velocity. Subsequent to chron 8 (25.8 Ma), probably at the Oligocene-Miocene boundary, fast convergence resulted in 92±22 km convergence from chron 8-6, 127±25 km from chron 6-5, and 72±17 km from chron 5 to the present measured at 11°N, 85°W, near the North Panama Deformed Belt. The Neogene convergence measured at the eastern Muertos Trough, at 17.5°N, 65°W, is 41±18 km from chron 8-6, 58±25 km from chron 6-5, and 22±17 km from chron 5-present day. The modeled convergence may correspond to the early Miocene onset of underthrusting of the Caribbean oceanic crust below the South America borderland in the Columbia and Venezuela basins, the onset of subduction in the Muertos trough, and folding and thrust faulting at the Beata Ridge and the Bahamas, and the breakup of the main part of the Caribbean plate into the Venezuela and Colombia plates, separated by the Beata Ridge acting as a convergent plate boundary (Mauffret and Leroy, this volume). The east-west shortening between the latter two plates may reflect the differential convergence between the two Americas, increasing from east to west.
The main differences with Pindell et al.?s (1988) model are:
1) Pindell et al.'s (1988) model implies relatively constant
convergence between the Americas of rates of about 5mm/y or less since chron
21, with the exception of faster convergence between chron 6 and 5.
Our model results in substantial variations in convergence rates from chron
25, as documented in Figure 10 and Table
6. In particular, we resolve an initial phase of fast convergence
between chron 8 (25.8 ma) and chron 6 (19.0 Ma) of nearly 10 mm/y, compared
with less then 4 mm/y from chrons 18-8 measured at 85°W. We suggest
that most of this convergence occurred after chron 6c (23.8 Ma), which corresponds
to the Oligocene-Miocene boundary, a plate reorganization in the South Pacific
(Cande et al., 1995), and the formation of the present-day deformed belts
north and south of the Caribbean area. Without identifying magnetic
anomaly 8, two stages of slow (chrons 13-8) and fast (chrons 8-6) convergence
are averaged.
2) We have computed uncertainties for our North America
- South America plate flow lines. Uncertainty ellipses for rotated data
points are especially helpful to evaluate whether or not we can resolve relatively
slow phases of relative motion.
3) We have put Caribbean plate reconstructions into the
Atlantic-Indian hotspot reference system from Müller et al. (1993).
This allows us to evaluate causes and effects of relative plate motions in
the Caribbean area.
Stéphan et al. (1986) put forward the hypothesis that both the Northern and the Southern Caribbean deformed belts are the result of the bending of the Caribbean continental frame related to east-west shortening. East-west shortening appears to have a variety of different causes. In the Panama area, east-west shortening is related to Nazca-South America convergence in that direction and collision of an east-west oriented arc with a north-south oriented margin (Mann and Corrigan, 1990; Wadge and Burke, 1983). In Hispaniola, northeast-southwest shortening is related to the interaction of the westward moving North American plate relative to a stationary Caribbean plate as shown before.
Our plate model shows well resolved north-south convergence between the Americas during the Neogene, and we argue that this plate convergence is likely the main cause for the formation of those deformed belts which cannot be attributed to east-west convergence as described above. We also suggest that the fast Neogene plate convergence between North and South America contributed to the late Miocene onset of the collision of the Costa Rica-Panama Arc with the western Cordillera of South America (Wadge and Burke, 1983; Eva et al., 1989; Mann et al., 1990; Mann and Corrigan, 1990). One of the main tectonic events affecting the Caribbean plate in the Neogene has been its breakup into the Venezuela and Colombia plates (see Mauffret and Leroy, this volume). The breakup may have been caused the observed east-west gradient in convergence between the Americas, the subduction of the buoyant Cocos plate under the Caribbean plate (Mauffret and Leroy, this vol.), or the north-northeastward motion of the Maracaibo block, which in turn may be related to differential North-South America convergence.
By combining Atlantic-Indian hotspots as a reference frame with
revised North America-African and South America-African relative plate motions,
and with a revised plate model for the Caribbean area we are able to show
the following:
1) The eastward escape of the Caribbean plate appears to
have ceased when the opening of the Cayman Trough started. This time
corresponds to the collision in central Cuba which prevented a further advance
of the Caribbean plate to the north-northeast above the Bahamas platform.
Only if the Cayman Trough would have opened later and spread considerably
faster than interpreted by Rosencrantz et al. (1988), and/or if contemporaneous
strike-slip between Jamaica and the Caribbean plate was considerably faster
than suggested by Burke et al. (1980) would the Caribbean plate have maintained
any eastward directed motion relative to the mantle after the mid-Eocene.
If we rely only on the interpreted post-chron 6 (19 Ma) spreading history
of the Cayman Trough by Rosencrantz et al. (1988), which is better constrained
than its previous opening, then the Caribbean plate is still found to have
been without any substantial motion relative to the mantle subsequent to chron
6 within the errors of absolute plate motion models.
2) It is not the case that North America-South America
plate motions had only minor effects on the development of the Caribbean region
after the Campanian, as suggested by Pindell et al. (1988). After the
breakup of the Caribbean plate into the Venezuela and Colombia plate, the
eastward driving force of the latter plate may have still been derived from
interactions with the Cocos plate. However, even if this is so, our
model suggests that that post-chron 8 (25.8 Ma) differential motion between
the Americas has resulted in a total of 291±64 km in convergence at
85°W near the North Panama Deformed Belt. In other words, the total
area available for the Colombian plate at 85°W was reduced by at least
~230 km in north-south direction in the last 25 million years. Surely
this reduction in space had a profound influence on the Colombian plate margin,
and has contributed to convergence along the North Panama Deformed Belt.
Van der Hilst and Mann (1994) show that the subducted Maracaibo slab underlying
northwestern South America extends up to 500 km from the Caribbean-South America
boundary to the south. The Maracaibo slab corresponds to subducted Caribbean
oceanic plateau crust. Our results suggest that about half of the north-south
extent of the Maracaibo slab under the South American continent may have resulted
from Miocene and later South America-Caribbean convergence, if most of the
North America-South America convergence was taken up at this boundary.
3) The Caribbean plate has been trapped between two larger
plates, and has been subject to edge-driven plate tectonic interactions since
then. It follows that the main control on North America-Caribbean and
South America-Caribbean plate interactions has not originated from Cocos plate-Caribbean
interactions, as these interactions have not resulted in any substantial motion
of the Caribbean plate relative to the mantle. Continuing oblique collision
along the passive margin of eastern Venzuela as reported by Algar and Pindell
(1993) must be attributed to the west-northwestward motion of South America
relative to the mantle and relative to a stationary Venezuela plate, rather
than to continuing eastward movement of the Caribbean plate. Equivalently
the Miocene and younger transpression observed in Hispaniola (Heubeck and
Mann, 1991) due to collision of arc rocks with the Bahamas platform is the
result of continuing westward motion of the North America plate (and the Bahamas
platform) relative to an approximately stationary Venezuela plate in a mantle
reference frame, rather than a continuing eastward "escape" of the Caribbean
plate. By the same token, the Gonave microplate, which has been detached
from the Caribbean microplate in the Pliocene and accreted to North America
has not been "left behind" (Mann et al., 1995), but has rather been broken
off the stationary Caribbean plate by the North American plate, leaving the
rest of the Venezuela plate behind.
4) The Tertiary tectonic history of the Caribbean plate
can be described by "tectonic escape" up to the mid-Eocene. Subsequently
the Caribbean plate came to a halt in the Atlantic-Indian mantle reference
system due to a progressive reduction in space between the two Americas for
the arc to move eastward and due to its collision with the Bahamas platform.
We show that the most severe reduction in space started at the Oligocene-Miocene
boundary, resulting in the gradual formation of many of today?s tectonic elements
and sedimentary basins in the Caribbean area.
Parameters are: r = sum of misfits, N = number of data points, s = number
of great circle segments, df = degrees of freedom, U = df/r (see text
for discussion). The parameters U and the misfit r in this table
are calculated with nominal uncertainties of 4 km for the magnetic anomaly
crossings and 5 km for the fracture zone crossings. The "true" data
uncertainties mag and fz
are related to their unknown estimates ()
(i.e. 4 and 5 km for magnetic and fracture zone data, respectively) by the
quality factor k={
/ }2.
Covariance matrices can be recontructed in the following way:
See parameter legend in Table 2.
The parameters rmaj and rmin are the semi-major and semi-minor axes of the
ellipse of confidence (95% level). The variable maj is the azimuth of
the semi-major axis.
Figure 3. New dense magnetic anomaly data in the Canary-Bahamas Transect area north of the Kane Fracture zone (Maschenkov and Pogrebitsky, 1992), combined with data from other sources.
Figure 4. Gravity anomalies from satellite altimetry from Sandwell and Smith (in press)
and interpreted and rotated magnetic anomaly and fracture zone identifications
in the central North Atlantic. The unrotated magnetic and fracture zone
identifications are identified by the following symbols: triangle (C5,
9.74 Ma; C18, 38.43 Ma; C30, 65.58 Ma); square (C6, 19.05 Ma, C21; 46.26 Ma;
C32, 71.59 Ma), upside down triangle (C8, 19.05 Ma; C24, 52.36 Ma; C33o, 79.08
Ma), circle (C13, 33.06 Ma; C25, 55.90 Ma; C34, 83 Ma). All rotated
data points are marked by crosses. Paleoridge or transform segments
as defined by magnetic anomaly or fracture zone identifications, approximated
as great circles in the inversion method used here, are denoted by alternating
small and large symbols.
Figure 5. Gravity anomalies from satellite altimetry from Sandwell and Smith (in press) and interpreted and rotated magnetic anomaly and fracture zone identifications in the South Atlantic. The symbols used for plotting magnetic and fracture zone identifications follow the same convention as in Figure 4.
Figure 6. Finite rotation poles and 95% confidence ellipses for North
America-Africa (upper right) and South America-Africa (lower left) plate motions
for 12 reconstruction times from chron 34 (83 Ma) to the present. As
a comparison we show the North America-Africa pole path from Klitgord and
Schouten (1986) and the South America-Africa pole path from Shaw and Cande
(1990) (gray triangles). Note the sharp cusp in our North America-Africa
pole path at chron 8 (25.8 Ma), resulting in a pole path for chron 5 to chron
8 that is distinctly different from Klitgord and Schouten's (1986) path.
The differences between ours and both Klitgord and Schouten's (1986) and Shaw
and Cande's (1990) models reflect a much improved accuracy in locating fracture
zones based on a dense gravity anomaly grid from satellite altimetry (Sandwell
and Smith, in press), and some new magnetic anomaly data.
Figure 7a. Gravity anomalies and South Atlantic symmetric plate flow lines in the equatorial Atlantic ocean. All flow lines have been constructed using both ridge-transform intersections as seedpoints. The area encompassed by the resulting dual flow lines approximates the amount of transpression or transtension that occurred during changes in spreading direction, assuming symmetric plate accretion. The Marathon and Mercurius fracture zones do not fit the computed flow lines well. However, the Four North and Doldrums fracture zones show a good fit to our flow lines. The latter observation confirms the validity of our South Atlantic plate motion model, whereas the former indicates that the North-South America plate boundary may be located in the vicinity of the Marathon/ Mercurius fracture zones, resulting in deviations from South America-Africa flow lines.
Figure 7b. Gravity anomalies and central North Atlantic symmetric plate flow lines in the equatorial Atlantic ocean. The Marathon fracture zone flow line clearly does not match the gravity anomaly expression of this fracture zone. The post-chron 6 flow line of the Mercurius fracture zone fits its eastern limb better than the South America-Africa flow line (compare with Figure 6a), but not its westen limb. Its is clear that this fracture zone would not be useful to constrain either North America-Africa nor South America-Africa plate motions, since it appears to have been subject to recent plate boundary deformation.
Figure 10. Gravity anomalies in the Caribbean area and North-American-South America plate motion through time of 3 points attached to the North America plate with respect to the South America plate for 8 stages from chron 34 (83 Ma) to the present. The relative motion vectors can be divided into four groups: 1) Slow sinistral strike-slip/transtension between chrons 34 and 25 (~2.8±0.8 - 4.8±1.1 mm/y at 85°W), 2) Northeast-southwest oriented convergence from chron 25-18 (6.5±1.5 mm/y at 85°W, 3) slow motion from chron 18-8 (3.6±2.1 mm/y at 85°W), and 4) fast north-south oriented convergence from chron 8-6 (9.6±3.1 mm/y for chrons 8-6 and 9.6 ±2.1 mm/y for chrons 6-5 at 85°W), followed by a decelaration in north-south oriented convergence post chron 5 (5.2±1.3 mm/y at 85°W). See Table 3b for a complete list of stage motion vectors.
Figure 11. North-American-South America plate motion through time of 3 points attached to the North America plate with respect to the South America plate computed for the same 7 stages as Pindell et al.'s (1988) model, shown by triangles. The general shape of both models is similar, reflecting the similarity of the magnetic anomaly data sets used to constrain both models. The main differences between the models are: 1) Pindell et al.'s (1988) model implies sinistral strike-slip between North and South America from chron 34-21, followed by convergence, whereas our model implies strike-slip until chron 25, followed by convergence. 2) Pindell et al.'s (1988) model implies relative constant convergence between the Americas of rates of about 5mm/y or less since chron 21. Our model results in substantial variations in convergence rates from chron 25, as documented in Figure 10 and Table 3b. In particular we resolve the onset of fast convergence after chron 8 of nearly 10mm/y, compared with les then 4 mm/y from chrons 18-8 measured at 85°W. This figure also shows the estimated motion of the Caribbean plate in a hotspot reference frame since chron 21 (open circles). See text for discussion.
Figure 12. North-American-South America plate motion through time of 3 points attached to the North America plate with respect to the South America plate for 8 stages from chron 34 (83 Ma) to the present and their simultaneous 95% confidence regions. Only the Barracuda Ridge area has been affected by North-South America plate motions as old as chron 34 (83 Ma). The ocean crust to the east becomes successively younger. The oldest relative motion vectors plotted correspond roughly to the age of the ocean crust south of the Fifteen-Twenty Fracture Zone. Relative plate motion in this area is not well resolved, except for oblique transtension in the Barracuda Ridge area from chron 34 to 25, and post chron 6 north-south compression, post-chron 6 extension in the Royal Trough area.
Figure 13. Plate reconstructions of the Caribbean area in an Atlantic-Indian mantle reference system for the following times: 55.9 Ma, 46.8 Ma, 25.8 Ma, 9.7 Ma.
Maintained by Bill ReidDivision
of Geology and Geophysics IT Support
31Jan, 2001