Late Cretaceous subduction initiation on the eastern margin of the Caribbean-Colombian Oceanic Plateau: One Great Arc of the Caribbean (?)

There is widespread agreement that the interior of the Caribbean Plate (Fig. 1) is largely underlain by Pacific oceanic crust that was subsequently thickened by oceanic plateau magmatism. Bathymetry and seismic imaging indicate that the Caribbean Plate is significantly thicker, ∼10–20 km in aggregate thickness, than normal oceanic crust (e.g., Diebold et al., 1981; Mauffret and Leroy, 1997; Driscoll and Diebold, 1998). Geologic and geochemical studies from Deep Sea Drilling Project (DSDP) sites, submersible collections from the Beata Ridge (Fig. 1), and subaerial exposures of widely dispersed basaltic igneous rocks believed to be accreted fragments of the Caribbean Plate, all indicate a within-plate oceanic plateau origin (e.g., Donnelly, 1973; Donnelly et al., 1990; Kerr et al., 1996; Sinton et al., 1998; White et al., 1999; Revillon et al., 2000; Kerr et al., 2009). Following Kerr et al. (2003), we refer to exposures of the plateau rocks as the Caribbean-Colombian Oceanic Plateau (CCOP) throughout the remainder of the paper. Relatively recent geochronological investigations (Alvarado et al., 1997; Sinton et al., 1997, 1998; Lapierre et al., 1999; Walker et al., 1999; Revillon et al., 2000; Luzieux et al., 2006) suggest the bulk of the CCOP formed in a relatively short amount of time (92–88 Ma; Turonian–Coniacian according to the time scale of Gradstein et al. [2004]), although there is evidence for younger magmatic additions (e.g., Revillon et al., 2000), and, as discussed in more detail below, there is also evidence on Curaçao for an older Albian period of plateau magmatism.

The Caribbean Plate is rimmed on the north, east, and south by islands of the Greater, Lesser, and Leeward Antilles, respectively (Fig. 1). Most of the Greater Antilles contain magmatic arc assemblages of Early Cretaceous through Eocene age (Pindell et al., 2005, and references therein). The Lesser Antilles, Aves Ridge, and Grenada Basin represent an arc/remnant arc/back arc basin triad formed during Late Paleocene–Eocene arc rifting and back arc basin formation (Fig. 1; e.g., Bouysse, 1988). The Lesser Antilles contain the modern arc produced by subduction of Atlantic Ocean floor beneath the eastern Caribbean.

The Leeward Antilles are isolated subaerial exposures of a submarine ridge that extends from Aruba on the west to La Blanquilla on the east at the southern terminus of the Aves Ridge (Fig. 1). These islands have played a pivotal role in models for the tectonic evolution of the Caribbean Plate because they contain onland exposures of CCOP rocks and Cretaceous magmatic arc assemblages (Beets et al., 1984; White et al., 1999; Kerr et al., 2003; Thompson et al., 2004). Many studies have implied or concluded that Cretaceous arc rocks in the Greater Antilles, Aves Ridge, and Leeward Antilles were originally continuous (Beets et al., 1984; Duncan and Hargraves, 1984; Bouysse, 1988; White et al., 1999; Kerr et al., 2003; Thompson et al., 2004; Jolly et al., 2006; Pindell et al., 2006) and formed a single “Great Arc” of the Caribbean (Fig. 2; Burke, 1988).

The origin of the Caribbean Plate and its fringing arc system is controversial. Most workers interpret the Caribbean Plate and associated “Great Arc” to have originally formed in the Pacific. Ultimately, subduction of proto-Caribbean seafloor led to insertion of the Caribbean Plate between the Americas. Some models suggest that the plateau originated on oceanic crust subducting to the east beneath the “Great Arc,” and that arrival of the plateau at the subducting boundary in the early Late Cretaceous led to polarity reversal, trapping the overthickened plateau crust behind an east-facing arc superimposed on the older west-facing arc (Fig. 2A; e.g., Burke et al., 1984; Duncan and Hargraves, 1984; White et al., 1999; Kerr et al., 2003; Thompson et al., 2004). In this model, Late Cretaceous magmatism in the “Great Arc” following polarity reversal would have occurred on a composite basement consisting of Early Cretaceous arc rocks and the margins of the CCOP, with evidence (structural or otherwise) for a change in the location of the plate boundary in the early Late Cretaceous. Other studies present several lines of evidence indicating that subduction polarity reversal more likely occurred in the Early Cretaceous, prior to development of the CCOP (e.g., Pindell et al., 2006; Jolly et al., 2006). In these models, there was no collision, the Aptian–Albian and younger “Great Arc” was continuously east-facing, and the CCOP developed west of the arc and on the same plate (Fig. 2B). There is also a body of literature that interprets the Caribbean Plate to have formed essentially in its present location between the Americas. This in situ model, most recently summarized by James (2009), has been challenged by numerous Caribbean researchers (e.g., Pindell and Kennan, 2009; Maresch et al., 2009; Stanek et al., 2009; Kerr et al. 2009; Diebold, 2009; Kennan and Pindell, 2009).

The two Pacific end-member models for the evolution of the Caribbean Plate make very different predictions for geologic relations in the Leeward Antilles. In order to test these models, we conducted a comprehensive study of the geology of the large western islands of the Leeward Antilles, Aruba, Curaçao, and Bonaire (ABC islands), and the small island of La Blanquilla to the east (Fig. 1). Our analysis is based on a combination of new detailed mapping, stratigraphic and structural studies, geochemical analyses, and geochronology on all four islands, and is integrated with previously published data. The study reveals pronounced differences in the Cretaceous to Paleocene geology of the islands that cannot be explained by either of the prevailing end-member models for Caribbean evolution, and that are inconsistent with the concept of a long-lived and continuous “Great Arc” system. Instead, our analysis suggests that the Greater Antilles Arc was terminated at its eastern edge by a transform boundary, that fragments of the Greater Antilles were translated into what is now the southern Caribbean realm along this boundary, and that subduction initiated beneath the CCOP (at the former transform boundary) at ca. 90 Ma, forming a Late Cretaceous magmatic arc that developed independently of the Cretaceous arc of the Greater Antilles. These results also provide a framework that serves to link the southern Caribbean Arc rocks to similar Late Cretaceous arc sequences exposed in Ecuador and Colombia.

In the following sections, we describe geologic relations and new data from the ABC islands and La Blanquilla. Our focus is on their Cretaceous to Paleocene evolution. For the most part we do not address the younger Paleogene to recent accretion and dispersal of Caribbean terranes along the Venezuelan margin. Throughout the text, we use the time scale of Gradstein et al. (2004) for the ages of Cretaceous and Paleogene stage boundaries. Analytical methods and data for geochronology (U-Pb zircon, ⁴⁰Ar/³⁹Ar) and geochemistry (major and trace elements) are presented as supplemental files. U-Pb zircon analyses (plutonic and detrital samples) were obtained either with the sensitive high resolution ion microprobe-reverse geometry (SHRIMP-RG) at the Stanford–U.S. Geological Survey Micro-Analytical Center, or by laser ablation-multicollector-inductively coupled plasma-mass spectrometry (LA-MC-ICP-MS) at the University of Arizona LaserChron Center (see Supplemental File 1¹). Geochemical analyses were performed by X-ray fluorescence (XRF) and ICP-MS at the Washington State University GeoAnalytical Laboratory (see Supplemental File 2²). ⁴⁰Ar/³⁹Ar analyses of detrital hornblende were obtained at the U.S. Geological Survey Thermochronology Laboratory in Denver, Colorado (see Supplemental File 3³).

Bedrock geology of Aruba consists of the Late Cretaceous Aruba Lava Formation (ALF) and crosscutting Aruba batholith, overlain unconformably by Eocene limestone and younger strata (Fig. 3; Westermann, 1932; Monen, 1980; Beets et al., 1977, 1996; White et al., 1999). The Aruba batholith forms most of the basement exposure; the ALF is primarily exposed in the central part of the island (Fig. 3A). The ALF, which consists largely of weakly metamorphosed mafic lavas, diabase intrusions, and associated volcaniclastic strata, is widely interpreted to be an exposed part of the CCOP on the basis of geochemical and age similarities (Kerr et al., 1997; Sinton et al., 1998; Beets et al., 1984; White et al., 1999). In particular, major element, trace element, and isotopic analyses (White et al., 1999) show that mafic rocks of the ALF are similar to other analyzed examples of the CCOP (e.g., Sinton et al., 1998; Revillon et al., 2000) and are chemically distinct from Caribbean arc magmatic associations. The dominantly tonalitic Aruba batholith, in contrast, has geochemical similarities to subduction-related magmas (Beets et al., 1984; White et al., 1999; Kerr et al., 2003; Thompson et al., 2004). Together, the ALF and Aruba batholith have played a prominent role in tectonic models of the southern Caribbean (e.g., Beets et al., 1984; White et al., 1999; Thompson et al., 2004). In order to better understand the history recorded by these Cretaceous rocks, we mapped the central part of the island (Fig. 3B) at a scale of 1:25,000. Our new work indicates important differences from prior studies in how the Cretaceous geology of Aruba should be interpreted.

Previous mapping of the island (summarized by Beets et al., 1984, 1996) led to interpretation of the ALF as an unbroken stratigraphic succession consisting of interlayered basalt flows, pyroclastic and volcaniclastic deposits, and epiclastic strata, intruded at all levels by diabase sills. Apparent interlayering of submarine and subaerial strata formed the basis for concluding that the ALF likely accumulated near sea level, on or near the flanks of an emergent volcano (Beets et al., 1984; White et al., 1999). In contrast, our mapping indicates an important unconformity on the island that separates the ALF into two very different stratigraphic packages, the entirely submarine lower ALF and the subaerial upper ALF (Fig. 3C).

The lower ALF forms the majority of the exposure area and can be divided into three mapable units (basalt, argillite, and diabase units) named for the dominant rock types within each (Fig. 3). The basalt unit consists mostly of pillowed to massive basalt flows (Fig. 4A), with only rare exposures of stratified rocks including basaltic tuff or volcaniclastic mudstone and sandstone. The overlying argillite unit (Fig. 4B) consists mostly of thinly bedded, variably silty argillite, with less common volcaniclastic siltstone and sandstone, pumiceous lapilli tuff, and pebble conglomerate. Graded bedding is common and indicative of deposition from turbidity currents. The diabase unit intrudes both the basalt and argillite units (Fig. 3) and consists of a complex assemblage of texturally variable dikes, sills, and small intrusions. Most common are fine- to medium-grained, equigranular rocks (Fig. 4C) with clinopyroxene as the dominant mafic phase.

The upper ALF, which is only locally present (Fig. 3B), consists of distinctive pyroclastic and epiclastic strata that overlie units of the lower ALF along an unconformity (Fig. 3). Two units can be distinguished: a locally preserved basal unit of basaltic tuff and an overlying polymictic conglomerate (Fig. 3C). Where present, the tuff unit is depositional on the lower ALF; elsewhere, the conglomerate unit rests directly on the lower ALF. In most places, the tuff unit consists of well-bedded accretionary lapilli tuff (Fig. 4D), although locally it is dominated by tuff breccias with abundant lapilli and bombs of basalt scoria. Accretionary lapilli show no evidence of fragmentation or abrasion (Fig. 4D). The conglomerate unit consists mostly of pebble to cobble conglomerate, with only local interbeds of finer-grained strata. Clasts in this unit are generally well-rounded and are derived entirely from underlying units of the ALF (Fig. 4E) including the accretionary lapilli tuff unit.

The unconformity between the lower and upper ALF reflects a period of exhumation and erosion, and a transition from marine deposition to subaerial conditions (Fig. 3C), as indicated by the following. (1) Diabase and gabbro clasts are common in the upper ALF conglomerate unit (Fig. 4E) and were clearly derived from erosion of the underlying lower ALF diabase unit. This requires some amount of exhumation and erosion. (2) There is abundant evidence of subaerial exposure and weathering along the unconformity, based on the presence of metamorphosed spheroidal weathering features in lower ALF basalts and diabase that immediately underlie the contact (Fig. 4F). For comparison, modern spheroidal weathering of the diabase unit on Aruba is shown in Figure 4G. Unlike the modern examples, the Cretaceous weathered rinds are composed of metamorphic minerals (pumpellyite and chlorite in most places, but also including metamorphic amphibole in outcrops near the Aruba batholith). The metamorphosed spheroidal weathering features are also locally cut by dikes emanating from the Aruba batholith, and are found exclusively in lower ALF rocks just below the unconformity. (3) Strata in the upper ALF appear to have been deposited in a subaerial setting: accretionary lapilli are generally believed to form in subaerial eruption columns (e.g., Schumacher and Schmincke, 1995), and sedimentary features of the conglomerate unit, including rounded clasts, suggest deposition in a fluvial setting and certainly indicate surface erosional processes. (4) Finally, the unusual local preservation and patchy outcrop pattern of the upper ALF (Fig. 3), in conjunction with the evidence for subaerial accumulation, suggests that it may have been deposited in canyons or river valleys, only remnants of which are now preserved. We were unable to structure contour the contact between the upper and lower ALF because there is not enough topographic relief on the contact. This, along with the subsequent folding of both the upper and lower ALF, prevents construction of a meaningful cross section across the unconformity.

The age of the exposed part of the ALF is constrained by the presence of ammonites collected from a pebbly mudstone within the argillite unit of the lower ALF that have been interpreted as Turonian in age (MacDonald, 1968) and by the 89 ± 1 Ma age of the Aruba batholith and associated dikes, which crosscut all units of the ALF (see below). Humphrey (2010) has obtained a preliminary U-Pb micro-zircon date of 97.3 ± 5.2 Ma on a gabbro from the lower ALF. This date is in agreement with other data that indicate a Late Cretaceous age for the lower ALF but does not help resolve a more precise age as it spans the Latest Albian to early Turonian within analytical uncertainty.

As mapped by Beets et al. (1996) and White et al. (1999), the Aruba batholith is a predominantly tonalitic intrusion that postdates deposition, eruption, and intrusion of the ALF. Our new mapping confirms that the pluton, and abundant associated dikes, cut discordantly across all units of the ALF (Fig. 3). The batholith has been the subject of a number of geochronological investigations, including that of White et al. (1999) who reported ⁴⁰Ar/³⁹Ar ages suggesting intrusion during the interval 85–82 Ma. We analyzed zircon separates from a sample of the pluton (location shown by the star in Fig. 3B; data and methods presented in Supplemental File 1 [see ^{footnote 1}]). Based on the weighted mean of nine individual analyses, our U-Pb data (originally reported in Wright and Wyld, 2004) indicate an emplacement age of 89 ± 1 Ma (Fig. 5; Supplemental File 1 [see ^{footnote 1}]). This is consistent with 87–90 Ma ages (⁴⁰Ar/³⁹Ar and U-Pb zircon) obtained recently from the batholith by van der Lelij (2008, 2010). Collectively, the data indicate that the Aruba batholith is close in age to the ALF wall rocks.

Chemically the batholith shows similarities to subduction-related magmas, including negative Nb anomalies, but isotopically the batholith is similar to basalts and diabases of the CCOP, including samples from the ALF (White et al., 1999). On a K-Ca-Na plot, samples from the batholith do not follow a typical calc-alkaline trend, but instead plot in the tonalite-trondhjemite-granodiorite (TTG) field (White et al., 1999). Elevated Sr/Y and modestly depleted heavy rare earth elements (HREE) in some samples also imply affinity of the batholith to TTG as well as Cenozoic adakites (White et al., 1999). Models for batholith petrogenesis include remelting of CCOP basaltic crust during injection of new plume-related magma, melting of the CCOP during incipient subduction (White et al., 1999), or melting of the CCOP by injection of subduction-generated magmas (Thompson et al., 2004). As explained later, we favor a subduction model.

Deformation and metamorphism of the ALF is well-known and has long been associated with intrusion of the Aruba batholith (Westerman, 1932; Monen, 1980; Beets et al., 1984, 1996; White et al., 1999). The principal structures noted in these previous studies are generally E-W–striking faults (of unspecified sense of offset) and apparently spatially associated zones of strong foliation (shear zones of White et al., 1999).

Our detailed mapping reveals a significantly different structural and metamorphic history for the ALF. First, we found little evidence for any of the faults shown on previous maps; most apparent offset contacts on older maps are actually irregular stratigraphic or intrusive contacts (Fig. 3B). Second, we found no evidence that foliation in the ALF is restricted to narrow zones. Intensity of foliation development varies with rock type: basalt, diabase, and conglomerate exhibit a variably developed cleavage, whereas fine-grained clastic rocks and tuffs have a more penetrative foliation (Fig. 6A), but the foliation is found in all ALF units and throughout the mapped area (Fig. 6C).

Our structural studies indicate clear evidence for an episode of regional deformation and metamorphism that entirely predates intrusion of the Aruba batholith. Structures formed during this event include open to tight folds at the outcrop and map scale (Figs. 6B and 6C), and the foliation, which is axial planar to folds and defined in part by preferred alignment of subgreenschist facies metamorphic minerals. These structures and metamorphic fabrics are crosscut sharply by dikes associated with the Aruba batholith (Fig. 3), overprinted by contact metamorphism (hornblende hornfels facies) in a narrow zone around the Aruba batholith (see dashed line in Figs. 3B and 6C), and progressively reoriented toward parallelism with the batholith contact in the thermal aureole (Figs. 6C and 6D). These relations require that batholith emplacement postdates and is unrelated to regional deformation of the ALF on Aruba. If the interpreted age of the ammonite collected from the lower ALF is correct, then the timing of deformation is constrained to the Turonian, by the Turonian age of the deformed ALF and the 89 ± 1 Ma age of the post-tectonic Aruba batholith.

Results of our study indicate that the Late Cretaceous (Turonian–Coniacian) stratigraphic, structural, and magmatic evolution of Aruba is much more dynamic than previously recognized. CCOP magmatism in the Turonian (ALF) was replaced abruptly by arc magmatism at 89 ± 1 Ma (Aruba batholith). Simultaneously, Aruba went from a submarine environment to subaerial conditions to depths associated with regional deformation and then batholith emplacement in a span of <4 m.y. Coincident timing between this remarkable vertical journey and the transition in magma genesis on the island suggests that the processes are likely related, as we explore in a later section.

Bedrock geology of Curaçao consists of the Cretaceous Curaçao Lava Formation (CLF) and overlying sedimentary strata of the Late Cretaceous to Early Paleocene (Danian) Knip Group and Midden-Curaçao Formation (Fig. 7; Beets, 1972; Klaver, 1976; Beets et al., 1977). These rocks are overlain unconformably by Eocene limestone and younger strata (Fig. 7). The CLF, which consists mostly of mafic submarine lava flows and is at least 5 km thick, is widely considered to be an exposed part of the CCOP and to be correlative with the ALF on Aruba (Beets et al., 1984; Donnelly et al., 1990; Kerr et al., 1996; Sinton et al., 1998; White et al., 1999). The Knip Group and Midden-Curaçao Formation, in contrast, have no counterpart on Aruba (compare Figs. 3 and 7). Finally, Curaçao was not affected by regional metamorphism and deformation until the Late Paleocene (Beets, 1972), at least 30 m.y. after the deformation on Aruba.

In order to better understand and interpret the Cretaceous evolution of Curaçao, especially in relation to that of Aruba, we examined the geology of the northwest part of the island, where pre-Eocene units are best exposed, and collected samples for geochronology. Our analysis builds on the detailed mapping of Beets (1972).

Bedrock geology of Curaçao is dominated by the CLF (Fig. 7), a sequence of mafic, submarine lava flows, commonly pillowed, with minor intercalations of basaltic hyaloclastite and some diabase sills. Compositionally, the CLF consists of picrites to olivine-phyric tholeiites in the lower part of the formation and olivine-phyric tholeiites to plagioclase-clinopyroxene tholeiites in the upper part (Beets, 1972). Only one nonvolcanic interval has been found in the CLF, a thin succession of pelagic limestone and siliceous shale interlayered with hyaloclastites, located in the upper part of the CLF in the southeast part of the island (Beets, 1972; Klaver, 1987).

The great thickness of the CLF and paucity of intercalated or interpillow sediments suggests that CLF lavas were erupted in a short time interval (Beets, 1972; Klaver, 1987). Sinton et al. (1998) reported five ⁴⁰Ar/³⁹Ar analyses of CLF basalts; two samples (one collected near the top of the unit and the other near the base) yielded plateau ages of 89.5 ± 1 Ma and 88.9 ± 0.8 Ma. These Turonian–Coniacian radiometric dates conflict with an Albian age defined by ammonites from the pelagic interval (Wiedman, 1978), leading Kerr et al. (1997) to suggest that the ammonites were either misidentified or reworked. Humphrey (2010) has obtained a 112.7 ± 7.3 Ma U-Pb micro-baddeleyite date from a diabase intruding the Upper CLF. This new result is in conflict with the ⁴⁰Ar/³⁹Ar ages and is in general agreement with the age call on the ammonites. The available age data indicate that plateau magmatism on Curaçao likely significantly predates that on Aruba.

The CLF is unconformably overlain in most places by the Campanian to Maastrichtian Knip Group (Fig. 7). As documented in detail by Beets (1972), a variety of features indicate that this unconformity marks a period of uplift and subaerial exposure of the CLF. First, weathering and soil formation processes are indicated by discontinuous zones of fragmented and mineralized CLF just below the contact. In these zones, CLF rocks display extensive brecciation with gaps between fragments filled by carbonate, jasper, and iron oxide; these “in-situ breccias,” whose fragments fit together like a jigsaw puzzle, extend downward and laterally into unbrecciated basalt (Beets, 1972). Second, the contact is also locally marked by thin, discontinuous lenses of shallow marine limestone that separate the underlying deep marine CLF from the overlying deep marine Knip Group (Beets, 1972). These limestones yield fossils of late Santonian to early Campanian age and most likely reflect renewed transgression following exposure.

Dikes and sills of dioritic composition are locally common on Curaçao, but no larger intrusions are present (Beets, 1972). Two general groups can be distinguished. An older group consists of semicircular dikes (necks of Beets, 1972) that are <50 m in diameter and intrude the CLF in the northwest part of the island. These have the composition of quartz diorite and contain variably abundant phenocrysts of plagioclase and hornblende. Another group of dikes and sills intrudes the Knip Group and Midden-Curaçao Formation but not the overlying Eocene and younger strata. These younger intrusions consist mostly of deeply weathered leucocratic diorite with hornblende phenocrysts.

In an effort to determine whether the older dikes could be related to the magmatic event which generated the nearby Aruba batholith, we collected samples for geochronology and geochemistry (for location, see CUR-17 on Fig. 7). U-Pb geochronology (data and methods contained in Supplemental File 1 [see ^{footnote 1}]) indicates that the older dikes are close in age to the Aruba batholith. Based on the weighted mean of 11 individual zircon analyses from one of the dikes, our U-Pb data indicate an intrusion age of 86.2 ± 0.8 Ma (Fig. 8; Supplemental File 1 [see ^{footnote 1}]). Trace-element data from dike samples (data and methods contained in Supplemental File 2 [see ^{footnote 2}]) are compared in Figure 9 with data from the Aruba batholith collected by White et al. (1999). Chemically, the dikes are quite similar to the Aruba batholith. Both produce nearly identical patterns, with Ta/Nb anomalies, on a multielement diagram normalized to primitive mantle (Fig. 9A), and both show moderate depletion in HREE on a chondrite-normalized REE plot (Fig. 9B). Likewise, Sr/Y values of the dikes and the batholith also overlap (Fig. 9C). REE patterns and Sr/Y values are similar to those of adakites and suggest melting of a garnet-bearing mafic source. Our interpretation of the data from Aruba and Curaçao is that this magmatic episode represents subduction initiation along the margin of the CCOP and that mantle-derived magmas may have ponded in the lower crust of the CCOP. Resultant melting of the CCOP led to the adakite-like magmatism recorded on Aruba and Curaçao. Thus, Curaçao, like Aruba, records a fundamental change from within-plate CCOP related magmatism to arc related magmatic activity at ca. 89–86 Ma.

We also collected a sample of one of the younger dikes for U-Pb zircon geochronology, but no zircons were found. Crosscutting relations, however, indicate that this group was most likely emplaced in the Paleocene (65–52 Ma; see Beets, 1972; Fig. 7).

The Knip Group consists of up to 2 km of deep marine strata that were deposited unconformably on the CLF in the Campanian to Maastrichtian (Beets, 1972). It consists mostly of pelagic chert, cherty limestone, and argillite, interspersed at various levels with epiclastic and volcanogenic deposits (Fig. 7C). The latter include discontinuous wedges of coarse-grained breccias derived from erosion of the CLF at the base of the group, and turbidite deposits with both continental and arc-derived detritus at higher stratigraphic levels (e.g., Lagoen Formation) along with local intervals of tuffaceous volcaniclastic sandstone (e.g., Koea Joeda Member of Lagoen Formation) (Fig. 7C; Beets, 1972). Volcanogenic material in the Knip Group is of andesitic composition and distinctly different from that found in the CLF (e.g., hornblende is abundant in the clastics). Siliciclastic detritus includes quartz and muscovite, plus quartzite, mica schist, and felsic plutonic lithics. Based on the stratigraphic analysis of Beets (1972), the Knip Group was deposited during a period of unrest on Curaçao, leading to large thickness variations of units from the southeast part of the island (very thin) to the northwest (much thicker; Fig. 7), and rapid lateral facies variations.

The overlying Early Paleocene (Danian) Midden-Curaçao Formation is a clastic succession, at least 1 km thick, of conglomerate, sandstone, and shale (Beets, 1972). It reflects continued submarine fan deposition of clastics with a mixed continental and arc provenance, but it lacks the pelagic and tuffaceous strata found in the underlying Knip Group.

Samples were collected for detrital zircon U-Pb geochronology from volcaniclastic sandstones of the Lagoen Formation of the Knip Group (sample CUR-21) and from siliciclastic sandstones of the Midden-Curaçao Formation (sample CUR-14; see Fig. 7 for locations and Supplemental File 1 [see ^{footnote 1}] for data and methods). The Midden-Curaçao sample was collected from the Jan Kok member (Beets, 1972) of the formation (Fig. 10A) and contains abundant quartz and detrital muscovite (Fig. 10B). The Lagoen sample was collected from the Koea Joeda member of the formation (Beets, 1972), which is a unit dominated by immature clastics containing an abundance of plagioclase and hornblende crystals with less abundant continentally derived material such as muscovite, quartz, and quartz-rich metasedimentary lithics (Fig. 10C). The geochronological results (Supplemental File 1 [see ^{footnote 1}] are similar for both units and consistent with a dual provenance, one from the continental margin of South America, as exhibited by early Mesozoic, Paleozoic, and Precambrian grains, and the other from a Cretaceous arc (Fig. 11). The intimate mixing of arc and continental detritus (Figs. 10C and 11) indicates a combined arc and continental source region. As discussed more fully below, we suggest that the detritus was derived from the ca. 75 Ma collision zone between an arc terrane constructed on the CCOP and the Ecuadorian/Colombian continental margin of South America (Vallejo et al., 2009; Cardona et al., 2008).

We also obtained ⁴⁰Ar/³⁹Ar geochronology on detrital hornblende separated from two samples of the Koea Joeda Member: sample CUR-21, the same sample from which we also analyzed detrital zircon, and sample CUR-22, which was collected from a separate but similar sandstone bed higher in the unit (Fig. 7; see Supplemental File 3 [see ^{footnote 3}] for data and methods). Both samples yield ⁴⁰Ar/³⁹Ar plateau and isochron ages that are ca. 74 Ma (Fig. 12; Supplemental File 3 [see ^{footnote 3}]), which corresponds closely to the main peak of Cretaceous ages in our detrital zircon data (Fig. 11, inset).

The deformation history of Curaçao differs markedly from that of Aruba. There is no evidence on Curaçao for the Turonian regional deformation that is so prominent on Aruba. Instead, Curaçao was affected by a single phase of regional deformation and metamorphism that affected all units older than the Eocene, including the Midden-Curaçao Formation (Beets, 1972). This Paleocene event is characterized by NW-trending folds at various scales, from the outcrop to the map scale where they control the map pattern of bedrock units (Fig. 7; Beets, 1972; Klaver, 1987). No tectonic foliation was formed, but mineral assemblages indicate metamorphism at phrenite-pumpellyite to zeolite grade.

The older history recorded on Curaçao is similar to that on Aruba although the CLF does appear to be older than the ALF. CCOP magmatism occurred in a deep marine setting in the Albian, followed by uplift and subaerial exposure at the end of plateau magmatism, and then an abrupt shift to arc magmatism by 89–86 Ma. Unlike Aruba, however, Curaçao was not affected by any regional deformation event during the transition in magma genesis. Instead, it experienced subsidence and accumulation of a thick sequence of submarine strata (Knip Group and Midden-Curaçao Formation) in the latest Cretaceous to Early Paleocene. These Late Cretaceous strata on Curaçao provide an important record of processes affecting the southern Caribbean during a time when there is no information preserved on Aruba. Three significant conclusions can be drawn from the available data. (1) Arc magmatism in proximity to Curaçao began no earlier than ca. 91 Ma, peaked at 74 Ma, and was over by ca. 66 Ma. (2) Curaçao made a paleogeographic transition during the Late Cretaceous from a setting far removed from any continental sediment influx in the Albian (during CCOP magmatism) to a setting that received sediment influx from the South American margin by the latest Cretaceous (Lagoen Formation, Knip Group). (3) During the latest Cretaceous to Early Paleocene, the tectonic setting of Curaçao allowed sediment influx from both an arc and the South American continent.

Bedrock geology on Bonaire consists of Late Cretaceous volcaniclastic, sedimentary, and intrusive rocks that have been traditionally grouped as part of a coherent stratigraphic succession named the “Washikemba Formation” (Pijpers, 1933; Klaver, 1976; Beets et al., 1977, 1984; Thompson, 2002; Thompson et al., 2004). These rocks, which are overlain unconformably by Maastrichtian and younger sedimentary strata, outcrop in two separate areas of the island but are much better exposed and more extensively studied in the northwestern area (Fig. 13). The “Washikemba Formation,” as defined in prior studies, consists of volcanic rocks with magmatic arc geochemistry and is thought to have accumulated over a time span from at least the Albian to the Coniacian (Klaver, 1976; Beets et al., 1984; Thompson et al., 2004). These rocks are widely viewed as representative of the southern end of a long-lived Cretaceous “Great Arc” of the Caribbean, and have played a pivotal role in tectonic models of the southern Caribbean region (Beets et al., 1984; Thompson et al., 2004).

We remapped much of the northwest part of the island and collected samples for geochronologic analyses in order to more quantitatively evaluate existing models and for comparison with data from the nearby islands of Curaçao and Aruba (Fig. 1). Our results indicate that the “Washikemba Formation” of previous studies actually consists of two very different units separated by a fault, and that the Cretaceous geology of Bonaire bears no relation to that of Aruba or Curaçao.

Based on our mapping, we propose that the term “Washikemba Formation” be abandoned and that the rocks formerly assigned to this formation be separated into two herein-named new units: the Washikemba Group and the Matijs Group. The distribution of the two new units, and their stratigraphies, are shown in Figures 13 and 14. The Matijs Group is dominated by pelagic and clastic strata and corresponds with the Salina Matijs assemblage of prior studies (Klaver, 1976; Beets et al., 1977; Thompson, 2002). The Washikemba Group, in contrast, is dominated by the intermediate to felsic composition volcanogenic and hypabyssal rocks most commonly associated with the “Washikemba Formation.” In the following sections, we describe the geology of Bonaire using this new nomenclature.

The WG comprises the Wecua, Slagbaai, and Branderis assemblages of prior workers (Klaver, 1976; Thompson, 2002). It consists of a thick (up to 4 km) succession of felsic, submarine volcanogenic strata intruded to varying degrees by shallow level dikes and sills of intermediate to felsic composition. These rocks have long been identified as an island arc assemblage on the basis of lithology and facies as well as major, trace-element, and isotope geochemistry (Klaver, 1976; Beets et al., 1984; Thompson, 2002; Thompson et al., 2004).

Volcanogenic strata throughout the WG are of dacitic to rhyodacitic composition and consist mostly of interbedded tuff breccia, lapilli tuff, and finer-grained ash tuffs (Figs. 15A and 15B), with local horizons of radiolaria-rich pelagic strata in the lower part of the group (Klaver, 1976; Thompson, 2002). Constituent fragments in the tuffs include various proportions of pumice, felsic volcanic lithics, crystal clasts of quartz and feldspar, and glass shards. Sedimentary features indicative of submarine deposition from turbidity currents and debris flows are well-defined, as noted by prior workers (Klaver, 1976; Thompson, 2002). These include normal grading, parallel lamination, and crossbedding in the finer-grained strata, and a combination of massive bedding to normal and reverse grading in the coarser-grained strata. Abundance of pumice, poor sorting, and angularity of clasts suggests reworking of primary pyroclastic deposits from a nearby arc volcano (see also Klaver, 1976; Thompson, 2002).

Volcaniclastic strata of the WG are intruded by two different suites of hypabyssal rocks (Figs. 13 and 14). Diorite dikes, sills, and small stocks (Figs. 15C and 15D) are prevalent in the lower part of the WG where they crosscut and/or surround selvages of volcaniclastic strata (mostly too small to show at the scale of Figure 13). Folds of bedding are locally found adjacent to the intrusive contacts, suggesting that intrusion took place before the volcaniclastic strata were fully lithified (see also Klaver, 1976; Thompson, 2002). Elsewhere, particularly in the upper 1500 m of the WG, the volcaniclastic strata are intruded by rhyodacite sills (Fig. 15E) that typically contain phenocrysts of plagioclase ± quartz.

In addition to the rocks described above, we also identified a distinctive conglomerate unit at the top of the WG in the northwestern part of the map area (Figs. 13 and 14), which was not specifically recognized in prior studies. This unit consists mostly of very poorly sorted conglomerate containing subrounded volcanic clasts up to boulder size (5 m diameter; Fig. 15F) that appear to be derived entirely from underlying rocks of the WG. The most common clast type is a flow-banded rhyodacite (shown in Fig. 15F) with phenocrysts of plagioclase ± quartz that is similar in composition to the rhyodacite sills of the upper WG. These conglomerates occur in thick (mostly >5 m) beds that are interlayered with much less common silty argillite and sandstone. Composition, sedimentary features, and map pattern suggest that the conglomerate unit may be a channel deposit cut into and sourced from the underlying WG.

Previous age constraints from strata that we now include in the WG are limited. Fossils from a pelagic interval in tuffs of the lower WG are mid- to late Albian (Beets et al., 1977), or ca. 108–100 Ma. Thompson et al. (2004) attempted ⁴⁰Ar/³⁹Ar analyses on volcanic rocks and their best data suggest an age of ca. 96 ± 4 Ma. We collected two samples for U-Pb geochronology from the WG (see Fig. 13 for location; and Supplemental File 1 [see ^{footnote 1}] for data and methods). One sample (BN-45) was collected from a felsic lapilli tuff in the upper part of the WG; the other sample (BN-26) was from the large block of rhyodacite in the conglomerate unit at the top of the WG (sampled block shown in Fig. 15F). Both samples gave Cenomanian ages: 98.2 ± 0.6 Ma for the tuff and 94.6 ± 1.4 Ma for the rhyodacite block (Fig. 16; Supplemental File 1 [see ^{footnote 1}]). Collectively, the available data indicate that the WG accumulated over a period from the mid- or late Albian to the Cenomanian.

The Matijs Group, as defined here, can be divided into three units, named for the dominant rock type in each (Fig. 13): an argillite unit that is locally crosscut by diabase stocks; a conglomerate unit that appears to be in depositional contact above the argillite unit; and a chert unit whose stratigraphic relations to the other units is ambiguous.

The Matijs argillite occurs at the base of the group (Fig. 14). It consists of pelagic and hemipelagic strata, including argillite, siliceous argillite, argillaceous chert, and cherty limestone, with less common fine-grained feldspathic sandstone and siltstone (Fig. 17A). Strata are generally laminated and thin bedded. Our mapping indicates that the diabase stocks, which have narrow contact aureoles and island arc trace-element characteristics (Supplemental File 2 [see ^{footnote 2}]; Fig. 18), are found exclusively in the argillite unit (Fig. 13). Preliminary in situ secondary ion mass spectrometry (SIMS) dating of seven individual microbaddeleyite grains from one of these intrusions has yielded a U-Pb age of 111.6 ± 5.1 Ma (Humphrey, 2010). The argillite unit is thus Albian or older based upon the microbaddeleyite geochronology.

The overlying conglomerate unit is characterized by the presence of distinctive polymictic conglomerate and breccia beds (referred to by prior workers as “boulder beds”; Klaver, 1976; Thompson, 2002). Coarser clastics are commonly very poorly sorted and locally contain boulders up to 2.5 m in length. Most of the larger clasts, and some finer-grained ones, consist of rock types identical to those in the underlying argillite unit, from which they were evidently derived (Figs. 17B and 17C). Some limestone boulders contain a shallow marine fauna including corals and algae (Klaver, 1976). The remainder of clasts, both large and small, consists of igneous detritus ranging from diabase and diorite to amygdular basalt and plagioclase-phyric volcanics. The upper part of the conglomerate unit also includes interlayered sandstone, siltstone, and argillaceous to cherty limestone beds, and is locally depositional on diabase and pillow basalt. Finer-grained strata in this part of the unit have figured prominently in prior studies because they contain a rich marine fauna including radiolaria, foramanifera, and inoceramids indicating a Coniacian age (Beets et al., 1977). The apparent disparity in age between the argillite unit (Albian or older) and overlying conglomeratic strata (Coniacian) suggest that the contact may be an erosional unconformity.

The chert unit consists of pelagic strata including dark gray to tan, laminated to thin-bedded chert, radiolarion chert, argillaceous chert, and minor shale (Fig. 17D). This unit is exposed only in Salina Matijs and has unclear contact relations with other units. It is tightly folded on the outcrop scale, unlike other members of the Matijs Group, and may actually be a large block in the conglomerate unit.

Previous workers have considered the rocks we include in the Matijs Group to be in stratigraphic continuity with rocks we include in the WG (Klaver, 1976; Beets et al., 1977; Thompson, 2002). This conclusion was based in part on the generally NE dip of bedding in all units (Fig. 13) coupled with the presence of Albian strata in the southwest and Coniacian strata in the northeast. Our new mapping and geochronology, however, indicates that this contact is not stratigraphic: (1) there are no obvious shared rock types between the two groups, nor is there any evidence of interlayering across the contact; (2) felsic intrusions of the upper WG are absent in the Matijs Group and mafic intrusions in the Matijs Group are not found in the WG; (3) conglomerates and breccias in the Matijs Group contain clasts indicative of erosion from the lower part of the group, but are lacking in clasts consistent with derivation from the underlying WG; (4) the thickness of the argillite unit in the lower Matijs Group varies substantially from northwest to southeast, while the conglomerate unit in the underlying Washikemba Group is truncated along the contact (Fig. 13); (5) the average orientation of bedding in the Matijs Group is actually distinctly different from that of the WG, as shown by stereonet plots of bedding (Fig. 14); and (6) finally, our new geochronologic results indicate that the lower part of the Matijs Group is actually older (mid-Albian or older) than the apparently underlying rocks of the upper WG (Cenomanian). Collectively, these relations suggest that the contact between the Washikemba and Matijs Groups is a fault that likely records significant offset. We name this the Bartol fault, for the name of the bay at its northwestern end (Fig. 13). More mapping is needed to determine what the sense of offset is along this fault.

The Matijs Group is unconformably overlain in two localities in northwest Bonaire by shallow marine limestone and calcareous sandstone of the mid- to Late Maastrichtian Rincon Formation (Fig. 13; Beets et al., 1977). Farther southeast on the island, rocks that have been mapped as “Washikemba Formation” (Thompson, 2002) and that appear to be similar to what we now call the WG, are unconformably overlain by up to 400 m of fluvial conglomerate, sandstone, and shale of the Soebi Blanco Formation (Fig. 13, inset map; Beets, 1972; Beets et al., 1977). Clasts in the Soebi Blanco Formation include a combination of volcanic lithics, likely derived from the WG, and lithics derived from an external continental source (presumably northern South America), including felsic gneisses, schists, and quartzites (Beets et al., 1977). Priem et al. (1986) obtained a ca. 1 Ga U-Pb zircon date from one of the gneissic cobbles. The Soebi Blanco Formation is not directly dated, but it is overlain by Eocene limestone and locally contains clasts of the Maastrichtian Rincon Formation, suggesting that it was likely deposited in the Paleocene (Fig. 14; Beets, 1972; Beets et al., 1977). Eocene limestone and other shallow marine strata also unconformably overlie the Washikemba and Matijs Groups in northwest Bonaire (Fig. 13).

Unlike Aruba or Curaçao, the principal manifestation of tectonism on Bonaire is tilting of strata and faulting in the latest Cretaceous to Paleocene. Tilting and faulting must be younger than the Cenomanian and Coniacian ages of the youngest affected rocks in each unit and older than the unconformably overlying Maastrichtian and Paleocene(?) Rincon and Soebi Blanco Formations. Offset along the Bartol fault must be younger than the Coniacian age of the youngest rocks of the Matijs Group and older than the Eocene strata that overlap it (Fig. 13).

The Cretaceous geology of Bonaire is completely unlike that of Aruba or Curaçao, and there is no evidence to suggest that it originated in proximity to the other islands. Instead, its record of intermediate to felsic, island arc magmatism spanning the Albian to the Cenomanian (WG) is more compatible with the history of arc magmatism in the Greater Antilles from which we speculate that it was tectonically derived. The paleogeographic setting of Bonaire changed significantly in the latest Cretaceous to Paleocene(?) as the faulted and tilted deep marine bedrock units were unconformably overlain by shallow marine strata of the Rincon Formation and fluvial strata of the Soebi Blanco Formation. By this time, Bonaire was situated in a position to receive sediment from the South American continental margin, whereas prior to this time it was in an oceanic arc setting removed from the influence of continental sedimentation. Within available age constraints the Soebi Blanco Formation may be temporally equivalent to the Danian Midden-Curaçao Formation on Curaçao (Figs. 7 and 14), which also contains abundant continentally derived detritus. We speculate that the Soebi Blanco Formation may represent a fluvial facies of the Midden-Curaçao Formation, and that Bonaire was situated in proximity to Curaçao by the earliest Paleocene.

The Aves Ridge is a largely submarine feature that forms the eastern boundary of the Caribbean seafloor (Fig. 1). It is widely considered to represent a subsided Cretaceous to Paleogene remnant arc that rifted in the early Paleogene as the extensional Grenada Basin opened and the locus of arc magmatism shifted eastward to the Lesser Antilles (Fig. 1). In models that invoke a long-lived “Great Arc” of the Caribbean (Fig. 2), the Aves Ridge is predicted to be underlain by arc igneous rocks of both Early and Late Cretaceous age, and to share elements of a common Cretaceous evolution with the islands of the Leeward Antilles. Prior geochronology of rocks from the Aves Ridge is very limited. K-Ar ages of 57, 58, 65, 67, 78, and 89 Ma were reported for granitic samples obtained in dredge hauls (Fox et al., 1971): although not explicitly stated, the analyses appear to have been performed on whole-rock samples and the actual data are not published. A U-Pb zircon age of 75.9 ± 0.7 Ma has also been determined on a granitoid dredged from the ridge (Neill et al., 2011).

The island of La Blanquilla is located at the southern end of the Aves Ridge and contains the only accessible subaerial bedrock exposures of the ridge (Fig. 1). We therefore examined the geology of La Blanquilla for comparison with the ABC islands and to collect samples for geochronology and geochemistry. Previous studies indicated that the island is underlain by intrusive rocks that were named the Garanton granodiorite by Maloney (1971) and later the Garanton tronjhemite by Schubert and Moticska (1973). Santamaria and Schubert (1974) noted that the tronjhemitic rocks grade into tonalitic rocks in the northwestern part of the island and reported K/Ar biotite ages of 64–66 Ma from the intrusive rocks.

We revisited La Blanquilla and discovered that it contains two plutons, an older biotite-hornblende granodiorite, and a crosscutting hornblende tonalite, exposed in the northwestern part of the island. Our U-Pb zircon analyses (Supplemental File 1 [see ^{footnote 1}]; Fig. 19) indicate a crystallization age of 75.5 ± 0.9 Ma for the older granodiorite (Fig. 19A) and 58.7 ± 0.5 Ma for the younger tonalite (Fig. 19B). Multielement primitive mantle normalized plots indicate that both plutons were formed in an arc environment (Fig. 20; Supplemental File 2 [see ^{footnote 2}]). The younger pluton also has very depleted HREE, positive Eu anomalies, and extremely high Sr/Y ratios (Fig. 20), which suggests an origin via melting of a garnet-bearing and plagioclase-free metabasalt (eclogite?) source.

Based on our study of La Blanquilla, the southern part of the Aves Ridge consists of a latest Cretaceous to Paleocene arc. This geology is very different from that found on the ABC islands to the west. The only point of similarity is the ca. 76 Ma age of the older pluton which is close to the ca. 74 Ma peak in detrital zircon ages and the ca. 74 Ma detrital hornblende ages from the Knip Group and Midden-Curaçao Formation on Curaçao.

Prevailing models for the Cretaceous evolution of the southern Caribbean Plate have emphasized the presence of both CCOP rocks and Early to Late Cretaceous volcanic arc rocks on the islands of the Leeward Antilles, and infer that the islands record interactions between the CCOP province and the “Great Arc” of the Caribbean (Fig. 1). Our study shows that there is considerable diversity in the Cretaceous history of the islands, which is not adequately explained by existing tectonic models. In this section we summarize the key similarities and differences between the islands and the limitations they pose for geodynamic models of the southern Caribbean region.

CCOP rocks are found on the islands of Aruba and Curaçao (Aruba and CLFs). Evidence on both islands for subaerial exposure and weathering at the end stages of magmatism provides a new tie linking the formations and suggests the possibility of a regional episode of uplift (thermal?) in the evolution of the CCOP province. There is no comparable early Late Cretaceous history recorded on Bonaire, neither CCOP magmatism nor subaerial exposure (Fig. 21).

Arc magmatism affected all three islands, but at different times. Bonaire was active as an arc from the mid-late Albian (or older) to the late Cenomanian (ca. 95 Ma). As previously discussed, we found no unequivocal evidence for younger volcanic activity on that island. In contrast, the first sign of possible arc magmatic activity on the other two islands is at 89 Ma (Aruba batholith) and 86 Ma (dikes on Curaçao). Detrital zircon ages, along with ⁴⁰Ar/³⁹Ar ages from Maastrichtian to Danian sandstones on Curaçao, provide proxy data for the age of volcanic source rocks or active volcanism in proximity to the island. These ages indicate a peak at 74 Ma, with only two grains with Cretaceous ages older than 91 Ma. From these data, we conclude that arc magmatism on and/or near Aruba and Curaçao spanned the Turonian to Maastrichtian, which entirely postdates arc magmatism on Bonaire.

There are also considerable differences in the depositional setting over time of the three islands, and in their deformational histories. Following subaerial exposure at the end of CCOP magmatism, Aruba was buried to depths sufficient for regional deformation with foliation development in all units of the ALF, metamorphism, and batholith emplacement. Available age data indicate that this change in environment may have taken place over a very short interval of time (<4 m.y.) in the Turonian. Rapid structural burial appears to be the only possible explanation. This record is in contrast with that on Curaçao, which was unaffected by any Late Cretaceous deformation and was instead buried by deep marine strata in the Campanian to Maastrichtian following a period of exposure and/or shallow marine conditions (Fig. 21). Bonaire presents yet another history; continuing marine sedimentation through at least the Coniacian, followed by uplift and tilting, and resumption of marine deposition only in the mid to Late Maastrichtian (Fig. 21).

In summary, Aruba and neighboring Curaçao share a similar history of CCOP and arc magmatism; they diverge in terms of their Late Cretaceous depositional setting and deformation history, but this can be explained by their different locations within an evolving convergent boundary as discussed in the next section. Bonaire, in contrast, appears to share no common Cretaceous history with the other islands, suggesting that it evolved elsewhere during this time. It is, in fact, much more similar in its history to islands of the Greater Antilles where Early Cretaceous arc volcanism is common and CCOP magmatic rocks are scarce or absent (e.g., Pindell et al., 2005, and references therein).

It was not until the latest Cretaceous to Paleogene that the three islands began to merge into a common tectonic history. During this time, there was exhumation of batholithic rocks on Aruba, strata derived in part from South America were deposited on Bonaire and Curaçao, and all islands were unconformably overlain by Eocene limestone (Fig. 21).

Models for the evolution of the Caribbean region include the concept of a “Great Arc” that extended from the Greater Antilles, Aves Ridge, through the Leeward Antilles and into Colombia and Ecuador, which was active in both the Early and Late Cretaceous (Fig. 2; e.g., Burke, 1988). Models differ in their interpretation of where the CCOP formed and on what tectonic plate, and whether collision of the CCOP with a west-facing “Great Arc” caused subduction reversal and entrapment of the plateau behind a new east-facing portion of the “Great Arc” (Fig. 2). These models are necessarily based on limited data because much of the CCOP lies underwater and is inaccessible to detailed study. The islands of the Leeward Antilles are thus crucial to tectonic interpretations because they contain on-land exposures of the CCOP.

Aruba, Curaçao, and Bonaire have been particularly important because of their large size relative to other Leeward islands. Their geology has been combined in tectonic models, but our study indicates that this approach is only warranted for Aruba and Curaçao; Bonaire has a very different history and likely evolved in a different part of the Caribbean during much of the Cretaceous.

Based on our study and analysis, the Cretaceous evolution of Aruba and Curaçao is not compatible with prevailing tectonic models. First, these islands contain no record of arc volcanism prior to the Turonian, and therefore do not represent the southern end of an Early to Late Cretaceous “Great Arc” of the Caribbean. Second, models invoking collision of the plateau with a west-facing arc, followed by subduction flip and establishment of a new east-facing arc might appear consistent with the record of Turonian deformation and structural burial on Aruba, but they are difficult to reconcile with the nearly coeval ages of CCOP and arc magmatism on Aruba, or with the lack of any Turonian/Coniacian deformation on Curaçao.

Finally, there is no evidence on Aruba, Curaçao, Bonaire, or La Blanquilla for construction of a Late Cretaceous arc on an amalgamated basement of CCOP and Early Cretaceous arc rocks. In order to fully develop a geodynamic model for the southern Caribbean we first review Late Cretaceous tectonics of Ecuador and Colombia.

The model we develop for the geodynamic evolution of the southern Caribbean is significantly influenced by recent investigations in Ecuador and Colombia. In Ecuador the ca. 85–72 Ma oceanic arc rocks of the Rio Cala Group overlie the equivalent of the CCOP which has been dated at ca. 88 Ma (Luzieux et al., 2006; Vallejo et al., 2006, 2009). In addition, the plateau rocks are intruded by arc related plutonic rocks locally dated at ca. 85 Ma in Ecuador (Vallejo et al., 2009). In Colombia, a ca. 91 Ma arc granitoid intrudes rocks correlated with the CCOP (Villagomez et al., 2008). It appears from these data subduction initiation along the CCOP in Ecuador and Colombia was temporally equivalent to the same event described here on Aruba and Curaçao (Fig. 22A). Thus Aruba and Curaçao appear to be a northern continuation of the Ecuadorian and Colombian arc terrane that was constructed on the CCOP. In the Campanian/Maastrichtian to early Paleogene the arc collided with the Ecuadorian and Colombia margins (Vallejo et al., 2009; Cardona et al., 2008). Following collision, subduction polarity reversed and the collided arc and its CCOP basement were partially subducted beneath the South American margin, resulting in the establishment of a postcollisional latest Cretaceous/Paleogene magmatic arc on the South American continental margin (Cardona et al., 2009). The Aruba/Curaçao section of the arc escaped collision as this arc segment was located to the north and east of the Ecuadorian and Colombian margins in the proto-Caribbean seaway.

We begin in the Late Albian (Fig. 22A) with a northwest-trending Greater Antilles arc terminated along its eastern boundary by a southwest-trending Subduction Termination Edge Propagator (STEP; Govers and Wortel, 2005) fault that is also linked at a triple junction to spreading in the proto-Caribbean. The STEP lengthens to the north as rollback occurs along the Greater Antilles subduction zone. We interpret Bonaire as a part of the Greater Antilles Island Arc. We also hypothesize that the Late Cretaceous Peruvian trench transitioned from a subduction boundary to a transform boundary to the northwest due to oblique convergence. The CCOP has already partly formed by this time based upon the Albian ammonites within the CLF as well as a 112.7 ± 7.3 Ma U-Pb microbaddeleyite date from the CLF (Humphrey, 2010).

By ca. 89 Ma subduction was initiated along the southeastern margin of the CCOP (Fig. 22B). This event is the initiation of the Ecuadorian-Colombian-Leeward Antilles arc (ECLA) built largely upon a basement of the CCOP. The ALF of Aruba was located on the southeastern edge of the plateau in the vicinity of the STEP and was partially subducted, which accounts for Cretaceous ductile deformation of the ALF. Eventually, the ALF was transferred to the hanging wall of the subduction zone where the Aruba batholith was emplaced. On the other hand, the CLF of Curaçao was located away from the newly initiated subduction zone (Fig. 22B) and did not undergo deformation associated with subduction initiation. The Late Cretaceous arc intrusions on Aruba and Curaçao represent the northernmost exposures of the ECLA arc, while the region between Aruba and Curaçao and the Greater Antilles arc was maintained as a transform boundary at this time. The section of the STEP between the Greater Antilles arc and the ECLA arc represents the ancestral Aves subduction zone prior to subduction initiation in the Campanian. We view the STEP/Greater Antilles arc intersection as a complicated zone from which fragments of the Greater Antilles like Bonaire (Fig. 22B) were transferred to the proto-Caribbean Plate. Thus Bonaire was originally located near the subduction/transform boundary at the eastern end of the Greater Antilles arc and subsequently transferred to the proto-Caribbean Plate as the Greater Antilles migrated past the northwestern South American margin (Fig. 22B). The presence of coarse fluvial continental detritus in the Soebi Blanco Formation indicates that Bonaire became attached to continental South America by the latest Cretaceous/early Paleogene (Fig. 22C).

Oblique collision in the Campanian (ca. 75 Ma), as previously discussed, led to the demise of arc magmatism along the ECLA arc, and the arc and its basement (CCOP) were translated north along the oblique collision zone (Fig. 22C). Aruba and Curaçao were located too far north along the ECLA arc to have been involved in the collision (Fig. 22C). For example, Curacao was a deep marine basin during the collision where turbidites derived from a volcanic arc and continental margin source were accumulating during the Campanian/Maastrichtian and Early Paleocene (see Figs. 7 and 10), as previously summarized. The occurrence of abundant detrital grains in the 1.0–1.2 Ga interval and a significant number of Triassic and Late Paleozoic grains all indicate an Andean source for these detrital components as rocks of these ages are present in the Andes of Colombia (e.g., Kroonemberg, 1982; Restrepo-Pace et al., 1997; Cordani et al., 2005; Vinasco et al., 2006; Molina et al., 2006). Thus, we place Curaçao in a position (Fig. 22C) to receive sediment derived from the collision zone between the collided part of the ECLA arc and the continental margin. Following collision of the Ecuadorian and Colombian part of the ECLA arc, oblique convergence along the collision boundary apparently resulted in subduction of the CCOP beneath the South American margin which led to the formation of a postcollisional arc (PCA, Fig. 22D; e.g., Cardona et al., 2009).

Finally, we hypothesize that following collision, subduction propagated to the north along the STEP producing the Aves Arc (Fig. 22D). In the absence of additional data from the Aves Ridge, we hypothesize that the two dated plutons on La Blanquilla may approximate the initiation of and termination of arc magmatism during the Campanian and early Paleogene, respectively. The 59 Ma pluton on La Blanquilla was emplaced shortly before the opening of the Grenada Basin when the Aves arc became an inactive remnant arc.

We suggest that the Cretaceous Great Arc of the Caribbean (e.g., Burke, 1988) may instead have evolved as three arcs now represented by the Greater Antilles, the Aves Ridge, and the ECLA arc. This geodynamic model is offered as a testable framework in which to view the Late Cretaceous tectonic evolution of the southern Caribbean region. Finally, progress in interpreting the tectonic significance of other problematic southern Caribbean terranes such as Tobago, the Villa de Cura Complex and Las Hermanas volcanics of mainland Venezuela, and additional bedrock exposures of the Leeward Antilles (Gran Roque and La Orchila; Fig. 1) might result when viewed within the context of the proposed model.

This work was supported by NSF grants EAR 0087361 and EAR 067533. This is a contribution to the Continental Dynamics BOLIVAR project. We particularly acknowledge conversations with Jim Pindell who has patiently educated us on Caribbean geology for many years. We also thank Dirk Beets (now deceased) and Gerard Klaver who generously gave of their knowledge of the ABC islands and supplied us with theses from the University of Utrect. We are also grateful to Pat Thompson who supplied us with pertinent unpublished portions of her Ph.D. dissertation on the island of Bonaire and to Rosalind White who provided us with a copy of the geologic map of Aruba by Beets et al. (1996). Transportation, via ship, to the Venezuelan military base on La Blanquilla was arranged by the Fundacion Venezolana de Investigaciones Sismologicas (FUNVISIS) and provided by the Armada de Venezuela. Luis Melo (FUNVISIS) acted as liaison between us and the military personnel onboard the ship and on La Blanquilla. Gustavo Rodriguez, the Comandante of the La Blanquilla military base, provided lodging, island transportation, and hospitality during our stay. The meeting hosted by Andrew Kerr, Jim Pindell, Iain Neill, and Alan Hastie at Cardiff University, Wales in September of 2009 contributed significantly to our knowledge of Caribbean tectonics and in particular made us aware of research projects in Ecuador and Colombia that are pertinent to the tectonic model proposed in this paper. Discussions with Roelant van der Lelij, Diego Villagomez, Iain Neill, Agustin Cardona, and Andrew Kerr have significantly influenced our ideas on Caribbean tectonics. Reviews by Art Snoke, Kevin Burke, Associate Editor Bob Hatcher, and particularly Jim Pindell, significantly improved the manuscript.