ratios than group 1. Such rocks are widespread in the
Caribbean Plateau (e.g., Kerr et al., 2003).
Explaining compositional differences between
the groups
High-MgO rocks have been found in accreted sections of
the Caribbean Plateau but not (thus far) in the Ontong Java
Plateau. Furthermore, the incompatible-element and isotopic compositions of high-MgO Caribbean Plateau lavas
are more heterogeneous than those of lower-MgO basalts
from both the Caribbean and Ontong Java plateaus.
It has been proposed that oceanic plateau source
regions are heterogeneous and that the homogeneity of
many plateau lavas and intrusions is a result of magma
mixing, either in the melting column or in magma chambers (Kerr et al., 1997). The source region of the Caribbean Plateau was clearly markedly heterogeneous over
much of its area and on a relatively small scale (Kerr
et al., 2003). Such widespread and small-scale heterogeneity is also consistent with evidence from present-day lavas
found at mid-ocean ridges and oceanic islands (e.g., some
oceanic islands display greater isotopic and elemental heterogeneity than observed in the entire Ontong Java
Plateau).
Why compositionally heterogeneous, high-MgO lavas
have not been found on the Ontong Java Plateau but are
relatively common in the Caribbean Plateau may well be
related to differences in the thickness of the plateaus and
in the depth to which the crust has been exhumed during
accretion (Kerr and Mahoney, 2007). High-MgO, heterogeneous lavas are most likely to be erupted where there
are few magma chambers and the magmatic plumbing system is poorly developed; in such areas, individual magma
batches are more liable to ascend quickly from the source
to the surface with little mixing in magma chambers.
Therefore, in large igneous provinces, high-MgO lavas
are most commonly found at the base of the succession.
Later magmas are more likely to become trapped in
magma chambers and so mix and fractionate (e.g., Kerr
et al., 2003).
Accreted crustal sections provide the best way of studying the internal structure of oceanic plateaus, because they
can expose deeper levels of the crust than those reachable
by drilling,. However, due to the thicker lithosphere and
the length of time between formation and accretion (Kerr
and Mahoney, 2007), only the topmost 3–4 km of the
Ontong Java Plateau basement crop out in large accreted
thrust slices in the Solomon Islands (Petterson et al.,
1997). Therefore, the deeper plateau sections that,
according to the model of Kerr (2003), will be markedly
heterogeneous, and the more Mg-rich lavas are not
exposed on the Ontong Java Plateau.
Tejada et al. (2004) has proposed an alternative model
to explain the homogeneity of the basalts of the Ontong
Java Plateau, which suggests that the mantle source of
the Ontong Java Plateau was essentially homogeneous at
the very large scale and extent of melting. Tejada
et al. (2004) argued that although mixing of more heterogeneous magmas in large open-system magma chambers
could partly explain this geochemical homogeneity,
mixing would always have to result in essentially the same
composition and so reasoned that this was unlikely given
that there is little evidence for magma mixing in the isotope and trace element data.
Given the great volume of melting and the large fractions of partial melting involved, it is probable that the
melting process played a key role in averaging out
smaller-scale source heterogeneity to produce the remarkable isotopic and incompatible-element homogeneity of
basalts from the Ontong Java Plateau. Perhaps significantly, the estimated extents of melting for most of the
Caribbean Plateau are lower than for the Ontong Java, a
feature that would have resulted in less homogenization
of melts in the mantle (Kerr and Mahoney, 2007).
A similar conclusion was also reached by Heydolph
et al. (2014) to explain the geochemical homogeneity of
initial volcanism at the Shatsky Rise LIP.
As noted above, some of the oldest lavas drilled on the
Kerguelen Plateau possess geochemical features consistent with contamination by continentally derived material.
These continental fragments are likely to have been incorporated within the Indian Ocean lithosphere during the
breakup of Gondwana prior to plateau formation. During
formation of the Kerguelen Plateau, partial melts from
these fragments would have contaminated the plateau
magmas. This model has been confirmed by the discovery
at ODP Site 1137 of a conglomerate containing Precambrian garnet–biotite gneiss cobbles (see review in Frey
et al., 2002). Furthermore, seismological evidence also
indicates the presence of continental crust in parts of the
plateau, although the overall amount is not well
established (e.g., Borissova et al., 2003).
Buoyancy and accretion of oceanic plateaus
The main reason for the elevated nature of oceanic plateaus (~2–3 km above the abyssal ocean floor) is the thickness of their basaltic crustal sections, which is generally
much greater than the 6–7 km of normal oceanic crust
generated at spreading centers. For instance, the Ontong
Java Plateau has an average crustal thickness of ~33 km
(e.g., Richardson et al., 2000), whereas the Caribbean Plateau crust varies from 8 to 15 km thick (see review in Kerr
et al., 2003).
This thicker basaltic crust results in oceanic plateaus
being more buoyant than the surrounding ocean floor
and is the key reason why plateaus often resist subduction
and so are often accreted to continental margins, thereby
increasing their likelihood of preservation in the geological record. The extent of the buoyancy depends on both
crustal thickness and plateau age. Plateaus that collide
with a subduction zone only a few million years after they
form will be more likely to resist subduction than a plateau
OCEANIC PLATEAUS
561
Caribbean Plateau (e.g., Kerr et al., 2003).
Explaining compositional differences between
the groups
High-MgO rocks have been found in accreted sections of
the Caribbean Plateau but not (thus far) in the Ontong Java
Plateau. Furthermore, the incompatible-element and isotopic compositions of high-MgO Caribbean Plateau lavas
are more heterogeneous than those of lower-MgO basalts
from both the Caribbean and Ontong Java plateaus.
It has been proposed that oceanic plateau source
regions are heterogeneous and that the homogeneity of
many plateau lavas and intrusions is a result of magma
mixing, either in the melting column or in magma chambers (Kerr et al., 1997). The source region of the Caribbean Plateau was clearly markedly heterogeneous over
much of its area and on a relatively small scale (Kerr
et al., 2003). Such widespread and small-scale heterogeneity is also consistent with evidence from present-day lavas
found at mid-ocean ridges and oceanic islands (e.g., some
oceanic islands display greater isotopic and elemental heterogeneity than observed in the entire Ontong Java
Plateau).
Why compositionally heterogeneous, high-MgO lavas
have not been found on the Ontong Java Plateau but are
relatively common in the Caribbean Plateau may well be
related to differences in the thickness of the plateaus and
in the depth to which the crust has been exhumed during
accretion (Kerr and Mahoney, 2007). High-MgO, heterogeneous lavas are most likely to be erupted where there
are few magma chambers and the magmatic plumbing system is poorly developed; in such areas, individual magma
batches are more liable to ascend quickly from the source
to the surface with little mixing in magma chambers.
Therefore, in large igneous provinces, high-MgO lavas
are most commonly found at the base of the succession.
Later magmas are more likely to become trapped in
magma chambers and so mix and fractionate (e.g., Kerr
et al., 2003).
Accreted crustal sections provide the best way of studying the internal structure of oceanic plateaus, because they
can expose deeper levels of the crust than those reachable
by drilling,. However, due to the thicker lithosphere and
the length of time between formation and accretion (Kerr
and Mahoney, 2007), only the topmost 3–4 km of the
Ontong Java Plateau basement crop out in large accreted
thrust slices in the Solomon Islands (Petterson et al.,
1997). Therefore, the deeper plateau sections that,
according to the model of Kerr (2003), will be markedly
heterogeneous, and the more Mg-rich lavas are not
exposed on the Ontong Java Plateau.
Tejada et al. (2004) has proposed an alternative model
to explain the homogeneity of the basalts of the Ontong
Java Plateau, which suggests that the mantle source of
the Ontong Java Plateau was essentially homogeneous at
the very large scale and extent of melting. Tejada
et al. (2004) argued that although mixing of more heterogeneous magmas in large open-system magma chambers
could partly explain this geochemical homogeneity,
mixing would always have to result in essentially the same
composition and so reasoned that this was unlikely given
that there is little evidence for magma mixing in the isotope and trace element data.
Given the great volume of melting and the large fractions of partial melting involved, it is probable that the
melting process played a key role in averaging out
smaller-scale source heterogeneity to produce the remarkable isotopic and incompatible-element homogeneity of
basalts from the Ontong Java Plateau. Perhaps significantly, the estimated extents of melting for most of the
Caribbean Plateau are lower than for the Ontong Java, a
feature that would have resulted in less homogenization
of melts in the mantle (Kerr and Mahoney, 2007).
A similar conclusion was also reached by Heydolph
et al. (2014) to explain the geochemical homogeneity of
initial volcanism at the Shatsky Rise LIP.
As noted above, some of the oldest lavas drilled on the
Kerguelen Plateau possess geochemical features consistent with contamination by continentally derived material.
These continental fragments are likely to have been incorporated within the Indian Ocean lithosphere during the
breakup of Gondwana prior to plateau formation. During
formation of the Kerguelen Plateau, partial melts from
these fragments would have contaminated the plateau
magmas. This model has been confirmed by the discovery
at ODP Site 1137 of a conglomerate containing Precambrian garnet–biotite gneiss cobbles (see review in Frey
et al., 2002). Furthermore, seismological evidence also
indicates the presence of continental crust in parts of the
plateau, although the overall amount is not well
established (e.g., Borissova et al., 2003).
Buoyancy and accretion of oceanic plateaus
The main reason for the elevated nature of oceanic plateaus (~2–3 km above the abyssal ocean floor) is the thickness of their basaltic crustal sections, which is generally
much greater than the 6–7 km of normal oceanic crust
generated at spreading centers. For instance, the Ontong
Java Plateau has an average crustal thickness of ~33 km
(e.g., Richardson et al., 2000), whereas the Caribbean Plateau crust varies from 8 to 15 km thick (see review in Kerr
et al., 2003).
This thicker basaltic crust results in oceanic plateaus
being more buoyant than the surrounding ocean floor
and is the key reason why plateaus often resist subduction
and so are often accreted to continental margins, thereby
increasing their likelihood of preservation in the geological record. The extent of the buoyancy depends on both
crustal thickness and plateau age. Plateaus that collide
with a subduction zone only a few million years after they
form will be more likely to resist subduction than a plateau
OCEANIC PLATEAUS
561