This is a draft of an article for publication in InterRidge News:
RAMESSES finds a Magma Chamber Beneath a Slow-spreading Ridge
Steven Constable$^1$,
Martin Sinha$^2$, Lucy MacGregor$^2$,
Deborah Navin$^3$, Christine Peirce$^3$,
Antony White$^4$ and Graham Heinson$^4$
$^1$Scripps Institution of Oceanography, La Jolla, CA 92093-0225,
USA
$^2$Bullard Laboratories, University of Cambridge, Cambridge CB3 0EZ, UK
$^3$Department of Geological Sciences, University of Durham, Durham DH13LE,
UK
$^4$School of Earth Sciences, Flinders University, Adelaide, 5001, Australia
In October 1993 we conducted a multi-component geophysical survey of
the Mid-Atlantic Ridge near the southern tip of the Reykjanes Ridge.
The experiment-- RAMESSES (Reykjanes Axial Melt Experiment:
Structural Synthesis from Electromagnetics and Seismics) included
wide-angle seismic refraction profiles, controlled source
electromagnetic soundings, magnetotelluric (MT) sounding, and some
normal incidence seismic reflection profiling, together with the
usual underway swath bathymetry, gravity and magnetic measurements.
This experiment provided a unique and remarkable opportunity to
undertake joint geophysical interpretations of these various data
sets which, taken together, provide the first compelling evidence of
a robust axial magma chamber (AMC) at a slow spreading ridge. The
experiment is described in Sinha et al. (1994) and preliminary
results have been published in Sinha et al. (1997).
Previous attempts to observe an AMC at a slow-spreading ridge have
generally confirmed their absence, suggesting that if they exist at
all they are ephemeral features. The only published evidence
(Calverts, 1995) for a slow-spreading ridge AMC is based on
Mid-Atlantic Ridge seismic data which had previously been used to
deny the existence of an AMC at that location (Detrick et al.,
1990). Given the global extent of slow-spreading ridges, it is
important to find and study whatever AMC's we can. Our understanding
of the systematics and dynamics of ridge systems will never be
complete until we understand the relationships between spreading rate
and the dynamics of crustal accretion.
We chose a study area at 57° 45' N on the southern Reykjanes
Ridge based on the following criteria:
a) At the southern end of the Reykjanes Ridge the broad bulge of
thick crust associated with the Icelandic hot spot gives way to a
median valley more typical of the remainder of the Mid-Atlantic
Ridge, and so although there are geochemical affinities with the hot
spot at this location, it can be argued that morphologically it is
representative of normal slow-spreading ridges. This was later
confirmed by the measurement of normal seismic oceanic crustal
thicknesses here.
b) The overall trend of the ridge axis is oblique to the spreading
direction, and consequently the Reykjanes Ridge is composed of a
series of axial volcanic ridges (AVR) arranged en echelon down the
ridge. These volcanic highs are clear evidence of magmatic activity
at the ridge, and hence an AVR was chosen which appeared to be in the
early stages of its constructional cycle. Using available
bathymetric and side-scan sonar data we chose the AVR
at 57° 45' N because it was still quite small, showed bright
backscatter on side-scan images indicative of fresh lava flows, and
showed no evidence of late-stage tectonic deformation.
c) The rough topography of the Mid-Atlantic Ridge causes significant
scattering of seismic energy incident on the seabed, and so the
slightly smoother seafloor of the Reykjanes Ridge (due to the
influence of the Iceland hot spot)
maximized our resolution at mid- and lower-crustal depths and also
our ability to establish a reliable measure of crustal thickness.
e) The water depths of 1-2 km at the chosen location were sufficient
to isolate the controlled source EM study from atmospheric effects
(both contamination from ionospheric sources and a propagated air
wave), yet sufficiently shallow that high-frequency ionospheric
signals could be detected in the magnetotelluric band.
d) This was a good site logistically, as it is relatively close to
the U.K. home port of the R.R.S. Charles Darwin, the ship used for
the study.
We shall now briefly outline the measurements and then describe the
resulting model shown in Figure 1.
SEISMIC REFRACTION. Eleven ocean-bottom seismometers were deployed
in two lines, along the ridge axis and across the axis (parallel to
spreading). Each line was shot using a series of 25 and 50 kg
explosive charges at 1 and 2~km intervals and re-shot using a large
volume airgun array fired at 100~m intervals. A clear characteristic
of all the across-axis record sections was a shadow zone of greatly
decreased amplitudes associated with the ridge location, and the
along-axis records exhibit a pronounced loss of energy at ranges
greater than 15~km.
The data were quantitatively modelled using Maslov ray-tracing and
synthetic seismogram calculations to produce along-axis and
across-axis models that fit the travel time picks to within 150~ms
for explosive shots and 100~ms for airgun shots,
and showed a qualitatively similar pattern of diminished
amplitudes. The low amplitudes and the associated travel time delays
require the existence of a broad axial low velocity zone capped by a
much thinner and narrower zone of extremely low velocity, presumably
melt. Off-axis, crustal thicknesses (7~km) and velocities are typical
of other slow-spreading ridges, and a Moho is observed in zero-age
crust immediately under the axis and the associated
low-velocity zone using PmP wide-angel reflections.
CONTROLLED SOURCE EM. A total of 14 electrometers were deployed
along the same profiles as the seismic experiment. Two transmitter
tows were completed, one along the ridge axis and one 5 km west of
the ridge, at frequencies of 0.1 to 11 Hz. Short range (1--5 km) data
on the ridge have amplitudes characteristic of low resistivities due
to seawater in a highly porous top 1~km of oceanic crust, and cannot
constrain the existence of an AMC. However, intermediate range data
(5--15 km) across the ridge exhibit a splitting of amplitudes for
the radial and azimuthal modes characteristic of a buried conductor
beneath the axis. Quantitative modelling was carried out using 1D
inversion and 2D forward modelling, using geometries partly
constrained from the seismic experiment. Long range data (20-50 km)
are not considered reliable because the transmitter failed to
maintain coherent phase for long enough to stack up signals at these
distances.
MAGNETOTELLURIC SOUNDING. Five electrometers and six three-component
fluxgate magnetometers were deployed to collect MT data (some of
the instruments were controlled-source electrometers collecting data
for both experiments). In view of the relatively short deployment
time of 20~days, the sample rate of the magnetometers was set to the
maximum possible (0.1~Hz) and this, combined with the shallow water
depth and active ionospheric source fields during the experiment
produced seafloor impedance estimates at frequencies as high 0.04~Hz,
an order of magnitude higher than is usually recorded in the deep
ocean. Skin-depths at the highest frequencies are less than the
crustal thickness, and consequently along-strike measurements were
sensitive to crustal resistivities which agree remarkably well with
1D and 2D models from the controlled source EM experiment. These data
also show a low resistivity zone in the mantle at a depth of 50 to
120 km, presumably associated with decompression melting. However,
the upper 40~km of the mantle is resistive, implying that the AMC
is isolated from the mantle melt source by a hot, but no longer
partially molten, uppermost mantle. This conclusion is supported
by the across-strike measurements, which are sensitive to the coast
of Greenland 650~km distant and the resistivity and thickness of the
lithosphere between the ridge and coast.
SEISMIC REFLECTION PROFILING. A limited amount of normal incidence
seismic reflection data were collected using an 8 channel streamer
and the airgun array, mainly for the purpose of mapping sediment
thicknesses. However, with only a minimum of processing the
along-axis reflection data show a bright reflection event at exactly
the two way travel times predicted for the top of the low velocity
zone in the refraction models, further supporting the existence of a
melt lens at the top of the larger low velocity zone. The reflector
is not visible in the across-axis line, either because it is
discontinuous or because the low stack-fold (4) cannot cope with the
rougher across-axis bathymetry.
Figure 1 shows the combined interpretation
from the various geophysical studies. The pillows and extrusives,
dikes, and gabbros are delineated in the seismic model as the usual
layers 2A, 2B, and 3, showing a slight thickening of layer 2 in a
broad region over the axis accompanied by slightly depressed
velocities within 10 km of the axis, suggesting an increased
porosity. Lowered resistivity, on the other hand, is restricted to
narrower zone within about 5 km of the axis, as the controlled
source EM method is sensing a region of hot seawater of lowered
resistivity directly over the AMC, rather than the increased
porosity alone. Resistivities imply a porosity of 30% in the
uppermost crust, decreasing to less than 1% in the dikes and
gabbros.
There is a broad region on-axis of reduced velocities in the gabbroic
section, presumably associated with increased temperatures and some
melting in a crystal mush zone. The thin melt lens that caps this
zone, required in the velocity model, is too small to be sensed by
the EM method, but the controlled source data help constrain the
lateral extent of the mush and its electrical resistivity at
around 2.5 Ohmm. Although at too low a frequency to delineate
structure, the MT data are sensitive to bulk electrical properties of
the crust, and by fixing layer 2 resistivities and mush zone size
from the EM model the MT data constrain the resistivity of the
crystal mush to be 2 to 5 Ohmm. From this we can infer a melt
fraction between 12 and 30%, or more accurately (since a bigger
mush zone could have a proportionately higher resistivity) a total of
2 to 4 cubic kilometers of melt per kilometer of ridge. For a crustal
thickness of 7 km and full spreading rate of 20 mm/yr this amounts to
15 to 30 thousand years' worth of melt. It is unlikely that the
residency time for an AMC is anywhere near this long, confirming the
ephemeral nature of this feature and suggesting that it was quite
recently created.
We see a normal Moho in the seismic model extending under the ridge.
The low velocities do not extend quite to the Moho, and both the
along- and across-strike MT data suggest that the uppermost mantle
resistivity is at least 500 \Ohmm . Laboratory models of olivine
conductivity versus temperature (Constable et al., 1992) predict
exactly this resistivity for a mantle temperature under the ridge of
1350°C. Because
the AMC has been recently emplaced the temperature must still be
close to but below the mantle solidus, and so there is little
evidence for a significant fraction of connected melt in the upper
40 km of mantle. Melt in this part of the uppermost mantle has
either been completely removed to form the AMC, or it forms
disconnected pockets that are either not migrating or migrating only
slowly. Finally, at depths between about 50 and 100 km we see low
resistivities again in the MT model, suggesting 1-3% melting,
presumably associated with decompression of the mantle.
In conclusion, the combination of geophysical techniques has not only
demonstrated the existence of an axial magma chamber at a
slow-spreading ridge, but placed fairly tight bounds on its size,
depth, and melt content. The relatively shallow depth of the AMC
combined with normal crustal thickness and slow spreading rate
provide strong constraints for geodynamical models of ridge
systems. Finally, the lack of connected melt in the uppermost
mantle provides an important message about the nature of melt
migration. We must think either in terms of efficient, episodic
extraction of melt from a substantial volume of mantle, or migration
of melt in relatively small veins or pods that do not maintain
connectivity after the initial onset of melt migration.
Finally, it is extremely encouraging that all the geophysical
techniques that were used are in dramatically good agreement, and
that the variety of techniques allow far more of the structure to be
well constrained than would be possible with any one method alone.
The success of this enterprise rests not just in the co-location of
instruments, but also in the close collaboration maintained by the
scientific party during the interpretation stages. The outcome
fully justifies the combination of many techniques in a single
integrated experiment on a carefully selected target.
Acknowledgements. The authors thank the captain and crew of the
R.R.S. Charles Darwin and the technicians and engineers of the
various research groups that collaborated on this project, including
the Research Vessel Services group (formerly) of Barry, U.K. This
research was supported with grants from the Natural Environment
Research Council, the National Science Foundation, and the Australian
Research Council.
Calvert, A., 1995: Seismic evidence for a magma chamber beneath the slow
spreading Mid-Atlantic Ridge, Nature, 377 410-414
Constable, S.C., Shankland, T.J. and Duba, A., 1992: The electrical conductivity
of an isotropic olivine mantle, J. Geophys. Res., 97, 3397-3404
Detrick, R.S., J.C. Mutter, P. Buhl, and I.I. Kim, 1990: No evidence for
multi-channel reflection data for a crustal magma chamber in the MARK area
on the Mid-Atlantic Ridge, Nature, 347, 61-64
Sinha, M.C., C. Peirce, S. Constable, and A. White 1994: RRS Charles Darwin
81 Cruise Report,
Sinha, M.C., D.A. Navin, L.M. MacGregor, S. Constable, C. Peirce, A. White,
G. Heinson, and M.A. Inglis, 1997: Evidence for accumulated melt beneath
the
slow-spreading mid-Atlantic ridge Phil. Trans. Roy. Soc. Lond. A, 355, 233-253