Xenolith constraints on seismic velocities in the upper mantle beneath southern Africa

[1] We impose geologic constraints on seismic three-dimensional (3-D) images of the upper mantle beneath southern Africa by calculating seismic velocities and rock densities from approximately 120 geothermobarometrically calibrated mantle xenoliths from the Archean Kaapvaal craton and adjacent Proterozoic mobile belts. Velocity and density estimates are based on the elastic and thermal moduli of constituent minerals under equilibrium P-T conditions at the mantle source. The largest sources of error in the velocity estimates derive from inaccurate thermo-barometry and, to a lesser extent, from uncertainties in the elastic constants of the constituent minerals. Results are consistent with tomographic evidence that cratonic mantle is higher in velocity by 0.5-1.5% and lower in density by about 1% relative to off-craton Proterozoic samples at comparable depths. Seismic velocity variations between cratonic and noncratonic xenoliths are controlled dominantly by differences in calculated temperatures, with compositional effects secondary. Different temperature profiles between cratonic and noncratonic regions have a relatively minor influence on density, where composition remains the dominant control. Low-T cratonic xenoliths exhibit a positive velocity-depth curve, rising from about 8.13 km/s at uppermost mantle depths to about 8.25 km/s at 180-km depth. S velocities decrease slightly over the same depth interval, from about 4.7 km/s in the uppermost mantle to 4.65 km/s at 180-km depth. P and S velocities for high-T lherzolites are highly scattered, ranging from highs close to those of the low-T xenoliths to lows of 8.05 km/s and 4.5 km/s at depths in excess of 200 km. These low velocities, while not asthenospheric, are inconsistent with seismic tomographic images that indicate high velocity root material extending to depths of at least 250 km. One plausible explanation is that high temperatures determined for the high-T xenoliths are a nonequilibrium consequence of relatively recent thermal perturbation and compositional modification associated with emplacement of kimberlitic fluids into the deep tectospheric root. Seismic velocities and densities for cratonic xenoliths differ significantly from those predicted for both primitive mantle peridotite and mantle eclogite. A model primitive mantle under cratonic P-T conditions exhibits velocities about 1% lower for P and about 1.5% lower for S, a consequence of a more fertile composition and different modal composition. Primitive mantle is also about 2% more dense at 150-km depth than low-T garnet lherzolite at cratonic P-T conditions. Similar calculations based on an oceanic geotherm are consistent with the isopycnic hypothesis of comparable density columns beneath oceanic and cratonic regions. Calculations for a hypothetical cratonic'' eclogite ( 50: 50 garnet/omphacite) with an assumed cratonic geotherm produce extremely high V-P and V-S (8.68 km/s and 4.84 km/s, respectively, at 150 km depth) as well as high density (similar to3.54 gm/ cc). The very high velocity of eclogite should render it seismically conspicuous in the cratonic mantle if present as large volume blocks or slabs. We discuss how the seismic velocity data we have compiled in this paper from both xenoliths and generic petrologic models of the upper mantle differ from commonly used standard earth models IASPEI and PREM.