Advances in understanding the thermal structure of the upper mantle have
been led by seismological studies that, for the first time, have produced
detailed maps of the global and regional variation of the depths to the
660- and 410-km seismic discontinuities.
In conjunction with the Clapeyron slopes of the relevant phase transitions
(see Bina and Helffrich [1994] for a recent analysis of the
experimental phase equilibria data), the temperature differences required
to produce such topography can be deduced. Revenaugh and Jordan
[1991] used shear wave reverberation data to measure topography variations
on the 410- and 660-km discontinuities and inferred that the lateral
temperature variations implied by this topography are 
200 K.
Short-period array data were also used to determine that the 660-km
discontinuity is depressed by 20-30 km and the 410-km discontinuity is
elevated by 
15 km beneath subduction zones, which together imply
300-400 K average temperature difference between subduction zones and
normal mantle [ Vidale and Benz, 1992].
In a high-resolution study
of the Izu-Bonin subduction zone, the 660-km discontinuity was found to
be depressed by 60 km, suggesting a thermal anomaly of 1000 K [ Wicks
and Richards, 1993]. A global study utilizing underside shear wave
reflections from the 660-km discontinuity found regional variations of
30 km with depressions correlated with subduction zones
[ Shearer and Masters, 1992]. The possibility that temperatures
in parts of the upper mantle may be sufficiently high to produce partial
melts at depths greater than 300 km has received support from a shear
wave reverberation study sampling the mantle beneath the Sea of Japan
[ Revenaugh and Sipkin, 1994]. As with the lower mantle, it now
appears that thermal anomalies on the order of several hundred K exist
in the upper mantle at the length scales (100-1000 km) sampled by these
seismic techniques.
Tomographic studies of upper mantle heterogeneity generally show good
correlation with surface tectonic features and can be used to infer the
depth extent of the thermal anomalies that are associated with mid-ocean
ridges and hotspots. In a high-resolution global surface wave study,
Zhang and Tanimoto [1993] found low velocities under hot spots
at 100-200 km depth while ridges showed very slow anomalies only in
the upper 100 km, with the low-velocity regions shifting away from
the ridge at greater depth.
In contrast, Su et al. [1992] found that very slow anomalies
under mid-ocean ridges extend continuously to at least 300 km, and
hotspots are not underlain by low-velocity anomalies.
Fast anomalies under continental shields were found to extend to 300-400
km depth [ Su et al., 1992, 1994]. Correlations between S-wave
velocity, bathymetry, and basalt chemistry were found beneath the
Mid-Atlantic Ridge at depths of 100-200 km, and the temperature
variations at these depths were estimated to be 100-300 K
[ Zhang et al., 1994].