Global tomography models are covered elsewhere in this volume, but some features of these models are of special interest to the upper mantle and transition zone. A recent concern has been to improve the depth resolution of the models in the uppermost mantle. Surface wave inversions [e.g. Tanimoto, 1991] have poor depth resolution below about 200 km. The use of SS - S differential times [ Woodward and Masters, 1991] can help separate heterogeneity above the 660-km discontinuity from that below, but provides only very limited depth resolution within the upper mantle. By combining body- and surface-wave observations (or free oscillation data), better depth resolution can be obtained in the models [e.g. Woodward and Masters, 1992; Su et al. , 1994].
The relationship between upper mantle velocity anomalies and surface tectonic features has been a source of controversy. Zhang and Tanimoto [1992, 1993] and Anderson et al. [1992] show that their surface-wave inversion model contains slow anomalies under ridges only to depths of about 100 km, whereas anomalies associated with hotspots extend somewhat deeper. Su et al. [1992] dispute the validity of this model by showing that it does not fit SS travel-time data; in their preferred models the ridge anomalies reach depths of 300 km or more [e.g. Su et al. , 1994]. The waveform modeling of Grand [1994] indicates that the anomaly beneath the East Pacific Rise extends to at least 200 km and possibly deeper.
Regional tomography studies in the western Pacific based on travel times available from the International Seismological Centre (ISC) have shown fast velocity anomalies that appear to represent horizontal slab extension into the transition zone behind subduction zones [ van der Hilst et al. , 1991, 1993; Fukao et al. , 1992]. The positions of these anomalies are in rough agreement with a depression in the 660-km discontinuity observed in SS precursor data (see below) and a large-scale transition zone anomaly in the global tomography models. Teleseismic ISC times have also been used to image the uppermost mantle beneath the western United States [ Humphreys and Dueker, 1994]; this study shows some association between surface tectonic and volcanic features and trends in the underlying seismic velocities.
Attenuation is intrinsically more difficult to measure than velocity;
nevertheless some progress has been made in resolving three-dimensional Q
variation in the upper mantle. Durek et al. [1993] inverted for an even-degree
expansion of Rayleigh wave attenuation and concluded that anelastic
heterogeneity is localized in the shallow mantle at 100 to 300 km depth. Romanowicz
[1994] inverted Rayleigh-wave amplitude data for an upper mantle Q model that
generally correlates with surface tectonics in the upper 250 km, and contains a
strong spherical harmonic degree-2 component below 300 km. In some
locations, observations suggest that anomalously large attenuation is present in
the uppermost mantle. Ding and Grand [1993], analyzing S and its
surface multiples, found shear-wave attenuation of 
in the
top 
70 km beneath the East Pacific Rise, while Flanagan and
Wiens [1994], measuring differential attenuation in sS versus S phases, found
generally high attenuation in back-arc regions with 
in the
uppermost 160 km of the mantle.