Geoscience Reference
In-Depth Information
continuous plume-like low-velocity upwellings
(Becker and Boschi, 2002; Ishii and Tromp, 2004;
Trampert
et al.
, 2004); this is consistent with
chemical stratification (Wen and Anderson, 1995,
1997).
Color 2D cross-sections are particularly
ambiguous; although certainly vivid -- and
impressive to nonspecialists -- they are not 'over-
whelming, compelling or convincing' evidence
for mantle dynamics or composition; they can
be overinterpreted. They cannot do justice to the
information content of a typical tomographic
study. There are also issues of physical, geody-
namic and petrological interpretations; are 'blue'
regions of the lower mantle, even if real, unam-
biguous indicators of cold, dense material that
started at the Earth's surface?
Geodynamic interpretations of tomographic
models based on quantitative analyses of tomog-
raphy (Scrivner and Anderson, 1992; Wen
and Anderson, 1995; Masters
et al.
, 2000; Gu
et al.
, 2001; Anderson, 2002a; Ishii and Tromp,
2004; Trampert
et al.
, 2004) are quite differ-
ent from the interpretations of QCT and color
images [
Probabilistic Tomography Maps
of Chemical Heterogeneities
]. Very few
slab-like features appear to be dense; hardly any
hotspots appear to be underlain by buoyant mate-
rial in the lower mantle. Long wavelength chem-
ical heterogeneities, however, exist throughout
the mantle (Ishii and Tromp, 1999, 2004).
Tomographic images are often inter-
preted in terms of an assumed velocity--density--
temperature correlation, e.g. high shear-velocity
(blue) is attributed to cold dense slabs, and low
shear velocity (red) is interpreted as hot rising
low-density blobs. There are many factors con-
trolling seismic velocity and some do not involve
temperature or density. Cold, dense regions of
the mantle can have low shear velocities (e.g.
Presnall and Gudfinnsson, 2004; Trampert
et al.
,
2004). Changes in composition or crystal struc-
ture can lower the shear-velocity and increase
the bulk modulus and/or density, as can be
verified by checking any extensive tabulation of
elastic properties and densities of minerals. Ishii
and Tromp (2004), for example, found negative
correlations between velocity and density in the
upper mantle.
For reviews of the situation regarding seis-
mic modeling -- including uncertainties and lim-
itations -- see Dziewonski (2005); Boschi and
Dziewonski (1999); Vasco
et al.
(1994) and Ritsema
et al.
(1999). The bottom line is that the mantle
is not similar to the one- and two-layer 1D struc-
tures that underlie the standard models of man-
tle geodynamics and geochemistry and tempera-
ture is not the sole parameter controlling seismic
velocities.
Spectrum of heterogeneity
Understanding how the power is dis-
tributed in the wavenumber domain
and
the distribution of the spectral power as a func-
tion of depth in the mantle is an extremely
valuable diagnostic in the interpretation of
tomographic models and assessing the appli-
cability of various mantle convection mod-
els [
www.mantleplumes.org/Convection.
html
]. Tanimoto (1991) was the first to point out
the significance of the predominance of the large-
scale heterogeneity in the Earth. Su and Dziewon-
ski (1991, 1992) showed that tomographic power
is approximately constant up to degree 6, but
then decreases rapidly. Mantle convection simu-
lations must satisfy this constraint.
Approaches
There are several approaches for interpreting
global seismic data. The regionalized approach
divides the Earth into tectonic provinces and
solves for the velocity of each. Application of
this technique shows that shields are the highest
velocity regions at shallow depths but conver-
gence regions are faster at greater depth, suggest-
ing that cold subducted material is being sam-
pled (Nakanishi and Anderson, 1983, 1984a,b).
Convergence regions, on average, are slow at
short periods, due to high temperatures and
melting at shallow mantle depths. The region-
alization approach is necessary when the data
are limited or when only complete great-circle or
long-path data are available. In the latter case the
velocity anomalies cannot be well isolated.
In the regionalized models it is assumed that
all regions of a given tectonic classification have
the same velocity. This is clearly oversimplified. It
is useful, however, to have such reference maps
