Geoscience Reference
In-Depth Information
mantle wedge above subducting slabs and lower
attenuation within the downgoing plate (e.g.
Flanagan &Wiens, 1990; Stachnik et al. , 2004). A
decrease of uppermost mantle attenuation with
increasing age of ocean floor was found, for in-
stance, by Canas & Mitchell (1978) and Sheehan
& Solomon (1992). These observations suggest
that attenuation is primarily temperature-related,
at least within the shallow mantle.
Global models of attenuation in the whole
mantle have also been derived mostly from long-
period surface wave amplitudes and amplitude
ratios of various body wave phases. These models
differ strongly in the treatment of elastic ef-
fects and source/receiver corrections. Romanow-
icz (1990) and Durek et al. (1993), for instance,
considered the amplitudes of four consecutive ar-
rivals of multiply-orbiting long-period Rayleigh
waves (e.g. R1, R2, R3, R4), in order to reduce the
effects of elastic focusing and source uncertainty.
While this approach helps to pronounce the con-
tribution of attenuation, it only yields informa-
tion on even-degree Q μ structure. An improved
data analysis scheme was used by Romanowicz
(1994), who selected both R1 and R2 amplitudes
that did not appear to be contaminated by source
errors and elastic effects. The selection criterion
enforced consistency between the attenuation co-
efficients inferred from four consecutive wave
trains, and those measured on minor and major
arc amplitudes only. This strategy led to maps
of both even- and odd-degree structure, that were
used to construct the first global model of shear
attenuation (Romanowicz, 1995).
Billien et al. (2000) and Selby & Woodhouse
(2000; 2002) explicitly treated elastic focusing
using a linear approximation derived from ray
theory (Woodhouse & Wong, 1986). The con-
tribution from focusing to the amplitudes of
Rayleigh waves was found to be considerable,
even for long periods between 70 s and 170 s.
This motivated Billien et al. (2000) to jointly
invert phase and amplitude measurements of
Rayleigh waves for degree-20 maps of phase ve-
locities and attenuation. However, being con-
cerned that focusing may not be predicted with
sufficient accuracy by current velocity models,
Selby & Woodhouse (2002) decided not to ac-
count for elastic effects in their inversion for 3D
shear attenuation. A similar approach was taken
by Gung & Romanowicz (2004) who neglected
focusing as well as source/receiver corrections,
because these factors seemed to have little effect
on their degree-8 model in synthetic tests.
As an alternative to surface waves, Reid et al.
(2001) estimated t from amplitude ratios of glob-
ally recorded S , SS and SSS waves. Also neglecting
the effect of focusing, the t measurements were
inverted for a degree-8 model of Q μ in the upper
mantle.
The currently most sophisticated study on
whole-mantle shear attenuation was initiated by
Dalton & Ekstr om (2006) who simultaneously
inverted Rayleigh wave amplitudes and phase
delays for maps of shear attenuation and phase ve-
locity, as well as for amplitude correction factors
for each source and receiver. They demonstrated
that attenuation estimates depend significantly
on each of the amplitude corrections, and that
the neglect of elastic focusing translates into
inaccurate Q structure. The phase delay maps
were then used to remove the focusing effect from
the amplitude data, which were then inverted
for a global degree-12 Q μ model (Dalton et al. ,
2008). More than previous models, the work of
Dalton et al. (2008) reveals a clear correlation
between shallow-mantle attenuation and surface
tectonics. In particular, high attenuation is found
around 100 km depth beneath the circum-Pacific
volcanic arc, the Lau basin, and along most
mid-ocean ridges, including the East Pacific Rise,
the Indian-Antarctic ridge and the Mid-Atlantic
ridge. Precambrian lithosphere is characterized
by lower than average attenuation.
In contrast to elastic P or S tomographies, there
is little agreement between attenuation models.
Dalton et al. (2008) compared their global Q μ
model to those of Romanowicz (1995), Reid et al.
(2001), Warren & Shearer (2002), Selby & Wood-
house (2002) and Gung & Romanowicz (2004).
When truncated at degree 8, the correlation of
these models was found to be mostly below 0.4
throughout the upper mantle. Furthermore, Dal-
ton et al. (2008) noted that global Q μ models differ
Search WWH ::




Custom Search