of 20
Theory of the Earth
Don L. Anderson
Chapter 13. Heterogeneity of the Mantle
Boston: Blackwell Scientific Publications, c1989
Copyright transferred to the author September 2, 1998.
You are granted permission for indi
vidual, educational, research and
noncommercial reproduction, distribution, display and performance of this work
in any format.
Recommended citation:
Anderson, Don L. Theory of the Earth.
Boston: Blackwell Scientific Publications,
1989.
http://resolver.calt
ech.edu/CaltechBOOK:1989.001
A scanned image of the entire book may
be found at the following persistent
URL:
http://resolver.caltech.edu/CaltechBook:1989.001
Abstract:
We must now admit that the Earth is not
like an onion. The lateral variations of
seismic velocity and density are as import
ant as the radial variations. The shape
of the Earth tells us this directly but
provides little depth resolution. The long-
wavelength geoid tells us that lateral density variations extend to great depth.
Lateral variations in the mantle affect
the orientation of Earth in space and
convection in the core; this property of th
e Earth is known as asphericity. In the
following chapters we further admit that the Earth is not elastic or isotropic.
Hetero
"
You
all
do
know
this
mantle.
"
-
SHAKESPEARE,
JULIUS
CAESAR
W
e must
now
admit that the Earth is
not
like
an
onion.
The lateral variations
of
seismic
velocity
and
den
-
sity are as important
as
the radial variations. The shape
of
the
Earth
tells
us
this directly
but
provides little depth reso
-
lution. The long
-
wavelength
geoid tells us
that lateral
den
-
sity variations extend to great depth. Lateral variations in
the mantle affect the orientation
of
Earth
in
space
and
con
-
vection
in the core;
this
property
of
the Earth is known
as
asphericity.
In the
following chapters
we
further admit that
the Earth
is not elastic
or
isotropic.
UPPER
-
MANTLE
HETEROGENEITY
FROM
SURFACE
-
WAVE
VELOCITIES
The
most
complete
maps
of
seismic heterogeneity
of
the
upper mantle are obtained from surface
waves.
By
studying
the velocities of
Love
and
Rayleigh
waves,
of different
pe
-
riods,
over many
great circles, small arcs, and long arcs, it
is possible to reconstruct
both
the radial
and
lateral velocity
variations. Although global coverage
is possible, the limi
-
tations imposed
by the
locations
of
long
-
period seismic sta
-
tions
and
of
earthquakes limit
the
spatial resolution.
Fea
-
tures
having half
-
wavelengths
of
about
2000
km,
as a global
average, can
be
resolved
with
presently available
data.
The
raw
data consist
of
average group
and/or
phase
velocity
over
many
arcs. These averages can
be
converted
to images
using
techniques similar
to
medical tomography.
Body
waves
have better resolution,
but
coverage, particularly
for
the upper mantle, is poor.
Even the
early surface
-
wave studies indicated that the
upper mantle
was
extremely inhomogeneous. Shield paths
were
fast,
oceanic
paths were slow,
and
tectonic
regions
were
also
slow
(Toksoz and
Anderson,
1966;
Anderson,
1967). The
most pronounced differences
are
in
the
upper
200
km,
but substantial differences
between regions extend
to about
400
krn.
A
high
-
velocity region
was
inferred
to
extend
to
depths
of
120
-
150
km
under stable continental
shields. This
thick
LID, or seismic lithosphere,
probably
represents
the thickness
of
the plate.
Body
-
wave
results also
give
a shield
LID
thickness
of
about
150
km.
On
average,
the shield
mantle
is also
faster
than average
oceanic
or tec
-
tonic
mantle
down
to 400
km,
but the
differences
below
200
km
are
much
less
than
above this
depth. Some
shield-
bearing continents
have
overriden oceanic lithosphere
in the
past
50
-
100
million
years,
and
others
were bordered
by
subduction
zones
prior to
the breakup
of
Pangaea.
Old
oce
-
anic lithosphere
has
a long
thermal time constant
and
serves
to cool
off
adjacent
mantle.
Below 200
km the
regions
of
higher than
average mantle velocity
may
represent the cool
-
ing effect
of
overriden oceanic lithosphere
andlor
the ab
-
sence
of
a partial
melt
phase. The rapid decrease
in
velocity
under shields
below
about 150
km
suggests
that this
is the
depth
of
decoupling. The suggestion that continental
roots
extend deeper
than
400
km
(Jordan,
1975)
rather
than
the
150
-
200
km
of
earlier studies
was
based
on
the
low-reso-
lution
ScS
phase rather
than
the higher resolution
P
-
waves
and
surface
waves
(Jordan, 1975;
Jordan and
Sipkin,
1976).
There are
two basic
approaches for interpreting global
surface
-
wave
data.
The regionalized approach introduced
by
Anderson (1967)
and
Toksoz
and
Anderson
(1966)
di
-
vides the Earth into tectonic
provinces
and
solves
for the
velocity
of
each. Applications
of this
technique
yielded
fast
shield
velocities at short periods
or shallow
depths
but
showed
that convergence regions
were
faster
at long pe
-
riods, or greater depth, suggesting
that cold
subducted
ma-
terial
was
being
sampled (Nakanishi
and
Anderson, 1983,
1984a,b).
Convergence regions, on average, are
slow at
short periods,
due
to
high
temperatures
and
melting at shal
-
low
mantle depths. The regionalization approach
is
neces
-
sary
when
the data are limited
or
when
only complete great
-
circle data are available.
In
the latter case
the velocity
anomalies cannot
be
well
isolated.
If short
-
arc
and
long
-
arc data are available, a complete
spherical harmonic expansion can
be
performed
with
no
as
-
sumptions required
about
tectonic regionalizations. Studies
of
this type
have
been
reported
by
Nakanishi and
Anderson
(1983,
1984a,b),
Tanimoto and Anderson
(1984, 1985)
and
Woodhouse and Dziewonski
(1984).
If only complete great
-
circle
data,
or free
-
oscillation
data,
are available,
then only
the even
-
order harmonics can
be
determined.
In
the regionalized
models
it is
assumed
that all
re
-
gions
of
a given tectonic classification
have
the same
ve
-
locity. This
is clearly oversimplified.
It
is useful,
however,
to
have
such maps
in
order to
find
anomalous regions.
In
fact, all shields are
not
the same, and velocity does
not
in
-
crease monotonically
with
age
with
the ocean. The region
around Hawaii,
for
example, is faster
than
equivalent age
ocean elsewhere at
shallow
depth
and
slower at greater
depth. There is also a
slow patch
in
the south
-
central
Pa
-
cific. The differences
between Love
wave
and Rayleigh
wave
maps
are primarily due to differences
in
penetration
depth
and
anisotropy
between
vertical
and
horizontal
S-
waves. The early surface
-
wave results
showed that
the shal
-
low
mantle
was slow, and
presumably hot, under
young
oceans, tectonic regions and subduction zones. Most
of
the
slow
regions are
in
geoid
highs. There
is a good
correspon
-
dence
of
surface
-
wave
velocity maps with heat
-
flow maps
but
generally poor correlation
with
the geoid. The geoid,
apparently,
has
little sensitivity
to
upper
-
mantle density
variations.
GLOBAL
SURFACE-
WAVE
TOMOGRAPHY
A global
view
of
the lateral variation
of
seismic velocities
in
the
mantle
can
now
be
obtained
with
surface
-
wave to
-
mography. In
the regionalization approach one assumes that
the velocities
of
surface
waves
are linearly dependent on the
fraction
of
time spent in various tectonic provinces. The
inverse problem
then
states that the
velocity profile
depends
only on the tectonic classification.
For
example, all shields
are
assumed to be
identical at
any
given depth. This as
-
sumption appears
to
be
approximately
valid
for
the
shallow
structure
of
the mantle
but becomes
increasingly tenuous for
depths greater
than
200
km.
However,
it probably provides
a maximum estimate
of
the depth
of
tectonic features,
and
it also provides a
useful standard
model
with which
other
kinds
of
results can
be
compared.
The second approach subdivides the Earth into cells or
blocks
or
by
some
smooth
function
such
as
spherical
har
-
monics.
Nataf
and
others (1984, 1986)
used
spherical
harmonics for the lateral variation
and
a series
of
smooth
functions joined at mantle discontinuities for the radial
variation.
In
this approach no a priori tectonic information
is built in.
In
both
of
these approaches
the number
of
parameters
that one
would
like
to
estimate far exceeds the information
content (the number
of
independent data points)
of the
data.
It is therefore
necessary to
decide
which
parameters are
best
resolved
by
the
data,
which
is the resolution, or
averaging
length,
which
parameters to
hold
constant
and
how
the
model
should
be
parameterized
(for
example as
layers or
srnooth
functions, isotropic or anisotropic).
In
addition,
there are a
variety
of
corrections
that might
be
made (such
as crustal thickness,
water
depth, elevation, ellipticity, at
-
tenuation).
The
resulting models are
as
dependent
on
these
assumptions
and
corrections as
they
are
on
the quality
and
quantity
of
the data. This
is
not unusual
in science:
Data
must always be
interpreted
in
a framework
of
assumptions,
and
the data are always,
to some
extent, incomplete
and
inaccurate.
In
the seismological
problem
the relationship
between the
solution, or the model,
including
uncertainties,
and
the data
can be
expressed formally. The
effects
of
the
assumptions
and
parameterizations,
however, are more
ob
-
scure,
but
these also
influence
the solution. The
hidden
assumptions are
the most
dangerous.
For
example,
most
seismic
modeling assumes
perfect elasticity, isotropy, geo
-
metric optics
and
linearity.
To
some extent all
of
these
as
-
sumptions are wrong,
and
their likely
effects must be kept
in
mind.
Nataf and
others (1986)
made an
attempt
to
evaluate
the resolving
power
of
their global surface
-
wave
dataset
and
invoked
physical a priori constraints
in
order
to
reduce
the
number
of
independent parameters
that needed
to
be
es
-
timated from the data. For example, the density, com
-
pressional
velocity
and
shear
velocity
are independent
pa
-
rameters, but their variation
with
temperature, pressure
and
composition
show
a high
degree
of
correlation; that
is,
they
are coupled parameters. Similarly, the
fact that
temperature
variations in the mantle are
not
abrupt
means that
lateral
antd
radial variations
of
physical properties
will
generally
be
smooth
except
in
the
vicinity
of
phase boundaries,
in
-
cluding partial melting. Changes
in
the orientation
of
crys
-
tals
in
the mantle
will
lead
to changes
in
both
the
shear-
wave and
compressional
velocity
anisotropies. These
kinds
of
physical considerations can
be
used in
lieu of the
standard seismological assumptions,
which
are generally
made
for mathematical convenience rather
than
physical
plausibility.
The studies
of
Woodhouse and Dziewonski
(1984)
and
Nataf and
others (1984, 1986)
give
upper
-
mantle
models
that
are
based
on
quite
different
assumptions
and
analyti
-
cal techniques.
Woodhouse and Dziewonski inverted
for
FIGURE
13
-
1
Group
velocity
of 152
-
s Rayleigh
waves
(kmls).
Tectonic
and
young
oceanic areas
are
slow
(dashed),
and
continental
shields
and
older oceanic areas
are
fast.
High
temperatures
and
partial melting are responsible
for
low
velocities.
These
waves
are
sensitive
to the upper
several
hundred
kilometers
of
the
mantle
(after
Nakanishi
and
Anderson,
1984a).
shear
velocity,
keeping the density, compressional
velocity
and
anisotropy fixed.
They
also
used
a very smooth
radial
perturbation function
that
ignores
the
presence
of
mantle
discontinuities
and
tends to smear out anomalies
in
the
vertical direction.
They
corrected for near
-
surface
effects
by
assuming
a bimodal
crustal thickness, continental
and
oceanic.
Nataf and
others corrected for elevation,
water
depth,
shallow
-
mantle velocities
and measured
or inferred crustal
thickness.
They inverted
for shear
velocity and
anisotropy,
the dominant parameters,
but included
physically plausible
accompanying changes
in density, compressional
velocity
and
anisotropy. Corrections
were
also
made
for
anelasticity.
The radial perturbation functions
were allowed
to change
rapidly across
mantle
discontinuities,
if
required
by
the
data.
In
spite
of
these differences, the resulting
models
are
remarkably
similar
above
about 300 km. The
main
differ
-
ences occur
below
400
km
and
seem to arise from differ
-
ences
in the
assumptions
and
parameterizations (crustal cor
-
rections, radial
smoothing
functions) rather
than
the
data.
The choice
of
an
a priori
radial
perturbation function can
degrade the vertical resolution intrinsic
to
the
dataset.
The
solution, in
this
case, is
overdamped
or oversmoothed.
Before
discussing the
inversion
results
-
the
earth
structures
themselves
-
I will
briefly
describe the distribu
-
tion
of
Love
and
Rayleigh
wave
velocities (Nakanishi
and
Anderson, 1983,
1984a,b).
Love waves
are sensitive
to the
SH
velocity
of
the shal
-
low
mantle,
above about
300
km
for the periods considered
here.
The
slowest regions
are at plate boundaries, particu
-
larly triple junctions.
Slow
velocities extend around the
Pa
-
cific
plate
and
include the East
Pacific
Rise, western
North
America, Alaska
-
Aleutian arcs, Southeast
Asia and
the
Pa
-
cific
-
Antarctic Rise. Parts
of
the Mid
-
Atlantic Rise
and
the
Indian
Ocean
Rise are also
slow.
The
Red
Sea
-
Gulf
of
Aden
-
East
Africa
Rift
(Afar triple junction) is one
of
the
slowest
regions. The upper
-
mantle
velocity anomaly
in
this
slowly
spreading region is
as
pronounced
as
under the rap
-
idly spreading East
Pacific
Rise. Since
it also shows up
for
long
-
period Rayleigh waves,
this
is a substantial
and
deep
-
seated
anomaly.
Shields are generally
fast,
particularly the
Brazilian,
Australian and
South African shields. The fastest
oceanic
regions
are the north
-
central Pacific, centered near
Hawaii,
and the
eastern Indian Ocean.
Rayleigh
waves
are sensitive to
shallow
P
-
velocities
and
SV
velocity
from about 100 to 600
km.
The fastest re
-
gions (see Figure
13
-
1)
are the
western
Pacific, western
Af
-
rica
and
the South Atlantic.
Western North
America, the
Red
Sea area, Southeast
Asia
and
the
North
Atlantic are the
slowest
regions. Slower than average
Rayleigh
-
wave phase
velocities
are
obtained at long
period
for
the stable conti
-
nental
areas. The
velocities in
the South
Atlantic and
the
Philippine
Sea plate are faster
than
shields.
Regions
of
convergence,
or
descending
mantle
flow,
are
relatively
fast
for
long
-
period
Rayleigh waves,
particu
-
larly
near
New
Guinea,
Sumatra,
the western
Pacific and
northwestern
Pacific.
These
regions have large
positive
ge-
oid
anomalies. Dense, cold material in
the upper
mantle
can
explain these results.
To
explain the
surface
-
wave
results,
the fast material
must
be
below
the depth sampled by
250
-
s
Love waves and 150
-
s Rayleigh waves, which is about
300
km depth,
and
must occupy a considerable
volume
of
the
upper
mantle. Conductive cooling
of
the oceanic litho
-
sphere
probably
extends
no
deeper
than about
150
km.
It is
doubtful that the
subduction
of
such a
thin high
-
velocity
plate could explain
these results.
Either slabs preferentially
subduct
in
cold
regions
of
the
mantle
or
they
pile up
in
the
upper mantle.
At
shorter periods the
velocity
of
the
mantle
beneath
subduction
zones is average
or slower
than
aver
-
age,
presumably reflecting
the
extensional tectonics and hot
shallow
mantle
under back
-
arc
basins.
A
map
synthesized
from the
1
=
2
coefficients at
250
s
is characterized
by
a broad
closed
high
-
velocity re
-
gion
centered in the
western
Pacific, northeast
of
New
Guinea,
and
a low
-
velocity region centered on
the East
Pa
-
cific
Rise. Because
of
symmetry the antipodal
regions
are
identical.
The
1
=
2
coefficients
are
just
part
of
the com
-
plete
set
that is
required to fully
describe the
Earth's
aspher-
icity and need
no
special
explanation.
The
1
=
2 pattern
of
the
associated
density
field
does
have special significance
because
it is
involved in
the
Chandler
wobble, tidal friction,
the
orientation
of
the
Earth's
spin
axis and
polar
wander.
Tectonic models based on
surface
tectonics have
a strong
1
=
2 component,
because
of
the ocean
-
continent
configu
-
ration,
the locations
of
island
arcs,
and the
thermal struc
-
ture
of
the
seafloor.
In particular, the
velocity highs
en
-
compass most
of
the
world's
old
oceanic lithosphere
and
convergence regions,
and
the
lows
include most
of
the
young
oceanic lithosphere
and many
of
the world's
hot-
spots.
The
correlation
between surface
-
wave velocity and
heat
flow and
tectonics,
and between velocity and
geoid is
relatively high
for
1
=
2,
particularly
at
long
periods.
If the
rotation axis
of
the Earth
is controlled by density
anomalies
in
the
upper
mantle,
and
if density correlates
with velocity,
then
the pattern
would
be symmetric about the equator.
The
overall
pattern
of
the
Love
-
wave
phase
velocity
variations shows a general correlation
with
surface tecton
-
ics. These
waves
are
most
sensitive
to
the
SH
velocity in
the upper
200
to
300
km
of
the mantle.
The
lowest velocity regions
are
located in
regions
of
extension
or
active volcanism:
the
southeastern
Pacific,
western North
America, northeast
Africa
centered
on
the
Afar
region, the central Atlantic, the central Indian
Ocean,
and
the
marginal
seas
in
the
western Pacific.
A
high-
velocity region
is located
in
the
north
central
Pacific,
cen
-
tered near
the
Hawaiian
swell.
Love
-
wave velocities
are
low
along parts
of
the
Mid
-
Atlantic
Ridge, especially near
the
triple junctions
in the North and South Atlantic.
High
-
velocity regions
in continents
generally
coin
-
cide
with
Precambrian shields
and Phanerozoic
platforms
(northwestern Eurasia,
western and
southern
parts
of
Af
-
rica, eastern
parts
of
North and South
America,
and Antarc
-
tica). Tectonically active
regions,
such
as
the
Middle East
centered
on
the
Red
Sea,
eastern
and
southern Eurasia,
eastern Australia,
and western North America
are
slow,
as
are
island
arcs
or
back
-
arc
basins such
as
the
southern
Alaskan
margin, the Aleutian, Kurile, Japanese, Izu
-
Bonin,
Mariana,
Ryukyu, Philippine, Fiji,
Tonga,
Kermadec,
and
New
Zealand
arcs.
Rayleigh
-
wave velocity
also correlates
well with
sur
-
face
tectonics,
particularly at
short periods. The periods
generally considered are
most sensitive
to
SV
velocity be
-
tween about
100
and
400
-
600 km
and thus
sample deeper
than
the
Love waves.
The
slowest
regions,
for
Rayleigh waves,
are
centered
on
the
Red Sea
-
Afar
region,
Kerguelen
-
Indian Ocean
triple junction,
western North
America
centered on the
Gulf
of California, the
northeast
Atlantic,
and Tasman
Sea-
New
Zealand
-
Campbell
Plateau.
The
fastest regions
are
the
western
Pacific,
New
Guinea
-
western Australia
-
eastern
Indian
Ocean,
west
Africa, northern Europe
and
the
South
Atlantic.
One
difference between
Love
waves and Rayleigh
waves
is evident
in
the
island
arcs
along
the
northwestern
margins
of
the
Pacific
Ocean,
such as
the Aleutian, Kurile,
Japanese, Izu
-
Bonin
-
Mariana, Ryukyu,
and Philippine re
-
gions. These
regions
are
characterized
by high velocity
for
Rayleigh
waves. The
difference between
Love
and Rayleigh
wave results
is partially explained
by
the difference in pene
-
tration depth.
The
Love
-
wave results
indicate that the
shal
-
low
mantle, in the
vicinity
of
island arcs and back
-
arc ba
-
sins,
is
slow. Rayleigh waves
sample the fast
material that
has subducted
beneath
the
island
arcs. Because
of
the large
wavelengths,
the lateral extent
of
the
fast
material
must be
considerable. Anisotropy
may
also be
involved:
In
regions
of
convergence
and
divergence,
the
mantle
flow can be
ex
-
pected
to
be
generally
vertical,
whereas in
the centers
of
convection cells,
the
flow
is
mainly
horizontal. The sense
of
anisotropy will
change
at
ridges
and subduction
zones.
For
Rayleigh waves
the
Atlantic
Ocean
is
generally
high
velocity
except
for
the northern part. The
low veloci
-
ties associated
with
the midoceanic ridge system,
which ap
-
pear
in the
Love
-
wave
maps,
are not
evident in
the
long-
period
Rayleigh
-
wave
maps. This
may
indicate
that many
ridge segments are
passive
or
that
SV
exceeds
SH
in regions
of
ascending
flow.
Maps
of
surface
-
wave velocity (Nakanishi and
Ander
-
son,
1983,
1984a,b)
provide,
perhaps,
the most
direct
dis
-
play
possible
of
the lateral
heterogeneity
of
the mantle.
The
phase
and
group
velocities
can be
obtained with high
pre-
cision
and
with
relatively
few
assumptions. In general, the
shorter period
waves, which
sample
only
the crust
and
shallow mantle, correlate
well and
as expected
with
sur
-
face tectonics. The longer period
waves, which
penetrate
into the transition region (400
-
650 km), correlate
less
well with
surface tectonics. The
inversion of
these results
for Earth structure
involves many more
assumptions
and
approximations.
REGIONALIZED
INVERSION
RESULTS
Figure
13
-
2
shows
vertical shear
-
velocity profiles, ex
-
pressed
as differences from
the
average Earth,
using
the
re-
gionalization approach
(Nataf
and
others,
1986).
Young
oceans, region
D,
have slower
than
average velocities
throughout the upper mantle
and
are particularly
slow
be
-
tween
80
and
200
km,
in
agreement
with
the higher reso
-
lution
body
-
wave
studies.
Old
oceans, region
A,
are fast
throughout the upper mantle. Intermediate
-
age oceans
(B
and
C)
are intermediate
in
velocity
at all depths.
Most
of
the oldest oceans are adjacent
to
subduction zones,
and
the
subduction
of
cold material
may
be
partially responsible for
the fast velocities at depth. Notice that velocities converge
toward
400
km but
that differences still remain
below this
Velocity
variation
(kmls)
FIGURE
13-2
Variation
of the
SV
velocity
with
depth
for various tectonic
provinces.
A
-
D,
oceanic
age
provinces ranging
from
old,
A,
to
young,
D;
S,
continental
shields;
M,
mountainous
areas;
T,
trench
and
island
-
arc regions.
These are regionalized results
(after
Nataf
and
others,
1986).
depth. The continuity of the
low
velocities
beneath
young
oceans,
which
include
midocean
ridges, suggests that the
ultimate source region for
MORB
is below
400
km.
Shields
(S) are faster than all other tectonic provinces except old
ocean from 100 to 250
km.
Below
220
km
the velocities
under shields decrease, relative to average Earth,
and below
400
km
shields are
among
the
slowest
regions.
At
all depths
beneath shields the velocities can
be
accounted for
by
rea
-
sonable mineralogies
and
temperatures without
any need
to
invoke partial melting.
Trench
and
marginal sea regions
(T),
on the other hand, are relatively
slow above
200
km,
probably indicating the presence
of
a partial melt,
and
fast
below
400
km,
probably indicating the presence
of
cold
subducted lithosphere. The large size
of
the tectonic regions
and
the long
wavelengths of
surface
waves
require that the
anomalous regions at depth are
much
broader
than
the sizes
of
slabs or the active
volcanic
regions at the surface. This
suggests very
broad upwellings
under
young
oceans
and
abundant
piling up
of
slabs under trench and old ocean
re
-
gions. The latter is evidence for layered mantle convection
and the
cycling
of
oceanic plates into the transition region.
Shields
and
young
oceans are still evident at
250
km.
At
350 km the
velocity
variations are
much
suppressed. Be
-
low
400
km,
most
of
the correlation
with
surface tectonics
has
disappeared,
in
spite
of
the regionalization, because
shields
and
young
oceans are
both slow, and
trench
and old
ocean regions are
both
fast. Most
of
the oceanic regions
have
similar velocities at depth. This is a severe test
of
the
continental tectosphere hypothesis
of
Jordan
(19751,
de
-
scribed later
in
this
chapter. Shields do
not have
higher
ve
-
locities
than some
other tectonic regions
below
250
km and
definitely
do
not have
"
roots
"
extending throughout the
up
-
per
mantle
or even below
400
km.
Results for other depths
are
given
by
Nataf and
others (1984, 1986).
In
high-
resolution
body
-
wave
studies, subshield velocities
drop
rap
-
idly
at 150
km
depth, although velocities remain relatively
high to
about
390
km.
These high velocities could represent
"roots"
physically attached
to the shield
lithosphere, over
-
ridden cold oceanic lithosphere
or
simply
"
normal
"
con
-
vecting mantle
weakly
coupled to the overlying shield litho
-
sphere
via
a boundary
layer
at 150
-
200 km depth. The
velocities
below 200
krn
can
be
explained
by
an
adiabatic
temperature gradient
and
therefore probably represent nor
-
mal
convecting mantle. Therefore,
it is the
slow mantle un
-
der ridges and tectonic regions that
is
anomalous,
and,
if
anything, these are the regions
with
the roots.
If the mantle
under shields is convectively stagnant, as implied
by
the
deep tectosphere hypothesis, a high thermal gradient
would
extend
over
a large
depth
interval. This could lead to
par
-
tial melting
and
a depression
of
the olivine
-
spinel phase
boundary under shields. I therefore prefer the 150
-
km
-
thick
plate hypothesis,
that
is,
a correspondence
of
the thickness
of
the plate
with
the seismic high
-
velocity
layer.
The
slightly higher than
average
subshield mantle
between about
200
km
and
350
-
390
km
may
be
a boundary layer
or
could