Download Detection of subducted crustal material in the mid

Detection of subducted crustal material
in the midmantle
from asymmetric PP reflections
Sebastian Rost1, Edward J. Garnero2 and Quentin Williams3
1 Institute of Geophysics and Tectonics, School of Earth and Environment, University of Leeds, Leeds, UK
2Dept. of Geological Sciences, Arizona State University, Tempe AZ, USA
3Dept. of Earth Sciences, University of California Santa Cruz, Santa Cruz CA, USA
School of
Earth and
Environment
University of
California
Santa Cruz
Introduction
Data
2000
12
PKKP
1800
10
8
PP
1400
6
PKP
Time (sec)
1200
4
PKiKP
1000
800
2
Pdiff
ScP
600
0
PcP
-2
P
400
-4
200
-6
0
20
40
60
80
100
120
Distance (deg)
140
160
YKA Array elements
0
180
Pdiff
PP
B. Vespagram
PKiKP
C. Vespagram Zoom
Panel C
10
10
Slowness [s/deg]
8
++ + + + + +
PP
8
PP
6
6
Pdiff
4
PKiKP
2
precursors
0
10
0
100
200
Relative Time (sec)
300
8
6
4
4
2
2
0
240
260
280
300
Power
Figure 1: Result of a "stacking stack" algorithm for data from the
short-period Yellowknife array (YKA) [Rost, Thorne and Garnero,
2006]. Shown in this traveltime-curve like display are Nth-Root stacks
from ~1400 events recorded at YKA. Color of the energy indicates
slowness. Traveltimes for PREM for major phases are shown as thin
lines. The PP precursor energy in the 90 to 110 deg distance range is
clearly visible with slownesses between P (Pdiff) and PP. No reflections
from the major upper mantle discontinuities that arrive as precursors
to PP (e.g. P410P and P670P) are visible in these stacks. The precursory
energy is likely produced by scattering along the path.
Figure 2: A Selected dataset and
data example. Arrivals for major
phases are marked. Note the energy
increase before PP. Sources are
shown as circles. All events are recorded at YKA in northern Canada.
Crosses denote the PP mid reflection
point.
B 4th-root vespagram of the recordings in A. Slowness and time for
major arrivals are shown. The PP
precursors show smaller slowness
than PP (circles). Crosses mark theoretical arrivals for symmetric underside reflections.
C. Zoom into precursor time window.
0
Relative Time (sec)
Results 1
PP
PdP
A
We detect 116 precursors from the 21 events. 110 of these allowed a successful backtracing to its orign,
while 6 precursor measurements were rejected and could not be traced back to a P*P scattering location.
More than half of the scatterers can be mapped to depths larger than 400 km. Scatterers in the uppermost
mantle are likely related to scattering in the oceanic lithosphere or shallow upper mantle. The deep
scatterers show a high correlation to fast seismic velocities from tomographic images. The reflectors form
sheetlike structures beneath two subduction zones.
The reflectors beneath the Tonga-Fiji and the Mariana subduction zone can be found to depths of 1000
km, thus indicating the penetration of these slabs into the lower mantle. Previous studies in the Mariana
region show evidence for mid mantle structure at depths of 1100 to 1500 km to the north-east of our study
area.
∆t
array
slowness
i
source
PP
u
PdP
time
δΘ
reflection
P*
P
array
∆t
i'
slowness
B
SKKP
1600
Method
We use the frequency-wavenumber analysis
(fk-analysis) [Capon 1973] to determine the full
slowness vector (slowness and backazimuth) of
the precursors. Short time windows (typical
4-6s) are chosen for the fk-analysis. We also
measure the differential travel time between
precursor and PP. Many of the precursors show
rather impulsive onsets. We estimate the
differential travel time uncertainty to be less
than 1s.
Using a backtracing algorithm it is possible to
locate the likely origin of the precursors. We fit
all three measurements (slowness, backazimuth
and traveltime) and find the likely reflector
(latitude, longitude and depth) by raytracing
through a 1D Earth model (e.g. IASP91). We
assume that the energy is P scattered to P
(P*P). Scattering combinations such as S*P and
PP*P or P*PP were also tested, but it has been
found that these combinations did not fit the
measurements. Most of the precursors show
backazimuths strongly off great circle path,
with slownesses smaller than PP, which
indicates an origin on the receiver side of the
path.
Data were collected from the CNSN datacenter through AutoDRM. YKA
is a small aperture (20 km) array with 20 array elements distributed along
two perpendicular lines. Interstation spacing is 2.5 km. The short-period,
vertical instruments record frequencies around 1 Hz. YKA was designed to
detect high-frequency P-waves from underground nuclear explosions and is
therefore well suited for the detection of high-frequency scattered waves.
For this study we selected 21 events from the catalog that show clear
precursors. Traces were individually inspected and obviously faulty traces
were discarded. We apply a bandpass filter with cutoff frequencies of 0.5
Hz and 1.4 Hz. Shown is a data example of an mb=5.7 event at 41 km
depth in the Talaud islands region (∆=100.14). Note the increase of energy
before the PP arrival. The slowness deviation in the vespagram indicates a
scattering origin for the precursors.
Slowness (sec/deg)
The precursor wavefield to PP, a P-wave once reflected at the free surface
between source and receiver, contains information about the structure of the upper
mantle. Underside reflections of PP (and SS) beneath the mid-reflection point off
the upper mantle discontinuities have been extensively used to map topography on
the discontinuities and to infer temperature and composition of the upper mantle.
However, additional precursory arrivals that are not related to reflections off
the global discontinuities can be detected in short-period recordings. These
precursors can be easily detected in the 95 to 110 deg distance range due to the
velocity structure of the upper mantle [King, Haddon and Husebye , 1975]. These
arrivals show slownesses strongly different from PP (several s/deg), therefore
excluding symmetric underside reflections as their origin. It has been proposed
earlier that this energy originates from asymmetric scattering of PP energy in the
upper mantle [Wright and Muirhead, 1969; Wright, 1972; King, Haddon and
Husebye, 1975].
Here, we study earthquakes from the SW Pacific and Indonesia recorded at the
Canadian, short-period Yellowknife array (YKA), that show similar precursors as
described in the earlier work. Using a frequency-wavenumber analysis we are able
to measure the full slowness vector of the precursory energy. Using this
information we can trace the energy back to the likely position of the scatterer or
reflector by fitting slowness (u), backazimuth (Θ) and differential traveltime (∆T)
with respect to PP.
We find that many of the precursors can be explained by scattering in or just
below the oceanic lithosphere in the Pacific. Nonetheless, more than half of the
scatterer positions are located at depths larger than 300 km and show strong
correlation to detected fast regions in seismic tomography. Our mineralphysical
modeling predicts an impedance contrast for the harzburgite to basalt (and pyrolite
to basalt) for these mid-mantle depths in agreement with our seismic data, so that
we likely detect the remains of subducted slabs in the mid-mantle.
i
source
PP
A
PP
500 - 600 km
0 - 100 km
2030 - 2130 km
B
0 - 144 km
722 - 867 km
∆u
P* P
100 - 200 km
600 - 660 km
time
144 - 289 km
867 - 1011 km
-4 -3 -2 -1 0 1 2 3 4
δVS [ % ]
Figure 3: Sketch of raypaths (left) and locations of PP and potential
precursors in slowness-time space (right). Precursors from mid point
reflections (e.g. underside reflections off the 410 and 670) will show
slownesses comparable with PP (A). Their slowness decreases slightly for
deeper reflectors (see Figure 2). In contrast, asymmetric reflections with
reflection points on the receiver side of the path, show a much smaller
slowness than PP, but a larger slowness than P (B). Their traveltime,
slowness (and horizontal incidence angle - backazimuth) can be used to
map the source of the reflection
200 - 300 km
660 - 730 km
289 - 433 km
1011 - 1156 km
300 - 400 km
730 - 830 km
433 - 578 km
1156 - 1300 km
400 - 500 km
830 - 930 km
578 - 722 km
2023 - 2168 km
-0.6 -0.3 0.0 0.3 0.6
Figure 4: Depth slices through the
tomography model by Ritsema and van Heijst
[2002] (A) and Karason and van der Hilst
[2001] (B). Each depth slice also includes the
detected scatterers (reflectors) marked as
crosses for the appropriate depth. The scatterer
locations at depths larger than 400 km show a
very good correlation with fast velocities down
to depths of 1000 km. These scatterer locations
are located beneath the Mariana and
Tonga-Fiji subduction zones.
Energy detected from shallow depths is most
likely related to scattering of velocity
heterogeneities in the oceanic lithosphere and
the uppermost oceanic mantle due to
incomplete mixing. The energy mapping the
deep scatterers could be reflected off the
oceanic crust transported to these depths by
subduction. This is the first time that PP
asymmetric precursors can be connected to the
mapping of subduction zones.
δV [ % ]
P
Discussion and Interpretation
Results 2
km
B
5
Impedance Contrasts
870
1
5
770
Refl. Depth [km]
1
970
660
2
6
2
520
7
6
5
4
3
2
1
6
400
240
120
0
3
7
3
4
Figure 5: A S-wave tomographic
cross sections [Ritsema and
Heijst, 2001] for the Mariana
region. Profile 3 contains seismicity (blue circles) from 1973-to
1995 in a 3 degree corridor
around the profile. All scatterers
(yellow circles) are projected
onto profile 4. B as A but for a
P-wave model [Karason and van
der Hilst, 2002].
7
4
-4 -3 -2 -1 0 1 2 3 4
-1
A
1
B
5
1
0
1
5
km
970
870
660
520
2
2
6
6
7
6
400
5
4
240
3
21
120
0
3
3
7
7
4
4
-4 -3 -2 -1 0 1 2 3 4
-2
-1
0
1
2
Figure 6: As Figure 5 but for
the Tonga region. Same models
are used. Reflectors are projected onto the closes profile.
The orange cirlce and the red
triangle indicate the beginning
and end of the profiles shown in
the insert.
Refl. Depth [km]
770
Basalt-Harzburgite
Basalt-Pyrolite
Shear
PP precursors contain information about Earth structure
besides what can be resolved from upper mantle
discontinuity structure.

Compressional
10
% Impedance Contrast
A
Using array methods it is possible to extract this
information of the precursor wavefield to resolve fine-scale
Earth structure.

8
Upper estimates
6
4
This experiment using YKA data shows evidence for
scattering from small-scale heterogeneities in the oceanic
lithosphere.
Lower estimates

2
0
-2
600
800
1000
1200
1400
1600
Depth (km)
Figure 4: Ranges of possible impedance contrasts between basalt and harzburgite and basalt and pyrolite.
Such calculations are particularly sensitive to (1) the
controversial bulk modulus of the calcium ferrite structured phase, for which estimates vary from 190 to 243
GPa (Ono et al., 2002; Guignot and Andrault, 2004),
and the precise pressure derivatives of the shear moduli
of the alkalic phases. No thermal contrasts between
basalt and underlying or overlying material are
included, as local thermal equilibration over the lengthscale of the reflected wave’s sampling is likely to have occurred. The decrease in these contrasts at depths below
~1300 km is generated by shear softening accompanying
the stishovite-CaCl2 phase transition in SiO2 (e.g. Carpenter et al., 2000). Determining the precise depth extent
(as opposed to the absolute amplitude) of this softening
is difficult because of the uncertain absolute temperature
of the subducted material: the estimate we show is likely
an upper bound on the depth range of this shear softening. For depths in the shallow lower mantle (above ~750
km depth), the impedance contrasts are strongly negative
due to the persistence of majoritic garnet within basalt
(e.g., Ono et al., 2001).
Additionally, many deep scatterers can be found mainly
beneath the Mariana and Tonga-Fiji subduction zones.

Deep scatterers have a high correlation to the location of
high seismic velocities from tomographic models.

These scatterers form a sheet-like structure that projects
to the surface location of the subduction zones.

Using P*P we are able to confirm that subducted
oceanic crust can penetrate into the mantle to depths of at
least 1000 km.

Mineralphysical modeling shows that both the
harzburgite-to-basalt and the basalt-to-pyrolite transitions
at mid mantle depths can produce the PP precursors.

Impedance contrast decrease rapidly around 1300 km
depth, which might make a detection in the lowermost
mantle difficult.
