Download Detection of subducted lithosphere in the midmantle from

Data and Method
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 (Figure 3). 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.
Figure 4 shows a data example of an mb=5.7 event at 41 km depth in the Talaud islands region (D=100.14). Note the increase
of energy in a beamtrace for PP slowness (Figure 5) indicating a scattering origin for the precursors. Precursors with varying
slownesses are clearly visible in a 4th-Root Vespagram of this event (Figure 6).
-30
110 120 130 140 150 160 170 180 190 200 210 220 230 240 250
B
YKA array elements
-20
Precursors
P
P-coda
A
Precursors
A
410
P
P
210
P
10
B
9
Slowness [s/deg]
PKiKP
B
1
1.0
PP
7
7
6
6
5
PP
P 210 P
8
2
4
5
Pdiff
4
4
3
3
2
0.8
0.6
Power
PP
10
9
8
Pdiff
50 s
PP
1
3
0.4
0.2
1
0
300
0
25
Time [ s ]
50
PKKP
PP
1400
PKP
1200
PKiKP
800
ScP
600
PcP
Pdiff
PP
20
40
80
100
120
Distance (deg)
140
160
Figure 1: Result of a "stacking stack" algorithm for data
from the short-period Yellowknife array (YKA) [Rost, Thorne
and Garnero, 2005]. Shown in this traveltime-like display are
Nth-Root stacks from ~1400 events recorded at YKA. The PP
precursor energy in the 90 to 110 deg distance range is clearly
visible. No reflections from the major upper mantle
discontinuities are visible in these stacks and the precursory
energy is likely produced by scattering along the path.
50
PdP
60
Pd P
10
20
30
0
0
-1
-4
60
40
70
0
-1
-6
0
30
0
-2
P
20
40
50
830 - 930 km
1930 - 2030 km
2030 - 2130 km
PP
C
D
E
F
P
time
Figure 2: Sketch of raypaths (left) and locations of P, PP and
potential precursors in slowness-time space (right). P and PP
show a strong slowness difference of some s/deg and traveltime
differences of some hundred seconds (top panel). Precursors from
mid point reflections (e.g. underside reflections off the 410 and
670) will show slownesses comparable with PP (center panel).
Their slowness decreases slightly for deeper reflectors. In
contrast, asymmetric reflections show a much smaller slowness
than PP, but a larger slowness than P (lower panel). Their
traveltime, slowness (and horizontal incidence angle backazimuth) can be used to map the source of the reflection
Conclusions
PP precursors contain information about Earth structure besides what can be resolved from upper mantle discontinuity
structure.
n Using array methods it is possible to extract this information of the precursor wavefield to resolve fine-scale Earth structure.
n This experiment using YKA data shows evidence for scattering from small-scale heterogeneities in the oceanic lithosphere.
n Additionally, many deep scatterers can be found mainly beneath the Mariana and Tonga-Fiji subduction zones.
n Deep scatterers have a high correlation to the location of high seismic velocitites from tomographic models.
n These scatterers form a sheet-like structure that projects to the surface location of the subduction zones.
n Using P*P we are able to confirm that subducted oceanic crust can penetrate into the mantle to depths of at least 1000 km.
n 3D raytracing through tomographic models can further improve the mapped location of the scatterers.
n Estimates of impedance contrast for the reflectors are difficult at present due to the requirement of full 3D synthetic
seismograms (reflector planes can vary in all 3 directions in space) at large distances up to ~2Hz frequency.
n
1000 750 2000
PP
Du
-2
-1
0
1
2
3
4
B
DT
P* P
-3
δVS [%]
time
70
P*P
Figure 9: Depth slices through the
tomography model by Ritsema [Ritsema and
van Heijst, 2002]. Each depth slice also
includes the detected scatterers (reflectors)
marked as black crosses (yellow for the
uppermost mantle for better visibility) 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.
The two reflectors at large depths (2100 km)
in the southern Pacific might sample the edge
of the slow anomaly in the central Pacific.
Precursors from 2 events define these points.
Energy detected from shallow depths is most
likely related to scattering in the oceanic
lithosphere and the uppermost oceanic mantle.
This is similar to the earlier work from the
1960s and 1970s. 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.
-4
P
60
120
source
0
10
110
Figure 7: Example waveforms of PP
and its precusors for event shown in
Figure 4. Shown are beam traces
calculated for the measured slowness
and backazimuth. Traces are numbered
according to precursor numbers in
Figure 4. The bottom trace shows the
beam trace for PP. The reflections show
different polarities, but deeper reflectors
generally show inverse polarity to PP.
4
2
200
0
6
100
P*
730 - 830 km
time
120
10 s
i
660 - 730 km
P
110
or
s
r
u
ec
r
p
P
600 - 660 km
500
90
PP
500 - 600 km
Reflector Depth [ km]
100
reflection
PC1
d=805 km
array
10
90
dQ
PP
80
PC2
d=27 km
Figure 8: Schematic of the backtracing
algorithm used to locate the origin of the
scattered energy. The precursors generally
show slowness and backazimuth different
from PP. From the array a ray is traced
through a 1D-Earth model using the
measured slowness and backazimuth.
Traveltimes for each point along this path are
also calculated. The possible reflector must
be located somewhere along this ray. For
each point along this path, starting at the
surface reflection point, rays are traced back
to the source. Traveltime is calculated for the
complete path from source to virtual reflector
to array until the measured PP-precursor
differential traveltime is fit. The resulting best
fitting point is defined by latitude, longitude
and depth in the 1D Earth model and
represents the most likely origin of the
scattered energy.
70
80
1000
We use the frequency-wavenumber analysis (fk-analysis) 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 (Figure 7). We estimate the differential travel
time uncertainty to be 1s.
Using a backtracing algorithm it is possible to locate the likely origin of the precursors (Figure 8). We fit all three
measurements (slowness, backazimuth and traveltime) and find the likely reflector by raytracing through a 1D Earth model
(IASP91). We assume that the energy is P scattered to P (P*P). We also tested variations such as S*P and PP*P or P*PP, but
these combinations did not fit the measurements and we can exclude them from our analysis.
PC3
d=604 km
400 - 500 km
PP
60
8
1600
400
PC4
d=91 km
SKKP
50
0
Pdiff
75
Figure 6: Example 4th-Root vespagram (slant-stack) calculated for event shown in Figure 4.
Shown is energy as function of time and slowness (incidence angle). A Full time window from
Pdiff to PKiKP. Arrival times and appropriate slownesses (for PREM ) are marked. B Zoom into
precursor time window. Four precursors were detected for this event (marked as yellow stars).
They generally show smaller slownesses than PP (but larger than P). For reference the slowness
and traveltime for the underside reflection from the 210-km discontinuity is also given. The close
arrival time of the two phases around precursor 3 poses a problem for the fk-analysis since it
measures the average slowness of the phases.
Figure 5: A Beam trace for PP slowness
and theoretical backazimuth. Arrival times
for Pdiff , PP and PKiKP are marked. Note
the increase of energy ~40 s before PP. B
Envelope of the trace shown in A. The
increase of energy seems to consist of some
discrete arrivals in a higher energy
background.
40
0
-1
1800
0.0
30
120
Time [ s ]
2000
20
110
100 s
200
12
10
100
100
300 - 400 km
A
90
0
0
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 mantle.
However, additional precursory arrivals that are not related to reflections off the global discontinuities can be detected in
short-period recordings (Figure 1). 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 (Figure 2). 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 (Figure 3-6). Using a frequency-wavenumber analysis we
are able to measure the full slowness vector of the precursory energy (Figure 7). Using this information we can trace the energy
back to the likely position of the scatterer or reflector (Figure 8) by fitting slowness (u), backazimuth (Q) and differential
traveltime (DT) 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
(Figure 9). 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 (Figure 9 and 11). It is likely that the PP precursors originate from
reflections off the former oceanic crust transported to depths of up to 1000 km due to subduction beneath the Mariana and
Tonga-Fiji subduction zone (Figure 10).
80
1
200 - 300 km
Ehime
University
Introduction / Overview
Precursors
2
PKiKP
100 - 200 km
slowness
-10
100 s
DI41A-1254
0 - 100 km
slowness
0
PKiKP
UC
Santa Cruz
slowness
10
PP
http://geophysics.asu.edu
-2
0
20
Pdiff
3Geodynamics Research Center, Ehime University, Matsuyama, Japan
-2
0
30
2Dept. of Earth Sciences, University of California Santa Cruz, Santa Cruz CA, USA
-2
0
40
1Dept. of Geological Sciences, Arizona State University, Tempe AZ, USA
Slowness (sec/deg)
50
We detected 116 precursors from the 21 events. 110 of these allowed a successful backtracing to its orign, while 6 precursor
measurement 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 (Figure 9
and 11). The reflectors form sheetlike structures beneath two subduction zones (Figure 10).
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 (Figure 12).
Sebastian Rost1, Edward J. Garnero1, Allen McNamara1, Quentin Williams2 and Dapeng Zhao3
Time (sec)
YKA
A
YKA Array elements
60
Figure 3: Dataset for this
study. Earthquakes from the
SW Pacific and Tonga-Fiji
(red stars) are recorded at
the small aperture (20km),
short-period
(1Hz),
Canadian Yellowknife array
(YKA - blue triangle). PP
raypaths (thin gray lines)
and PP midpoint reflection
points (black circles) are also
shown. For orientation, the
plate boundaries are shown
as thin black lines. 21 events
are used with epicentral
distances from 95 to 105 deg
and depths from 10 to 60 km.
Their magnitudes range from
mb = 5.5 to 6.6.
Figure 4: Data example for
event on 02 Apr 1996,
h=41km, D=100.14 deg.
Recordings from all available
YKA elements are shown.
Data were filtered with a
bandpass
with
cut-off
frequencies of 0.5 and 1.4 Hz.
A P to PKiKP time window
that includes PP and its
precursors. B Zoom into the
PP precursors time window is
shown in the lower panel,
showing the coherency of the
arrivals in the precursors
wavefield. No underside
reflection energy from the
210- and 410-km discontinuities is observed in
these data (traveltimes for
these phases are given for
reference).
Results
Detection of subducted lithosphere in the
midmantle
from asymmetric PP precursors
Figure
11:
Vertical cross section
through the Ritsema
tomography model in
the Mariana vicinity.
The location of the
cross section is
shown in the globe
insert. Blue colors
denote high seimic
-4 -2 0 2 4
velocities, and red
δVS [%]
colors low seismic
velocities relative to a 1D Earth model. The line shows the 0.7%
isocontour. Red circles show the location of the detected reflectors
(depths > 400 km) in this region projected onto the cross section.
References
Figure 10: Perspective view of the distribution of scatterers (see also
computer demonstration). Box includes the mantle (from 0 to 2891 km). Shown
on top is topography for reference. The depth of the scatterers is colored
according to depth. A View from South to North (West is on the left hand side).
Viewpoint elevation is approximately 15 deg. B View from North to South (West
on the right hand side). C View from West to East (North on the left hand side).
D View from East to West with negative (-15deg) viewpoint elevation. E View
from above (North to the top; West to the left). Sources are shown as circles
(depth color-coded). Circle diameter shows event magnitude. F View from
below (South is at top and west to the left).
Two main deep groups of scatterers can be detected in close proximity to the
Mariana and Tonga-Fiji subduction zones. The correlation of these deep
scatterers to fast seismic velocities can be better seen in Figure 9. The
shallow scatterers do not show any geographic coherency and are most likely
related to small-scale heterogeneity in the oceanic lithosphere. The scatterer
locations are generally off great circle path and within a distance of 25 deg
from the sources. Only a few receiver side scatterers have been found. The
Mariana group is in close proximity to detections of heterogeneities in the mid
mantle as shown in Figure 12.
30
20
10
120
130
0 200
140
600
1000
1400
Reflector Depth
150
1800
km
Figure 12: Zoom into the Mariana
vicinity. Reflectors from P*P are shown
as squares colored according to depth.
Slab positions in the upper mantle
(RUM-model) are shown as lines.
Earlier work finds evidence for
structure in the mid mantle east of this
region. Symbols: stars - Niu et al
(2003); circle - Castle and Creager
(1999); diamond - Kruger (2001);
hexagon: Kaneshima and Helffrich
(1998); triangle - Kaneshima and
Helffrich (2003). These locations are
all deeper than the scatterers of this
study.
Castle, J.C. and K.C. Creager (CC99), A steeply dipping discontinuity in the lower mantle beneath Izu Bonin, J. Geophys. Res. 104, 7279-7292, 1999.
Kruger, F. et al. (K01), Mid mantle scatterers near the Mariana slab deteted with a double array method, Geophys. Res. Lett., 28, 667-670, 2001.
Kaneshima, S. and G. Helffrich (KH98), Detection of lower mantle scatterers northeast of the Mariana subduction zone using short-period array data, J. Geophys. Res., 103, 4825-4838, 1998.
Kaneshima, S. and G. Helffrich (KH03), Subparallel dipping heterogeneities in the mid-lower mantle, J. Geophys. Res., 108, doi:10.1029/2001JB001596, 2003.
King, D.W., R.A.W. Haddon and E.S. Husebye, Precursors to PP, Phy. Earth Planet. Int., 10, 103-127, 1975
Niu, F. H. Kawakatsu and Y. Fukao (N03), Seismic evidence for a chemical heterogeneity in the midmantle" A strong and slightly dipping seismic reflector beneath the Mariana subduction zone, J. Geophys. Res.,
108, doi:10.1029/2002JB002384, 2003
Ritsema, J. and H.J. van Heijst, New constraints on the P velocity structure of the mantle from P, PP and PKPab. Geophys. J. Int., 149, 482-489, 2002.
Rost, S. M. Thorne, and E.J. Garnero, Imaging the short-period wavefield using stacked Stacks, in preparation, 2005.
Wright, C., Array studies of seismic waves arriving between P and PP in the distance range 90 to 115 deg, Bull. Seismol. Soc. Am., 62, 385-400, 1972.
Wright, C. and K.J. Muirhead, Longitudinal waves from the Novaya Zemlyz nuclear explosion of October 27, 1966, recorded at the Warramunga array, J. Geophys. Res., 74, 2034-2047, 1969.