Transcript Shear wave splitting beneath the East European Craton

1
Methods and applications of
shear wave splitting
An example of the East European Craton
Soutenance de Thèse
Andreas Wüstefeld
27 Sept. 2007
2
Outline
 Introduction
 Part 1: Splitlab
A graphical interface for the splitting process
 Part 2: Null criterion
Synthetic test reveals characteristic differences of two splitting techniques
 Part 3: Splitting Database
Access splitting measurements publications online
 Part 4: The East European Craton
Application to stations on the old EEC
3
Geodynamics: study of deformation
[Illustration by Jose F. Vigil. USGS]
4
Causes of seismic anisotropy
Horizontal layering
Upper and lower crust, transition zone, D ’’
Vertically aligned cracks
Crust
Alignment of minerals
Lower crust, upper mantle,
inner core
5
What causes mineral alignment?
“Vertically coherent deformation”
The last tectonic deformation is frozen-in into the lithosphere
“Simple asthenospheric flow”
Mainly present day mantle flow causes anisotropy
6
Shear-wave splitting:
the phenomenon
If initial polarisation coincides with a
anisotropy axis, the shear wave is
not split (Null case)
Anisotropic layer
Invert the splitting by grid-searching for
combination of fast axis and delay time
which best removes the splitting
7
Shear-wave splitting:
the techniques
Remove splitting:
Radial
1. Minimum Energy on Transverse: Remove transverse Energy
2. Rotation-Correlation: Searching for maximum correlation
3. Eigenvalue criteria: Searching for most linear particle motion
Transversal
8
European Anisotropy
% velocity perturbation
Tomography of Europe at 150km depth (Debayle et al., Nature, 2005)
Splitting results of various authors
Is there mantle flow around the East European Craton?
How does the anisotropy continue beneath the Craton?
9
Part I
Shear-wave splitting in Matlab
10
Configuration
- A shear wave splitting environment in Matlab
11
Seismogram Viewer
Select splitting window and filter
Minimum Energy
Rotation Correlation
SKS
12
Diagnostic Viewer
13
ResultViewer
www.gm.univ-montp2.fr/splitting
Splitlab efficiently
compares different
techniques
[Wüstefeld et al., in press]
14
Part II
Synthetic test
Null Criterion
15
Null Criterion
Synthetic test
 Comparison of two splitting techniques
Rotation correlation method
Minimum energy method
90
45
45
0
0
-45
-45
fast axis
90
delay time
-90
-90
-45
0
45
90
-90
-90
4
4
3
3
2
2
1
1
0
-90
-45
0
45
Backazimuth
90
0
-90
Model parameters:
Fast axis: 0°
Delay time: 1.3sec
SNR:
15
-45
-45
0
45
90
0
45
Backazimuth
90
16
Null Criterion
Why is there a 45° difference?
 The Rotation-Correlations seeks for maximum wave-form similarity
 If the initial energy on Transverse is small (Null case), the
maximum correlation is found for a test system 45° rotated:
Q' cos  sin  Q Q cos
T '   sin  cos   T    Q sin  
  
   

 This also results in small delay time estimates
17
Null Criterion
Synthetic test
 Comparison of two splitting techniques
Rotation correlation method
Minimum energy method
90
45
45
0
0
-45
-45
fast axis
90
delay time
-90
-90
-45
0
45
90
-90
-90
Model parameters:
Fast axis: 0°
Delay time: 1.3sec
SNR:
15
-45
0
45
90
4
4
3
3
Is this a common feature?
2
2
5 SNR between 3 and 30
1
1
7 delay times between 0 and
2 sec
0
-90
-45
0
45
Backazimuth
90
0
-90
-45
0
45
Backazimuth
90
18
Null Criterion
Null criterion
3185 measurements:
5 SNR between 3 and 30
7 delay times between 0
and 2 sec
NULL:
|ΦSC - ΦRC| > 22.5º
dtSC/dtRC ≤ 0.3
The comparison of two techniques objectively and automatically
- Detect Nulls
- Assign a quality to the measurement
[Wüstefeld & Bokelmann, BSSA, 2007]
19
Automated splitting?
 Perform splitting to a set of test windows around
theoretical SKS arrival
=> No manual phase picking needed!
 Skip Null measurements
 Stack (non-normalized) energy map
[Wolfe & Silver, 1998]
 Repeat for different filters!
 Determine global energy minimum (of each event)
20
Example station ATD
330 earthquakes
9 start times
6 end times max = 162
3 filter sets
}
Barruol & Hofmann [1999]
Automatically detected global minimum
Φ = 48°; dt = 1.59sec
Φ = 42°; dt = 1.6sec
21
Automated splitting
 Possible with SplitLab
 Reduced processing time
 Objective and repeatable
 Uniform database
22
Part III
Shear wave splitting database
23
Shear wave splitting database
http://www.gm.univ-montp2.fr/splitting/DB
24
Shear wave splitting database
http://www.gm.univ-montp2.fr/splitting/DB
ECH
48.216
7.158
85
0.88
Barruol, G., Hoffman, R.
Upper mantle anisotropy beneath the Geoscope stations
J. Geophys. Res.
1999
104
10757-10773
http://www.gm.univ-montp2.fr/PERSO/barruol/
Silver & Chan method
25
Shear wave splitting database
http://www.gm.univ-montp2.fr/splitting/DB
ECH
48.216
7.158
85
0.88
Barruol, G., Hoffman, R.
Upper mantle anisotropy beneath the Geoscope stations
J. Geophys. Res.
1999
104
10757-10773
http://www.gm.univ-montp2.fr/PERSO/barruol/
Silver & Chan method
26
Global mean: 1sec
SKS database:
 2286 measurements
 122 references
27
Comparison with surface waves
Predicted splitting parameters
28
Coherence
of predicted
and observed
splitting
 Good global
coherence
 Splitting in western
US occurs above
200km depth
 In Central Europe
best coherence at
200-350km km
depth interval
29
Part IV
- The real world -
Shear wave splitting beneath the
East European Craton
30
The East European Craton
31
Results
16 stations analyzed
Delay times between
0.4 sec and 1.1 sec
Variable fast orientations,
but similar within a block
32
Comparison with other datasets
33
Comparison with other datasets

Weak correlation with plate
motion vectors

Anisotropy not related to
present day asthenospheric
processes

Regionally good correlation with
predicted splitting

Short scale variations, but
consistent within a block

Anisotropy within the
lithospheric blocks
34
Polish-Lithuanian-Belarus Terrane
[after Bogdanova et al., 2006]
35
Excursus:
Magnetic structures and seismic anisotropy
Magnetic structures
reflect tectonic events.
Rocks are magnetic up to a temperature
of 580° (Currie Temperature)
This temperature is generally
reached at depths close to the moho
The crustal contribution to
splitting is presumeably small (<0.2sec)
Parallelism between magnetic structures and
fast orientations indicates that observed
anisotropy is in the lithosphere
36
Polish-Lithuanian-Belarus Terrane
NE51
PUL
TRTE
NE52
Fast orientations follow magnetic
structures
NE53
SUW
Lithospheric anisotropy
Magnetic intensity anomaly
37
Fennoscandia
Results in Fennoscandia are in good
agreement with the SVEKALAPKO
experiment
Continous rotation of fast
orientations supports single-block
hypothesis
[after Vecsey et al., 2007]
38
Ural mountains
ARU
AKTK
39
Ural mountains
Magnetic intensity map
ARU and AKTK show fast orientations
perpendicular to trend of mountain chain.
Distance to deformation front might
indicate out of reach for compressive
deformation of orogeny.
Anisotropy possibly related to
ancient subduction processes
40
Sarmatia
No clear magnetic structures
Fast orientations in the west align
with TTZ
Lateral erosion due to mantle
flow along western edge of the
craton?
[modified after Thybo et al., 2003]
41
The EEC shows
 Weak correlation with plate motion vectors
 Variable fast orientations, but consistency within a tectonic block
 Short scale variations across the borders of the blocks
 Rather good correlation of (crustal) magnetic anomalies and
(upper mantle) seismic anisotropy
Anisotropy is frozen-in into the lithosphere
 Stations in the West align with TTZ
Mantle flow around the craton?
42
Conclusions
 Splitlab:
-
User friendly, efficient
Simultaneous comparison of methods
 Null criterion
-
Detect Nulls and assign quality
Allow for automatic splitting
 Splitting database
-
Central and interactive depository of splitting publications
Generally good correlation with surface waves
 East European Craton
-
Weak anisotropy (delay times between 0.4 - 1.1sec)
Comparison of splitting with magnetic structures possible
Lithospheric frozen-in anisotropy
Possible mantle flow around the craton
43
Thank you …
44
Can the depth of splitting be constrained?


Lines: Comparisson with predicted splitting orientations [0° < misfit < 90°]
Background: relative predicted splitting [0 < strength < 1]
45
Null Criterion
Model Delay time: 0.7sec
45
90
-45
0
45
90
90
45
0
-45
Rotation correlation method
-45
0
45
90
-45
0
45
90
45
0
-45
Minimum energy method
90
delay time
0
-90
4
90
-45
0
45
90
delay time
SNR = 3
45
2
0
-90
4
-45
0
45
90
delay time
0
0
2
2
0
-90
4
-45
0
45
90
delay time
-45
-45
SNR = 5
90
4
SNR = 10
45
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
Delay time comparison
SNR = 20
Fastaxis
0
Fastaxis
-45
Fastaxis
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
Fastaxis
SNR = 20
SNR = 10
SNR = 5
SNR = 3
Fast axis comparison
2
0
-90
90
45
0
-45
Rotation correlation method
4
2
0
-90
4
-45
0
45
90
-45
0
45
90
-45
0
45
90
2
0
-90
4
2
0
-90
4
2
0
-90
45
0
-45
Minimum energy method
90
46
Null Criterion
Model Delay time: 1.3sec
45
90
-45
0
45
90
90
45
0
-45
Rotation correlation method
-45
0
45
90
-45
0
45
90
45
0
-45
Minimum energy method
90
delay time
0
-90
4
90
-45
0
45
90
delay time
SNR = 3
45
2
0
-90
4
-45
0
45
90
delay time
0
0
2
2
0
-90
4
-45
0
45
90
delay time
-45
-45
SNR = 5
90
4
SNR = 10
45
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
Delay time comparison
SNR = 20
Fastaxis
0
Fastaxis
-45
Fastaxis
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
Fastaxis
SNR = 20
SNR = 10
SNR = 5
SNR = 3
Fast axis comparison
2
0
-90
90
45
0
-45
Rotation correlation method
4
2
0
-90
4
-45
0
45
90
-45
0
45
90
-45
0
45
90
2
0
-90
4
2
0
-90
4
2
0
-90
45
0
-45
Minimum energy method
90
47
Null Criterion
Model Delay time: 2.0sec
45
90
-45
0
45
90
90
45
0
-45
Rotation correlation method
-45
0
45
90
-45
0
45
90
45
0
-45
Minimum energy method
90
delay time
0
-90
4
90
-45
0
45
90
delay time
SNR = 3
45
2
0
-90
4
-45
0
45
90
delay time
0
0
2
2
0
-90
4
-45
0
45
90
delay time
-45
-45
SNR = 5
90
4
SNR = 10
45
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
Delay time comparison
SNR = 20
Fastaxis
0
Fastaxis
-45
Fastaxis
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
90
45
0
-45
-90
-90
Fastaxis
SNR = 20
SNR = 10
SNR = 5
SNR = 3
Fast axis comparison
2
0
-90
-45
0
45
90
Rotation correlation method
4
2
0
-90
4
-45
0
45
90
-45
0
45
90
-45
0
45
90
2
0
-90
4
2
0
-90
4
2
0
-90
-45
0
45
Minimum energy method
90
48
Shear-wave splitting
Theory:
The resulting radial and transverse components after anisotropic layer are
u~Radial (t )  u R (t  t / 2) cos2   u R (t  t / 2) sin 2 
u~
(t )   1 sin 2 [u (t  t / 2)  u (t  t / 2)]
Transversa l
2
R
R
uR,T = initial radial and
transverse particle motion
~
u
Radial ,Transversa l = particle
motion after splitting
α = angle between fast
direction and backazimuth
δt = delay time between fast
and slow component
The splitting can be inverted by a search for a singular covariance matrix

Cij ( , t )   u~i ( , t )u~ j ( , t  t )dt; i, j  Radial, Transverse

Search for combination of fast axis and delay time
which gives most singular Covariance matrix to
remove the splitting of the shear wave
49
Null Criterion
Data example LVZ
50
Null Criterion
10º; 1.1sec
Result of LVZ:
Minimum Energy
90
90
45
45
LVZ
Fast axis
Rotation-Correlation
0
-45
-45
-90
-90
0
45
90
135 180 225 270 315
good splitting
delay time
0
360
fair splitting
0
45
90
135
weak
4
4
3
3
2
2
1
1
0
180 225 270 315
good Null
360
fair Null
0
0
45
90
135 180 225 270 315
Backazimuth
360
0
45
90
135
180 225 270 315
360
Backazimuth
44 events
51
Splitting projected to depth of CMB
54
Polish-Lithuanian-Belarus Terrane
55
East European Craton
(after Wikipedia)

The East European craton is the core of the Baltica proto-plate and consists
of three crustal regions/segments: Fennoscandia to the northwest, VolgoUralia to the east, and Sarmatia to the south. Fennoscandia includes the
Baltic Shield (also referred to as the Fennoscandian Shield) and has a
diversified accretionary Archaean and Early Proterozoic crust, while Sarmatia
has an older Archaean crust. The Volgo-Uralia region has a thick
sedimentary cover, however deep drillings have revealed mostly Archaean
crust. There are two shields in the East European Craton: the
Baltic/Fennoscandian shield and the Ukrainian shield. The Ukrainian Shield
and the Voronezh Massif consist of 3.2-3.8 Ga Archaean crust in the
southwest and east, and 2.3-2.1 Ga Early Proterozoic orogenic belts.

The intervening Late Palaeozoic Donbass Fold Belt, also known as part of
the Pripyat-Dniepr-Donets aulacogen, transects Sarmatia, dividing it into the
Ukrainian Shield and the Voronezh Massif.
56
The thick & cold EEC
% velocity
perturbation
(fast)
(slow)
Surface wave tomography
after Debayle et al [2005]
57
Excursus:
Magnetic field and seismic anisotropy
Depth of the 550°C isotherme
(after Artemieva [2006])