Teleseismic imaging of the Slave craton and
implications for the assembly of cratonic lithosphere
by
Chin-Wu Chen
M.Sc., Geophysics, National Taiwan University, Taipei, Taiwan, 2000
B.Sc., Earth Science, National Central University, Chungli, Taiwan, 1998
Submitted to the Department of Earth, Atmospheric and Planetary Sciences
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
February 2010
ARCHIVES
© Massachusetts Institute of Technology 2010. All rights reserved.
MASSACHUSETTS INST TfUTE
OF TECHNOLOGY
MAY 0 5 2010
A uthor ................
...........................................
LIBRA R IES
Department of Earth, Atmospheric and Planetary Sciences
November 2, 2009
Certified by .
.......
Stephane Rondenay
Associate Professor of Seismology
Thesis Supervisor
A ccepted by ....................... ).........................
Maria T. Zuber
E.A. Griswold Professor of Geophysics
Head, Department of Earth, Atmospheric and Planetary Sciences
2
Teleseismic imaging of the Slave craton and implications for the
assembly of cratonic lithosphere
by
Chin-Wu Chen
Submitted to the Department of Earth, Atmospheric and Planetary Sciences
on November 2, 2009, in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy
Abstract
In this dissertation, I investigate the assembly and evolution of the Archean cratonic lithosphere by using two complementary seismological methods to image the lithospheric structure of the Slave craton in Canada. First, I perform surface wave tomographic inversions
to constrain the depth dependence of Rayleigh wave phase velocity, shear wave velocity,
and azimuthal anisotropy of the Slave cratonic lithosphere. The tomographic images reveal
high shear wave velocities associated with a particularly depleted, cold, and unperturbed
Archean cratonic lithosphere. Furthermore, the inversions reveal distinct anisotropic domains in the crust, the lithospheric mantle, and the sub-lithospheric mantle. These results
reflect the evolutionary history of the cratonic lithosphere. Secondly, I image seismic discontinuities in the lithosphere using receiver-function analysis of converted P-to-S waves. The
resulting seismic profile shows a pronounced low velocity discontinuity at -100 km depth
beneath the central Slave craton. This seismic discontinuity shows striking spatial correlation with both an electrical conductive anomaly derived from magnetotelluric sounding,
as well as a petrologically-defined ultra-depleted layer. The synthesis of coincident seismic,
electrical, and petrological evidence supports that this geophysical and petrological boundary represents a compositional interface marked by alteration minerals. I suggest that this
mineralization resulted from relict metasomatism associated with an Archean subduction
event, which played an important role in the assembly of the Slave craton. Finally, to improve the efficiency and automation of receiver function calculation and data preprocessing
workflow, I develop an application of an array-conditioned deconvolution technique for effectively processing large amounts of seismic array data. I demonstrate that this technique is
readily applicable to teleseismic array data. This technique is especially effective in turning
noisy traces from earthquakes with smaller magnitudes into usable data.
Thesis Supervisor: Stephane Rondenay
Title: Associate Professor of Seismology
4
Acknowledgmentst
I used to think that the acknowledgments section was the easiest part of the thesis to write.
Reflecting upon the past six years of study and life at MIT, I find myself overwhelmed and
humbled by all the help, support, and love I have received; it made this thesis possible and
enriched my life so much that I don't know where to start expressing my gratitude. Now I
realize writing acknowledgments is no easy task.
I am deeply indebted to my advisor, Stephanie Rondenay. He bravely took me under
his wing six years ago and convinced me that doing fieldwork is fun, when I wanted to be
an armchair seismologist working on the core-mantle boundary. While my physical fitness
was intensively challenged, I experienced some of the most memorable and unique moments
in my life when in the field. Yes, the all-time favorite was seeing the Christmas-tree-like
lights from the GMC Yukon after snowshoeing for hours on the darkened Big Skidder Hill in
Washington State. He showed me how doing serious science does not have to conflict with
enjoying life, when we relaxed with a beer in the Black Knight Pub in Yellowknife after a long
day, sampled whisky in snow-stormed Seattle before embarking on a tough field season, or
jumped into the glassy Aegean Sea under the Mediterranean sunshine after rushing through
narrow streets on the island of Syros to retrieve a seismic station. I hold these memories
dear.
Stdphane's encouragement and infinite patience were instrumental in the completion of
this dissertation. I would not have come this far without his insight, openness, generosity,
and sense of humor. He taught me to be meticulous on details yet always keep the big picture
in mind. His enthusiasm is contagious. His deep knowledge of seismological techniques and
broad interests in earth sciences led the multi-disciplinary approach in my research. He has
more than the characteristics needed to be a successful scientist and mentor; he is a genuinely
good guy and good friend. I hope this thesis marks not the end but a continuation of our
collaboration and friendship.
Sam Bowring was my second general project advisor and my thesis committee chair. He
n MIT
sty (1'*andv Ires1(ea(Irch (Ie suppolted by a;I
I
F(mdatin grants EAR-04090aUd EAR-0544996.
i
sPreiden t ial Fel
lship an(d Nationl
Science
broadened my horizons by stimulating me to explore the geological and geochemical aspects
of the Slave craton and the evolution of the Earth in general. The breadth and depth of
his knowledge is unparalleled. And lie never wears a bug net in the field! He was also a
most serious and generous reader, offering insightful comments and discussions along the
way, which reshaped my understanding of craton assembly.
I started working with Doug Miller during my 2008 summer internship at SchlumbergerDoll Research, on the deconvolution project that eventually became an important part of my
thesis. Working with Doug has been a privilege and a pleasure. His mathematical insight
always pointed to the core of a problem. I appreciate very much the many afternoons we
spent in front of the computer scratching our heads over the behavior of the wiggles. I also
enjoyed the Shakespeare and lute songs lie introduced me to. Doug is a sincere mentor and
the ideal of a curious, enthusiastic, and versatile scientist. I look forward to the opportunity
of future continued collaboration.
I owe much appreciation to Hugues Djikpesse, whom I first met at the career fair at
MIT. He later offered me the internship at SDR and was an effective and caring supervisor,
always ready to help. His thoughtful suggestions greatly improved the clarity of our paper.
I also thank Jakob Haldorsen for his help with my presentation, insightful discussions on
deconvolution, and heartfelt support and encouragement.
Many thanks to Rob van der Hilst for his helpful suggestions that improved my thesis. He
always welcomed my questions and gave useful advice. Thanks also go to Alison Malcolm.
I benefitted much from her stimulating questions on my work.
David Snyder is the co-leader of our fieldwork in the Slave province. I learned much from
his intensive experience in the field and in logistics management. His input to our papers
and open-minded discussion of new ideas are much appreciated. I thank Rob Evans for the
exciting collaboration on the comparison of receiver function image and the magnetotelluric
model. His input was critical to this work. Thanks also go to Wiki Royden for her help on
my second general project, Lindy Elkins-Tanton for helpful discussion on melting of basalt,
and Nafi Toks6z for his encouragement.
Science becomes more fun when you are not doing it alone. In my first few years of Ph.D.
study, I was fortunate to interact with the seismology group at Brown University. Dayanthie
Weeraratne and Yingjie Yang generously spent much time helping me understand and run
the surface wave inversion. Dayanthie keeps sharing with me her thoughts and experience
regarding research and life. Occasional consultation with Donald Forsyth and Aibing Li were
also appreciated. In addition, I thank Sonja Aulbach, Max Moorkamp, Fiona Darbyshire for
sharing their results with me, John Carson and Peter Holman of Gelogical Survey of Canada
for giving me access to the Slave craton radiometric data, and Laura Mackenzie and Zhu
Zhang for their help with Antelope and Cascadia dataset.
The big family of students and postdocs in EAPS is an amazing source of support and
friendship. In particular, I thank Huajian Yao for his last minute and generous help with
the neighborhood algorithm and many discussions on surface waves, Fred Pearce for sharing
stimulating ideas and his codes, Lili Xu for helping me get started with receiver functions,
Jenny Suckale for the wonderful adventure in Greece, Xander Campman for his pleasant company in the Slave, Ping Wang and Chang Li for many discussions and help on seismology
in general, and everyone above for numerous fun conversations. I thank Einat Lev, Erwan
Mazarico, James Dennedy-Frank, Sarah Johnson, and Krystle Catalli for being awesome
officemates at the original 521 and 610, Christy Till and Jay Barr for being my petrology
consultants, and Danny MacPhee and Becky Flowers for advice on my second general project.
In addition, I thank Maureen Long, Emily van Ark, Kyle Straub, Florian Bleibinhaus, Caroline Beghein, Rongrong Lu, Samantha Grandi, Darrell Coles, Burke Minsley, Andrea Llenos,
Noah McLean, Scott Burdick, David Forney, Qin Cao, Kang Hyeun Ji, Aleksandra Hosa,
Jiangning Lii, Shane McGary, Neil de Laplante, Andrey Shabelansky, Mike Krawczynski,
Min Chen, Gabi Melo, Junlun Li, Alejandra
Quintanilla
Terminel, Xin Zhan, Hui Huang,
Bongani Mashele, Xuefeng Shang, Haijiang Zhang, Tim Seher, and Sadi Kuleli for being part
of the family that I am very fond of. Furthermore, the administrative team Terri Macloon,
Beth MacEachran, Carol Sprague, Roberta Allard, Jacqui Taylor, Sue Turbak, and Vicki
Mackenna make EAPS such a pleasant and orderly place to work and study. Joe Hankins is
the nicest and most helpful librarian. Linda Meinke and Scott Blomquist are the IT squad
without whom my computer would probably not listen to me.
I appreciate Debbie Levey and her family for warmly welcoming me to join them at
various holiday occasions over the years in which I learned about American culture and
tradition. I also thank the instructors in the MIT Writing Center, especially Susan Spilecki,
for their help on my thesis writing.
Special thanks are due to two of my true comrades and best friends, Yang Zhang and
Taiwei Wang. Yang is a fellow student in EAPS and my housemate for the past five years.
We shared all the ups and downs in research and in life. We cheered each other up and
also procrastinated together. He is a man of integrity. Yang, I know you will be great in
whatever you choose to do. When you get rich, don't forget to buy me a winery in Napa!
Taiwei is a sincere and caring person, and was courageous enough to pursue his dreams of
travel in the middle of his doctoral study. Now he has become a successful manager in the
travel business. Taiwei, when I need to go check out my winery, give me free flight tickets!
Outside EAPS, I have been fortunate to have a big group of great friends, collectively
known as 'Salon Chaos,' with whom I share laughs and worries, thoughts and nonsense.
Without them my life in Boston would not have been so full and so memorable that leaving
becomes so difficult; they are: I-Fan Lin, Mo-Han Hsieh, Chia-Ling Pai, Ling Chao, Lisa
Kang, I-Tsung Tsai, Hung-Chi Miao, Tsung-Ren Huang, Li-Yuan Wang, Shwu-Jen You,
Hsien-Ho Lin, Ching-An Meng, Pei-Ru Jian, Ching-Ting Pi, Chia-Hui Chen, Jiao Guo,
Liang-Yu Chen, Wan-Yi Huang, Chih-Chao Chuang, Yu-Chen Yeh, Hsin-Pei Shi, Carl Hu,
Su Wu, Hsiao-Shan Lin, Iris Chang, Wei-Chuan Shi, Wan-Chen Lin, and Bernard Yen. I
cherish our friendships and hope to continue sharing with you my journey of life. I am
also grateful for the encouragements from my former advisors Ban-Yuan Kuo and Ling-Yun
Chiao, and colleagues Shu-Chuan Lin, Wen-Tzong Liang, Teh-Ru Song, and Wen-Che Yu.
My deepest gratitude is due to my family. My treasured parents Ching-Ying Cheng and
Roger Chen, my brother Chin-Hsueh, Aunt Angela, and my grandmother Su-Jen Hsii are
always there looking after me and support my choices although they did wonder what was
taking me so long. I also extend my thanks to Yow-Sheue Chen. Finally, my great appreciation to Julia Shu-Huah Wang, my anchor and my muse, whose love, kindness, funniness
and silliness make my life full of joy and hope.
Contents
17
1 Introduction
1.1
Archean cratons and the cratonic lithosphere . . . . . . . . . . . . . . . . . .
17
1.2
The Slave craton and the POLARIS-MIT portable seismic array . . . . . . .
19
1.3
Outline of the thesis
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
1.4
F igure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
2 Upper mantle seismic structure beneath the Slave craton from Rayleigh
23
wave tomography
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
2.2
Data and processing
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
2.3
Phase velocity inversion
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
2.4
Phase velocity results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
32
2.4.1
Inversion for 1-D phase velocities
. . . . . . . . . . . . . . . . . . . .
33
2.4.2
2-D phase velocity maps . . . . . . . . . . . . . . . . . . . . . . . . .
33
2.5
Anisotropy Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
34
2.6
1-D S-wave velocity structure
. . . . . . . . . . . . . . . . . . . . . . . . . .
35
2.6.1
1-D velocity profile . . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
2.6.2
Synthetic tests of 1-D models . . . . . . . . . . . . . . . . . . . . . .
36
2.6.3
Damping parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . .
37
Inversion for 3-D S-wave velocity structure . . . . . . . . . . . . . . . . . . .
37
2.7.1
The NA optimization . . . . . . . . . . . . . . . . . . . . . . . . . . .
37
2.7.2
S-wave velocity lateral variations
. . . . . . . . . . . . . . . . . . . .
39
2.7
2.8
D iscussion . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.8.1
The cold cratonic block in the central Slave craton
2.8.2
The thickness of the cratonic lithosphere . . . . . . . . . . . . . . . .
40
2.8.3
Depth-dependent azimuthal anisotropy . . . . . . . . . . . . . . . . .
42
2.8.4
A mid-lithosphere low velocity layer? . . . . . . . . . . . . . . . . . .
45
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
45
2.10 T ab le . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
46
2.11 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
2.9
3 Geophysical detection of relict metasomatism from an Archean (ca 3.5 Ga)
subduction zone
63
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . .
64
3.2
Seismic observations
. . . . . . . . . . . . . . . . . . .
65
3.2.1
D ata . . . . . . . . . . . . . . . . . . . . . . . .
65
3.2.2
M ethod
3.2.3
Single station profiles . . . . . . . . . . . . . . .
3.2.4
Common-conversion-point (CCP) profiles . . . .
3.2.5
The robustness and nature of the central Slave s eismic discontinuity
3.3
3.4
. . . . . . . . . . . . . . . . . . . . . .
Seismic vs. magnetotelluric observations
. . ......
3.3.1
The magnetotelluric observations . . . . . . . .
3.3.2
Coincident seismic and electrical anomalies . . .
Causes of the seismic and electrical anomalies . . . . .
3.4.1
Thermal anomaly . . . . . . . . . . . . . . . . .
. . . .
73
3.4.2
Compositional anomaly . . . . . . . . . . . . . .
. . . .
74
3.4.3
W ater . . . . . . . . . . . . . . . . . . . . . . .
. . . .
76
3.4.4
Partial m elt . . . . . . . . . . . . . . . . . . . .
. . . .
77
3.4.5
Sum m ary . . . . . . . . . . . . . . . . . . . . .
. . . .
77
3.5
Implications for the assembly of cratonic lithosphere . .
. . . .
78
3.6
Conclusions . . . . . . . . . . . . . . . . . . . . . . . .
. . . .
80
3.7
Table . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . .
81
3.8
4
Array-conditioned deconvolution of multiple component teleseismic record91
ings
5
81
Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
92
4.2
Methodologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
93
4.2.1
Waterlevel deconvolution . . . . . . . . . . . . . . . . . . . . . . . . .
93
4.2.2
Array-conditioned deconvolution. . . . . . . . . . . . . . . . . . . . .
95
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
97
4.3
Synthetic experiments
4.4
Application to the Slave craton array data . . . . . . . . . . . . . . . . . . . 100
4.5
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.6
Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Conclusions
113
A Tables
117
Bibliography
123
12
List of Figures
1.1
A simplified geological map of the Slave craton.
2.1
Simplified geological map of the Slave craton . . . . . . . . . . . . . . . . . .
47
2.2 Receivers and earthquakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 8
2.3
Great circle paths at 50 s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 8
2.4 Example of dispersion data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 9
2.5
Rayleigh wave vertical sensitivity . . . . . . . . . . . . . . . . . . . . . . . . 50
2.6
Model parameterization
2.7
1-D dispersion curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.8
2-D Phase velocity maps 22-40 s . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.9
2-D Phase velocity maps 50-100 s
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
. . . . . . . . . . . . . . . . . . . . . . 53
. . . . . . . . . . . . . . . . . . . . . . 54
2.10 2-D Phase velocity perturbations
2.11 Anisotropy results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.12 1-D S-wave velocity profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.13 Synthetic tests of 1-D S-wave velocity m odels
. ..... .... ..... ..
57
. .... ..... .... .... ...
58
2.15 S-wave velocity models from NA . . . . . . . . . . . . . . . . . . . . . . . . .
59
2.14 Synthetic tests of damping parameters
2.16 1-D marginals from NA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 0
2.17 S-wave velocity lateral variations
. . .. .... ..... .... .... ... 61
2.18 S-wave velocity perturbation profile . . . . . . . . . . . . . . . . . . . . . . .
3.1
Synthesized map of the Slave craton . . . . . .
3.2
RF profiles of station ACKN . . . . . . . . . .
62
83
3.3
RF profiles of station EKTN . . . . . . . . . . . . . . . . . . . . . . . . . . .
84
3.4
Ps conversion points at 100 km depth . . . . . . . . . . . . . . . . . . . . . .
85
3.5
Receiver-function and magnetotelluric profiles . . . . . . . . . . . . . . . . .
86
3.6
SV- and SH-component receiver-function profiles . . . . . . . . . . . . . . .
87
3.7
Depth and amplitude uncertainty of the central Slave seismic discontinuity .
88
3.8
Subduction m odels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
89
4.1
An example earthquake array data and the deconvolved results . . . . . . . . 104
4.2
Construction of synthetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.3
Summary of synthetic experiments
4.4
Comparison of amplitude spectra of deconvolved traces . . . . . . . . . . . . 107
4.5
Earthquakes and receivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.6
The raw data of four example earthquakes . . . . . . . . . . . . . . . . . . . 109
4.7
The deconvolved data of the four example earthquakes
4.8
The deconvolved data sections of five receivers . . . . . . . . . . . . . . . . . 111
4.9
The deconvolved data plotted as a function of earthquake magnitude
. . . . . . . . . . . . . . . . . . . . . . . 106
. . . . . . . . . . . . 110
. . . .
112
List of Tables
2.1
Model parameters used in the Neighborhood Algorithm . . . . . . . . . . . .
46
3.1
End-member velocity models for uncertainty evaluation . . . . . . . . . . . .
81
A. 1 Seismic stations used in this thesis
. . . . . . . . . . . . . . . . . . . . . . . 118
A.2 Events used in the surface wave tomography . . . . . . . . . . . . . . . . . . 119
A.3 Events used in the receiver-function analysis and the array-conditioned deconvolution
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
A.4 Events used in the receiver-function analysis and the array-conditioned deconvolution (continued) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A.5 Events used in the receiver-function analysis and the array-conditioned deconvolution (continued) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
16
Chapter 1
Introduction
1.1
Archean cratons and the cratonic lithosphere
Archean cratons form the nuclei of Precambrian continental shields and represent remnant
fragments of the oldest continents that once existed on the Earth's surface [Bleeker, 2001].
These cratons have remained stable for billions of years; therefore, they provide a unique
window into the tectonic processes that took place when they were formed. The investigation
of the formation of cratons is thus pertinent to address questions such as whether plate
tectonics as we see it today played a role in the assembly of the earliest continents, and if so,
when it started operating. Archean cratons possess a number of unique characteristics that
are thought to be associated with their longevity. These characteristics include low surface
heat flows [Pollack et al., 1993], cold and thick underlying continental 'roots' comprising meltdepleted mantle lithosphere [Jordan, 1978, 1988; Grand, 1994; Griffin et al., 1999], and the
protection afforded by surrounding Proterozoic orogens [Lenardic et al., 2000]. In particular,
the unique combination of thermal and compositional structure of the cratonic lithosphere
is viewed as the chief factor contributing to the long-term stability and preservation of
Archean cratons. The most long-standing and influential hypothesis relating these effects
to the stability of cratonic lithosphere is the 'tectosphere' proposed by Jordan [1978]. This
hypothesis invokes that the negative buoyancy resulting from the cooler thermal state of the
old cratonic roots is exactly compensated by the intrinsic compositional buoyancy imposed
by the low densities of iron-depleted peridotites which made up the roots.
The tectosphere hypothesis successfully explains the stable state of cratonic lithosphere.
Evidence from a variety of geophysical, petrological, and geochemical investigations has also
emerged to support the presence of deep, cold, and iron-depleted roots underlying Archean
cratons. However, the mechanisms responsible for the assembly of the cratonic lithosphere
remain a topic of ongoing debate. Over the past few decades, a number of notable hypotheses have been proposed to explain the formation of cratonic roots, including (1) advective
thickening of collisional orogenesis [Jordan, 1978]; (2) island arc aggregation [Ashwal and
Burke, 1989]; and (3) shallow subduction and underplating of oceanic lithosphere [Helmstaedt and Schulze, 1989]. Geological and geochemical investigations can help discriminate
between these models, as they provide direct constraints on the composition of crustal and
mantle rocks that can help determine their origins, and thus point to how they were formed.
On the other hand, deep probing geophysical methods enable us to image the inaccessible
lithosphere, allowing us to delineate the structures pertaining to tectonic processes. In particular, seismic imaging has proven to be a powerful tool for this purpose. The primary goal
of this thesis work is to use seismic imaging techniques to investigate the structure of the
cratonic lithosphere, focusing on three outstanding questions: (1) What is the depth extent
of the cratonic lithosphere? (2) Is the mineral/structural fabric of the cratonic lithosphere
uniform, or is it distributed in domains that suggest several episodes of assembly and deformation? (3) Is there evidence of seismic discontinuities within the lithosphere and, if so,
what is the association of those structures with the tectonic history of the craton?
In the next section, I introduce the subject of my investigation, the Archean Slave craton,
which is located in the northwestern Canadian Shield, and then discuss the data I use in
this thesis.
1.2
The Slave craton and the POLARIS-MIT portable
seismic array
The Slave craton is the ideal natural laboratory in which to conduct multi-disciplinary investigation of the Archean cratonic lithosphere (Fig. 1.1). It is well exposed, has abundant
xenolith-bearing kimberlites which provide petrological constraints on the entire lithospheric
column, and it has been insulated from plate margin processes for over 2 Ga, due to the
protection from adjacent Proterozoic orogens. Its old age and high degree of preservation
have long attracted geologists' attention [e.g., Bowring et al., 1990]. Due to the discovery
of diamonds in this region in the past decade, the Slave craton has since been under more
intense exploration and study. These studies have resulted in an extraordinarily comprehensive petrological and geochemical mapping of its crustal and mantle composition revealing
distinct mantle zonation and layering [Griffin et al., 1999; Grutter et al., 1999; Kopylova and
Caro, 2004].
Similarly, several reconnaissance seismic investigations have also delineated
multiple seismic discontinuities in the lithosphere and revealed varying anisotropic signatures [Bostock, 1998; Bank et al., 2000; Snyder et al., 2002].
However, these results are
based on data recorded at single stations, and thus do not provide sufficient constraints on
the extent and short wavelength structure of these observed discontinuities.
In light of the need for denser seismic sampling, the POLARIS consortium of Canada
(Portable Observatories for Lithospheric Analysis and Research Investigating Seismicity)
undertook the deployment of 22 temporary broadband seismic stations on the Slave craton
from 1999 to 2006. In the first project described in this thesis, I use data from the POLARIS
network to characterize the structure and fabric of the Slave craton using surface wave
analysis.
To facilitate the application of high-resolution seismic imaging techniques, we also deployed 10 broadband seismometers from the MIT instrument pool that supplemented the
existing POLARIS network. The combined network was deployed for a period of two years,
from July 2004 to July 2006, but only resulted in six months' worth of data, due to the
expected power shutdown during the Canadian winter and the unexpected failure of the
recording device. Nevertheless, the combined POLARIS-MIT array affords an average station spacing of -20 km, providing the dense array data necessary for two other projects
described in this thesis: a receiver function analysis of the central Slave craton and the
application of a new deconvolution technique to teleseismic array data.
1.3
Outline of the thesis
Chapter 2 presents the first surface wave tomography of the Slave craton derived from the
inversion of Rayleigh wave phase and amplitude data recorded by the POLARIS network.
I employ a recently-developed array inversion technique to take into account multipathing
and finite frequency effects of surface waves to better constrain the lateral variations and
depth dependence of the Rayleigh wave phase velocity and shear wave velocity structures,
as well as azimuthal anisotropic signatures in the lithosphere.
In Chapter 3, I construct a seismic receiver function image of the central Slave craton
using data from the combined POLARIS-MIT array. I synthesize the seismic image with
results from magnetotelluric, petrological, and geological analyses to explain the nature of
a coincident seismic and electrical conductive anomaly beneath the central Slave craton. I
then use these results to discuss possible models of assembly of the Slave craton.
An essential data processing step in the receiver-function analysis of Chapter 3 consists in
source-normalizing the signals so that they may be stacked. This processing step involves the
deconvolution of the recorded data by the source wavelet estimate, an operation that requires
searching for an optimal regularization parameter and thus requires extensive user input.
This limitation motivates the work in Chapter 4, in which I introduce an array-conditioned
deconvolution method that constructs a deconvolution operator based on the noise present in
the seismic array data, and therefore relaxes the ad-hoc choice of regularization. I investigate
the applicability of the method to teleseismic data processing, and discuss the performance of
the method using synthetic data and field data from the Slave craton. This array-conditioned
deconvolution presents an effective and automatic means for processing large amounts of
array data. Finally, Chapter 5 summarizes the key results of the work in this dissertation.
...............................................................
1.4
.....
.........................
. .......................
..........................................
..........
- - - = M NWWM
M
6M
6 -
Figure
-115
-110
64
6
6
62
-115
* kimberlite pipes
-110
central Slave basement complex
* Slave craton, juvenile crust
Figure 1.1: Simplified geological map of the Slave craton and surrounding regions. The
distribution of major kimberlite fields are shown. The mantle geochemical boundaries are
indicated by the dashed blue lines [Griitter et al., 1999].
22
Chapter 2
Upper mantle seismic structure
beneath the Slave craton from
Rayleigh wave tomography
Abstract
The knowledge of seismic structure and depth extent of cratonic lithosphere is crucial to
understanding the dynamics of cratonic evolution and stabilization, a subject that remains
poorly understood. In this study, Rayleigh wave phase and amplitude data are analyzed to
constrain velocity structure in the upper mantle beneath the Slave craton, in the northwestern Canadian Shield. We measure phase velocities at periods between 20 s - 142 s (with
greatest sensitivity at depths of 28 - 200 km) using crossing ray paths from events recorded
by the POLARIS broadband seismic network and the Yellowknife array. To resolve lateral
velocity variations under this region, we employ a recently-developed inversion technique that
takes into account the multipathing and finite frequency scattering effects of surface waves.
The inversion also solves for azimuthal anisotropy at each period, thus providing constraints
on the depth of anisotropic characteristics that other methodologies such as shear-wave splitt Parts of this chapter were published as: Chen, C.-W, S. Rondenay, D. S. Weeraratne, and D. B. Snyder,
New constraints on the upper mantle structure of the Slave craton from Rayleigh wave inversion, Geophys.
Res. Lett., 34, L10301, doi:10.1029/2007GL029535, 2007.
ting cannot recover. Phase velocities obtained for the Slave province are comparable to those
from other cratons at shorter periods, but exceed global average by ~2% at periods above
60 s (-80 km depth). The one-dimensional inversion of phase velocities yields high shear
wave velocities (4.72+0.2 km/s) representative of cratonic lithosphere between 50 - 200 km
depths, which are attributed to the cratonic lithosphere. The depth to the base of the cratonic lithosphere is estimated at 220±65 km. Two-dimensional phase velocity maps indicate
noticeable lateral variations. In the western part of the Slave province, which corresponds
to the oldest portion of the craton, high velocities are consistently observed from 25 to 33
s (~33 - 44 km). In contrast, lower velocities are present at periods up to 50 s (-70 km)
in the central part of the craton, a region that hosts a cluster of kimberlite pipes. Inversion
for anisotropy reveals two robust, distinct trends in the Slave lithosphere. First, at shortest
periods 20 s - 25 s (~28 - 33 km), the fast direction is N1O0 E ± 200, corresponding mainly to
crustal structure. Second, at periods 29 s - 100 s (-37 - 150 km), the fast direction averages
N59 0 E ± 200, which is consistent with shear-wave splitting measurements and is attributed
to fossil lithospheric fabric. Anisotropy is also detected at longer periods corresponding to
sub-lithospheric depth, but its strength and principal direction are poorly resolved. These
anisotropic characteristics indicate that anisotropy exists at all levels in the cratonic lithosphere from crust to lithospheric mantle, as well as in the asthenosphere. Anisotropy in the
lithospheric mantle may be representative of subduction episodes that have contributed to
the growth and stabilization of cratonic roots in the region. Deeper anisotropy is attributed
to asthenospheric flow.
2.1
Introduction
Cratons form the stable nuclei of continental plates and have survived tectonic activities
since Precambrian time. Knowledge of the formation and evolution of cratons is important
for better understanding tectonic processes active during Earth's early history, which have
remained a subject of much debate over past decades and for which a variety of dynamic
models has been proposed [e.g., Jordan, 1978; Hoffman, 1990, and the references therein].
Although much evidence shows that thick continental roots exist beneath cratons, a situation
that results in their stability and longevity through geological time [e.g., Polet and Anderson, 1995; Rudnick et al., 1999; Simons et al., 2002], controversies remain regarding the
mechanism of stabilization and the depth extent of cratonic lithosphere. The Slave craton
in the northwestern Canadian Shield provides a unique laboratory for detailed examination
of the structure of cratonic lithosphere. This well-preserved Archean craton hosts the oldest
rocks that have been found on Earth, the -4.03 Ga Acasta gneisses [Bowring et al., 1990],
and numerous kimberlite pipes that are rich in mantle xenoliths [e.g., Griffin et al., 1999]. In
addition, the favorable location of the Slave craton with respect to global seismicity affords
a comprehensive data coverage that is ideal for seismological studies. In the past decade,
the Slave craton has been exhaustively instrumented by passive seismic sensors, yielding a
seismic dataset of unprecedented volume and quality. In this study, we investigate the upper
mantle of the Slave craton using new surface wave data and a recently-developed, two-step
inversion technique to obtain tomographic images that further constrain lithospheric thickness and anisotropic structure in this region, and to gain new insight into the dynamics of
continental evolution.
The Slave province is a small (-600 x 400 km) Archean granite-greenstone craton located
in the northwestern corner of the Canadian Shield (Fig. 2.1). To first order, the province
can be subdivided into western and eastern parts based on an N-S trending boundary in
geochronological, Pb and Nd isotopic signatures [see, e.g., Isachsen and Bowring, 1994]. The
western portion, the so-called Central Slave Basement Complex [CSBC; Bleeker et al., 1999],
is a composite gneiss and granitoid complex aged 4-2.8 Ga, whereas the eastern half consists
of isotopically juvenile (<2.7 Ga) rocks [Isachsen and Bowring, 1994]. The E-W dichotomy
is explained by a tectonic model consisting of a series of consecutive terrane accretion events
[Kusky, 1989]. The Hackett River (volcanic) arc in the east was formed as an island arc by
east-dipping subduction at 2.67 Ga. In front of this arc, the Contwoyto terrane was formed
as an accretionary complex. This pair then collided with a microcontinent in the west (the
CSBC), before 2.6 Ga.
The Slave craton is flanked by two Proterozoic provinces, the Thelon magmatic zone to
the east and the Wopmay orogen to the west. The Thelon magmatic zone is a result of the
collision between the Slave and Rae provinces at -1.9 Ga, before which there may have been
episodes of east-dipping subduction beneath the Rae province [Hoffman, 1989]. The Wopmay
orogen is an accretionary belt that resulted from the docking of terranes to the west of the
Slave during successive events of subduction between 1.97-1.84 Ga [Bowring and Grotzinger,
1992].
The northern part of the Slave craton has been affected by the emplacement of
the Mackenzie dyke swarm at ~1.26 Ga, which has been attributed to the impingement
of a mantle plume beneath present-day Victoria Island [117'W, 71'N; LeCheminant and
Heaman, 1989]. Kimberlite magmatism has occurred sporadically over the last 540 Ma, from
the Cambrian to the Tertiary, with youngest events taking place between Late Cretaceous
(75 Ma) and mid-Tertiary (45 Ma) and being concentrated on the Lac de Gras region, which
is now host to several diamond mines [Pell, 1997; Snyder and Lockhart, 2005].
Geochemical and petrological studies of mantle xenoliths have yielded important insight
into the mantle composition and structure of the Slave craton. Griitter et al. [1999] analyzed
garnet crystals from surface till samples and kimberlitic xenoliths to infer the existence of
an ultra-depleted mantle zone centered on the Lac de Gras region. Under the same region,
thermobarometric analyses [Griffin et al., 1999] revealed a sharp boundary between ultradepleted and less depleted mantle layers, confining the former to a depth of 140-150 km.
Similar mantle stratification has also been suggested by studies of xenoliths from the northern
Slave craton [Kopylova and Russell, 2000]. These studies all indicate a distinct stratification
in the Slave lithosphere that is marked by variable degrees of iron-depletion and suggests an
assembly by vertical stacking of subducted slabs.
A number of regional and local seismological investigations have helped characterize the
deep lithospheric structures of the Slave province. Regional seismic tomography provides
constraints on velocity structure in a broad area of the upper mantle.
Both body- and
surface-wave tomographic models [e.g., Grand, 1994; Li et al., 2009; van der Lee and Frederiksen, 2005] have revealed the existence of deep continental root beneath the Canadian
Shield extending to -300 km depth. At a smaller scale, Bostock [1998] analyzed converted
P-to-S (Ps) wave data from the permanent Yellowknife seismic array in the southwestern
edge of the craton, and imaged well-developed mantle stratigraphy between the Moho and
the transition zone, which he inferred as evidence for cratonic assembly by shallow subduction and underplating. Cook et al. [1999] also reported a remnant east-dipping slab imaged
by a seismic reflection profile across the Wopmay orogen to the southwestern edge of the craton. Although promising, these results were derived over limited areas of the Slave province
and more comprehensive, craton-wide studies are necessary to validate these interpretations.
A previous teleseismic study by Bank et al. [2000] was carried out using data from a reconnaissance experiment of 13 temporary broadband stations on the Slave craton. The resultant
P-wave travel-time tomography revealed higher velocities underlying the oldest part of the
craton (i.e., CSBC), while a low-velocity anomaly was observed in the upper mantle at >
350 km depth beneath the Lac de Gras kimberlite field. Based on the strength of the fast
velocity anomaly, this model yielded an estimate of the lithospheric thickness to be 250±50
km. More recently, Straub et al. [2004] combined travel-time data from the above experiment and the POLARIS broadband seismic network to improve the resolution of P-wave
velocity model for the upper 700 km of mantle beneath the Slave craton. A low-velocity
anomaly was observed centered to the south of Lac de Gras extending from 50 to >300 km,
leading Straub et al. [2004] to speculate on the link between the low-velocity structure and
kimberlite magmatism.
In addition, azimuthal anisotropy in the Slave craton has been investigated in a few
shear-wave splitting studies.
Bank et al.
[2000] found an average splitting time of 1 s
from the reconnaissance experiment, with a fast direction of N60 0 Ei20' consistent with the
present-day absolute plate motion (APM) of North America. They consequently attributed
the lithospheric anisotropy to the influence of plate motion. In another study, Eaton et
al. [2004] conducted an experiment along the Great Slave Lake shear zone to analyze colocated SKS-splitting and magnetotelluric data, and found a two-layer anisotropic structure
indicative of the presence of fossil lithospheric anisotropy. Their model yielded a crustal
anisotropic layer localized near the shear zone. Snyder et al. [2003] also suggested a twolayer scenario from inversion of azimuthally-varying SKS-splitting data of a single station
in Lac de Gras region, with an upper layer corresponding to crustal and uppermost mantle
structure (0-90 km depth), and a lower lithospheric layer influenced by plate motion.
To reconcile inconsistent results among studies mentioned above, including estimates of
lithospheric thickness and the existence and depth locations of anisotropic layers, we undertake a surface-wave study from which we obtain higher resolution tomographic images of
the Slave's lithosphere than previously available. So far, body-wave tomography has yielded
good lateral resolution of mantle structure, but it lacks comparable vertical resolution due
to smearing along steep incident ray paths. Surface wave studies provide better vertical
resolution in the upper 200 km, but in large-scale regional or global tomographic models,
short-wavelength heterogeneities are often compromised by smoothing. Here, we take advantage of Rayleigh wave data recorded at a local seismic network that focuses on the target
region, and employ an array-analysis inversion technique that accounts for complexity of
incoming wave field influenced by heterogeneities outside the array [Forsyth and Li, 2005].
The dataset and methodology used in this study are first discussed in Section 2.2 and 2.3,
followed by a description of our results in terms of phase velocity maps in Section 2.4, and
anisotropy in Section 2.5. The 1-D and 3-D S-wave velocity structures are discussed in
Section 2.6 and 2.7, respectively. We conclude with a discussion on the implications of our
findings for the evolution of cratonic lithosphere.
2.2
Data and processing
We use fundamental mode Rayleigh wave data recorded at the POLARIS seismic network,
operated by the Geological Survey of Canada, and the Yellowknife array (Fig. 2.2a). The
POLARIS network consists of 22 stations, 15 of which form a ~200-km long NNW-SSE
array centered on the Lac de Gras region, while the remaining sites are located in the
western Slave. The network encompasses a -600x200-km area in the central and southern
part of the craton. All stations use TRD data acquisition systems to record teleseismic
data at 40 Hz sampling rate, from three-component broadband seismometers. Two types
of seismometers are used: six of them are Giiralp CMG-40T with flat response to ground
velocity between 0.016 and 50 Hz, and the other 16 are CMG-3ESP (flat response 0.033 -
50 Hz). The Yellowknife array comprises 18 short-period sensors deployed along a crossshaped network centered to the west of Yellowknife, and four broadband stations, one at
each arm of the cross, whose data are included in our dataset. These four stations use
Streckeisen STS-1 seismometers (flat response 0.0027 - 50 Hz). Therefore, data from a total
of 25 broadband stations are used, but not all the stations were operational until late 2003.
Instrument responses from different types of seismometers have been normalized to that of
the CMG-3ESP before data processing. Although grid-like array geometry would be more
advantageous for three-dimensional surface wave tomography, the crossing ray path coverage
obtained in the current network still offers sufficient resolution of lateral heterogeneity in the
central and southern regions of the craton.
The dataset used in this study consists of 42 events (Appendix Table A.2) at epicentral
distances ranging from 30' to 120' with magnitude Ms > 5.5 (Fig. 2.2b), that occurred
during 2003. Earthquake depths range from 7.6 to 175 km; among these events only one
occurred deeper than 100 km. The fairly comprehensive azimuthal distribution of events
provides dense crossing of ray paths that insure the resolvability of both lateral heterogeneity
and azimuthal anisotropy. Fig. 2.3 shows the event-station path distribution of data at 50 s
period, which is representative of the pattern of ray path coverage at most periods analyzed
in this study. The white box outlines the region of densest ray coverage, where the highest
resolution is expected (hereafter referred to as the target region).
The dispersive Rayleigh waves, which are observed here on the vertical component of the
seismograms (Fig. 2.4), provide constraints on vertical variations of S-wave velocity. Vertical
sensitivity kernels (Fig. 2.5) show that sensitivity of Rayleigh waves spans a broad depth
range that is dependent upon the frequency considered; the distribution in S-wave velocity
within this range contribute to the changes in phase velocities we observe in dispersive
Rayleigh waves.
The peak sensitivity occurs at a depth of approximately one-third the
wavelength of Rayleigh wave analyzed. We measure Rayleigh wave phase velocities at 12
frequencies ranging from 7 mHz to 50 mHz (20 s to 142 s). We first filter the Rayleigh wave
using a 10-mHz-wide band-pass filter centered at the target frequency. We then isolate the
Rayleigh wave in a selected time window to eliminate body-wave phases and noise outside the
window. The same window length is applied to all seismograms for a given event-frequency
pair.
2.3
Phase velocity inversion
We adopt a two-step procedure for the inversion of Rayleigh wave for S-wave velocity structure. First, Rayleigh wave amplitude and phase data are inverted for phase velocities. Second, phase velocities are inverted for S-wave velocity structure. In this section, we describe
the inversion for phase velocities.
Most surface wave inversion methods are based on the assumption that the surface wave
propagates on the great circle path, and that the wave front is planar when traveling across
a seismic array. However, surface waves may deviate from great circle path due to Earth's
heterogeneous structure along the ray path from source to receiver, resulting in changes in
direction of propagation across an array, multipathing, focusing/defocusing of energy, and
scattering [e.g., Alsina et al., 1993; Friederich et al., 1994]. In recent years, such non-greatcircle-path effects have been taken into account in newly developed methods for the analysis
of surface waves [Friederich and Wielandt, 1995; Forsyth et al., 1998].
In this study, we employ the array-analysis inversion technique developed by Forsyth
and Li [2005] and recently modified by Yang and Forsyth [2006] to include finite-frequency
sensitivity kernel in determining the amplitude of the incoming waves. This method models the incoming wave field with a two-plane-wave approximation and simultaneously solves
for the characteristics of incoming wave field and phase velocity, as well as anisotropic parameters, taking into account multipathing effects caused by velocity anomalies outside the
seismic array. The wave field is represented by the sum of two plane waves with unknown
phase, amplitude, and propagation direction. Thus, a total of six parameters are used to
describe the initial incoming wave field at each frequency. Although a simple approximation
of a possibly more complex interference pattern, this two-plane-wave assumption provides a
fairly good description of the spatial pattern of amplitude variation across an array [Li et
al., 2003] by using a minimum number of model parameters to account for the non-planar
energy in the incoming wave field.
The surface-wave phase velocity, C(w, 0), in an azimuthally anisotropic medium can be
expressed as:
C(w, 0) = Bo(w) + B1(w) cos 2w + B 2 (w) sin(20) + B 3 (w) cos40 + B 4 (w) sin 40,
where w is frequency and 0 is azimuth.
(2.1)
Bo is the azimuthally averaged phase velocity;
B1 and B 2 are anisotropic phase velocities. Higher-order terms (the 40 terms in eq. (2.1))
are usually ignored because theory indicates that they should be small for Rayleigh waves
[Smith and Dahlen, 1973]. The variation of phase velocities across the array are described
by Gaussian weighted average of velocities at model grid nodes. The grid used in this study
region comprises 377 nodes (Fig. 2.6). The density of nodes is higher in the middle of the
model with a spacing of 0.50 both in latitude and longitude, while two rows of sparser nodes
surround the edges with a spacing of 1'. These outside nodes have less damping, and help
to absorb travel time variations not accounted for by the two-plane wave approximation.
In the inverse problem, Rayleigh wave amplitudes and phases are determined through
Fourier analysis of the filtered and windowed seismograms; each seismogram provides two
pieces of information for a given frequency. Before inverting for phase velocities, we first solve
for best fitting initial plane wave parameters for each event using the simulated annealing
method [Press et al., 1992]. We then employ a generalized linear inversion [Tarantola and
Valette, 1982] that simultaneously solves for phase velocities on grid nodes and adjustments
to plane wave parameters for each event. The solution to the general non-linear least-squares
problem is given by Forsyth and Li [2005]:
Tm=(GTC
+-1
- 1 (GT C- 'Ad - C-
[m - mo])
(2.2)
where
Am= [6A ,6A ,6# ,
,6
66
6v]T
(2.3)
and
C-
=
0
0
0
0
0 l2
0
0
0
0
0
0
0
0
0
0
0
V
0
0
0
0
0
(2.4)
'-
m is the current model, mo is the starting model, Am is the change to the current model, in
which 6A', 6q,
and 60 represent the perturbations of the amplitude, phase, and incident
angle of the ith plane wave for the kth event; and 6v is the perturbation of phase velocity.
6d is the difference between the observed and predicted data for the current model, G is
the partial derivative or sensitivity matrix relating predicted changes in d to perturbations
in m; CnJ and C-,' are the a priori data and model covariance matrices, respectively. vii,
vg, and
Vkk
denote diagonal elements in the inverse of the model covariance matrix, and o-
is the damping parameter (see Section 2.6.3 for more discussion). In the inversion process,
we iterate over eq. (2.2) until the root-mean-squares (rms) mistfit changes less than 0.05
% between two consecutive iterations. Previous experience has shown that in general the
convergence is reached within 10 iterations. In the inversion, the standard errors of phase
velocity measurements are estimated from the model covariance matrix as the square root
of the variance at a given grid node.
2.4
Phase velocity results
In practice, the inversion for phase velocities is performed in two steps. First, we solve for a
best uniform, i.e., average, phase velocity model of the target region at each period. These
uniform models are then used as starting models in the inversions for phase velocities in each
grid node.
2.4.1
Inversion for 1-D phase velocities
Fig. 2.7 shows the average Rayleigh wave phase velocities as a function of period obtained
for the Slave craton, along with those from selected global and regional cratonic studies. At
shorter periods between 20 and 33 s, the phase velocities increase from 3.7 to 4.05 km/s, with
sensitivity in both crustal and uppermost mantle structures. These values are comparable
to those of other cratons, e.g., Tanzanian craton [Weeraratne et al., 2003, and references
therein]. Between 33 and 60 s, a decrease in slope is observed. The sensitivity of Rayleigh
waves at these periods covers a depth range comprising both crust and upper mantle from
about 40 to 70 km. At longer periods from 60 to 140 s, the dispersion curve is in general flat
with an average velocity of 4.25 km/s, but error bars do not preclude a steady increase from
4.25 km/s at 60 s to 4.4 km/s at 140 s. Only slight fluctuation in phase velocities from 60 to
80 s may be suggested, but the velocities generally increase with period within error. The
average value of ~4.25 km/s is significantly higher than the average Rayleigh wave phase
velocity obtained by Brune and Dorman [1963] for the whole Canadian Shield in the same
period range. It is also higher (-2%) than average cratonic values from other regions, e.g.,
eastern North America [Li et al., 2003], Superior craton [Darbyshire et al., 2007], South
Africa [Li and Burke, 2006], and global phase velocity models [Ekstr6m et al., 1997].
2.4.2
2-D phase velocity maps
Figs. 2.8 and 2.9 show the 2-D phase velocity maps for periods from 22 to 100 s. The color
scale is shifted at 50 s to accommodate the increase in phase velocity at longer periods.
Each phase velocity map is accompanied by a map representing twice the standard error.
All maps are masked using 0.05 km/s error contour of the 50 s phase velocity map showing
information within 95% confidence limits; areas outside the contour have error >0.05 km/s
and are regarded as not well resolved. The standard error at periods up to 50 s is less than, or
equal to, 0.05 km/s throughout the target region; the error becomes larger at longer periods
within the same region. Checkerboard synthetic tests indicate that our inversion successfully
resolves lateral velocity variations with scalelengths of -200 km but smaller anomalies are
not well resolved.
While there are no dramatic lateral variations observed in the maps, average phase velocities are well resolved and generally increase with increasing period. Between periods 22
and 33 s, lateral velocity variations are up to 0.08±0.05 km/s. In the southwestern corner,
a region of higher velocity is observed. To better assess the range of lateral phase velocity
variations, we display velocity perturbations relative to average velocity (Fig. 2.9). In this
representation, some dominant features appear to be consistently present between periods.
In particular, a high velocity anomaly is observed in the southwestern to western corner of
the target region for period 25-33 s. Low velocities are also observed to the northeast and
may persist to periods of up to 50 s. We note that these dominant high and low velocity
perturbations are located in the western and eastern parts of the craton, which correspond
to the CSBC and the region comprising juvenile crust, respectively. Therefore, the phase
velocity perturbations appear to closely reflect the west-east division in surface geological
features.
2.5
Anisotropy Results
As stated in Section 2.3, the inversion technique used here also solves for azimuthal anisotropy.
Due to the frequency-dependent sensitivity of surface waves, the inversion provides estimates
of azimuthal anisotropy as a function of depth. In the uniform anisotropic inversions (Section 2.4.1), the fast direction is required to be constant laterally within the craton but it is
independently solved for at each period. The best fitting anisotropic parameters are shown
in Fig. 2.11. We note that some level of anisotropy is detected at nearly all periods analyzed.
The anisotropy appears to increase from an average of 1.37% at periods below 40 s to 2.7%
for periods above 60 s. At periods below 29 s, the principal fast direction is N1
0 E±20'.
The azimuth of anisotropy changes to a NE-SW direction of N59 0 Epm200 at longer periods
between 29 and 100 s. At periods >100 s, the fast direction rotates towards a NNE-SSW
direction of N23 0 E±33'. Note that error bars become large at these longest periods, and
therefore the recovered values are not as robust as those obtained at shorter periods.
2.6
1-D S-wave velocity structure
We invert for a one-dimensional S-wave velocity profile using the uniform phase velocity
results obtained in the first inversion stage (Section 2.4). In the inversion, we use the ak135
reference model [Kennett et al., 19951 as the starting model, and search for the perturbed Swave velocity model that best fits the observed phase velocities in a least-squares sense. The
inversion process resembles the representation of eq. (2.2) in Section 2.3, but note that the
data and model parameters used here are different. Here, the model (m) refers to the S-wave
velocity model, and the starting model (mo) is ak135. We follow the layer parameterization
in ak135, i.e., 24 layers between the surface and the upper mantle transition zone. The
iteration criterion also resembles that in Section 2.3, i.e., we iterate over eq. (2.2), and
consider that convergence has been reached when the rms misfit changes by less than 0.05%
from one iteration to the next. The predicted phase velocities are computed from S-wave
velocity models using partial derivatives obtained through the DISPER80 package [Saito,
1988].
To examine the robustness of the resulting 1-D velocity profile, we also systematically
carry out a series of synthetic tests, in which we consider the effects of smoothing and
damping parameter in the inversion process.
2.6.1
1-D velocity profile
The inverted 1-D S-wave velocity profile is shown in Fig. 2.12. A high-velocity lid averaging
4.72±0.2 km/s is observed to extend from 50 to -200 km depth, and is inferred to represent
the cratonic lithosphere. The high-velocity lid is underlain by a negative velocity gradient at
depths 150-250 km. Such negative gradient has been observed in other cratonic lithospheres
[e.g., Debayle and Kennett, 2000; Weeraratne et al., 2003], and is perhaps associated with
the transition to a low-viscosity asthenospheric layer. Synthetic tests (see Section 2.6.2)
show that S-wave velocities are well resolved down to ~200 km, below which the results
become more dependent on the starting model. In light of this observation, we use a set
of different starting models to test the robustness of the principal features observed in the
resulting model. In all the inversions, high velocities from 50 to ~200 km depth are resolved
independently of the starting model, indicating that this structure is required by the data.
The presence of a high-velocity lid in this depth range is supported by continental/globalscale tomographic results [e.g., van der Lee and Frederiksen, 2005], which show a ~4%
positive (fast) anomaly in the upper mantle beneath the Slave craton relative to reference
Earth models. Finally, we also observe an apparent velocity reduction at -100 km depth.
Synthetic tests in Section 2.6.2 show that the inversion tends to smooth rapid velocity
contrasts.
Therefore, that the low velocity layer is able to be retained in the resulting
model indicates that this feature is also required by the data.
2.6.2
Synthetic tests of 1-D models
To evaluate the robustness of the inversion, we invert predicted phase velocities of known Swave velocity models (initial models) from the literature and compare the resultant models
to the initial models; the results of this exercise are shown in Fig. 2.13.
For cratonic
regions (Northern Superior, Kaapvaal, and India) where high velocities are present in a thick
lithosphere, the inverted models capture well the general velocity structures of the initial
models, but fail to recover sharp edges and introduce smoothing in the overall structure.
This smoothing effect is even more significant in the Tanzanian, Hawaiian, and Iceland
profiles, where the lithospheres are thinner and clear upper mantle low velocity layers are
present. These tests indicate that the inversion may result in underestimated velocities by
an average of 2%, especially in the vicinity of sharp velocity contrasts, and that a smoothed
velocity gradient observed in the inverted model may represent a stronger velocity contrast
in the initial model (i.e., recovered contrasts are on average 5% smaller than those present
in the original models).
2.6.3
Damping parameter
In the inversion, the damping parameter is assigned as a dividing factor in the diagonal of
the inverse of the model covariance matrix (see eq. (2.4)). Thus, a smaller parameter value
imposes more damping on the resulting model, forcing the resulting model to remain close
to the starting model. In contrast, a larger parameter value results in less damping, and
the resulting model may vary more freely, sometimes causing artificial fluctuations especially
when the data are poorly constrained. We vary the damping values from 0.05 to 0.3 and
show the inversion results in Fig. 2.14. We observe that the velocity contrast between the
high and low velocity regions indeed increases as the damping value increases. Conversely,
when the damping value is smaller, the resulting model is more influenced by the starting
model. In this study, we choose a damping parameter of 0.1 because the resulting model
achieves the smallest rms misfit between the predicted and observed phase velocities.
2.7
Inversion for 3-D S-wave velocity structure
As described in Section 2.4.2, we obtain phase velocities as a function of period for each grid
point in the study region. In this section, we invert these 1-D phase velocity profiles for Swave velocities as a function of depth at all grid points. We adopt the neighborhood algorithm
[NA; Sambridge, 1999a, b] to search in the complete model space for the S-wave velocity
models that best fit the phase velocity observations. This approach has been successfully
applied to a number of global and regional surface wave tomographic studies [e.g., Beghein
et al., 2002; Yao et al., 2008].
2.7.1
The NA optimization
The NA is a two-stage direct search technique that surveys the parameter space to find an
ensemble of 'good' data-fitting models, rather than seeking a single optimal model. The
first stage, sampling, uses a geometrical construct, the Voronoi cells, to approximate the
misfit function and drive the search towards the best data-fitting regions, while continuing
to explore a wide variety of different models. In each sampling step, the algorithm calculates
the misfit in the model space and uses this information to direct the next sampling, therefore
it is derivative-free and self-adaptive. It only uses two 'tuning parameters': n, the number
of models generated at each iteration, and nr, the number of 'best' data-fitting Voronoi cells
in which random walks are performed at each iteration. The result of this sampling stage
is a distribution of misfits, which approximates the posterior probability density function
(PPDF).
The second stage, i.e., the appraisal stage, generates a 'resampled ensemble' using the
information in the available ensemble derived from the sampling stage. Then, this resampled
ensemble is integrated numerically to compute the likelihood (the 1-D marginal PPDF, or
1-D marginals) associated each model parameter, whose width can be used to infer the model
uncertainty. One can also compute the covariance matrix (2-D marginals) to infer trade-off
between two model parameters.
The model parameterization used in the NA is listed in Table 2.1. We illustrate the
performance of the NA using results at the grid point [110.25 0 W, 65.25 0 N]. A total number
of 13,680 models were generated in the sampling stage of NA. Then, the 1-D marginals are
computed from these models and their misfits, and are used to determine the posterior mean
value and standard error for each model parameter. A most likely (best data-fitting) model
can be determined by the peaks of the 1-D marginals. Fig. 2.15 shows the posterior mean
model and its standard error of the example grid point. The corresponding 1-D marginals are
shown in Fig. 2.16. The width of the 1-D marginal distribution indicates that the crustal and
the two uppermost mantle layers are well constrained. In these layers, the posterior mean
S-wave velocities (the black vertical line in each panel in Fig. 2.16) are close to the most
likely values. Due to the depth-dependent sensitivity, fundamental mode Rayleigh waves
provide weaker depth resolution toward longer periods, hence the poorer constraints on the
lower upper mantle layers. The standard errors are in average 0.24 km/s in the crustal layers,
0.25 km/s in the top two upper mantle layers, and 0.28 km/s in the other mantle layers. The
Moho depth is poorly constrained at this grid point, determined at 38.12 km depth, with a
standard error of 2.8 km.
The posterior mean model is determined for each grid point, and these models are com-
bined to infer 3-D velocity variations.
2.7.2
S-wave velocity lateral variations
Fig. 2.17 shows the resulting S-wave lateral variations of the study region at depths of 10,
30, 50, 70, 100, and 150 km. The geochemical boundaries are also outlined in the figure.
Throughout the crust and the lithospheric mantle, the lateral variations in S-wave velocity
are -<3%. In the central to western craton, there appears to be a prominent high velocity
region extending from the lower crust (-30 km depth) to the uppermost mantle (-80 km
depth), which is spatially coincident with the CSBC. The lateral extent of this region also
appears to be confined by the outline of the mantle geochemical boundary by Griltter et al.
[1999] (blue dashed lines). To the east, a low velocity zone is observed at 50-80 km depth,
centered roughly on the petrologically-defined ultra-depleted layer (UDL) (green box; Griffin
et al. [1999]) in the mantle and on the Lac de Gras kimberlite field. Below 100 km depth, the
lithospheric mantle is characterized by high velocities, with the highest velocities observed
to be coincident with the UDL. Fig. 2.18 shows a velocity cross-section (orange line in Fig.
2.17f) perpendicular to the geochemical boundaries. The confined high velocity structure is
3-5% faster than the reference 1-D model. In the uppermost mantle from Moho to 90 km
depth, the S-wave velocity is -2.5% lower than the reference model.
2.8
Discussion
We have derived Rayleigh-wave phase velocity variations and S-wave velocity structure for
the lithosphere of the Slave craton. The Slave's lithosphere exhibits a thick, high-velocity
region, and anisotropy is observed within the crust, the lithospheric mantle, and asthenosphere. In this section, we discuss the determination of the thickness of cratonic lithospheric
and the implication of depth-dependent azimuthal anisotropy for the Slave's evolutionary
history.
2.8.1
The cold cratonic block in the central Slave craton
The lateral extent of the high S-wave velocity structure observed from the S-wave velocity in
the Lac de Gras region at -100-200 km depth is spatially coincident with the geochemically
distinct mantle layer. Petrological evidence from mantle xenoliths in the Lac de Gras region
also indicates the presence of the ultra-depleted layer extending from -100 km to -140-150
km depth. In addition, the 1-D S-wave velocity profile shows a low velocity layer centered
at -100 km depth, overlying a high velocity lithospheric lid down to -220 km depth. Taken
together, we interpret that the high velocity structure represents the ultra-depleted cratonic
block in the central Slave craton, and the low velocity layer likely marks its upper boundary.
In Chapter 3, we draw on additional petrological evidence showing the presence of a compositional front at -100 km depth in this region, and suggest that the central Slave craton
may be a remnant fragment of a once larger continent.
2.8.2
The thickness of the cratonic lithosphere
The depth extent of cratonic roots has long been a source of debate, especially since the
proposal of the tectosphere hypothesis [Jordan, 1975], which suggests that an old, chemically
distinctive continental keel persists to >300 km depth beneath cratonic regions. Thermal,
petrological, and electrical studies on the continental lithosphere all indicate, however, that
the conductive continental root is at most 200-250 km thick [Jaupart et al., 1998; Rudnick
et al., 1998; Hirth et al., 2000]. Global/large-scale regional seismic tomography has shown
high velocity structures representative of cratonic roots that may extend to >-250 km depth
[e.g., Grand, 1994; van der Lee and Nolet, 1997; Ritzwoller et al., 2002; Zhou et al., 2006; Li
et al., 2009]. In the North American continent, van der Lee and Frederiksen [2005] suggested
that the average lithospheric thickness is -250 km based on S-wave velocity model obtained
from surface-wave inversion. Similar depth ranges have also been observed in Australian
continent and Kaapvaal craton [e.g., Simons et al., 1999; Debayle and Kennett, 2003; Fouch
et al., 2004]. Gung et al. [2003] pointed out, however, that seismic anisotropy induced by the
low-viscosity asthenospheric channel can account for the discrepancy between the >300 km
and the 200-250 km continental lithospheric thicknesses inferred by global versus regional
studies, respectively.
One important objective of this study is to determine the depth extent of the Slave
craton's lithosphere.
Estimates of lithospheric thickness from seismic studies often vary
depending on the criteria used to define the lithosphere and asthenosphere. In body-wave
tomography, the depth extent of high-velocity regions (i.e., 0.5-2% above global velocity
average) in the uppermost mantle has been used to define the extent of the lithosphere
[e.g., Grand, 1994]. This definition may, however, overestimate the lithospheric thickness
due to vertical smearing. In surface-wave tomography, on the other hand, the base of the
lithosphere is inferred from the strongest negative gradient in S-wave velocity marking the
base of a high-velocity lid [e.g., Debayle and Kennett, 2000]. Weeraratne et al. [2003] also
noted that, if the lithosphere-asthenosphere transition is due to a gradual thermal effect,
then the top of the negative gradient may mark the lithosphere/asthenosphere transition.
However, if the velocity contrast at the base of the lithosphere is due to other factors such as
onset of melting, increase in water content, or compositional change, the top of the negative
gradient could underestimate the thickness.
Synthetic tests (Section 2.6.2) show that, for cratonic regions, the recovered negative
gradient marking the base of the lithosphere is usually shifted toward depth shallower than
where it really is. Thus, the depth at which the negative gradient reaches the slowest velocity
may approximate the depth of transition from the lithosphere to asthenosphere in the initial
model. In our S-wave velocity profile through Archean stable lithosphere, high velocities
of 4.7±0.2 km/s observed from 40-150 km depth are up to 5% higher than global averages
(i.e., ak135) in this depth range. Below 150 km, we observe a negative gradient that reaches
a minimum of -4.68 km/s at -275 km depth, consistent with the ak135 velocity at this
depth (within error). If we use the center of the negative velocity gradient to infer the base
of the lithosphere, we obtain a depth of 220+65 km, as derived in the 1-D S-wave velocity
profile. However, reduced depth sensitivity of long period Rayleigh waves prevents us from
uniquely resolving structures below -200 km depth. Therefore, we suggest that the base of
the Slave's lithosphere occurs at no less than 150 km, and that a thicker lithosphere that
extends to 250 km depth [van der Lee and Frederiksen, 2005] cannot be ruled out at this
stage.
The thickness of the Slave craton's lithosphere appears to be similar to that found in
other Archean cratons, such as Kaapvaal [-250 km; Fouch et al., 2003] and Siberia [~220
km; Priestley and Debayle, 2003]. Our results generally agree with that from previous studies
on the S-wave velocity structure of the Canadian Shield [Frederiksen et al., 2001], and Pwave travel-time tomography of the Slave craton [Bank et al., 2000], both of which suggest
a -250-km thick lithosphere. In addition, the thickness agrees with the thermal thickness
estimate in a global study by Artemieva and Mooney [2001], which is also -250 km.
Our results agree with the idea that the Slave lithosphere observed at present time has
remained stable since its assembly. It has been suggested that a shallow subduction/underplating event associated with the Wopmay Orogeny, in the Proterozoic (-1.9 Ga), represents
the final stage of the Slave craton's stabilization [Bostock, 1998; Davis et al., 2003; Snyder,
2008]. Since then, the Wopmay orogen has likely protected the craton from being affected
by the asthenospheric flow at the western edge of the North American craton. It has been
reported in Australia [Simons et al., 1999] that the Archean cratonic regions to the west of
this continent may have experienced destabilization during continental breakup, resulting
in a thinner cratonic root (-150 km thick) than that beneath the Proterozoic region in the
center of Australia (-300 km thick). Such kind of cratonic root destruction has also been
suggested for the Tanzanian craton, but there it is attributed to the upwelling of an upper
mantle plume [Weeraratne et al., 2003]. Although the emplacement of dyke swarms in the
northernmost part of the Slave craton has been attributed to plume activity (see Section
2.1), there is no geophysical or petrological evidence that the current Slave's lithosphere
experienced significant basal erosion associated to either continental breakup or plume interaction. Therefore, the Slave craton likely represents an example of cratonic lithosphere
with undisrupted long-term stability.
2.8.3
Depth-dependent azimuthal anisotropy
The cause of continental seismic anisotropy has remained a contentious issue, for which two
end-member hypotheses have been proposed: one suggests that anisotropy reflects the inher-
ent lithospheric fabric associated with past deformation [e.g., Silver and Chan, 1991], and
the other favors a correlation between lithospheric anisotropy and the present-day absolute
plate motion (APM), thus recording current asthenospheric flow direction [e.g., Vinnik et
al., 1992]. Recent advances in analysis method and observation of seismic anisotropy result
in much evidence that points to the combination of both [see, Fouch and Rondenay, 2006,
for a review]. For the Canadian Shield as a whole, the direction of APM is roughly parallel
to the orientation of the lithospheric fabric, so both causes would give similar azimuth of
fast polarization [Bokelmann and Silver, 2002]. For the Slave craton, the APM direction is
reported as 2250 (or S450 W) [Minster and Jordan, 1978].
Until now, vertical distribution of anisotropy has only been elusively inferred in the Slave
craton. Bank et al. [2000] suggested a solely sub-lithospheric origin; however, splitting data
only provide limited constraints on the vertical extent of anisotropy. Conversely, here we
find that anisotropic signatures are robustly observed at most periods between 20-100 s,
indicating that anisotropy is present in the lithosphere as well. The anisotropy with APMparallel fast direction is observed between 29 s and 100 s periods, an interval with greatest
depth sensitivity ranging between sub-Moho to lower lithosphere (50-150 km). These results
are consistent with those of Snyder and Lockhart [2005], who suggested solely on the basis
of SKS-splitting results that the main component of mantle anisotropy may be caused by
large-scale, shape-preferred orientation (SPO) geometry within the lithosphere.
In their
interpretation, this SPO is attributed to a concentration of parallel kimberlite dykes that
are aligned with the direction of APM, i.e., perpendicular to the direction of minimum
horizontal stress, in which fractures would have opened to accommodate kimberlite dyke
intrusions.
Our results also indicate variations of anisotropy with period, but due to the broad
depth range of surface wave sensitivity kernels, we only suggest the existence of a change
in azimuth from a N-S fast direction at short periods below 29 s to a NE-SW fast direction
at longer periods. The shallow N-S trending anisotropy may correspond to regional crustal
and uppermost mantle structure resulting from Archean tectonic processes (e.g., sutures)
that predated the stabilization of the Slave cratonic lithosphere [Davis et al., 2003], which,
as described in Section 2.1, involved a series of E-W accretionary events. An independent
study by Snyder and Bruneton [2006] reached similar conclusions by jointly analyzing SKSsplitting and Rayleigh wave dispersion data. In our study, the anisotropy results at periods
>100 s are less well constrained, but we observe a consistent change in fast direction that
may result from APM-oblique asthenospheric flow due to topography at the base of the
lithosphere.
Some first-order calculation can help assess the contribution of anisotropy by comparing
the delay time of the splitting data with that from the anisotropy constrained by our surface
wave analysis. The calculation is described as follows: We have three known variables: the
thickness (h) of the layer through which the seismic waves propagate, the average seismic
velocity (V) and the anisotropy (a). Given a fast- and a slow-polarized velocity, Vf and
Vs, respectively, the splitting time is 6t = h/(Vf - Vs), where Vs = 2V/(a+2), and Vf =
(a+1)Vs. From our anisotropy results, the 40-km thick crust with 1%anisotropy and average
S-wave velocity of 3.8 km/s would result in a delay time of 0.11 s. This value is consistent
with that of the shear-wave splitting analysis by Silver and Chan [1991], who concluded that
the anisotropy contributed by the entire crust in the Canadian Shield is on the range of
0.1-0.3 s. If we assume a 200-km-thick homogeneous layer as the lithospheric mantle with
average S-wave velocity of 4.75 km/s and 1.8% anisotropy, its contribution in delay time
would be 0.8 s. As a result, the whole lithosphere would give the delay time of 0.91 s. Some
additional contribution from the sub-lithosphere may therefore be required to fully reconcile
the average 1 s delay time derived from Bank et al. [2000]'s splitting analysis (Fig. 2.11).
Debayle et al.
[2005] reported global azimuthal seismic anisotropy and argued that
the Australian continent is the only region where its anisotropy strongly correlates with
present-day plate motion, probably because it is the fastest moving continent (-8 cm/yr).
In comparison, the North America plate is moving with an average speed of -2 cm/yr to
the southwest [Gripp and Gordon, 2002]. Nevertheless, the depth constraint presented here
suggests that the sub-lithospheric contribution to the observed anisotropy is present for the
Slave craton, although the amount may be less than that of the Australia.
2.8.4
A mid-lithosphere low velocity layer?
In our 1-D S-wave velocity profile, we observe a velocity reduction centered at ~100 km depth.
Furthermore, the lateral variations of S-wave velocity also indicate a significant low velocity
region (-3% lower relative to the reference model) occurring from below Moho to
-100
km
depth beneath the Lac de Gras region. In Chapter 3, we use receiver-function analysis of
P-to-S (Ps) converted waves to image the lithospheric structure beneath the central Slave
craton, and observe a pronounced seismic discontinuity at ~100 km depth in the same region.
In conjunction with results of magnetotelluric soundings and mantle xenolith analysis, we
interpret that this seismic discontinuity represents a metasomatic boundary associated with
an Archean subduction zone.
2.9
Conclusions
The surface-wave study described in this chapter has yielded new insights into the depth and
velocity structure of the Slave's cratonic lithosphere, as well as into the vertical distribution
of anisotropy in the upper mantle. The Slave cratonic root persists to a depth of -220 km,
which is comparable to lithospheric thickness found in other Archean cratons. The observed
high velocities in both 1-D S-wave velocity profile and 3-D structure indicate a particularly
depleted and/or cold, unperturbed Archean continental lithosphere, and comparisons with
other cratonic environments suggest that the Slave craton may be an end member in this
respect. Furthermore, distinct anisotropic domains are revealed in the crust, the lithospheric
mantle and the sub-lithospheric mantle. The layered anisotropic signatures may reflect the
Slave craton's assembly and evolution: the crustal anisotropy reflects the recent deformation;
the lithospheric mantle anisotropy indicates possible ancient subduction processes, and the
anisotropy in the sub-lithosphere may depict the present-day dynamics of mantle convection. The Archean lithosphere of the Slave craton exemplifies long-term stability due to the
protection of surrounding Proterozoic orogens.
2.10
Table
Model parameter
Moho depth
V, in the upper crust
V, in the lower crust
V, in the Moho-90 km
V, in 90-130 km
V, in 130-170 km
V, in 170-220 km
V, in 220-280 km
Reference value
38 (km)
3.5 km/s
4.0 km/s
4.7-4.63 km/s
4.63-4.78 km/s
4.78-4.8 km/s
4.8-4.68 km/s
4.68 km/s
Perturbation range
[-5 51 km
[-0.8 0.4] km/s
[-0.8 0.4] km/s
[-0.6 0.6] km/s
[-0.6 0.6] km/s
[-0.6 0.6] km/s
[-0.6 0.6] km/s
[-0.6 0.6] km/s
Average perturbation
-0.48 km
0.017 km/s
-0.027 km/s
-0.031 km/s
0.051 km/s
0.079 km/s
0.094 km/s
0.039 km/s
Table 2.1: Model parameters as well as their reference values, perturbation ranges, and
average perturbation in NA. The reference values are according to the 1-D S-wave velocity
profile (Fig. 2.12). The perturbation ranges are with respect to the reference values. The
average perturbation is the average perturbation for all grid points in the study region.
2.11
Figures
62
6
-115-110
* kimberlite pipes
-105
central slave basement complex
Slave craton, juvenile crust
Figure 2.1: Simplified geological map of the Slave craton and surrounding regions. The
brown area is denoted as the central Slave basement complex (CSBC) by Bleeker et al.
[1999]. Dashed blue lines indicate geochemical boundaries identified by Griitter et al. [1999].
(a)
-120*
68" 4
-116'
-112*
-108"
-04'
64*
62"
-720"
-112'
-108'
Figure 2.2: Receivers and earthquakes. (a) POLARIS and Yellowknife seismic stations used
in this study (red triangles). Dotted white lines outline the boundaries of the Slave craton
and CSBC. (b) Azimuthal distribution of 42 events in the current dataset. The epicentral
distances range from 30' to 120'. The yellow box indicates the location of the Yellowknife
array.
-120"
60- f4
-720"
-1_6
-11'
-108"
-104
4
-116*
-112'
-108.
60'
-104'
Figure 2.3: Great circle paths at 50 s are representative of the ray path coverage in this
study. The white box outlines the target region where highest resolution is expected due to
the optimal density of crossing ray paths.
LGSN 2003/05/21 depth=10 km; distance=7360 km; mb=5.7; baz=52.5
Figure 2.4: Dispersion of Rayleigh wave is shown for an event in northern Algeria recorded
at station LGSN located to the south of Lac de Gras region. Filters have bandwidth of 10
mHz centered at 12 periods ranging from 20 s to 142 s (7 mHz to 50 mHz).
50 100
t
-
150 c
200 -
250
300
350
400 -
0.000
o.o5
0.010
0.015
0-020
0.025
0.030
6c/Sp
Figure 2.5: One dimensional sensitivity kernels for phase velocities from 20 s to 100 s. The
depth of peak sensitivity occurs at 1/3 of the wavelength and is affected by structures in a
broader depth range as the period increases.
-120"
-116'
-112*
-108*
-104*
-12 0*
-116*
-112*
-10 8'
-10 4'
Figure 2.6: Distribution of grid node locations within the study area for phase velocity
inversions.
-~
,
I
I
I
I
4.4
4.3
4.2
zE.4.1
o
4.0
Slave
' 3.9
.2
Shield
-Can.
3.8 -
[Bre & Dorm n, 1963]
3.7
[Weeraratne et aL, 2003]
3.6 -
[.strom et al., 1997]
Tanzanian
Cmton average
3.5-
,
0
i
20
40
60
80
100
i
120
i
140
160
Period (s)
Figure 2.7: Uniform Rayleigh wave phase velocities as a function of period for the study
area. Red line shows the phase velocities obtained for the Slave craton in this study, whereas
other color lines denote phase velocities of other cratonic entities for comparison (see legend).
Uniform anisotropy is simultaneously solved in the inversion and is presented in Fig. 2.11.
Error bars indicate 2o- errors.
.......
...
phase velocity
--116" -112'
66*
-108'
......
std error
-116' -112" -108"
22s.16
4.12
4.04
4.00
62'
64* -
km/s
L.7
3.96
1
3.92
<
M
O
3.88
3.84
3.-/6
3.72
62"
66'
3.856 km/s
s8.7
64'
62'
66-
3.951km/s
33 s
64'
62*
66"
.045 km/s
40 s
64"
62'
N.097 km/s
0.02 0.04 0.06 0.08 0.10
2 x Standard error (km/s)
Figure 2.8: Two-dimensional phase velocity maps for periods from 22 s to 40 s (left column).
The color bar for velocity is to the right; contour intervals are 0.02 km/s. The starting model
for inversion is the 1-D phase velocity profile shown in Fig. 2.7 (red line), whose values are
also indicated in the lower right corner of velocity maps for each period. The trade-offs from
anisotropy are considered in the inversions and will be discussed later. The map in right
column shows 2- error of the recovered phase velocities for each period. All maps are masked
to show the information within 95% confidence level (the target region). White triangles are
locations of seismic stations, and boundaries of the Slave craton are outlined in gray dashed
lines.
phase velocity
-116'
-112'
-108'
std error
-116'
-112'
-108'
-
4.32
-4.28
-4.24
*
0
- 4.20
3
-4.16
-
4.12
0.02 0.04 0.06 0.08 0.10
2 x Standard error (km/s)
Figure 2.9: Phase velocity maps for periods from 50 s to 100 s as described in Fig. 2.8.
The range of lateral velocity variations decreases somewhat with increasing period, as the
sensitivity spreads to a broader depth extent and results in a decrease in depth resolution.
As shown in the right column, the standard error increases within the target region at longer
periods. Contour intervals for the standard error are 0.01 km/s. Also shown are the strikes
of the Bathurst fault system to the northeast and the Great Slave Lake shear zone to the
southeast (gray lines in the velocity maps)
-116'
66"
-112'
-108'
-116'
-112*
-1080
-116* -112
-108*
22ss6
/ \
64*
62
-
.045 km/ 0
.777 km/
h25as
.278 km/
80vse
e'
64o
64"
Fiur
62w
.951 km/vle
Mapsof2D'hseveoct
6610
a a1.0
-0.8 in. th
.59 km
.28 kmS
Te
petubtin forpeios 22s-10'
pesm
-0.2
2s to 0
etr
.
Phase velocity perturbations (%)
Figure 2.10: Maps of 2-D phase velocity perturbations for periods 22 s - 100 s. The perturbations are calculated with respect to the 1-D phase velocity profile of Fig. 2.7 (red line),
with reference values indicated in the lower right corner in each panel. The Slave craton is
depicted by gray dashed lines, whereas the east boundary of the CSBC is outlined in white.
Contour intervals are 0.1%. The degree of lateral perturbations decreases generally with
increasing period, while some prominent features are consistently seen. In particular, we
observe a pronounced high velocity anomaly from 25 s to 33 s in the west part of the craton,
and a low velocity anomaly in the east, persisting from 22 s to 50 s.
(a)
I
10
N
I
|
20
30
I
I
|
I
I1
40
50
60
I
I
70
I
I
80
90
period (s)
I
I
I
I
I
l
I
I
I
I
100
110
120
130
140
150
(b)
-120'
-116'
-112'
-108'
~62*
60-104*
Figure 2.11: Anisotropy results. (a) 1-D anisotropy model found for the study area in the
inversion for each period. Fast directions are indicated by red bars in map view, with gray
vectors denoting 2o errors. Vertical blue error bars indicate 2o errors on estimated strength
of anisotropy. (b) Anisotropy results from SKS-splitting analysis of Bank et al. [2000]. Black
bars point in the direction of the fast split wave; the length of bars represent the delay time
between two split waves as indicated in the legend.
50 -
100
150
E
.c
200
41-J
250
300
350
---
Slave
---- ak135
. i .
i
I , I . I . I .
400 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2
Vs (km/s)
Figure 2.12: 1-D S-wave velocity profile (black line) obtained from inversion of phase velocity
in Fig. 2.7. The starting model (dashed line) is ak135. A moderate damping parameter of
0.1 is used in the inversion. See text for further discussion.
0
50
50
-
~
-
~50
100
100
100
E
150
150
150
c
200
200
200
250
250
250
-
300
-300
-350 -
.1
N.Superior
3.2
3.6
4.0
4.4
4.8
.400-,
5.2
3.2
350
:aapvaal
SI I . .
3.6
I. .
4.0
.I .
1.
400 - ,
4.8
5.2
3.2
4.4
3.6
4.4
4.8
5.2
Vs (km/s)
Vs (km/s)
Vs (km/s)
4.0
0
50
100
E
.
150
200
0.
250
-350 -
Tanzania
I
r-1,
3.2
3.6
4.0
4.4
Vs (km/s)
4.8
3.2
3.6
4.0
4.4
Vs (km/s)
4.8
-400
5.2
- ,
3.2
Iceland
3.6
4.0
4.4
4.8
Vs (km/s)
Figure 2.13: Synthetic tests of 1-D S-wave velocity profile inversions. In each panel, the
dashed curve denotes the velocity model obtained from individual regional studies. The solid
curve denotes the resulting model of the 1-D inversion. The thin, grey curve denotes the
starting model in the inversion. The final models are able to capture the velocity structures
of the initial models, but the sharp edges or discontinuities in the initial models are often
smoothened in the resulting models.
......
.
...
.....
MMONNOWL
50
100
150
.; 200
damp
= 0.05
damp
0.1
= 0.2
-
0.
250 -
300 -
-
damp
-
damp =0.25
-
damp =0.3
---
ak135 (starting model)
350 -
400 , I
3.2
3.4
3.6
I
I
3.8
4.0
4.2
4.4
I
4.6
4.8
5.0
5.2
Vs (km/s)
Figure 2.14: Summary of the resulting models obtained from using different damping parameters in the inversion. The damping parameter controls the magnitude of variation in the
velocity profile. In the inversion, the damping parameter is assigned as a dividing factor in
the diagonal of the model covariance matrix, thus a smaller value imposes stronger damping
effect and results in a final model with less deviations from the starting model.
Shear Wavespeed(km/s)
0
3.5
4
4.5
20
40
60
80
100
120
140
CL 160
180
200
220
240
260
280
Figure 2.15: S-wave velocity model at grid point [110.25W, 65.25N] derived from NA. The
solid curve shows the posterior mean velocity at each layer, and the shaded area indicates the
corresponding standard error. The dashed curve shows the reference model from Fig. 2.12.
1-D Marginals
I I I IA l,
33
,
40
Moho depth
,
,
43
AV:130-170 km
AV:90-130 km
A .A~.!
0
AV: 170-220 km
-0.6
0
AV:220-280 km
0.6
Figure 2.16: The 1-D marginal posterior probability density functions of the Moho depth
and the S-wave velocity at each layer at the grid point [110.25W, 65.25N]. The horizontal
axis indicates the variation range of the Moho depth (km) or the velocity variation range
(AV; km/s). The vertical axis shows the normalized probability density. The solid line
indicates the posterior mean values of each model parameter. The peak in the probability
density distribution indicates the most likely (best data-fitting) parameter value.
(a)
depth:-10km
66
depth
=30km
1A
65.5
3.45
IS
3S5
16
Longitude
(deg)
Longitude(deg)
4S
tss
4.6
4,65
t7
4.7S
4,6
4.62
4.64
4.66
4.68
4.7
4.72
4.74
4.76
4.78
4.8
Longitude(deg)
Longitude(deg)
Figure 2.17: S-wave velocity variations inferred from NA: (a) 10 km; (b) 30 km; (c) 50 km; (d)
70 km; (e) 100 km; (f) 150 km. Color bars indicate Vs (km/s). The blue dashed lines outline
the ultra-depleted layer in the mantle lithosphere inferred from garnet xenocrysts [Griitter
et al., 1999]. The green box outlines the UDL inferred from mantle xenoliths [Griffin et al.,
1999].
0
100
1500
200
0
100
200
300
5
Distance(km)
Figure 2.18: The S-wave velocity perturbation across the Lac de Gras region (line A-B in
Fig. 2.17(f)). The color bar denotes perturbation percentage relative to the reference velocity
model (the 1-D S-wave velocity model in Fig. 2.12). The high velocity structure is observed
at -100 km to 200 km depth beneath the Lac de Gras region.
Chapter 3
Geophysical detection of relict
metasomatism from an Archean (ca
3.5 Ga) subduction zone
Abstract
The origin of Archean cratonic lithosphere is the subject of much debate. Geological and geochemical data are consistent with a plate tectonic origin involving assembly of predominantly
oceanic terranes and micro continents into large stable cratons. Deep probing geophysical
surveys of Archean cratons provide a means of understanding the processes responsible for
the formation of these earliest continents, although the signals are complex and present-day
cratons are only the surviving remnants of once larger entities. Here we present a unique
geophysical view of structure within the Archean Slave craton, in northwestern Canada,
that shows clear evidence for subduction processes frozen into Archean lithosphere. New
seismic imaging results from the central Slave are synthesized with a coincident magnetotelluric model, and interpreted using petrological and geochemical constraints from mantle
xenoliths. We find the most striking correlation between seismic and electrical structures
tParts of this chapter have been accepted for publication in Science as: Chen, C.-W., S. Rondenay,
R. L. Evans and D. B. Snyder, Geophysical detection of relict metasomatism from an Archean (ca 3.5 Ga)
subduction zone, Science, in press.
ever observed in a continental setting in the form of a coincident seismic discontinuity and
electrical conductor at -100 km depth. The magnitude of both anomalies, in conjunction
with the occurrence of phlogopite rich xenoliths originating at the same depth, point to a
metasomatic origin. We believe that fluids were released from a subducting slab and altered
the mantle directly below the base of a pre-cratonic lithosphere. Our model suggests that
cratons are formed by subduction underplating and accretion of pre-existing fragments, and
that these processes were active as early as 3.5 billion years ago.
3.1
Introduction
Archean cratons form the core of many of Earth's continents. By virtue of their longevity,
they offer important clues about plate tectonic processes during early geological times. A
question of particular interest is whether subduction is the main mechanism of cratonic
assembly and, if so, when this process began to operate. Answers to these questions may be
found in the lithosphere of the Archean Slave craton, which is located in the northwestern
Canadian Shield (Fig. 3.1) and represents the oldest known fragment of the North American
continent [Bowring et al., 1989]. The Slave craton has been studied extensively to reveal
processes of cratonic assembly as it is well exposed and has been insulated from plate margin
processes for over 2 Ga. In the center of the Slave craton, inclusions of rock fragments from
the mantle indicate that the lithospheric mantle is separated into two distinct layers at 140150 km depth, with a harzburgitic upper layer exhibiting a very high degree of iron depletion
[Griffin et al., 1999; Griitter et al., 1999] and a lherzolitic lower layer that is less depleted in
iron [Griffin et al., 1999]. Previous seismic studies identified remnants of subducted oceanic
lithosphere [Bostock, 1998; Snyder, 2008; Moorkamp et al., 2007], believed to be linked to the
final stages of assembly of the Slave craton during the late Archean and to its stabilization
during the Proterozoic. Despite these results, there remains a great deal of uncertainty over
the timing and mechanisms of the formation of the Slave craton. For example, it is unknown
when its >200 km thick lithospheric root was in place, and if it was constructed solely
by subduction accretion [Helmstaedt and Schulze, 1989] or by successive accretion-collision
episodes [Jordan, 1978]. Further insight into these questions requires new and integrated
datasets. In this study, we address this objective by investigating coincident seismic and
electrical profiles across the Slave craton.
3.2
Seismic observations
In this section, we first introduce the seismic data and methods used to image the discontinuities in the lithospheric mantle of the Slave craton. This is followed by a presentation of
the imaging results and evaluation of the robustness of the results.
3.2.1
Data
The dataset used in this study was acquired using 23 three-component broadband seismometers deployed across the Slave craton over a 24-month period (July 2004-August 2006). However, due to expected shutdown of power during the Canadian winter, and unexpected failure
of the recording device, at the end only six months' worth of data was recovered. Among
the 23 seismic stations, 13 of them were part of the POLARIS Slave network. The other 10
stations are part of the MIT instrument pool (Fig. 3.1; Table Al). The POLARIS Slave
network was designed primarily as a linear array with a few outlier stations to the west
of the array for additional lateral data coverage; these outlier stations are not used in this
study. The linear array is oriented perpendicularly to major geochemical boundaries in the
mantle [Griitter et al., 1999]. The MIT array was deployed in July 2004, and remained in
the field for two years. Each station was equipped with a CMG-3TD broadband seismometer
(flat response 50 Hz- 120 s) and a Giiralp recording unit. The MIT array was designed to
supplement the existing POLARIS stations so as to render the station spacing suitable for
high-resolution seismic imaging. The combined POLARIS-MIT array results in an average
station spacing of -20 km.
We recorded 62 teleseismic events with magnitude mb > 5.8 within the epicentral distance
range of 30'-100'. These events illuminate the study region from various backazimuths (Fig.
3.1), but most events come from western Pacific, Fiji-Tonga, and Central and South America.
Thus, the best backazimuthal coverage is in the NW and the SE quadrants.
3.2.2
Method
We analyze P-to-S (Ps) converted waves in the coda of teleseismic P waves using an approach
developed by Bostock [1998], which is similar to the traditional receiver function method [e.g.,
Langston, 1979], but with some modifications, including the incorporation of the wavefield
decomposition and the use of simultaneous deconvolution. In the following, we outline the
key steps in the data processing and imaging.
First, all three-component [radial (R), transverse (T), and vertical (Z)] seismograms for
each event are transformed into upgoing P, SV, and SH wavefields using the inverse freesurface transfer matrix [Kennett, 1991], which is expressed as
0
P
SV
=
SH
2p2
0
-?
I
0 p#
0
R
T
(3.1)
Z
where a and # are the P- and S-wave velocities at the free surface, p is the horizontal slowness,
and q,
=
-
p-
2
and qO
= X/-2
_ p-2 are the corresponding vertical slownesses. In
a layered, 1-D Earth model, this transformation, in theory, isolates completely the incident
P wave from the converted Ps wave at horizontal discontinuities, and suppresses the effect
of signal distortion caused by downgoing waves reflected at the free surface. Secondly, a
time-domain window is applied to the P component, so as to exclude undesirable signals
that are considered not part of the source time function, and that could cause degradation
of the subsequent deconvolution process, such as PP. The resultant windowed P component
thus represents the approximation of the effective source time function.
The next key step is the normalization of source signature for each event. This step
is necessary in order to separate the structural impulse response in the subsurface from
the extended earthquake source time functions. It also allows for coherent stacking of the
converted waveforms and thus enhancing the signal-to-noise ratio. The normalization process
requires the construction and application of a deconvolution operator that is able to solve
the forward problem:
(3.2)
d(t) = w(t) * r(t),
in which the observed signal d(t) is represented by the convolution of the Earth's impulse
response r(t) with the combined source time function and instrument response w(t), and
where r(t) is the unknown. In the most commonly used method, the frequency-domain
deconvolution, an approximate solution of the impulse response, i(w), can be expressed as
[e.g., Berkhout, 1977]
P(w)
w
*(w)
(3.3)
d(w)
w(W)w*(W) + 6
where the asterisk denotes the complex conjugate, w is angular frequency and 6 is a regularization factor, sometimes termed waterlevel (Clayton and Wiggins, 1976), which controls
the trade-off between model variance and data misfit. In Chapter 4, we will discuss and
introduce an alternative approach to the deconvolution problem in more detail.
Bostock [1998] refined this process by grouping several wavefield-transformed seismograms to estimate a single impulse response (?(t)) in a least-squares simultaneous deconvolution of N seismograms in the frequency-domain as
(t)= F-1
N
.
1)
,
(3.4)
where dn and wn represent the nt' S-wave component and corresponding windowed P-wave
component, respectively, and F-
1 denotes
an inverse Fourier transform.
Finally, to map the energy of stacked converted phases into mantle "reflectivity" as a
function of conversion depth, R(z), a moveout correction is required and is achieved by
including the term of phase delay times in eq. (3.4) as
R(z) = F
REzN d1[(w)w* (w)exp(WT(,
N
En
n(3.5)
wn (W)w* (W) +
z))
where T(pn, z) is the delay time of the converted phase originating at an interface at depth
z, and pn is the horizontal slowness of the nth P-component seismogram. These stacked
traces can be plotted as a function of time or conversion depth, and they form the basis of
1-D receiver function (RF) imaging. In the following, we will show two types of RF profiles.
First, for individual stations, the stacks are plotted as a function of epicentral distances, and
thus provide insight into the structures underlying the station. Second, for array data, the
RF traces are binned according to common conversion points at the depth of interest. The
resulting stacks are then plotted side-by-side to form a pseudo-2-D profile, illuminating the
structure underlying the array in a continuous manner.
3.2.3
Single station profiles
Figs. 3.2 and 3.3 show the SV- and SH-component RF profiles, as a function of epicentral
distance, of two exemplary stations, ACKN and EKTN (Fig. 3.1). Note that the epicentral
distance profiles are constructed using only events from the NW quadrant because the dense
data coverage in this backazimuthal range provides the most continuous sampling. On the
plots of SV-component RF stacks versus epicentral distance (Figs. 3.2a, 3.3a), three main
features are apparent for both stations.
First, the crust-mantle boundary, or Moho, is
observed as a strong positive (downward slow-to-fast) velocity gradient at ~5 s. Second, two
free-surface reverberations are observed as positive and negative impulse responses at -16
and -21 s, respectively. Third, we observe a pronounced negative impulse response at -12
s, corresponding to a depth of -100 km. This feature exhibits negative differential slowness
with respect to direct P wave, i.e., the apparent arrival time of the phase decreases as the
distance increases, which is characteristic of primary converted phases, indicating that it
originates from a mantle discontinuity. In contrast, free-surface reverberations display an
increase in the apparent arrival time as the distance increases.
Further insight into the nature of the Moho and the mantle discontinuity is provided by
examining the SH-component profiles (Figs. 3.2b, 3.3b). The lack of coherent signals at
-5 s on the SH components indicates that the Moho is a flat velocity discontinuity with
minimal or no anisotropy beneath the two stations. The same characteristics persist in all
other stations in this study. Moreover, a number of impulse responses are observed in the
SH components beneath these two stations. These include a weak negative response and
a positive one at -8 s and -18
s, respectively, beneath ACKN, and a positive-negative
dipolar response at 12-14 s beneath EKTN. These features have been previously identified
and interpreted as caused by anisotropy [Snyder, 2008]. The lack of consistent observation
of these features in nearby stations suggests that these are likely localized heterogeneous
structures.
3.2.4
Common-conversion-point (CCP) profiles
The examination of RF profiles as a function of epicentral distance reveals heterogeneous
layered structures in the lithospheric mantle beneath the stations of interest. To gain a better
understanding of the lateral extent of these structures and their possible association with
the tectonic history, we take advantage of the data of the densely-spaced POLARIS-MIT
array and construct a pseudo-2-D profile using the CCP stack [Dueker and Sheehan, 1997].
The most pronounced and consistent structure observed in the single station profiles is the
negative response at -100 km depth. Therefore, we focus our image at this particular depth
and to achieve better lateral resolution of this discontinuity.
We calculate the Ps conversion points at the depth of 100 km by tracing the converted
S wave back into that depth (Fig. 3.4), and then perform geographical binning to group
seismograms with conversion points within the same bin. Each bin is 0.10 in length along
the latitude. The resulting RF stacks for each bin are then juxtaposed to each other, and
therefore form a pseudo-2-D profile projected in an N-S direction (see Fig. 3.1 for projection
line).
We construct CCP profiles using data from different quadrants. The data from the NW
quadrant yield the clearest and most stable RF stacks, and we therefore use the corresponding
profile for the following analysis, shown in Fig. 3.5. In this profile, two main features reveal
the structure of the underlying lithosphere. First, the Moho appears as a positive (downward
slow-to-fast) velocity gradient across the entire profile at depths ranging from 34 to 41 km,
marking a southward thickening of the crust. Second, a pronounced negative (downward fastto-slow) velocity gradient dips southward from 96 km to 124 km depth beneath the center
of the craton. We refer to this signal as the central Slave seismic discontinuity. Portions
of this seismic discontinuity have been identified previously in our single station profiles
and by other investigators [Snyder, 2008; Moorkamp et al., 2007], but here it is seen as a
continuous profile. The lack of SH-response by the seismic discontinuity (See Section 3.2.5)
suggests that it is not caused by anisotropic parameters [Bostock, 1998]. To estimate the
velocity contrast of this discontinuity, we compared its amplitude with that of the Moho. The
receiver function signals are caused by seismic impedance (product of density and seismic
velocity) contrasts in Earth, but sensitivity tests show that in the context of lithospheric
structure they are mostly sensitive to velocity variations [Rychert et al., 2007]. Therefore,
their amplitudes effectively represent seismic velocity contrasts. The velocity contrast of the
Moho across the Slave craton is determined to be -20% from a seismic refraction experiment
[Fernandez Viejo and Clowes, 2003]. Accordingly, the amplitude of the central Slave seismic
discontinuity corresponds to a 9-21% velocity contrast. These values corroborate the results
from modeling of the converted SV signals, which imply a velocity contrast of ~6-23%
[Moorkamp et al., 2007].
3.2.5
The robustness and nature of the central Slave seismic discontinuity
To evaluate the nature and robustness of the central Slave seismic discontinuity, we construct
SV and SH receiver function profiles using data from three different backazimuthal ranges
(Fig. 3.6): 0-90' (NE), 90-180' (SE), and 270-360' (NW). The southwest quadrant (1802700) contains only one event that does not yield a continuous and stable receiver function
profile, thus it is excluded from the discussion. The SV-component profiles from all three
quadrants indicate the presence of a low velocity discontinuity at -100 km depth beneath
the central Slave craton. The amplitude and lateral extent of the discontinuity varies from
one quadrant to the other. This may reflect (i) variability in the number and quality of
the events - the NW quadrant comprises 36 events, versus 18 and 4 for the SE and NE
quadrants, respectively, (ii) rapid lateral variations in the properties of the discontinuity,
or (iii) a combination of these two factors. The scale of possible lateral variations can be
constrained by the locations of the Ps conversion points of Ps waves that are incident from
different quadrants (Fig. 3.4). This exercise suggests that lateral variation must occur over
scale-lengths of the order of ~20 km to account for discrepancies between quadrants. The
SH-component profiles for the various quadrants (as well as for smaller backazimuthal bins not shown) do not reveal any clear, coherent signals near 100 km depth beneath the central
Slave. This suggests that the discontinuity observed in the SV profiles is not due to seismic
anisotropy.
The uncertainty on the depth of the central Slave discontinuity is related mainly to the
background velocity model and to the coherence between individual waveforms that are
combined into stacks. To evaluate the depth uncertainty due to changes in the background
velocity model, we generate depth-migrated receiver functions using models with different
crustal and mantle Vp and Vs/Vs (Table 3.1), and determine the corresponding changes in
depth of the 100 km discontinuity signal. A realistic range of Vp and Vp/Vs for the crust
is determined on the basis of the aforementioned seismic refraction experiment [Fernandez
Viejo and Clowes, 2003]. Results from this experiment show that the crust comprises two
layers: an upper layer (0-20 km depth) with Vp ranging between 6.35-6.55 km/s, and a
lower layer (20-40 km depth) with Vp ranging between 6.7-6.9 km/s. The Vp values in the
upper crustal layer correspond to a rock composition of metagabbro and greenschist with
Vp/Vs of 1.78 [Holbrook et al., 1992]. For the lower crust, the values of Vp correspond to
felsic to mafic granulite with Vp/Vs of 1.78-1.9. For the upper mantle, the seismic velocities
are estimated based on properties of lherzolitic and harzburgitic lithologies at pressuretemperature condition of the Slave lithosphere, giving a range in Vp of 7.7-8.3 km/s and
a Vp/Vs ratio of 1.77-1.78 [Hacker et al., 2003]. We derive four end-member models based
on the estimated values of Vp and Vp/Vs (Table 3.1), and use these models to re-compute
the receiver functions for the 8 bins in which the central Slave discontinuity is most clearly
detected. The corresponding depth range of the discontinuity (i.e., the minimum of the
negative impulse) is shown by a green line in Fig. 3.7, and implies an uncertainty of <10
km on the depth of the discontinuity.
To evaluate the depth uncertainty related to waveform coherence, we use the bootstrapping technique [Efron, 1979 and estimate the 95% confidence interval on the depth of the
discontinuity (Fig. 3.7). The bootstrap test is conducted on the bins that contain at least
10 converted waves. A population of receiver function waveforms is generated by simultaneously deconvolving groups of traces randomly selected from the original dataset (with
replacement, up to the size of the original samples) and repeating the operation 100 times.
For those bins that sample fewer than 10 waves, we simply display the individual receiver
functions in conjunction with the simultaneously deconvolved waveform. This analysis indicates that variability between waveforms causes uncertainties of 2-10 km on the depth of the
discontinuity. This test also yields an estimate of uncertainty on the amplitude of converted
signal. From the bins that are amenable to bootstrapping, we find that the amplitude of the
signal exhibits an error bar of 0.017 (normalized units) based on the 95% confidence interval,
which translates into an uncertainty of 2.5-6% on the magnitude of the S velocity anomaly.
3.3
3.3.1
Seismic vs. magnetotelluric observations
The magnetotelluric observations
The magnetotelluric (MT) technique measures time variations of the magnetic and electric
fields at the Earth's surface to infer electrical resistivity structure in the subsurface. Three
MT experiments were conducted on the Slave craton during 1996-2000 by a Canada-US
team [Jones et al., 2001]. The acquired MT response data were then processed and inverted
to infer the electrical resistivity structure of the Slave craton's lithosphere. The resulting
electrical image reveals a highly conductive anomaly in the central Slave mantle [Jones et
al., 2001] (Fig. 3.5). The anomaly has a conductivity of 0.01-0.03 S/m, which is two orders
of magnitude higher than that of the surrounding mantle and therefore prevents penetration
of the signal (and imaging) at greater depths. It has a depth ranging from -80 to 120 km
with a southward dip, and may be caused by interconnected graphite films deposited along
grain boundaries, as interpreted by Jones et al. [2001].
3.3.2
Coincident seismic and electrical anomalies
When displayed simultaneously in vertical cross-section (Fig. 3.5), the seismic discontinuity
and the conductive anomaly exhibit a remarkable spatial coincidence. Although the conductive anomaly appears slightly shallower (by ~4-9 km) than the seismic discontinuity, this
difference is within reasonable error of the depth estimates associated with each technique namely 10 km for the seismic profile, as discussed in Section 3.2.5, and 15 km for the MT
image [Jones et al., 2001]. In map projection, the regions over which these structures are
observed overlap with the lateral extent of the upper-mantle harzburgitic layer [Griffin et al.,
1999] (Fig. 3.1). In the next section, we provide a review of possible causes and explanation
for both anomalies.
3.4
Causes of the seismic and electrical anomalies
Possible causes for both discontinuities include the existence of local gradients in temperature, composition or water content, and the presence of partial melt. In this section we
investigate the effect of each of these factors on the seismic and electrical properties of the
upper mantle.
3.4.1
Thermal anomaly
Experimental studies [Jackson et al., 2002] have shown that the thermal dependence of
velocity can be described by the following expression:
aln Vs
aT
ln Vsu
= aT
T
-F(a)
Q- 1(w,T) (E+PV)
7
73
~
R(3.6)
RT2
where
Q
=Ad-
m T' exp
o(E±PV)
(3.7)
RT
ira
ira
F(a) = - cot 2
2
where Vs is the shear-wave velocity, T is the temperature,
(3.8)
Q is the seismic quality
F(a) is attenuation-dependent coefficient in which F(a) --+ 1 as a -*
factor, and
0. The parameter val-
ues are chosen from previous studies [Jackson et al., 2002; Isaak, 1992; Rychert et al., 2005]
and are conservative so as to yield the minimum possible temperature change: change in the
unrelaxed shear-wavespeed with respect to temperature,
In VsU/&T
=
-8.6 x 10-5K-1;
activation energy, E=424,000 J/mol; A=730 s-a"mm ; grain size, d=1 mm; frequency (w)
dependence of attenuation, a=0.26; grain size sensitivity of attenuation, m=0.26; activation
volume, V = 6 x 10- 6m 3 /mol; pressure, P=3.16 GPa. The background temperature is set
at T=750'C [Kopylova and Caro, 2004] and the central period is To=2 s. If the observed
seismic anomaly of 9-21% Vs reduction is to be explained solely by thermal effect, the required temperature increases are therefore calculated to be 1000-2200 C. Lab measurements
[Bagdassarov et al., 2007] on the Slave peridotite xenoliths also indicate that a temperature
increase of -400'C (assuming the background temperature to be ~750'C) would be needed
to explain the electrical anomaly. Such gradients are implausible in the context of a cold,
depleted mantle root with a petrologically-inferred thermal gradient of < 6 C/km [Kopylova
and Caro, 2004], unless they were caused by a transient heating event such as a sudden
injection of kimberlitic magma. However, the last kimberlite eruption occurred -50 million
years ago [Griffin et al., 1999; Snyder and Lockhart, 2005], much too long ago to explain the
required, current day thermal anomalies.
3.4.2
Compositional anomaly
The main causes of compositional anomalies in cratonic environments are melt depletion,
juxtaposition of different mantle terranes, and mineral alteration/deposition. Xenolith stud-
ies suggest that the lithosphere of the Slave craton exhibits varying degrees of depletion in
iron and other magmaphile elements [Griffin et al., 1999]. However, the largest depletion
interface inside the lithosphere, which occurs at -150 km in the central Slave, could account
for at most ~1% shear-wavespeed anomaly [Kopylova et al., 2004]. Moreover, the effects of
iron depletion on electrical conductivity are negligible (only about one third of an order of
magnitude) [Jones et al., 2009]. Therefore, melt depletion alone is not a viable candidate to
explain the anomalies.
Juxtaposition of different mantle terranes is associated with possible mechanism of cratonic assembly and modification, such as underplating by flat subduction, plume head subcretion, and foundering of the lithospheric root. However, interfaces between such adjacent
terranes (if one omits the effects of metasomatic alteration for the time being) result either
in varying degrees of melt depletion, which as seen above is not a viable explanation, or
in transitions between mantle peridotites and eclogitized oceanic crust, which are quasiindistinguishable seismically [Rondenay et al., 2001; Rondenay et al., 2008] and electrically
[Heinson and Constable, 1992]. Interfaces between underplated terranes have been invoked to
explain anisotropic discontinuities observed in the SW Slave craton [Bostock, 1998]. However, the central Slave discontinuity does not require seismic anisotropy (Fig. 3.6), and
therefore an alternative mechanism may be required.
There is probably no single type of mineral deposit or alteration that can simultaneously
explain both anomalies. However, the anomalies can be due to different minerals that stem
from a unique process of metasomatic alteration. For the seismic anomaly, hydrated minerals
such as phlogopite, chlorite and talc are the most likely candidates, because they are stable
at P-T conditions corresponding to the location of the anomaly [Frost, 2006; Grove et al.,
2006; Pawley and Wood, 1995], and are known to reduce seismic velocity [Hacker et al., 2005;
Mainprice et al., 2008]. Strong supporting evidence for this hypothesis comes from analyses
of xenoliths from the central Slave that identify a horizon of phlogopite bearing eclogites (-50
vol% phlogopite) at ~100 km depth [Aulbach et al., 2007]. Although phlogopite has been
invoked to explain conductive anomalies in the mantle [Boerner et al., 1999], these anomalies
have much smaller magnitude than that observed here and there is no experimental data that
shows that phlogopite may generate the 0.01-0.03 S/rn anomaly observed in the central Slave
[Jones et al., 2001]. Alternatively, a more favored mineral to explain conductive anomalies
in the mantle is graphite [Jones et al., 2001; Mareschal et al., 1992]. Very small quantities of
graphite deposited as interconnected films along grain boundaries can increase the electrical
conductivity by several orders of magnitude [Mareschal et al., 1992]. There are a few mantle
xenoliths from the central Slave that contain small concentrations of graphite, although they
are not the phlogopite-rich xenoliths [Aulbach et al., 2007]. However, it is possible that a
thin lithospheric layer exhibiting high concentration of graphite may not be well-represented
in the population of xenoliths entrained to the surface due to its high level of alteration
and resulting friability [Aulbach et al., 2007].
Therefore graphite films can be retained
as a possible candidate for the electrical anomaly. Mineral deposits can therefore explain
the seismic and electrical anomalies, whereby the two types of minerals invoked could be
related to a same process of metasomatic alteration by a fluid rich in carbon. This would
be consistent with fluid cycle within a subduction zone complex where large quantities of
sediments (source of C) are being subducted [Kerrick and Connolly, 2001].
3.4.3
Water
Water can lower both seismic velocities and electrical resistivity [Karato, 1995; Karato, 1990].
Liquid water is unlikely to remain stable over periods of millions of years in a cold, stable
cratonic lithosphere, as it would eventually react with surrounding rocks to form alteration
minerals. Therefore its effects are not considered further. Alternatively, water present as hydrogen in the structure of nominally anhydrous minerals is a factor that must be considered.
In terms of seismic observations, water primarily reduces seismic velocities through anelastic
effects [Karato, 1995] - theoretical calculations suggest that increasing water content by an
order of magnitude results in an increase in seismic attenuation by a factor of two. Based
on the observed range in values that attenuation can take in the continental lithosphere
[Gung and Romanowicz, 2004] and the relationship between attenuation and seismic velocities (considering water as the only cause of attenuation, i.e., omitting additional effects from
temperature), it has been argued that water may cause at most 4.3% velocity reduction in
the lithosphere [Rychert et al., 2005]. This is not enough to explain the anomaly observed
here, but it may very well contribute to it, especially in the framework of the hypothesis
that the discontinuity is due to an alteration front caused by metasomatism. With regards
to electrical conductivity, the Archean lithosphere is considered to be cold and dry [Hirth et
al., 2000], and the Slave's upper mantle temperature is insufficient to provide high enough
activation energy to induce hydrogen diffusion that would enhance electrical conductivity to
the observed levels.
3.4.4
Partial melt
The last mechanism we investigate as a possible source for both anomalies is a layer of partial
melt at 100 km depth in the lithosphere of the central Slave. Theoretical calculations show
that a very small quantity of partial melt (<3%) is sufficient to explain both the observed
seismic velocity drop [Hammond and Humphreys, 2000] and electrical conductivity increase
[Roberts and Tyburczy, 1999]. However, the Slave mantle temperature inferred from steadystate geotherm [Kopylova and Caro, 2004] is much below the minimum solidus temperature
under water saturation conditions [Grove et al., 2006]. Moreover, as observed in section 3.4.1,
it appears unlikely to sustain a thermal perturbation required to induce partial melting at
the depth of observed anomalies.
3.4.5
Summary
In summary, we find that the most probable mechanism that can explain the coincident
seismic and electrical anomalies below the central Slave craton is a metasomatic alteration
front. Although no single mineral can produce both anomalies, secondary minerals that
formed from metasomatic alteration may be responsible. For the seismic anomaly, phlogopite
is the most likely candidate, as it is stable at P-T conditions corresponding to the location
of the anomaly [Boerner, 1999] and represents up to -50% of the bulk composition of the
eclogitic mantle xenoliths that originated ~100 km below the central Slave craton [Aulbach
et al., 2007]. This quantity of phlogopite could reduce S-velocities by -17% relative to the
host eclogite and surrounding mantle peridotite [Hacker et al., 2005], which is in agreement
with the observed seismic discontinuity. Phlogopite also increases electrical conductivity,
but experimental data shows that it can contribute at most 1% of the observed conductive
anomaly [Guseinov et al., 20051. We attribute the additional conductivity increase to thin,
interconnected films of graphite deposited along grain boundaries, the presence of which has
been invoked to explain this and other conductivity anomalies in cratonic lithosphere [Jones
et al., 2001; Mareschal et al., 1992].
3.5
Implications for the assembly of cratonic lithosphere
We propose that the two minerals, phlogopite and graphite, are related to metasomatic alteration, consistent with dehydration within a subduction zone complex where large quantities
of sediments (source of carbon) were subducted [Kerrick and Connolly, 2001]. The existence
of a metasomatized layer at 100 km depth beneath the central Slave fits well within the
context of subduction related to the assembly of Archean cratons [Snyder, 2008; Helmstaedt
and Schulze, 1989; Lee, 2006].
Two possible subduction models can cause the observed
metasomatic front and underlying ultra-depleted layer. In one model, the discontinuity is
due to the underplating of a subducted oceanic slab (Fig. 3.8a), and marks the interface
between the overriding plate and a layer of eclogitized and metasomatized oceanic crust. In
this model, the ultra-depleted layer represents harzburgitic oceanic mantle [Griffin et al.,
1999]. An alternative scenario of assembly involves an episode of shallow subduction (Fig.
3.8b), in which the mantle wedge becomes stagnant and freezes due to low subduction angle
- a process that has been documented in Phanerozoic times, for example in parts of the
Andes [Wagner et al., 2006]. In this case, the discontinuity marks the interface between the
overriding plate and a layer of ponded melt that has crystallized into eclogite and has been
metasomatized at the top of a rapidly cooling mantle wedge. Here, the ultra-depleted layer
would correspond to a mantle wedge that has undergone extensive partial melting prior to
becoming isolated from mantle circulation. We note that regardless of the exact subduction model, the discontinuity appears to represent the base of an ancient lithosphere that
preceded cratonization.
The role played by this subduction event in the assembly of the Slave craton may be
constrained by the age of minerals found in local xenoliths and by the lateral extent of the
observed discontinuity. Lu-Hf dating of clinopyroxenes in the phlogopite-bearing eclogite
indicates a minimum age of ~2.7 Ga [Aulbach et al., 2007], which approaches closely the
timing of collision between the younger eastern Slave terranes and the older Central Slave
Basement Complex (Fig. 3.1) [Davis et al., 2003]. However, this event caused a N-S suture
that runs nearly parallel to the seismic line [Davis et al., 2003], and a related subduction
should therefore be observed over the entire length of the profile, which is not the case.
Moreover, robust Re-Os age constraints on peridotitic diamonds from the ultra-depleted
layer suggest that the central Slave's lithospheric mantle down to 150 km was formed and
stabilized by 3.5 Ga [Westerlund et al., 2006]. Therefore, the central Slave discontinuity is
likely caused by metasomatized mantle from a Paleoarchean subduction zone. The presence
of a Paleoarchean cratonic lithosphere in the Slave does not conflict with the occurrence
of craton-wide arc volcanism and plutonism during ulterior Neoarchean time [Davis et al.,
2003]. It is rather an indicator that the blocks that aggregated in the final stage of cratonic
assembly may themselves have been fragments of older, thick proto-cratonic entities. This
model is consistent with the limited lateral extent of the discontinuity, which may mark the
outline of the broken block (Fig. 3.8), and is supported by the unique geochemical signature
and high concentration of diamondiferous kimberlites found in the same region [Griffin et al.,
1999]. If subduction has been active since the Paleoarchean, similar or related structures with
limited lateral extent should be detected in other parts of the craton and in other cratons
worldwide. Indeed, anisotropic seismic discontinuities have been detected below the SW edge
of the Slave craton [Bostock, 1998] and the adjacent Wopmay orogen [Mercier et al., 2008].
These have been inferred to mark Proterozoic collision-subduction events associated with
the final assembly and stabilization of a large part of the Laurentian continent [Hoffman,
1988]. Away from cratonic edges, similar seismic discontinuities have been detected by reanalysis of data from the dense Kimberley array in the Kaapvaal craton of South Africa
[Snyder et al., 2004]. Moreover, a global compilation of long-aperture seismic refraction data
indicates the existence of a widespread (albeit not ubiquitous) negative discontinuity in the
90-120 km depth range beneath several Archean cratons [Thybo and Perchuc, 1997]. This
negative discontinuity could signal the presence of structures similar to that observed in the
central Slave under other cratons worldwide. Our model predicts that subduction-related
discontinuities would have limited lateral extent. Indeed, the character of discontinuities
(i.e., their spatial extent, depth, dip, and physical properties) depends on the details of each
subduction event, including the amount of metasomatism, subduction geometry, and the
history of much younger events of accretion and (kimberlite) magmatism, which could have
altered or erased evidence of earlier subduction episodes, making their geophysical detection
difficult. Given the limited detailed geophysical sampling of cratons and the non-uniform
nature of these subduction-related discontinuities, it is thus not surprising that not many of
them have been observed to date.
3.6
Conclusions
We have synthesized coincident seismic, magnetotelluric, and petrological evidence to detect
the presence of a geophysical/petrological boundary in the lithosphere of the central Slave
craton. The spatial overlap between the seismic discontinuity and the conductive anomaly,
in conjunction with the occurrence of mantle xenoliths rich in alteration minerals representative of a metasomatic front, supports cratonic assembly by subduction and accretion of
lithospheric fragments. Although evidence of cratonic assembly is rarely preserved, these results suggest that plate tectonics was operating as early as Paleoarchean times. This study
demonstrates that further studies combining high-resolution seismic, magnetotelluric, petrological and geochemical data are key for mapping and characterizing similar discontinuities
in cratons worldwide.
3.7
Model
1
2
3
4
Table
UC
VP
6.35
6.35
6.55
6.55
UC
VP/VS
1.78
1.78
1.78
1.78
LC
VP
6.7
6.7
6.9
6.9
UM
LC
VS P
VP
7.7
1.9
7.7
1.9
8.3
1.78
1.78
8.3
UM
VP /VS
1.77
1.78
1.77
1.78
Discontinuity depth (km)
100
98
108
106
Table 3.1: End-member models used to estimate the uncertainty on the discontinuity depth
due to the background velocity model. Last column shows values for the bin centered at
64.97'N. The model values are from Fernandez Viejo and Clowes [2003], Hacker et al.
2003], and Holbrook et al. [1992]. UC, upper crust. LC, lower crust. UM, upper mantle.
3.8
Figures
.........
.
....
........
...........
-115
-110
Seismicdiscontinuity
0 POLARISstation
Electricalconductive
anomaly
0 MIT station
layer
Outlineof the harzburgitic
(from mantlexenoliths)
Outlineof the harzturgiticlayer
*-(from garnetxenocrysts)
A-A: Seismicprofile
B-B':Electricalprofile
Figure 3.1: Map of the Slave craton. The boundaries of the exposed craton are outlined in
red. Inset shows a polar projection of the world centered on the craton (green square), with
the 62 earthquakes (red circles) used for the analysis. Crustal topology and geochemical
signatures broadly subdivide the Slave craton into two distinct regions [Bleeker et al., 1999]:
The older (4.03-2.83 Ga) Central Slave Basement Complex to the west, filled in brown, and
isotopically juvenile rocks (2.67-2.6 Ga) to the east, filled in blue. The lateral extent of an
ultra-depleted harzburgitic layer of the mantle lithosphere has been inferred from petrological
analysis of mantle xenoliths [Griffin et al., 1999] (outlined in green) and geochemical analyses
of garnet xenocrysts [Griitter et al., 1999] (region between the two black dashed lines). The
seismic stations used in this study are denoted by white circles and squares for POLARIS
and MIT stations, respectively, which form a -400 km long linear array with ~20 km average
station spacing. The line A-A' marks the location of the seismic profile shown in Fig. 3.5.
The lateral extent of the central Slave seismic discontinuity is indicated by blue shading, and
that of the conductive anomaly is outlined in red. The line B-B' is the nominal projected
location of the MT array [Jones et al., 2001].
ACKNSVcomponent
Epicentral
distance
[degrees]
4-- Moho
Lithospheric
discontinuity
Moho
Id
Reverberations
ACKNSHcomponent
40
distance
[degrees]
Epicentral
80
60
70
50
-
100
90
-
s
mug-
Anisotropic
structures?
Uw
Figure 3.2: (a) SV- and (b) SH-component RF profiles of station ACKN for phases arriving
from 0-40 s after the direct P wave, plotted as a function of epicentral distance. Red and blue
colors effectively represent positive (downward slow-to-fast) and negative (downward fast to
slow) seismic velocity contrasts, respectively. Arrows point to observed consistent signals
indicating the Moho and its reverberations, as well as possible lithospheric structures.
EKTNSVcomponent
Epicentral distance [degrees]
100
4M
Moho
Lithospheric
discontinuity
I Moho
Reverberations
EKTNSHcomponent
Epicentral distance [degrees]
Anisotropic
structure?
Figure 3.3: The same as in Fig. 3.2 but for station EKTN.
84
OUR
68*
-116'
-112'
-108*
68
66*-
64'
64*
62
62*
-116'
-112
-108'
Figure 3.4: Ps conversion points at 100 km depth for events from the NW (magenta), NE
(blue), SW (yellow), and SE (green) quadrants. The red points denote the conversion points
from NW events for which the low-velocity discontinuity is observed.
(b),
(a)
100
CL
ic
ci)
8~
B i-
34
116
400
1350
4600
16000
54000 B
Resistivity (i'm)
160
180
200
220
Horizontal distance (kin)
Figure 3.5: Teleseismic receiver-function (RF) and magnetotelluric (MT) profiles across the
Slave craton. (a) Color-coded seismic profile along the line A-A' (Fig. 3.1). Red and
blue colors effectively represent positive (downward slow-to-fast) and negative (downward
fast to slow) seismic velocity contrasts, respectively. M, Moho; RI and R2, free-surface
reverberations. (b) MT-derived resistivity profile [Jones et al., 2001] across the line B-B'
shows the central Slave conductive anomaly (red to yellow) at 80-120 depth. The solid black
line highlights the base of the conductive anomaly within the region of interest (dashed black
box), repeated to scale in RF profiles for comparison. (c) Close-up RF traces for bins in the
region of interest.
0
E
SH component
SV component
-2
100
200
200
300
300
400
400
500
0500
600
600
700
700
80000
250
200
150
100
Horizontal distance (kin)
50
300
250
200
150
100
Horizontal distance (kn)
50
01
0L
2
mI LI
10
E
400
W
-
*
300
g
o
WAR
i
c
r
EmE
500
)0
60 10
--
700
200
250
150
100
Horizontal distance (kn)
50
u
__-
-
-
-
70
-..
0
~
via
i
50
600
800
40 0
10*
300
l
10
200
-
-:
0
350
-
M--
100
300
350
80
0
50
250
150
200
100
Horizontal distance (kin)
300
350
300
350
0-
0
300
100
200
200
-
300
E
r-n
-
-
E 700
4500
300
-~400
500
600
-
-
-
-
--
600
L
50
-ii
700-
700
Ofnn
0
.
-
-
-
--
250
200
150
100
Horizontal distance (kn)
800
0
300
50
250
150
200
100
Horizontal distance (kn)
Figure 3.6: SV- and SH-component receiver function profiles constructed by simultaneous
deconvolution of converted P-to-S (Ps) waves binned according to common conversion points
along line A-A' (Fig. 3.1), using data from (a) NW; (b) NE; and (c) SE quadrants. Red
and blue colors represent positive (downward slow-to-fast) and negative (downward fast to
slow) seismic velocity contrasts, respectively. In all three SV components, a coherent positive
discontinuity corresponding to the Moho is observed across the entire array at -40 km depth,
and a negative discontinuity is observed at -100 km depth below the central Slave (green
arrows). No clear coherent signal is observed across the array in the SH profile.
87
bin 64.57 IN(3)
bin 64.67 *N(44)
0
-
01
00
<
0-
0
50
100
150
200
-0.1
250
0
50
100
bin6477*N(26)
150
200
250
200
250
200
250
200
250
bin64.87
N(26)
0
E
0
0
50
<-0.1
1
100
150
200
250
0
50
100
bin64.97
*N(5)
150
bin65.07
IN(17)
S0.1
EE
0
50
100
0'
E
51
150
200
250
0
.
50
bin65.17
ON
(29)
0.1
100
150
bin65.27
IN(6)
-01'
0
<--0.10
0
50
100
150
Depth (kn)
200
250
0
50
100
150
Depth (k1)
Figure 3.7: Depth and amplitude uncertainty of the central Slave seismic discontinuity. Each
panel shows SV receiver functions (in blue), simultaneously deconvolved and stacked, for the
8 bins at 100 km depth in which the discontinuity is observed. The central latitude of the
bins is indicated above each panel, along with the number of waves contributing to each bin
(in parentheses). In bins containing more than 10 events, a 95% confidence interval for the
resulting receiver functions was obtained by bootstrapping, and is indicated by grey lines.
For those bins that sample fewer than 10 waves, we display the individual receiver functions
(grey dashed lines) to provide a visual assessment of the inter-trace variability. Note that
these receiver functions are individually normalized. The red triangles indicate the depth of
the discontinuity estimated from the receiver functions. The green lines denote the range of
depths obtained by recalculating the receiver functions for a range of possible background
velocity models (Table 3.1). The magenta lines represent the 95% confidence interval on the
depth estimate of the central Slave discontinuity based on bootstrapping.
Ba
Crust
Uppermost
Mantle
(a)
(a
Proto-lithosphere
r
Metasomatic
\
boundary
)
Oceanic crust
Edogite Layer1 _
-
UltraLdepleted
Layer-
Layer2
c
antle
1
-
Lherzolitic
Layer
-
Present
LAB
(b)
/Metsomatic(
Crust
budary
Uppermost
Mantle
Eclogite Layeri1
Ultra-depleted
Layer
Edlogite
Layer2
Lherzolitic
Layer
Present
LAB
Proto-lithosphere
--
oceanic Crust
Oencmnl
Oceanic m
Figure 3.8: Two models of cratonic assembly by subduction that explain the seismic-electrical
discontinuity in the central Slave's lithosphere. The tectonic elements of each model are
assigned to petrological horizons (labeled axis to the left of both panels) inferred from mantle xenoliths [Griffin et al., 1999; Aulbach et al., 2007]. LAB: Lithosphere-asthenosphere
boundary. (a) Accretion of a subducted slab at the base of a pre-cratonic lithosphere. The
seismic-electrical discontinuity is associated with the interface between the proto-lithosphere
and the altered crust of the subducted slab, while the ultra-depleted layer (UDL) represents
the subcrustal lithospheric portion of this slab (13). An additional subcretion event involving
subduction [Bostock, 1998] and/or a mantle plume [Griffin et al., 1999] is required in this
model to account for the lower layers of the cratonic lithosphere. (b) Low-angle subduction
beneath the pre-cratonic lithosphere. The seismic-electrical discontinuity is associated with
the interface between the base of the proto-lithosphere and a layer of crystallized and metasomatized partial melt. The UDL is due to extensive melting in the mantle wedge [Davis
et al., 2003]. The lower eclogitic and lherzolitic layers represent the crust and mantle of the
subducted slab. The inferred subduction event (either (a) or (b)) led to the assembly of a
proto-cratonic lithosphere that was at least 150 km thick by 3.5 Ga, based on peridotitic
diamonds found in the UDL [Westerlund et al., 2006]. We propose that the central Slave
comprises a broken fragment of this proto-craton (see vertical broken lines), which was aggregated to other continental fragments during the Neoarchean assembly of the Slave craton
[Davis et al., 2003; Snyder, 2008].
90
Chapter 4
Array-conditioned deconvolution of
multiple component teleseismic
recordings
Abstract
We investigate the applicability of an array-conditioned deconvolution technique, developed
for analyzing borehole seismic exploration data, to teleseismic receiver functions and data
preprocessing steps for scattered wavefield imaging. This multichannel deconvolution technique constructs an approximate inverse filter to the estimated source signature by solving
an overdetermined set of deconvolution equations in a least-squares optimal sense, using an
array of receivers detecting a common source. We find that this technique improves the
efficiency and automation of receiver function calculation and data preprocessing workflow.
We apply this technique to synthetic experiments and to teleseismic data recorded in a dense
array in northern Canada. Our results show that this optimal deconvolution automatically
determines and subsequently attenuates the noise from data, enhancing P-to-S converted
phases in seismograms with various noise levels. In this context, the array-conditioned detThis chapter is in review at Geophys. J. Int. as: Chen, C.-W., D. E. Miller, H. A. Djikpesse, J. B. U.
Haldorsen, and S. Rondenay, Geophys. J. Int., in review, 2009.
convolution presents a new, effective and automatic means for processing large amounts of
array data, as it does not require any ad-hoc regularization; the regularization is achieved
naturally by using the noise present in the array itself.
4.1
Introduction
A number of methodologies have been developed over the years to analyze converted seismic
waves, ranging from single station applications to high-resolution imaging using dense arrays
of broadband seismometers. Such developments have been made possible by the increased
availability of teleseismic data recorded at dense broadband seismic arrays. We refer the
reader to Rondenay [2009] for a comprehensive review of processing steps that have been
developed to obtain images of discontinuities in the Earth's subsurface from data consisting
of seismograms sampled by dense arrays of recorders. Of particular interest are methods
focused on P-to-S (Ps) conversion in the coda of teleseismic P waves, due to its generally
high signal-to-noise ratio and lack of contamination from later arriving primary phases. Such
signal was first used for direct imaging in landmark studies by Vinnik (1977) and Langston
(1979).
To increase the signal-to-noise ratio of converted phases, these authors combined
records from multiple sources by stacking traces that were source-normalized and time-shifted
according to incidence angle. The term receiver function (RF) was introduced by Langston
(1979) to describe these normalized records of converted waves and their stacks.
A key step in the RF processing chain is the 'source-normalization' which requires the
construction and application of a deconvolution operator to remove the extended earthquake
source function, replacing it with an approximate impulse. Fig. 4.1 illustrates this step using
data from the POLARIS-MIT seismic array in the Slave province, Canada. Fig. 4.1a shows
the P and SV component data from a single earthquake recorded at 18 stations, after
application of the free-surface transfer matrix method [Kennett, 1991] to partition the threecomponent records into P-SV-SH wavefields. The effective source function clearly rings for
more than a minute, mainly due to reverberation in the crust near the source. Fig. 4.1b
shows the same data after application of a deconvolution operator derived by the method of
Haldorsen et al.
[1994,
1995], as discussed herein. The deconvolved SV data show a clear
arrival at -4.8 seconds, resulting from P to SV conversion at the Moho discontinuity. It is
the purpose of this paper to discuss this deconvolution method in the context of teleseismic
data and to describe its application to data from the POLARIS-MIT array.
4.2
Methodologies
Our study focuses on investigating the effectiveness of the array-conditioned deconvolution,
in comparison with conventional frequency-domain deconvolution method, i.e., the waterlevel
deconvolution. Thus, in this section, we first provide a review of the waterlevel deconvolution
method, and then introduce the array-conditioned deconvolution.
4.2.1
Waterlevel deconvolution
Deconvolution is usually cast as a solution to the forward expression (c.f. Rondenay, 2009,
Section 5):
d(t) = w(t) * r(t) + n(t)
(4.1)
in which the observed signal d(t) is expressed as the convolution of an Earth impulse response
r(t) with a source signature w(t). In eq. (4.1), n(t) represents residual energy, typically assumed to be Gaussian random noise with zero-mean. The normalization process to solve for
r(t) involves deconvolving w(t) from d(t). For the ideal case, i.e., there is no noise, the source
signature and the observed signal are known and not frequency band-limited, this problem
may be solved directly by division in the frequency domain. However, the deconvolution
procedure is usually ill-posed because of the presence of random noise, frequency bandwidth
limitation, and inaccuracies in estimation of source signature. Therefore, the process has to
be regularized. This is usually achieved in the frequency domain by prewhitening the amplitude spectrum of the source wavelet, to avoid small amplitudes that would cause numerical
instabilities and ringing in the deconvolved signal. Hereafter, we will only be using signals
in the frequency domain. For simplicity, we shall keep the same notation for the variables
in eq. (4.1).
An approximate solution of the impulse response f is expressed as [e.g., Berkhout, 1977]:
f(w) =
w*d(w)
w(w)w*(w) +
(4.2)
6
where the asterisk denotes the complex conjugate, w is angular frequency and 6 is a regularization factor. The factor, sometimes termed waterlevel [Clayton and Wiggins, 1976],
represents the expected noise power. When 6 is zero, eq. (4.2) is a simple spectral division
solving the equation d(w) = w(w) r(w). When 6 is large, the denominator in eq. (4.2) is
approximately constant and eq. (4.2) becomes a correlation with the estimated source.
The method assumes that the noise spectrum is white and requires either independent
knowledge of the noise power or a search for the 'best' parameter that stabilizes the deconvolution process. This is usually done on a trial and error basis, and thus is subjective
and labor-intensive. It is desirable to introduce more objective means to estimate the regularization parameter. For example, Bostock [1998] considered a family of recorded traces
dm(w) and associated source estimates wm(w) and proposed choosing 6 by minimizing the
generalized cross-validation function GCV(6) shown as
GCV(6)
MLZh [d, (wi)
=M
E
-
Wm,(Wi) T(Wi)]12
1
[ML -E'_Xo)]
(4.3)
where
X(w)
rnzziWrn(W)Wn(W)
=
E
(4.4)
-1 Wm(W)W*(W) + 6
with M denoting the number of traces, and L is the number of frequencies represented in
the discrete Fourier transform. This process does not require any assumption concerning the
noise level in the data, but it still assumes a white noise spectrum and requires an iterative
grid search to obtain the value for 6 (within a given range) that results in the minimal GCV.
4.2.2
Array-conditioned deconvolution
Haldorsen et al. [1994, 1995] described a method for exploiting the redundancy in seismic
array data to obtain an optimized deconvolution filter by using the data to estimate both
the source and noise spectra without assuming that either is white. That method may be
summarized as follows.
Suppose we are given data recorded at an array of receivers and time-shifted and normalized such that each observed trace dm(t) can be assumed to contain a common source
signature w(t), superposed with a variable 'noise' nm. That is, we are given a subscripted
array of equations, like eq. (4.1):
dm(t) = w(t) + n,,(t)
(4.5)
Here r(t) from eq. (4.1) is assumed to be an impulse. Thus, all aligned signals contributing
to the source estimation are assumed to be part of the source signature. Additional copies
shifted and misaligned (e.g., multipath signal arriving obliquely across the array) are formally
part of the 'noise', but will be preserved and spiked insofar as they carry the same signature
as the aligned signal. Similarly, the filter derived from the aligned P data can be applied to
SV data to compress and enhance the converted signal carrying the same source signature,
yielding a compressed arrival with the delay relative to the aligned signal preserved by the
deconvolution operator.
In the frequency domain, this data model is written as a set of equations:
dm(w)
=
w(w) + nm(w).
(4.6)
A deconvolution filter W can be determined, independently for each w, as the solution
to the set of eq. (4.6) augmented by the equations
W(w)dm(w) = 1.
(4.7)
These equations have the least-squares solution
W(w)
(W)
=
Er
(4.8)
(w)'
where the caret denotes estimate, and ET(w) is the average total energy of the raw traces:
M
dm(w)| 2 .
ET(w) = M 1
(4.9)
m=1
Substituting from eq. (4.6), eq. (4.8) can be rewritten as
W(w) =(W)
jz(w)j2 + EN(W)
(4.10)
where
M
EN(w)
m
- w (w)
(4.11)
m=1
This agrees with eq. (4.2) when EN(w) is a constant, independent of w, and thus represents
a data-adaptive solution to the filter regularization problem, which is applicable in a wider
context than is the waterlevel deconvolution.
The properties of this optimum filter are discussed in detail in Haldorsen et al. [1994].
For example, one can rearrange eq. (4.8) to give
D(w),
W(w) =
(4.12)
where the frequency-domain semblance D(w) is given by
D(w)
()
E (u)
96
.
(4.13)
The optimum filter in eq. (4.12) is thus recognized as a spectral division filter, multiplied
by the semblance, which acts as a data adaptive, band-limiting filter attenuating frequencies
where the signal-to-noise ratio is small.
In the original discussion, the source estimate and the filter construction were derived
together, assuming that all the data from a single recorded component were used in constructing both the numerator and the denominator of the filter (eq. (4.8)).
Subsequent
experience has shown that these two aspects of the filter construction can be uncoupled
and that the traces used to estimate 7i may be distinct from those used in estimating ET.
Moreover, the filter itself may be applied to traces that are distinct from the design traces.
In particular, when, as in the case of teleseimic data, it may be reasonably assumed that a
complicated packet of energy is converted from P to S somewhere near the receiver array,
the P arrivals can be aligned and used to estimate the signature while the complete ensemble
of component data is used in estimating the total energy. Note, however, that stability is
only guaranteed if the source estimation traces are included in the estimate for total energy.
In the next section, we carry out synthetic experiments to evaluate the performance of
the array-conditioned deconvolution and to compare the results with those using waterlevel
deconvolution.
4.3
Synthetic experiments
We construct the synthetic waveforms by using forward-modeled Earth impulse responses,
as well as observed seismograms from the 16 August 2005 earthquake (mb=6.5) in Japan,
recorded at 18 stations of the POLARIS-MIT array in the Slave province, Canada. We
perform deconvolution on this synthetic dataset with the addition of various levels of noise.
The procedure of the synthetic waveform construction is as follows:
(1) We compute the synthetic P and SV impulse responses using Zoeppritz reflection
and transmission coefficients [e.g., Aki and Richards, 2002] calculated for a simple two-layer
velocity model and a single horizontal slowness representative of the field data. Fig. 4.2a
shows the result of this computation. The P component has the direct P wave (P) and the
first order multiples that end with P (PPP,PSP, SPP, $$#). The S component has the
converted S wave (S) and the first order multiples that end with S (PPS, PSS, SPS, $$$)
Note that each P arrival has a corresponding S arrival obtained by replacing the last P
segment with an S segment, hence the relative time delay is the same in all cases.
(2) We align the P-component seismograms of the Japan event and derive a 'synthetic'
source signature through diversity stack [Embree, 1968] of the aligned seismograms. This
synthetic source signature thus represents the noise-free common source signal (Fig. 4.2b).
(3) We convolve the synthetic source signal with the synthetic P and SV impulse responses to yield the noise-free synthetic data (Fig. 4.2c).
(4) For each trace, we extract 300-second long data before the P arrival from the P- and
SV-component seismograms recorded by the POLARIS-MIT array, to be representative of
background noise. We also subtract the synthetic source signal from the respective observed
P-component seismogram, and the residuals obtained are representative of additional incoherent noise between traces. We combine these two types of noise, randomly shift them in
the time domain, and add a scaling factor A for controling the amplitude, before adding
them to the noise-free synthetic data. As such, the complete synthetic data model for the
P-component (d,(t)) and SV-component (d,,(t)) can be described as, respectively,
dp(t)
tb (t) * gp(t) + AN,(t);
(4.14)
and
dsv(t) = f(t) * gsv(t) + ANsv(t ),
(4.15)
where gp(t) and gev(t) are the synthetic P and SV impulse responses, and N,(t) and Nsv(t)
are the total (combined and shifted) noise in P and SV components. By changing the scaling
factor A, we are able to generate synthetic data with various noise levels so as to test the
effectiveness of the deconvolution methods.
Fig. 4.3 summarizes the results of the synthetic experiments. Fig. 4.3a shows the synthetic array data (P and SV components) with noise level A=1. Fig. 4.3b and c show the
deconvolution results using the waterlevel method with the GCV-derived 6 parameter and
with waterlevel of 1% of the maximum amplitude of the source signature estimate, respectively. Fig. 4.3d shows the result using the array deconvolution. This synthetic test allows
us to make the following observations. First, the GCV yields trace-dependent 6 values that
are equivalent to 0.001 to 0.01 percent of the maximum amplitude of the source estimate.
Second, while the waterlevel method in general recovers the impulse response in most SV
traces, it fails to resolve traces that are anomalously noisy, for instance, traces 3 and 17.
Furthermore, as the waterlevel factor increases, the deconvolved signal broadens and loses
resolution. This is expected because using a higher waterlevel amounts to prewhitening more
high frequency signals. In a sense, it becomes a low-pass filter, removing high frequency content in the data. Conventionally, this process of iterating over a number of waterlevel factors
is conducted and visual inspection is required until a 'best' waterlevel is determined. On
the other hand, the array deconvolution (Fig. 4.3d) does not require any iterative process or
human intervention, and stabilizes noisy traces while better resolving the impulse response
consistently across the array. Here, ET(w) is calculated using P-component data.
Note that, in the deconvolution process, (v(t)*gp(t) becomes the effective source signature,
and that relative amplitudes in the deconvolved SV data are slightly altered from those of
g8 (t). This is an issue for any deconvolution process. The consistency achieved by using
a single deconvolution operator for all receivers should enable further analysis beyond the
scope of this paper.
Similar results are observed when we increase the noise in the synthetic data. The
waterlevel deconvolution becomes unstable, i.e., the deconvolved traces are more ringing,
whereas the array deconvolution still achieves similar resolution.
One way to evaluate the performance of the deconvolution filters is to measure the variance between the deconvolved signals across the array. We calculate the variance by summing
the square of the difference between each trace and the mean trace. The corresponding variance of each deconvolved data section is shown as the number in the parentheses above
each panel in Fig. 4.3. The array deconvolution yields a much better, i.e., smaller, variance
than those from the other two approaches. For waterlevel deconvolution, we note that there
appears to be a trade-off between variance and broadening of the deconvolved signal; larger
waterlevel results in smaller variance but less sharp impulse. In contrast, array-coditioned
deconvolution always achieves small variance and sharper impulse. Fig. 4.4 shows the comparison of the amplitude spectrum of deconvolved signals of trace 3 (Fig. 4.3), from the three
approaches. The 'raw' synthetic data is dominated by low frequency noise, and the array
deconvolution, compared with the other two approaches, achieves a better resolution of the
impulse without sacrificing much high frequency content.
In this section, we have demonstrated the effectiveness of the array-conditioned deconvolution, especially for noisy data. In the following section, we will apply this deconvolution
to a field dataset of the Slave province. In this example, we focus our demonstration on the
P- and SV-component seismograms, but note that the method is readily applicable to SH
components as well.
4.4
Application to the Slave craton array data
We use seismic array data recorded in the Slave province, an Archean craton which is located
in the northwestern Canadian Shield. The Slave craton has been the subject of intensive
geophysical and petrological studies due to its longevity and the presence of abundant diamondiferous kimberlites. The POLARIS-MIT seismic array (Fig. 4.5) in the Slave craton
consists of 30 seismic stations, each equipped with a three-component broadband seismometer. A previous receiver-function study [Chen et al., 2008] identified a distinct crust-mantle
boundary, or Moho, at ~4.8 s across the array, using waterlevel deconvolution and commonconversion-depth stacks of high quality data from 62 teleseismic events with magnitude mb
5.8 recorded during 2004-2006. Now, using the new array-conditioned deconvolution method,
we are able to analyze data from 135 events with magnitude mb > 5.5 (Fig. 4.5) during the
same recording period. We use the event locations provided by the USGS PDE catalog, and
rotate the horizontal-component data to radial and transverse components (vertical component remains the same). We subsequently partition the components into P, SV, and SH
wavefields by the free surface transfer matrix [Kennett, 1991].
100
After wavefield partition,
we align the data by the predicted arrival times calculated in a 1-D global reference model
(e.g., iasp9l, Kennett et al. 1991). The source signature is estimated from the P-component
by diversity stack [Embree, 1968], and the noise energy is calculated from the P-component
data. The deconvolution is then performed to yield deconvolved P and S signals. We observe
that the deconvolved P impulses across the array often show time differences between each
other, indicating inaccurate original reference alignment. Therefore, in practice, the deconvolved P impulses are iteratively realigned by adjusting their time lags, and a subsequent
deconvolution is performed to yield the final results.
Fig. 4.6 shows the raw data of four example earthquakes. These raw data show different
characteristics of the coherently aligned signals in the P components, marking the various
earthquake source signatures, as well as different patterns and amplitudes of the background
noise in the SV components. Fig. 4.7 shows their deconvolved results from array deconvolution, compared with those from waterlevel deconvolution. The results of earthquake
data are consistent with those of synthetic tests. Both deconvolution methods result in
delta-function-like and well-aligned P signals; however, the array-deconvolved ones appear
sharper, indicating the effectiveness of the array deconvolution in collapsing the signal into a
spike. On the SV components, coherent signals at -4.8 s can be observed in all data sections,
representing the conversion at the Moho. However, the array-deconvolved data appear more
stable and consistent throughout, while the corresponding waterlevel-deconvolved data are
less so. In addition, a number of differences are worth noting. First, the array-deconvolved
traces contain more high frequency energy than do the waterlevel-deconvolved ones. Second,
there are traces that cannot be well resolved by waterlevel deconvolution and that result in
anomalously low frequency signal (e.g., in the Costa Rica event, SV traces 1 and 8; in the
Tonga event, traces 3, 6, and 8). In contrast, array deconvolution in general achieves more
stability. We also calculate the variance, as defined in the synthetic tests, of the deconvolved
data (shown as the number in the parentheses above each panel). In these four examples,
the array-deconvolved data all have much smaller variances (at least one order of magnitude
smaller) than those of the waterlevel-deconvolved ones. This shows the advantage of array
deconvolution in extracting coherent signals across array while attenuating noise. An addi101
tional advantage can be noted by examining the Tonga event (Fig. 4.7d). This event has a
magnitude mb
=
6.3, but the noisy SV components with anomalous low frequency patches
has prevented it from being used in the previous receiver function analysis. Using the array
deconvolution, however, we are able to attain more stable and thus usable signals from this
event.
The processing procedure is implemented for the whole dataset of 135 events. In Fig. 4.8,
we show the deconvolved SV traces as a function of backazimuth at five receivers. We
observe that, in addition to coherent signals corresponding to the Moho, there appears to
be various coherent signals at different times between the surface (t
(t
=
=
Os) and the Moho
4.8s) from receiver to receiver. These variations suggest the presence of local crustal
heterogeneities beneath each receiver, and were not observed before when only limited high
quality seismic records were utilized. We also plot the deconvolved data as a function of
earthquake magnitude. An example using data from station ACKN is shown in Fig. 4.9. We
observe that the Moho signal appears consistently visible in the entire magnitude range, and
does not degrade at smaller magnitudes (5.5 < mb < 5.8). This promising result indicates
that the array deconvolution can be readily applied to earthquakes with smaller magnitudes.
Further analysis of this application is the topic of a separate paper.
In closing, we note that, traditionally, the deconvolution has been achieved in an iterative
manner, whether it is to find a 'best' regularization parameter in the frequency-domain
deconvolution, or to minimize the difference between observed and modeled data in the
time-domain deconvolution (e.g., Gurrola et al., 1995; Ligorria and Ammon, 1999). In this
context, the array-conditioned deconvolution presents a new, effective and automatic means
for processing large amounts of array data, as it does not require any ad-hoc regularization;
the regularization is achieved naturally by using the noise present in the array itself.
4.5
Conclusions
The application of the array-conditioned deconvolution improves the efficiency and automation of the deconvolution process that is an essential step in receiver function analysis and in
102
data preprocessing for imaging of scattered waves. Synthetic experiments demonstrate the
effectiveness of the deconvolution technique, especially for noisy data. Application of this
technique to a teleseismic dataset from the Slave craton yields a deconvolved data section
that clearly identifies the Ps conversion at the Moho, and suggests the presence of local
crustal heterogeneities beneath each receiver. The performance of the array deconvolution
with noisy data promises the potential of exploiting earthquakes with smaller magnitudes,
which would increase the number of usable sources, thus providing more comprehensive
azimuthal coverage than was possible before.
103
4.6
Figures
(a)
(b)
Event:20052280246
Event:20052280246
300
Time (sec)
Time (sec)
Figure 4.1: (a) The P- and SV-component data of a Japan (mb=6.5, 36-km deep) earthquake. (b) The deconvolved P- and SV-component data of (a).
104
0.2'
-0.6-
-100
-150
-50
0
Time(se)
50
100
150
50
0
Time(sec)
50
100
150
(C)
-150
1
Figure 4.2: (a) The synthetic P- and SV-component impulse responses for an incident P
wave of ray parameter p=0.0 6 s/km, sampling an isotropic two-layer model. The model
cosists of a 40 km-thick horizontal layer (ao=6.6 km/s, #0=3.7 km/s, po= 26 0 0 kg/m 3 ) over
a half space (a 1 =8.1 km/s, /1=4.5 km/s, po= 3 5 00 kg/m 3 ). (b) The source signature estimate
used to construct the synthetic array data. (c) The noise-free synthetic data constructed
from convolving (a) with (b).
105
(a) Synthetics A = 1
(c) Waterlevel 6
=
Time
(
.0)
(d) Array decon (0.0003)
1 % (0.164)
-5
0
5
10
15
20
25
30
-5
0
5
10
15
Time(sec)
20
25
30
15
-10
-5
0
5
10
15
Time
(se)
20
25
30
35
-10
35
Figure 4.3: Summary of the synthetic experiments. (a) The synthetic array data of A=1.
The deconvolved data section using (b) waterlevel deconvolution with GCV-derived 6; (c)
waterlevel deconvolution with the factor of 1%; and (d) array-conditioned deconvolution.
The number in the parentheses indicates the corresponding variance.
106
Waterlevel
GCV
E
2.5
Array decon
Raw
0
0.2
0.4
0.6
1.2
1
0.8
Frequency (Hz)
1.4
1.6
1.8
Figure 4.4: Comparison of the amplitude spectra of the deconvolved SV signals of trace 3,
along with that of the 'raw' synthetic trace.
107
(b)
(a)
68
-
68
Figure 4.5: (a) The earthquake event distribution projected with the Slave craton in the
center (yellow diamond). The red circles denote events used in the previous receiver function
study [Chen et al., 2008]. The white circles denote the additional events that are analyzed
by the array deconvolution. The green circles denote the four exemplary events whose data
are shown in Fig. 4.6. The combined dataset includes a total of 135 events. (b) Simplified
geological map of the Slave craton (outlined in red). The yellow shaded area is the central
Slave basement complex [CSBC; Bleeker et al., 1999], which is the oldest portion (2.6-4 Ga)
of the craton. The blue shaded area denotes the eastern Slave craton where is covered by
juvenile crust. The seismic stations used in this study are denoted in squares (MIT stations)
and circles (POLARIS stations). The five stations denoted in blue are those whose data are
shown in Fig. 4.8. From south to north, these stations are BOXN, LGSN, LDGN, EKTN,
and ACKN.
108
(b) Crete, Greece (mb=5.9; d=25 km)
(a) Honshu, Japan (mb=6.5; d=36 km)
Event:20040770521
Event:20052280246
Time (sec)
(d) Tonga Islands (mb=6.3; d=130 km)
(c) Costa Rica (mb=5.8; d=9 km)
Event:20041810701
Event:20040251143
300
Time(sec)
300
Time (sec)
Figure 4.6: The raw data (P and SV components) of four exemplary earthquakes from (a)
Honshu, Japan; (b) Crete, Greece; (c) Costa Rica; and (d) Tonga Islands. Note the traces
are individually normalized. The magnitude (mb) and depth of each event are also indicated.
109
(a) Japan (0.0004)
6 = 0.1 % (0.032)
(b) Crete, Greece (0. 0023)
8w
-'c -s
1 % (0. 013)
-52
-
-0Q702
t
6
s
1
15 8)
-6
0
10
15
(c) Costa Rica (0.0018)
6-=0.1 % (0.033)
(d) Tonga Islands (0.0025)
6 = 0.01 % (0.129)
2)
Figure 4.7: The deconvolved data of the four earthquakes shown in Fig. 4.6. The left
column shows the results from array-conditioned deconvolution. The right column shows
the corresponding results from waterlevel deconvolution, denoted with the waterlevel factor
used. The number in the parentheses indicates the corresponding variance.
110
(a)
(b)
for
receiver:
9 : LGSN
Deonolved
SV
6: BOXN
Deoolved SVo receiver:
100
60--
0
-5
10
5
0
2-
-
---
20--
15
0
-
20
(d)
(c)
recev
d SVfor,,
Do
11:LDGN
0
80
50
400
1~10
200
10
-5
0
10
5
15
(-)
20
(S~OO
0
--
(e)
0
100
5P
0
-5
10
15
2
80
so
0
-5
0
10
5
15
20
(Sec)
0
0
10
10
azim h (*)
Back
Figure 4.8: The deconvolved SV data sections of five receivers. (a) BOXN; (b) LGSN;
(c) LDGN; (d) EKTN; (e) ACKN. (f) The representative backazimuthal distribution of the
teleseismic events recorded at this array. A majority of earthquakes are located at the
western Pacific subduction zones (around 300').
111
(a)
(b)
Deconvolved
SV for receiver:15 : ACKN
Magnitudedistribution
for eventsrecordedat receiver:15 ACKN
120
100
80
---.....
-. ....... .
... .-.-.-.-.-- .-- -.--.
. ..
-
.-.
.-.- --
E
60 --..
..- - ....
-. ....
-.-...
...
-.
40 --..-.-
.-
10
15
20
Time (sec)
5.4
. . . ..
--
. ....
. . . -. . .
..
-.
. .- . .
. . .
-
-
. . . .
-
. -. . .
Iw
20
5
-....
....
5.6
5.8
6
6.2
6.4
Magnitudein mb
6.6
6.8
7
7.2
Figure 4.9: (a) The deconvolved SV data section of station ACKN plotted as a function of
earthquake magnitude. (b) The distribution of traces according to magnitude.
112
Chapter 5
Conclusions
In this dissertation, I used a number of seismological techniques to investigate the lithospheric structure of the Slave craton and used the results to infer the mechanisms that led
to its assembly. The results from surface wave tomography provide new constraints on the
thickness of the lithosphere, its lithology and its mineral/structural fabric. The receiverfunction analysis identifies a seismic discontinuity at -100km depth. This discontinuity is
interpreted, in conjunction with magnetotelluric, petrological and geochemical results, as
the remnant of a subduction front that was active during the Archean (< 3.5 Ga) and led
to the assembly of the central Slave craton. In parallel to these applied seismic studies, I
implemented an exploration-based deconvolution technique that improves the efficiency of
receiver function calculations. The key findings and contributions of this dissertation are
summarized as follows.
Chapter 2 presents the first surface wave tomography analysis of the Slave province to be
done at craton-scale. The study outlines lateral variations of Rayleigh wave phase velocities
that can be associated with surface geological expressions. The one-dimensional shear-wave
velocity profile reveals the existence of a lithospheric lid with an average velocity that is
4% higher than the global average. This reflects a particularly depleted and unperturbed
cratonic lithosphere, which corroborates petrological and geochemical constraints. Furthermore, the depth dependence of azimuthal anisotropy reveals distinct domains throughout the
lithospheric column, reflecting the influences of various tectonic processes on the evolution
113
of the cratonic lithosphere. Finally, a laterally-confined high shear-wave velocity region is
observed below ~100 km depth in the central Slave craton. This region is spatially coincident with both the geochemically-defined ultra-depleted mantle layer and the diamondiferous
kimberlite cluster. These results point to a distinct origin of the lithosphere underlying the
central Slave craton.
In Chapter 3, I produce a high-resolution receiver-function image across the Slave craton
that resolves fine structures in the lithosphere. A prominent seismic discontinuity is observed
at ~100 km depth below the central Slave, spatially coincident with an electrical conductive
layer observed in magnetotelluric models. Analyses of mantle xenoliths collected in the
central Slave identify a layer of mineral alteration at the same depth as the seismic/electrical
discontinuity and place an age constraint of >3.5 Ga on the emplacement of the underlying
lithospheric mantle. The synthesis of these geophysical and geological observations suggests
that the -100 km boundary represents a metasomatic front associated with an Archean
subduction event. This finding suggests that plate tectonic processes, i.e., subduction, were
operating as early as Paleoarchean times.
The finite lateral extent of the seismic/electrical discontinuity further coincides with the
region of high shear-wave velocity indentified in the central Slave by surface wave tomography. This reinforces the hypothesis proposed in Chapter 3 that this region represents the
broken block of a thick (>150 km) continent with ultra-depleted roots that was formed by
subduction ~3.5 billion years ago and was later broken apart.
Chapter 4 presents an array-conditioned deconvolution technique that constructs an optimal deconvolution filter based on the estimate of a common source signature, and is naturally
regularized using the noise present in the data itself. Without any ad-hoc choice of regularization parameters, this technique proves especially suitable for analyzing large amounts of
array data, such as those from the USArray.
This dissertation demonstrates that multi-disciplinary investigations using geophysical,
geochemical, and petrological analyses are key to better constraining the origin of the
Archean cratonic lithosphere. Pursuing such integrated approaches in combination with
the development of new effective data processing technique will further enable us to char114
acterize the interior structure of the whole planet, from the ancient stable cratons to active
subduction zones, and to better understand the tectonic processes that shape the Earth.
115
116
Appendix A
Tables
117
Station name
ACKN
BOXN
CAMN
CARNa
COKN a
COWN
DSMN
DVKNb
EKTN
EXLN a
FINNS
GBLN
GLWN
IHLN
ILKN
KNDN
KNSN a
LDGN
LGSNb
LUPN
MCKN
MGTN
MLON
MRDNa
NODN
NORMa
NUNN a
ROCN a
SNPN
TACNa
YMBN
YNENb
YKW1
YKW2
YKW3
YKW4
Latitude
(ON)
64.991
63.852
63.732
65.868
63.269
65.268
63.179
64.509
64.698
64.857
65.558
64.199
64.725
63.305
64.224
63.419
63.357
64.578
64.333
65.738
64.198
63.685
63.969
63.568
63.960
64.779
66.149
65.991
63.517
66.286
64.874
65.088
62.482
62.433
62.561
62.492
Longitude
(OE)
-110.870
-109.717
-110.899
-111.783
-108.728
-111.186
-113.900
-110.309
-110.609
-110.827
-111.430
-112.993
-109.330
-110.891
-115.130
-109.201
-109.001
-110.522
-110.130
-111.256
-110.213
-109.591
-109.895
-109.543
-110.960
-110.786
-112.151
-111.750
-110.907
-112.201
-111.532
-111.050
-114.484
-114.603
-114.610
-114.743
Elevation
(m)
Data used in
SW
469
436
420
513
366
513
307
420
468
450
520
392
449
417
266
421
418
416
440
505
431
411
418
434
435
468
529
512
458
568
439
463
171
144
170
161
Data used in
RE
_V
V
V
_V
V
V/
_V
V
V
V/
V,
V
V/
V
V
V
V
A/
V
V
V
V
V
V
Table A.1: The seismic stations used in this thesis. aThe MIT stations. Each was equipped
with a CMG-3TD seismometer. bThe POLARIS stations that were using CMG-40T seismometers. The remaining stations are POLARIS stations with CMG-3ESP seismometers.
Also listed are four broadband stations of the Yellowknife array (YKW1-4). SW, surface
wave tomography. RF, receiver-function analysis.
118
611
W1~j"
TUOJJ aslIA3OP01 2Ui4UllO 'JOAT030.
oLJ4 o4 paKdsau q4iW qu~ oij4 jo 14nTrnize3q si zpg -UiAiao. a114 04 4TJaAa 014 WUOIJ
jva4uaoida sT V (Z Ta;deqD) ATjdv@oTUo4 @A1'AX 00Wj.flS OM4 UT P@Sfl S4U@Aa OflJ
DOUV4SIp
:Z-V
a1qej
N9 )/6/CooZ
9Wf7:f7
9V:60~3 (9oZ)9Z/zO/Woo
(96T)9T/zO/AooN
(TST)oC/9o/:ooZ
(6LOW)/9o/o
(zz1/o/voo
LLL'61
9*9
00T
£L9OL'6Z1
f77
SZ90fZ,
V69*6V1
£*96
9*9
S*LLZ
zs,9
86970*01
L*0 9*L1 9*L
0VL1
£f
0171990OZ1 6'L9 L*9
OUT-f
01
0PF*9v1
E*Y0Z V*6
£:9
098TZ
o Jg
976
Z9
0"1
6*90
Of:yP' 9
08Z*9L1
Oz
6ZLf 0L
6*VS7,
0M9*L0090OE
V01
8'9 ML
C'9V1
T21
066*691
06-v9
WO0M
E'lP E*9
01U
0Z6*9L1
0z
6*iV8Z 8'Lf V9
09T1fd
0fS-8C
L*9
89
97Mf
079
ll:To:iO
t*9f4
6*9
V*L
1,9
L*801
8*9
Lf76Z 8*901
0LZ
VF8E
V6LI
676 T'9
TV9
9TL
Z*L
fY9
8*8
19
6*96 0*9
C6f%
87Z
LL'M
£f7LZ
06f%
ZU
L*96
~Z
6W1
'VL9V
9*8
976
T66
9
VfM9
9*ZLZ
9*90
6*18Z
YPT9
EUT1 V
T9
OUTI
Z*9
6*8
Zf6
6*f9Z
8*86
(0)
zvq
T9
8*9
E9
9*9
Z*9
1*9
S*
9*9
9*9
9'L
Z*g E9
S*L
V
001
qui
1V8
1Lvzss
0fS*6Z1
VUf
9fp
06Z*LT-
M~
01
III
Lf IF
1F7
OILTLT09V0V,
0LV96
ozv1lg1
OTUT-
08TO011
01
of,
09E-9089-m-
9q
069*17
019'6E
01P9T
09T*N
0 9T
0WvL
09V*
0~018*
(a.)
q~doUl
Ofv'i0006fR
0Z9LE
081 90ZE*91
0
0UT1
0fZ*LL
I)f79f7910L0TZ1
0tvzH71v
0WfM
11
6T
01
01
01T
£f1
01T
OWf01
00V1701
f7Z
0LL*06IV z
001T
0fL8S
(JL&(J9T
MZ
(U-130
(0)
apn~luorj
OZS81
0LL*81
0Z9 T
0691U6f<0
(N,)
(69t)9Z/oi/tOOZ
0:LZ:0Z
LZ:L0:00
U6Z:T
vv: I :-I~
EZTZ
TWVfM:T
ZO:80:Z
£fNV6T
M£P:60
1r:WZ:0
9f:80:0z
f:%0EO90
V0:L :00
WWO
MW901
L:1T:L1
Z:
17:
MLZ:L0
TfV:E0:Z0
9Wf0
6TT:10:1
oM:Mzz
Z :sL:1
9::Ly
8tfltV61
T:90:Z0
LVY9V:Z
0 :P0:61
90fEtf780
(7L1)EZ/9o/f:ooZ
(Iz1)OZ/90/COOZ
(L91)91/90/VOOZ
(991)91/9o/EOOZ
(9'v1)9Z/9o/EooZ
(LM)LT/90/WOoZ
(,vI)vo~o/oz
(M)MAO/E0o
(TIi)TO/NCO/OZ
(LOT)LT17/CO/OoZ
(06)TE/WWO/Z
(L8)8z/EO/Eooz
(TL)ZT/Co/tOoN
(oL)11/mo/mo
NO
Z/o
(o0)61/WWO/ Z
(9v)X1/mo/mo
(E~IzoE
(ZOT/o/ooz
(9Z)9O/zo/wo
(zz)z/io/ooz
(zz)z/o/fooz
(IZ)TZ/Io/tOO9
(oZ)oZ/1o/moz
(oz)oz/1o/tooz
~uapnjpww
aTull
qw
a4,ec[
031
-@?vd 4x@u aq4 uo sanuquoo ajqP4 aqjL -t, ja4delqD ui pa4uasaid poTpam
uoi4nloAuooop pauoi4jpuoo-Xe.Tje aq4 Xq P@zXI'BUC @.I@Ak 4qeq4 Olaq Pa4SIJ S4U@A@ g j JO IP404
,e 2uoT-ue 's4uaAo Zq pasn E ia4deTIO ui pa4uasaid 2Ul2VU-ll U0140MIJ-10AT030.1 Dql :E-V alqel
/A
/A
/11
A
/A
/A
/A
/11
/A
/A
60WP9
9 9s*Tloc
C6Sg*VOC
ST19119
9EVL*v9
Z908*18
V06L*99
LZ06*OTC ZVL:P'Z6
IL09"LL
698T*Zl
89VUSU
U00*09
CT89*86Z VW9119
MSTOC C986*89
6 UTOC
9ML*O:vc
89:VT*09T
68LZTOC
c9fT*zLz
lv8TT*SOC 6900'SL
TOOO'SOC 6L96'LL
19L*68Z
99:pg*LL
9T96*9:Pg 9966*96
696T8
TVLTVC
TU*OOC
EETT9
LLMM
KW99
600Z*69Z 9V96*96
VOWNT L9TT*6v
:VSOL*TOC 9zog*T:v
U66*00C 899F99
969C70C EL9TT9
gTgo*8L
Z9Z9*FT
89Z*6CZ
TM76
8699'88Z TL:PL,98
UzTzoc 9LZ8*Lg
9Z VLU
98TT*96
6ZTL'09T ZZE9*99
K,86U
EUTITE
8M*66Z VZL8*OTI
90687OC VT99*Lg
LUL'89C
9T6L*SL
MC70C TE99*119
961,T*08
zz9L*ZOC :V869*89
99C9*66Z T,69L*Zt,
88XTIL
C:Vvo*sc
8806*8L
99TF 9
8669*69
6969T6
899L*O:p
G)
v
fv
/A
/A
A
AH ul pasfl
617TTT
VOW89C
v6LL'69
88T8*LtT
ZTLC'6CZ
9TO0,89
(1)
ze a
9
01
808,8ET
M*M
T*9 TOT
V9
9T
6LL*SET
V6
V9
V69*ZZT
9
9t
999*9890T
:PE9*OZT
'9
9*9
9TV
LLVV9
OT
:VT9*L9L*9
9
8
so:V*TgT
Z*TZ
ZUT
8T9*9ET
tT
OT
6176
679
:V:V6*:V6L*69
996*TVT
8
M*99T8"M
t,9L*RZT
T *9 WOZ
V68'8ZT
VVT
L's
808*9:PT
6T9*8LTT9
099
U
L*S
ZL9*E8
6*9
TRZ T
EOE*T9T
6
E:VO*LSS*Z9T
69Z*99T
9*9
0T
9V8*L6T*9 9*88T
E00*09T
9*S
9T
90V*TVT
T88*T:VT
L *9
LT
T9*T9
9*9
LT9
UVELT:pz
Ltlz*:V:VT
9*
9*8Z
TZO*9VT
09
99*ZLT0T
9*9
966'98Z*ZT
L:VL*LE*T9
9
B979T
STV
9E8*:V:VT
T *L81
6ZO*TL
TE
800*Ttli
8
6LT*68
9
VZ9
60Z*P:VT
97
89V09T
6*9
Tflz
9ZETZ
STZ
L, 9
:VLL*OL
Z*9
0
L66*EZ*6Z
9*9
LL8*z88*6ZT
V9
96T*tLT9*9
OT
6LT*OEOml)
(H -)
q4d@a
apn4j 3uo-j
qut
STE*LE
SMLE
9ZZ*LE
Mvz
zz:V*Tl
9UTT
9ZT*LE
90T*EZ
866TV
TTEE
LOU
UTU
LZV*91
EMU
E9L*T9
88:V,9z
68V9Z
60*6T
6E9*LTV69W
86T*Lt
8UOT
Z06*OTLEE*9T
989*99
T9z*:K
MU
6Z*9E
:VE*gTE:VL*TT
9zTzv
11110*8T908*OT
L90,TL
9zz*
8TWU
ZTS*9E
SZF*9E
VWEE
EWTV
6WES
689TE
L6E*9E
ZFT*9E
89E*8
WKZZT*M
(N -)
apn4!4,el
1,04T:60
ZT:EO:60
00:99:80
09:80:110
E9:9z:lz
90:9ETT
oz:9tl:lz
8:V:LO:Lo
ZT:00TO
9Z:89::VT
LO:LO:OT
K:90:01
OTTO:60
EETT:90
9T:OE:60
99::V9:ST
VT:9V60
U:9T:VT
ZT:LZ::VO
:Vtl:80Tz
6V:OE:01
OE:TO:LO
EP:vO:60
TZT :ZZ
M:6T:9T
60:9 :OZ
OT:LVEO
V:V:8E:ZT
Moq:Lo
9E::VE:LT
ZTTVTT
9TTE:6T
TZ:L :00
6E:LO:U
60::V9:TO
ZT:90:8T
:PO:'VZ:TZ
oo:zo:Ez
69:L:P:8T
90:N:9T
9T:M80
00: TZ:90
6T:9:P:ZZ
9:P:Lz:zo
LV6TTT
T T:MV:
TT
TO: ET
917:
aluljL
(L6Z)W0T/:VOOZ
U6Z)WOT/VON
UWWOT/VOOZ
(6SZ)9T/OT/VOOZ
(NZ)60/0T/VOOZ
(w)SO/m/poo
(TSZ)LO/OT/vooz
(Z9Z)ST/60/VOOZ
(L9Z)CT/60/-vOOZ
(ZgZ)80/60/VOOZ
(6VZ)90/60/VOOZ
(az),Vz1so1vooz
(Tu)RT/so/vooz
(w)OT/so/vooz
(OZZ)LOlsolvooz
(qoz)n1Lo1,vooz
(voz)zz1Lo1vooz
(zoz)oz12,o1Fooz
(L6T)gT/LO/:VOOZ
(C6T)TT/LO/VOOZ
(061)S0/LO/VON
(IST)6Z/90/VOOZ
(:VLT)zz/go/vooz
(99T)VT/go/vooz
(ZgT)OT/go/vooz
(09OW90400Z
(09T)6Z/90/VOOZ
(6VOW90400Z
(nT)qT/go/-vooz
(fv)m/go/vooz
UZ090/go/vooz
(vu)CO/go/voog
(OZT)6Z/VO/-vOO9
NOOVT/vo/vooz
(gol)vl/vo/vooz
(ZOT)TT/VO/VOOZ
(960)90/VO/VOOZ
(V60)CO/VO/VOOZ
USO)a1colvooz
(980)n/co/vooz
(oso)oz/co/vooz
(2,M)LTIMIVOOZ
(ZLO)ZT/CO/VOOZ
(ggo)vz/zo/vooz
Nco)vo/zo/vooz
(9ZO)9Z/TO/VOOZ
(VZO)VZ/TO/VOOZ
04vcl
-@2pd SnOlAoid aq4 Tuoij sanupuoD :t-V alqpl
f,97L70E
MJVT
ZT:PO*TSI
TIV08,MT
98LVEK
LEZ9*T:VT
ZZ6*60E
'v006*OYVT
99MSEE
6889*Syvl
1A
/A
E90WEE
/A
66L:V*ZM
to%*UE
LM*LK
UTWIM
R,0970E
MWITH
9LO*091
918'UT
/A
/A
T996'99
K99*99
/1"
/A
/A
/A
A
A
/A
/A
/A
/A
/A
/A
/A
/A
/A
III posfl
9ZLV8TT
9VLO'68Z
99:v9*88Z
:PEZ6'OOE
98T9*911T
9z97*zoE
98ZL*LVE
ZE06*60E
UNTOE
NLE*L9E
Z006'OT
IFLEZ*Z E
:VEYO*96Z
EZZO*68Z
SEWME
E88E*T6Z
TEWME
PLUZOE
T Z*ZZT
9ff6'Sf,1
:PFL6*Z:VT
Z999U
Z8811*VOE
Z YWTIM
(1)
ZLq
M:9*08
61L'OS
6TLL'YPL
U'U'99
szSC78
60N*T9
97O79
IVIVF979
80K*06
996T*99
!R6L6*:VS
90z *flv
9998*99
MU'RR
ETT9*88
E 90*99
L9z*LC
9TZV*06
IUVR
TZ90*96
TLn*1,8
68K*TL
96879
998179
9996*98
8ZTT*:Vs
88X L
619*99
STLT*tls
LZ6L*OL
86K*09
ZLFLL
ZFF0118
V:V6Z*EL
ZN*ZL
ZZTZ*T8
z9Tg*Lg
LZ98*LS
9
6'9
L *9
9
8'9
9,9
9
9*9
L*9
9
T'9
9*9
9
9*9
T*9
W9
9*
L*
F9
6*9
L*9
6*9
L*9
9*9
9*9
T*9
T*9
9
6*
8*9
T*9
9
F9
8
ITL
88TU
9z
SU*k9TI
WT
9:VT*9111
1617ZI
U
NV7801
ol
L86*Z:Vl
6tO*Z:Vl
K
WETT
86*89IV9*98Z96*ZLOT
ztv*LT9
ZRWRT
60T*OL6*6L
89*TTTOT
ZLI*98LZ
:VTI:*6LI
LT
L61'699* Tl
OT
L98*ttli
too*L9Z*96T
9Z9*:V6
8,9z
TR
L86*089oVz8OT
ST
9v8*Zs6LL*tVT
OT
990*9VT
8'6t,
168*Stl
8*9z:p
6*6ZI
988*LLlo:v*o:vl
EV
IT
Z99'CS
MWI
OT
9XZVT
89
6*:VII
LOL*ZL
VT
918*99
VV6L
zz
L66* :Pl
:vz
L98*9K
L*ZFT
9V
6':VZ
9
ff
6TT*9:VT
TvI
90L*199T
ZLT*T,891
809*LL8*96
ZWN
T*:VT
988*8CI
IOL*8cl
II
(a.)
(IHN)
-q4daC[ opn4j iuoq
998*17:Vl
z:v8*0:vl
szz*g:vl
T,986*8
6L60*89
MVV9
VSL J,9
ztl 9119
L691,9
G)
v
9*9
8*9
L*9
9
qu-1
M*vc
6MIT
SL9*9Z68TT,
vso*vz
LTS*9
f,99*8c
9LZ'SC
8L*6TL'vZ*TT
9K*9TTTT*TL
86V C
LWLTST8*U:VZJT
6Cg*Tg
L86*6T68L*99
W VgTSS*SZ
98Z*CZ80*9
TL*
TL*TT
6L VC I
ZT,9*6Z
C6Z*L:P09*gc
T6v*OC
LOWU
TCL*OV
LOT*8C
tgL*OC
SZL*TV
880*9z
TTO*9T
9v6*Zf,
VWOT
6'Z:V
900*E:v
U9* T
Z09*6
969*V
LSL*g'v
NZL
80E*L
(N -)
@PR414prl
sz:9v:ol
OV:09TO
Lt:99:10
80:gz:zo
ZO:9T:10
oz:Wsl
SZ:gT:OT
sz:9:v:zo
0T,:R:ZO
9T:to:ll
9t:TT::Vl
LO:ZO:9T
99:vt:Lo
tT:90:ZT
t:IWZT
tv:91:zo
ZI:OT:LT
MVV:zz
M:LT:Fo
TO:Wol
IV:90:oz
9t: 11: 90
MTVtZ
TZ:ZI:61
90:8v:vo
8v:It:Iz
99:9VTO
LO:V Jvl
gl:zz:zz
ltl:tlo:oz
T:V:tg:lo
ov:LE:zl
tlo:vo: z
zz:gz:zo
80W z
60:9VSI
9z:vMo
90:60:Vl
zg:Ll:oz
TT: T:Vl
tT:ZE:ST
LO: TV I I
zz:LO:80
9S:90:60
WK:0Z
M:ofl:Io
LT:Po:Iz
(TWWOT/woz
(TWWOT/goog
(69Z)9Z/60/90OZ
(V9Z)TZ/60/90OZ
(W)90/60/9009
(6U)LZ/80/9009
(qu)vz/80Aooz
(8zz)qT/so/gooz
(gzz)FT/80/gooz
(gTz)to/80Aooz
(Loz)qz1Lo1qooz
(90z) ZILOA00Z
(VOZ)WLo/gooz
(V6T)U/LO/qOOZ
(Z6T)TT/LO/900Z
( Wgo/Lo/goog
NqT)vT/q0Aooz
(VqT)m/q0Aooz
(EgOu/90/gooz
(m)go/qoAooz
(ZqT)TO/q0Aooz
(TVT)Tz/go/goog
NZ090/go/woz
NzOgo/w/gooz
(VgT)VO/goAoog
(TTT)Tz/VoAooz
(60T)6T/:VO/900Z
(TOT)TT/ro/gooz
(00T)OT/voAooz
(L60)LO/VO/90OZ
(6LO)OZ/EO/90OZ
UW9ZIZOA00Z
(qgo)gz/zo/qooz
(no)zz/go/gooz
(qvo)VT/zo/qooz
(OVO)60/ZO/900Z
*O)go/go/gooz
(810)8T/To/gooz
(910)gT/To/gooz
(Tv0q0/zI/vooz
(EVOWIT/Voog
(gzt)TZ/TT/pooz
NZOOZAT/Vooz
(OZ09TAT/voog
(To0LZ/OT/Vooz
(TOOLZ/OT/vooz
(86Z)VZ/OT/VOOZ
@TUlj,
ZZZV*66Z
ZZI
-&ed SnOIAO.Td aq4 u-oij sanupuoD :g*V alqejL
9VEO*66Z
vzN*LM
TvVTOM
UWLM
9M*96Z
1001'8 E
11017T
TZ*6M
tZVL*gsz
ML'80E
MWTOE
888LTOE
:VTZT*86Z
Z86Z*M
T600*SEZ
6MT* VT
Lvz* M
SZT:V*ZOE
A
/A
/A
/A
/A
/A
/A
TZTIZTM
NUTTE
E8 9*8 E
Z66T*LOE
E9:V6*LOE
TEVE'ZOE
9TOL*06Z
E9zTzT
Y 98*80E
91ITFUZ
L6tT*TIT
C89T*OOC
8LZ8*LU
LLTO*LgE
SIT6'LOE
988z*gLz
T06Z*ZT
6T9:V*86Z
:V :V,99E
968*99E
:Vg'6:V
TOVLV
UzVg
19MOV
Lz9fI*9L
986L*ZL
M I '9L
T'9
8*9
9
T*9
9*9
9
T:VLL*LS
98LL*L9
6908U
EgLO*TL
969TTV
LTEL*gE
6LLS*OV
6ST6*68
TL99*L6
960L*96
ZZ6L*L9
Lssz*g :
908STE
T:VLS*9E
TITUgE
VTTWES
6gv6*SL
99ZZ*69
MSTL
908L, Z9
869Z*LL
9EZ6*LS
VZOVEL
UWK
LEVOU
609L*T9
SLWT
ZLTV*08
VE90*99
LOWTE
9898*08
T608'09
906*08
869'08
6*
6*
Z*9
9
Z'L
9
L*9
V,9
8*9
9
9
L*9
T*9
T*9
9
T*9
6*
L*
9
L*9
6*
9*9
L*
9*9
Z*9
9
K
9z
VT
zz
oz
L
TT
6
VT
OT T
L6T
6T
OE
OE
99
ZT
L
TT
ZT
6
zz
6
9*fIZZ
TI*9z
9z
6Z
TT
OT
9VT
OE
LVT
TT
9
OT
L*OSE
VgE
Z6T
871,
OT
L
(LU31)
du Ul pasfl
qv-t
ZVE[
iq4d@C[
9vz*vgl
9TV99T
806'T:PT
989*T91
9TOWT
TTO*YVIIT
9v8*TL
998*99
Vgt*L6
Z80*LLT
9VUUT
LvV89T
E:VL*99T
LWNT
LOET6
EZT*VLTZZO*TLT6E*ZOT
9T9*L9T
EZE*L9T
9TS,99T
680*L91
8LZ*TZT
E60*TL
6WRT
6Z*OET
898*99
L66*6ZT
E60*v9T968*L9968*:VK
989*OZT
ZVOU
999TEI
:PO*691T9E*EZT
ETIT9T
LU
ZSTTL
(a.)
apn4!2uol
18z'st
EWR
ZV9*9r
L19* 9
68L*SZ
6TT*9Z
6TT*6E
9Z6*9Z
6W9V
M*H
ITTTE
TLZ*Ig
ZLL*09
LZ9-F9
60ETZ
L8T*OZLTO*LZSTT:Vt
T6T,*09
L6T'T9
LZ9*09
OWN
898*ZZ
LWK
90ZTV
MW
68O*8E
6Z*OZ
:PLL*9z
6TO*TE
CINT9
T9t*ZZLOT*SE
LV*L9
V98TE
LLUE
90z*zg
6M"9Z
9T8"9V
UKTE
LWK
(N -)
apnq!4v-l
Z:V:LZ:Tz
z0:10:W
V:oz: I
C:IMT
LT:6t:ST
6C:9T:8T
09:L9:W
60:ZO:TZ
8:V:6:V:90
ztl:8T:vo
9Z:To:oz
WROTT
oo:ZT:TT
flIv:lvo:vz
T9:zz:LT
of,:9z:91
:VT:LT:6T
OT:E:V:oo
90:89:9T
89:lz:Lo
ETI:NTO
zo:9zTz
6T:ZO:OT
9t,:LTI:Tz
ov:TO:Tz
6E:OT:91
60:ET:ET
sz:TII:9T
61:ZZ:OT
OE:9E:gl
zo:E9:ZT
:P9:9Z:6T
ITWU
:Pg:6Z:61
oz:vo:gT
VT:80:OT
TE:9z:gT
80:19:91
LT:90:01
LE:OZ:6T
lo:OE:80
@-Lull
(9TOZT/11/gooz
(zu)oz/so/qooz
(6ZZ)LT/80/90OZ
(6ZZ),LT/80/90OZ
(Vzz)zI/80/qooz
(SIZ)go/so/gooz
UST)qolmlqoo
(6LT)SZ/90/90OZ
(99T)gT/90/90OZ
N9T)VT/90/900Z
(Z9011/90/gooz
(ZVT)ZZ/90/90OZ
(ZVT)zz/so/gooz
(8cT)8T/qo/qooz
(TWITAO/gooz
(m)CO/go/gooz
(OZT)OC/1'0/gooz
(OZT)OE/VO/90OZ
(6TT)6Z/VO/90OZ
(ZIOZZ/vo/gooz
(TIT)TZ/:PO/90OZ
(01T)OZ/vo/gooz
(160)10/VO/90OZ
(qvv)ZT/ZT/qooz
(9VOWZT/gooz
W:0:O/ZTAOOZ
(9COZO/UA00Z
(ZE&ZATA00Z
(TEOLZATA00Z
NZOTOT/900Z
(VZOWIT/gooz
(TZOLT/11/gooz
(8TOVI/11/900Z
(VTOOT/11/9009
(96Z)WOT/900Z
(96Z)EZ/OT/90OZ
(C6Z)OZ/OI/SOOZ
(88z)qT/01Aooz
(88z)qT/01Aooz
(Z8Z)60/01/900Z
(ZSZ)60/OT/90OZ
a4p(l
Bibliography
Aki, K., and P. G. Richards, Quantitative Seismology, 2nd Ed., University Science Books,
Sausalito, 704 pp., 2002.
Alsina, D., R. Snieder, and V. Maupin, A test of the great circle approximation in the
analysis of surface waves, Geophys. Res. Lett., 20, 915-918, 1993.
Anderson, D. L., The deep structure of continents, J. Geophys. Res., 84, 7,555-7,560, 1979.
Artemieva, I. M., and W. D. Mooney, Thermal thickness and evolution of Precambrian
lithosphere: A global study, J. Geophys. Res., 106, 16,387-16,414, 2001.
Ashwal, L. D., and K. Burke, African lithospheric structure, volcanism, and topography,
Earth Planet. Sci. Lett., 96, 8-14, 1989.
Aulbach, S., N. J. Pearson, S. Y. O'Reilly, and B. J. Doyle, Origins of xenolithic eclogite
and pyroxenites from the central Slave craton, Canada, J. Petrol., 48, 1843-1873, 2007.
Bagdassarov, N. S., M. G. Kopylova, and S. Eichert, Laboratory derived constraints on
electrical conductivity beneath Slave craton, Phys. Earth Planet. Inter., 161, 126-133,
2007.
Bank, C.-G., M. G. Bostock, R. M. Ellis, and J. F. Cassidy, A reconnaissance teleseismic
study of the upper mantle and transition zone beneath the Archean Slave craton in NW
Canada, Tectonophys., 319, 151-166, 2000.
Beghein, C., J. Resovsky, and J. Trampert, P and S tomography using normal mode and
surface wave data with a Neighborhood Algorithm, Geophys. J. Int., 149, 646-658, 2002.
Berkhout, A. J., Least-squares inverse filtering and wavelet deconvolution, Geophysics, 42,
1369-1383, 1977.
Bleeker, W., L. Ketchum, V. Jackson, and M. Villeneuve, The central Slave basement complex, part I, its structural topology and autochthonous cover, Can. J. Earth Sci., 36,
1083-1109, 1999.
Bleeker, W., Evolution of the Slave craton and the search for supercratons, Geol. Assoc.
Can., Abstract, 26, 14, 2001.
123
Bokelmann, G. H. R., and P. G. Silver, Shear stress at the base of shield lithosphere, Geophys.
Res. Lett., 29, 2091, doi:10.1029/2002GL015925, 2002.
Boerner, D. E., et al., Electrical conductivity in the Precambrian lithosphere of western
Canada, Science, 283, 668-670, 1999.
Bostock, M. G., Mantle stratigraphy and evolution of the Slave province, J. Geophys. Res.,
103, 21,183-21,200, 1998.
Bowring, S. A., and J. P. Grotzinger, Implications of new chronostratigraphy for tectonic
evolution of Wopmay orogen, Northwestern Canadian Shield, Am. J. Sci., 292, 1-20, 1992.
Bowring, S. A., I. S. Williams, and W. Compston, 3.96 Ga gneisses from the Slave province,
Northwest Territories, Canada, Geology, 17, 971- 975, 1989.
Bowring, S. A., T. B. Housh, and C. E. Isachsen, The Acasta gneisses: remnant of Earth's
early crust, in Origin of the Earth, H. E. Newsom and J. H. Jones, eds., Oxford Univ.
Press, New York, 319-343, 1990.
Bruneton, M., et al., Complex lithospheric structure under the central Baltic Shield from surface wave tomography, J. Geophys. Res., 109, B10303, doi:10.1029/2003JB002947, 2004.
Chen, C.-W., S. Rondenay, R. L. Evans, and D. B. Snyder, Lithospheric discontinuity in the
central Slave craton: A new perspective from seismological, electrical, and petrological
constraints, EOS Trans. AGU, 89(53), Fall Meet. Suppl., Abstract U43B-0067, 2008.
Clayton, R. W., and R. A. Wiggins, Source shape estimation and deconvolution of teleseismic
bodywaves, Geophys. J. R. astr. Soc., 47, 151-177, 1976.
Clowes, R. M., P. T. C. Hammer, G. Fernandez-Viejo, and J. K. Welford, Lithospheric
structure in northwestern Canada from Lithoprobe seismic refraction and related studies:
a synthesis, Can. J. Earth. Sci., 42, 1277-1296, 2005.
Conrad, C. P., and C. Lithgow-Bertelloni, Influence of continental roots and asthenosphere
on plate-mantle coupling, Geophys. Res. Lett., 33, L05312, doi:10.1029/2005GL025621,
2006.
Cook, F. A., and P. Erdmer, An 1800 km cross section of the lithosphere through the
northwestern North American plate: lessons from 4.0 billion years of Earth's history, Can.
J. Earth Sci., 42, 1295-1311, 2005.
Cook, F. A., A. J. van der Velden, K. W. Hall, and B. J. Roberts, Frozen subduction in
Canada's Northwest Territories: Lithoprobe deep lithospheric reflection profiling of the
western Canadian Shield, Tectonics, 18, 1-24, 1999.
Davis, W. J., A. G. Jones. W. Bleeker, and H. Griitter, Lithosphere development in the
Slave craton: A linked crustal and mantle perspective, Lithos, 71, 575-589, 2003.
124
Debayle, E., and B. L. N. Kennett, The Australian continental upper mantle: Structure and
deformation inferred from surface waves, J. Geophys. Res., 105, 25,453-25,450, 2000.
Debayle, E., B. Kennett, and K. Priestley, Global azimuthal seismic anisotropy and the
unique plate-motion deformation of Australia, Nature, 433, 509-512, 2005.
Dueker, K. G., and A. F. Sheehan, Mantle discontinuity structure from midpoint stacks
of converted P to S waves across the Yellowstone hotspot track, J. Geophys. Res., 102,
8313-8327, 1997.
Eaton, D. W., A. G. Jones, and I. J. Ferguson, Lithospheric anisotropy structure inferred
from collocated teleseismic and magnetotelluric observations: Great Slave Lake shear zone,
northern Canada, Geophys. Res. Lett., 31, L19614, doi: 10.1029/2004GL020939, 2004.
Efron, B., Bootstrap methods: Another look at the Jacknife, Annals Statistics, 7, 1-26, 1979.
Embree, P., Diversity seismic record stacking method and system, U.S. Patent, 3,398,396,
1968.
Fernandez Viejo, G., and R. M. Clowes, Lithospheric structure beneath the Archaean Slave
Province and Proterozoic Wopmay orogen, northwestern Canada, from a Lithoprobe
refraction/wide-angle reflection survey, Geophs. J. Int., 153, 1-19, 2003.
Forsyth, D. W., and A. Li, Array-analysis of two-dimensional variations in surface wave
phase velocity and azimuthal anisotropy in the presence of multipathing interference, in
Seismic Earth: Array Analysis of Broadband Seismograms, AGU Geophys. Monogr., 157,
A. Levander and G. Nolet, eds., 81-98, 2005.
Fouch, M. J., and S. Rondenay, Seismic anisotropy beneath stable continental interiors, Phys.
Earth Planet. Inter., 158, 292-320, 2006.
Frederiksen, A. W., M. G. Bostock, and J. F. Cassidy, S-wave velocity structure of the
Canadian upper mantle. Phys. Earth Planet. Inter., 124, 175-191, 2001.
Friederich, W., and E. Wielandt, Interpretation of seismic surface waves in regional networks: Joint estimation of wavefield geometry and local phase velocity-Method and tests,
Geophys. J. Int., 120, 731-744, 1995.
Friederich, W., E. Wielandt, and S. Stange, Non-plane geometries of seismic surface wavefields and their implications for regional surface-wave tomography, Geophys. J. Int., 119,
931-948, 1994.
Frost, D. J., The stability of hydrous mantle phases, Rev. Mineral. Geochem. 62, 243-271,
2006.
Goes, S., R. Govers, and P. Vacher, Shallow mantle temperatures under Europe from P and
S wave tomography, J. Geophys. Res., 105, 11,153-11,169, 2000.
125
Grand, S. P., Mantle shear structure beneath the Americas and surrounding oceans, J.
Geophys. Res., 99, 11,591-11,621, 1994.
Griffin, W. L., et al., Layered mantle lithosphere in the Lac de Gras Area, Slave Craton:
Composition, structure and origin, J. Petrol., 40, 705-727, 1999.
Gripp, A. E., and R. G. Gordon, Young tracks of hotspots and current plate velocities,
Geophys. J. Int., 150, 321-361, 2002.
Grove, T. L., N. Chatterjee, S. W. Parman, and E. Medard, The influence of H2 0 on mantle
wedge melting, Earth Planet. Sci. Lett., 249, 74-89, 2006.
Griitter, H. S., D. B. Apter, and J. Kong, Crust-mantle coupling: evidence from mantlederived xenocrystic garnets, in Proceedings of the 7th InternationalKimberlite conference,
J. B. Dawson v. 1, J. J. Gurney et al., eds., Red Roof Design, Cape Town, 307-313, 1999.
Gung, Y., M. Panning, and B. Romanowicz, Global anisotropy and the thickness of continents, Nature, 422, 707-711, 2003.
Gurrola, H., G. E. Baker, and J. B. Minster, Simultaneous time-domain deconvolution with
application to the computation of receiver functions, Geophys. J. Int., 120, 537-543, 1995
Guseinov, A. A., I. 0. Gargatsev, and R. U. Gabitova, Electrical conductivity of phlogopite
at high temperatures, Phys. Solid Earth, 41, 670-679, 2005.
Hacker, B., et al., Near-ultrahigh pressure processing of continental crust: Miocene crustal
xenoliths from the Pamir, J. Petrol., 46, 1661-1687, 2005.
Haldorsen, J. B. U., D. E. Miller, and J. J. Walsh, Multichannel Wiener deconvolution of
vertical seismic profiles, Geophysics, 59, 1500-1511, 1994.
Haldorsen, J. B. U., D. E. Miller, and J. J. Walsh, Walk-away VSP using drill noise as a
source, Geophysics, 60, 978-997, 1995.
Hammond, W. C., and E. D. Humphreys, Upper mantle seismic wave velocity: Effects of
realistic partial melt geometries, J. Geophys. Res., 105, 10975-10986, 2000.
Heinson, G., and S. Constable, The electrical conductivity of the oceanic upper mantle,
Geophys. J. Int., 110, 159-179, 1992.
Helmstaedt, H. H., and D. J. Schulze, Southern African kimberlites and their mantle sample:
implications for Archean tectonics and lithosphere evolution, in Kimberlites and related
rocks: Their composition, occurrence, origin, and emplacement, v. 1, Spec. Pub. 14, J.
Ross, ed., Geol. Soc. Australia, 358-368, 1989.
Hirth, G., R. L. Evans, and A. D. Chave, Comparison of continental and oceanic mantle electrical conductivity: Is Archean lithosphere dry? Geochem. Geophys. Geosys. 1,
doi:10.1029/2000G000048, 2000.
126
Hoffman, P. F., United plates of America, the birth of a craton: Early Proterozoic assembly
and growth of Laurentia, Ann. Rev. Earth Planet. Sci., 16, 543-603, 1988.
Hoffman, P. F., Precambrian geology and tectonic history of North America, in The Geology
of North America - An overview, W. W. Bally and A. R. Palmer, eds., Geol. Soc. Am.,
Boulder, 447-512, 1989.
Hoffman, P. F., Geological constraints on the origin of the mantle root beneath the Canadian
Shield, Phil. Trans. R. Soc. London, Ser. A, 331, 523-532, 1990.
Holbrook, W. S., W. D. Mooney, and N. I. Christensen, in Continental Lower Crust, D. M.
Fountain, R. Arculus, and R. W. Kay, eds., Developments in Geotectonics 23, Elsevier,
Amsterdam, 1-43, 1992.
Isaak, D. G., High-temperature elasticity of iron-bearing olivines, J. Geophys. Res., 97,
1871-1885, 1992.
Isachsen, C. E., and S. A. Bowring, Evolution of the Slave craton, Geology, 22, 917-920,
1994.
Jackson, I., J. D. Fitz Gerald, U. H. Faul, and B. H. Tan, Grain size sensitive sesmic wave
attenuation in polycrystalline olivine, J. Geophys. Res., 107, doi:10.1029/2001JB001225,
2002.
Jordan, T. H., The continental tectosphere, Rev. Geophys. Space Phys., 13, 1-12, 1975.
Jordan, T. H., Composition and development of the continental tectosphere, Nature, 274,
544-548, 1978.
Jones, A. G., I. J. Ferguson, A. D. Chave, R. L. Evans, and G. W. McNeice, Electric
lithosphere of the Slave craton, Geology, 29, 423-426, 2001.
Jones, A. G., R. L. Evans. and D. W. Eaton, Velocity-conductivity relationships for mantle
mineral assemblages in Archean cratonic lithosphere based on a review of laboratory data
and application of extremal bound theory, Lithos, 109, 131-143, 2009.
Karato, S., The role of hydrogen in the electrical conductivity of the upper mantle, Nature,
347, 272-273, 1990.
Karato, S., Effects of water on seismic wave velocities in the upper mantle, Proc. Japan
Academy, 71, 61-66, 1995.
Kennett, B. L. N., The removal of free surface interactions from three-component seismo-
grams, Geophys. J. Int., 104, 153-163, 1991.
Kennett, B. L. N., and E. R. Engdahl, Traveltimes for global earthquake location and phase
identification, Geophys. J. Int., 105, 429-465, 1991.
127
Kennett, B. L. N., E. R. Engdahl, and R. Buland, Constraints on seismic velocities in the
Earth from traveltimes, Geophys. J. Int., 122, 108-124, 1995.
Kerrick, D. M., and J. A. D. Connolly, Metamorphic devolatilization of subducted marine
sediments and the transport of volatiles into the Earth's mantle, Nature, 411, 293-296,
2001.
Kopylova, M. G., and G. Caro, Mantle xenoliths from the southeastern Slave craton: Evidence for chemical zonation in a thick, cold lithosphere, J. Petrol., 45, 1045-1067, 2004.
Kopylova, M. G., and J. K. Russel, Chemical stratification of cratonic lithosphere: Constraints from the Northern Slave craton, Canada, Earth Planet. Sci. Lett., 181, 71-87,
2000.
Kopylova, M. G., J. K. Russel, and H. Cookenboo, Upper-mantle stratigraphy of the Slave
craton, Canada: Insight into a new kimberlite province, Geology, 26, 315-318, 1998.
Kopylova, M. G., J. Lo, and N. I. Christensen, Petrological constraints on seismic properties
of the Slave upper mantle (Northern Canada), Lithos, 77, 493-510, 2004.
Kusky, T. M., Accretion of the Archean Slave province, Geology, 17, 63-67, 1989.
LeCheminant, A. N., and L. M. Heaman, Mackenzie igneous events, Canada: middle Proterozoic hotspot magmatism associated with ocean opening, Earth Planet. Sci. Lett., 96,
38-48, 1989.
Langston, C., Structure under Mount Rainier, Washington, inferred from teleseismic body
waves, J. Geophy. Res., 84, 4749-4762, 1979.
Lee, C.-T. A., Compositional variation of density and seismic velocities in natural peridotites
at STP conditions: Implications for seismic imaging of compositional heterogeneities in
the upper mantle, J. Geophys. Res., 108, doi:10.1029/2003JB002413, 2003.
Lee, C.-T. A., Geochemical/petrologic constraints on the origin of cratonic mantle, in
Archean geodynamics and environments, AGU Geophys. Monogr., 164, K. Benn, J.-C.
Mareschal, and K. C. Condie, eds., 89-114, 2006.
Lenardic, A., et al., What the mantle sees: the effects of continents on mantle heat flow, in
The history and dynamics of global plate motions, AGU Geophys. Monogr., 121, 95-112,
2000.
Lewis, T. J., R. D. Hyndman, and P. Fliick, Heat flow, heat generation, and crustal temperatures in the northern Canadian Cordillera: Thermal control of tectonics, J. Geophys.
Res., 108(B6), 2316, doi:10.1029/2002JB002090, 2003.
Li, A., D. W. Forsyth, and K. M. Fischer, Shear velocity structure and azimuthal anisotropy
beneath eastern North America from Rayleigh wave inversion, J. Geophys. Res., 108(B8),
2362, doi: 10.1029/2002JB002259, 2003.
128
Li, C., R. D. van der Hilst, E. R. Engdahl, S. Burdick, A new global model for
P-wavespeed variations in Earth's mantle, Geochem. Geophys. Geosys., 9, Q05018,
doi:10.1029/2007GC001806., 2009.
Ligorria, J. P., and C. J. Ammon, Iterative deconvolution and receiver-function estimation,
Bull. Seismol. Soc. Am., 89, 1395-1400, 1999.
Mainprice, D., Y. Le Page, J. Rodgers, and P. Jouanna, Ab initio elastic properties of talc
from 0 to 12 GPa: Interpretation of seismic velocities at mantle pressures and prediction
of auxetic behaviour at low pressure, Earth Planet. Sci. Lett., 274, 327-338, 2008.
Mareschal, M., W. S. Fyfe, J. Percival, and T. Chan, Grain-boundary graphite in Kapuskasing gneisses and implications for lower-crustal conductivity, Nature, 357, 674-676, 1992.
Mercier, J.-P., M. G. Bostock, P. Audet, J. B. Gaherty, E. J. Garnero, and J. Revenaugh,
J. Geophys. Res., 113, doi:10.1029/2007JB005127, 2008.
Minster, J. B., and T. H. Jordan, Present-day plate motions, J. Geophys. Res., 83, 5331-5354,
1978.
Moorkamp, M., A. G. Jones, and D. W. Eaton, Joint inversion of teleseismic receiver functions and magnetotelluric data using a genetic algorithm: Are seismic velocities and electrical conductivities compatible? Geophys. Res. Lett., 34, doi:10.1029/2007GL030519,
2007.
Nielsen, L., H. Thybo, A. Levander, and L. N. Solodilov, Origin of upper-mantle seismic
scattering - evidence from Russian peaceful nuclear explosion data, Geophys. J. Int., 154,
196-204, 2003.
Pawley, A. R., and B. J. Wood, The high-pressure stability of talc and 10 A phase: Potential
storage sites for H2 0 in subduction zones, Am. Mineral., 80, 998-1003, 1995.
Pell, J. A., Kimberlites in the Slave craton, Northwest Territories, Canada, Geoscience
Canada, 24, 2, 77-91, 1997.
Polet, J., and D. L. Anderson, Depth extent of cratons as inferred from tomographic studies,
Geology, 23, 205-208, 1995.
Pollack, H. N., S. J. Hurter, and J. R. Johnson, Heat flow from the Earth's interior; analysis
of the global data set, Rev. Geopys., 31, 267-280, 1993.
Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in
FORTRAN: The Art of Scientific Computing, 2nd ed., Cambridge Univ. Press, Cambridge,
963 pp., 1992.
Priestley, K., and E. Debayle, Seismic evidence for a moderately thick lithosphere beneath
the Siberian Platform, Geophys. Res. Lett., 30(3), 1118, doi:10.1029/2002GL015931, 2003.
129
Ritsema, J., and H. van Heijst, New seismic model of the upper mantle beneath Africa,
Geology, 28, 63-66, 2000.
Ritzwoller, M. H., N. M. Shapiro, M. P. Barmin, and A. L. Levshin, Global surface wave
diffraction tomography, J. Geophys. Res., 107(B12), 2335, 2002.
Roberts, J. J., and J. A. Tyburczy, Partial-melt electrical conductivity: Influence of melt
composition, J. Geophys. Res., 104, 7055-7065, 1999.
Rondenay, S., Upper mantle imaging with array recordings of converted and scattered teleseismic waves, Surv. Geophys., 30, 377-405, doi:10.1007/s10712-009-9071-5, 2009.
Rondenay, S., M. G. Bostock, and J. Shragge, Multiparameter two-dimensional inversion of
scattered teleseismic body waves, 3, Application to the Cascadia 1993 data set, J. Geophys.
Res., 106, 30,795-30,808, 2001.
Rondenay, S., M. G. Bostock, and K. M. Fischer, Multichannel inversion of scattered teleseismic body waves: practical considerations and applicability, in Seismic Earth: Array
Analysis of Broadband Seismograms, AGU Geophys. Monogr., 157, A. Levander and G.
Nolet, eds., 187-204, 2005.
Rondenay, S., G. A. Abers, and P. E. van Keken, Seismic imaging of subduction zone metamorphism, Geology, 36, 275-278, 2008.
Rudnick, R. L., and A. A. Nyblade, The thickness and heat production of Archean lithosphere: Constraints from xenolith thermobarometry and surface heat flow, in Mantle
Petrology: Field Observations and High Pressure Experimentation, Y. Fei, C. M. Bertka,
and B. 0. Mysen, eds., Geochem. Soc. Spec. Pub., 6, 3-12, 1999.
Rudnick, R. L., W. F. McDonough, and R. J. O'Connell, Thermal structure, thickness and
composition of continental lithosphere, Chem. Geol., 145, 395-411, 1998.
Rychert, C. A., K. M. Fischer, and S. Rondenay, A sharp lithosphere-asthenosphere boundary imaged beneath eastern North America, Nature, 436, 542-545, 2005.
Rychert, C. A., S. Rondenay, and K. M. Fischer, P-to-S and S-to-P imaging of a sharp
lithosphere-asthenosphere boundary beneath eastern North America, J. Geophys. Res.,
112, doi:10.1029/2006JB004619, 2007.
Saito, M., DISPER80: A subroutine package for the calculation of seismic normal-mode
solutions, in Seismological Algorithms: Computational Methods and Computer Programs,
D. J. Doornbos, ed., 293-319, Academic Press, San Diego, 1998.
Sambridge, M., Geophysical inversion with a neighbourhood algorithm- I. Searching a parameter space, Geophys. J. Int., 138, 479-494, 1999a.
Sambridge, M., Geophysical inversion with a neighbourhood algorithm- II. Appraising the
ensemble, Geophys. J. Int., 138, 727-746, 1999b.
130
Savage, M. K., Seismic anisotropy and mantle deformation: What have we learned from
shear wave splitting? Rev. Geophys., 37, 65-106, 1999.
Silver, P. G., and W. W. Chan, Shear wave splitting and subcontinental mantle deformation,
J. Geophys. Res., 96, 16,429-16,454, 1991.
Simons, F. J., A. Zielhuis, and R. D. van der Hilst, The deep structure of the Australian
continent from surface wave tomography, Lithos, 48, 17-43, 1999.
Simons, F. J., R. D. van der Hilst, J.-P. Montagner, and A. Zielhuis, Multimode Rayleigh
wave inversion for heterogeneity and azimuthal anisotropy of the Australian upper mantle,
Geophys. J. Int., 151, 738-754, 2002.
Smith, M. L., and F. A. Dahlen, The azimuthal dependence of Love and Rayleigh wave
propagation in a slightly anisotropic medium, J. Geophys. Res., 78, 3321-3333, 1973.
Snyder, D. B., Stacked uppermost mantle layers within the Slave craton of NW
Canada as defined by anisotropic seismic discontinuities, Tectonics, 27, TC4006,
doi: 10. 1029/2007TC002132.
Snyder, D. B., and G. D. Lockhart, Kimberlite trends in NW Canada, J. Geol. Soc., London,
162, 737-740, 2005.
Snyder, D. B., and M. Bruneton, Seismic anisotropy of the Slave craton, NW Canada, from
joint interpretation of SKS and Rayleigh waves, Geophys. J. Int., 169, 170-188, 2007.
Snyder, D. B., M. G. Bostock, and G. D. Lockhart, Two anisotropic layers in the Slave
craton, Lithos, 71, 529-539, 2003.
Snyder, D. B., S. Rondenay, M. G. Bostock, and G. D. Lockhart, Mapping the mantle
lithosphere for diamond potential using teleseismic methods, Lithos, 77, 859-872 , 2004.
Sobolev, S. V., H, Zeyen, G. Stoll, F. Werling, R. Altherr, and K. Fuchs, Upper mantle
temperatures from teleseismic tomography of French massif Central including effects of
composition, mineral reactions, anharmonicity, anelasticity and partial melt, Earth Planet.
Sci. Lett., 139, 147-163, 1996.
Straub, K. M., S. Rondenay, and D. B. Snyder, Upper mantle structure of the Slave craton
from teleseismic body-wave tomography, Eos Trans. AGU, 85(17), Jt. Assem. Suppl.,
Abstract S33A-06, 2004.
Tarantola, A., and B. Valette, Generalized non-linear problems solved using the least-squares
criterion, Rev. Geophys. Space Phys., 20, 219-232, 1982.
Thybo, H., and E. Perchu6, The seismic 8' discontinuity and partial melting in continental
mantle, Science, 275, 1626-1629, 1997.
131
Ulrych, T. J., M. D. Sacchi, and S. L. M. Freire, Eigenimage processing of seismic sections,
in Covariance Analysis for Seismic Signal Processing, Geophys. Development, 8, Soc.
Exploration Geophysicists, 14, 978-997, 1999.
van der Lee, S., and A. Frederiksen, Surface wave tomography applied to the North American upper mantle, in Seismic Earth: Array Analysis of Broadband Seismograms, AGU
Geophys. Monogr., 157, A. Levander and G. Nolet, eds., 67-80, 2005.
van der Lee, S., and G. Nolet, Upper mantle S velocity structure of North America, J.
Geophys. Res., 102, 22,815-22,838, 1997.
Vinnik, L. P., Detection of waves converted from P to SV in the mantle, Phys. Earth Planet.
Inter., 15, 39-45, 1977.
Wagner, L. S., S. Beck, G. Zandt, and M. N. Ducea, Depleted lithosphere, cold, trapped
asthenosphere, and frozen melt puddles above the flat slab in central Chile and Argentina,
Earth Planet. Sci. Lett., 245, 289-301, 2006.
Weeraratne, D. S., D. W. Forsyth, and K. M. Fischer, Evidence for an upper mantle plume
beneath the Tanzanian craton from Rayleigh wave tomography, J. Geophys. Res., 108(B9),
2427, doi: 10.1029/2002JB002273, 2003.
Westerlund, K. J., et al., A subduction wedge origin for Paleoarchean peridotitic diamonds
and harzburgites from the Panda kimberlite, Slave craton, evidence from Re-Os isotope
systematics, Cont. Mineral. Petrol., 152, 275-294, 2006.
Yang, Y., and D. W. Forsyth, Regional tomographic inversion of amplitude and phase of
Rayleigh waves with 2-D sensitivity kernels, Geophys. J. Int., 166, 1148-1160, 2006.
Yao, H., C. Beghein, and R. D. van der Hilst, Surface wave array tomography in SE Tibet
from the ambient seismic noise and two-station analysis- II. Crustal and upper-mantle
structure, Geophys. J. Int., 173, 205-219, 2008.
Zhou, Y., G. Nolet, F. A. Dahlen, and G. Laske, Global upper-mantle structure
from finite-frequency surface wave tomography, J. Geophys. Res., 111, B04304,
doi:10.1029/2005JB003677, 2006.
132