Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Nature of a low-velocity zone atop the transition zone in northwestern Canada Schaeffer, Andrew John 2009

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


24-ubc_2009_fall_schaeffer_andrew.pdf [ 9.02MB ]
JSON: 24-1.0052946.json
JSON-LD: 24-1.0052946-ld.json
RDF/XML (Pretty): 24-1.0052946-rdf.xml
RDF/JSON: 24-1.0052946-rdf.json
Turtle: 24-1.0052946-turtle.txt
N-Triples: 24-1.0052946-rdf-ntriples.txt
Original Record: 24-1.0052946-source.json
Full Text

Full Text

Nature of a low-velocity zone atop the transition zone in Northwestern Canada by Andrew John Schaeffer B.Sc., The University of British Columbia, 2006 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master of Science in The Faculty of Graduate Studies (Geophysics) The University Of British Columbia (Vancouver) August 2009 c© Andrew John Schaeffer 2009 Abstract Seismic studies over the past decade have identified a S-wave low-velocity zone (LVZ) above the transition zone at various locations around the globe. This layer is hypothesized to be a lens of dense, fluid-rich silicate melt pond- ing atop the 410 km discontinuity, beneath the silicate melt-density crossover predicted to exist within the upper mantle. We have assembled a P - and S-receiver function (PRF and SRF, respectively) dataset from the CNSN Yellowknife Array (YKA), the CANOE array, and the POLARIS-Slave ar- ray, to quantify the physical properties and geographical extent of the layer in Northwestern Canada. In order to compute the Poisson’s ratio, an impor- tant discriminant of possible composition and/or fluid content, we generated a suite of 1-D velocity models based on IASP91, but with varying thicknesses and velocity ratios for a hypothetical layer above the 410 km discontinuity. From these models we computed moveout curves for the range of slowness represented in the YKA data. A grid search was performed over the model space of interval thickness and Poisson’s ratio to obtain an estimate of the model that best accounts for the data. In addition, we performed a linearized inversion of transmission coefficient amplitudes to estimate the shear veloc- ity contrast at the bounding interfaces of the LVZ. Results indicate a LVZ of thickness ∼36 km with a shear velocity contrast of -7.8%, and Poisson’s ratio of 0.42. In combination, these two results require an associated increase in compressional velocity into the LVZ. The Poisson’s ratio lies well above the IASP91 average of ∼0.29-0.3 for this depth range and favours the presence of high melt or fluid fractions. Geographic profiles of PRFs and SRFs 1-D migrated to depth from CANOE and POLARIS-Slave arrays reveal 410 km and 660 km discontinuities at nominal depths with little variation in transi- tion zone thickness. PRF results from the Slave craton indicate a potential LVZ beneath many stations at an average nominal depth of ∼340 km, high- lighted by events from the northwest. The CANOE array SRF profile images an emergent LVZ beginning at ∼280 km depth dipping eastwards to 310 km approaching YKA. ii Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . vi Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Statement of Co-Authorship . . . . . . . . . . . . . . . . . . . . . viii 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Background and Objectives . . . . . . . . . . . . . . . . . . . 1 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2 Slave low-velocity zone . . . . . . . . . . . . . . . . . . . . . . 5 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2 Mantle Setting . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3 Physical Properties of the LVZ at YKA . . . . . . . . . . . . 11 2.3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.3.2 Receiver Functions at YKA . . . . . . . . . . . . . . . 11 2.3.3 Kinematic Constraints . . . . . . . . . . . . . . . . . 18 2.3.4 Dynamic Constraints . . . . . . . . . . . . . . . . . . 23 2.4 Geographic Controls . . . . . . . . . . . . . . . . . . . . . . . 26 2.4.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 2.4.2 CANOE . . . . . . . . . . . . . . . . . . . . . . . . . 28 SRF profile . . . . . . . . . . . . . . . . . . 28 PRF profile . . . . . . . . . . . . . . . . . . 30 2.4.3 POLARIS Slave Array . . . . . . . . . . . . . . . . . 32 SRF profile . . . . . . . . . . . . . . . . . . 33 PRF profile . . . . . . . . . . . . . . . . . . 33 iii Table of Contents 2.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 2.5.1 Implications for Physical Conditions . . . . . . . . . . 37 2.5.2 Possible Sources of Error . . . . . . . . . . . . . . . . 40 2.5.3 Geographical Variations . . . . . . . . . . . . . . . . . 43 2.6 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . 45 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 3.1 Summary of Results . . . . . . . . . . . . . . . . . . . . . . . 57 3.2 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . 59 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 iv List of Figures 1.1 Mantle Transition Zone Mineral Phases . . . . . . . . . . . . 2 2.1 Dry Silicate Melt Density Crossover . . . . . . . . . . . . . . 6 2.2 Location Map . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.3 YKA Earthquake Distribution . . . . . . . . . . . . . . . . . 12 2.4 YKA NW Corridor raw receiver functions . . . . . . . . . . . 15 2.5 YKA SE Corridor raw receiver functions . . . . . . . . . . . . 16 2.6 Incident P wavefield and associated scattered phases . . . . . 17 2.7 Velocity Model Parameterization . . . . . . . . . . . . . . . . 20 2.8 Model Grid Search Results . . . . . . . . . . . . . . . . . . . 22 2.9 Final Model Grid Search Results . . . . . . . . . . . . . . . . 23 2.10 Model-Data Comparison . . . . . . . . . . . . . . . . . . . . . 24 2.11 Earthquake Distribution for Portable Arrays . . . . . . . . . 29 2.12 CANOE SRF Reflectivity Section . . . . . . . . . . . . . . . . 30 2.13 CANOE PRF Reflectivity Secion . . . . . . . . . . . . . . . . 31 2.14 POLARIS Slave SRF Reflectivity Section . . . . . . . . . . . 34 2.15 POLARIS Slave PRF Reflectivity Section . . . . . . . . . . . 35 2.16 Bulk sound velocity for Hydrated vs Anhydrous solid Olivine 39 2.17 Synthetic Modelling Results . . . . . . . . . . . . . . . . . . . 44 2.18 Summary cartoon of geographic results . . . . . . . . . . . . . 46 v Acknowledgments I would like to extend my thanks to all of the people who have helped me over the course of my graduate studies. First, I would like to thank Michael Bostock for giving me the opportunity to work with him. He has provided invaluable expertise, insight and encouragement over the years. I would also like to thank the members of my supervisory committee, Ron Clowes and Dominique Weis, and my external examiner Nikolas Christensen for their helpful suggestions throughout my degree. I would also like to thank Garry Clarke, Mark Jellinek, and Christian Schoof. My life at UBC has been enriched through the interaction with many great people. I would like to extend special thanks to Pascal Audet, Jean- Philippe Mercier, Brendan Smithyman, Hideharu Uno, and Ralf Hansen, who have all contributed thoughtful advice and support throughout my project. Finally, I would like to thank my family, Jo, Ron, Emily, and Stephanie, all of whom I am indebted to for their undying support during my graduate studies. vi For Steph, Jo, Ron, and Emily vii Statement of Co-Authorship Chapter 2 in this thesis will be submitted for publication to an international Earth science journal. This was a co-authored publication with Michael Bostock at UBC. Michael identified the research program. I performed the research, data analysis, and modelling. Michael and I interpreted the results. I prepared the manuscript in close collaboration with Michael. Schaeffer, A.J., and Bostock, M.G., 2009. Nature of a low-velocity zone atop the transition zone in Northwestern Canada, In preparation. The format has been slightly altered from the intended published version to meet the thesis format requirements of FoGS. viii Chapter 1 Introduction 1.1 Background and Objectives The Earth’s mantle is divided into two main regions, the depleted upper mantle and relatively enriched lower mantle. These two reservoirs are sep- arated by the mantle transition zone, which is defined by a series of phase transformations in the major mantle minerals. At nominally anhydrous con- ditions, olivine (α-(Mg,Fe)2SiO4) in the upper mantle alters to wadsleyite (β-(Mg,Fe)2SiO4) at 410 km, followed by a further transformation to ring- woodite (γ-(Mg,Fe)2SiO4) at 520 km. The base of the transition zone is defined by the dissociation of ringwoodite to perovskite ((Mg,Fe)SiO3) and magnesiowüstite ((Mg,Fe)O) [Frost , 2008]. The upper mantle, transition zone, and lower mantle minerals can incorporate at least minor amounts of water in their crystal structures, either as point defects or bound hydroxyl groups [Kohlstedt et al., 1996]. Recent mineral physics studies have demon- strated that contrasts in the water storage and solubility exist between the three regions, and that transition zone minerals have a greater storage ca- pacity than both upper and lower mantle materials [Kohlstedt et al., 1996; Litasov et al., 2003]. Therefore, upwelling material from the transition zone may be super-saturated in H2O, and initiate dehydration melting upon en- tering the upper mantle [Huang et al., 2005]. Seismic evidence consistent with such a process has been observed in several locations around the globe; in particular, through the identification of low-velocity zones in close prox- imity to the 410 km discontinuity [e.g., Revenaugh and Sipkin, 1994]. The objectives of this thesis are firstly to characterize the physical properties of a LVZ observed atop the mantle transition zone beneath Yellowknife, NWT [Bostock , 1998], and secondly, to examine the variability and extent of the 1 Chapter 1. Introduction Figure 1.1: Mantle transition zone structure and the dominant olivine-group phase transformations. The following study focuses primarily on pressures less than 14 GPa, which corresponds to the region above (and including) the 410 km discontinuity. Figure modified from Frost [2008]. LVZ across northwestern Canada from the Cordilleran deformation front in the west to the Slave craton in the east. We accomplish these objectives through teleseismic receiver function (RF) analysis, a technique for characterizing receiver-side Earth structure using waves generated by earthquakes at epicentral distances >30◦ [Bostock , 2007]. The incoming wavefield interacts with discontinuities, resulting in the generation of a scattered wavefield consisting of conversions and reverber- ations whose timing and amplitude depend upon the depths and elastic property contrasts of discontinuities beneath the receiver. This set of pulses is contained within the coda of teleseismic P - and S-waves recorded at the Earth’s surface, and can be recovered through deconvolution. 2 Chapter 1. Introduction In Chapter 2 of this thesis, we present our analysis of the seismic sig- nature of an LVZ atop the mantle transition zone in northwestern Canada. For this analysis, P receiver fucntions computed at the Yellowknife Array (YKA) were utilized for kinematic modelling of the Poisson’s ratio and thick- ness of the LVZ, and dynamic modelling for the shear velocity contrasts at the interfaces. Subsequently, we examine the geographic variability of the LVZ across northwestern Canada through the use of P and S receiver functions computed from two portable arrays, the Canadian Northwest Ex- periment (CANOE) and Portable Observatories for Lithospheric Analysis and Research Investigating Seismicity (POLARIS) Slave, deployed across the study region. In Chapter 3 we present a summary of our results and highlight avenues for future work. 3 Bibliography Bostock, M., Teleseismic Body-Wave Scattering and Receiver-Side Struc- ture, in Seismology and the Structure of the Earth, Treatise on Geophysics, vol. 1, edited by G. Schubert, pp. 219–246, Elsevier, 2007. Bostock, M. G., Mantle stratigraphy and evolution of the Slave province, Journal Of Geophysical Research, 103 (B9), 21,183–21,200, 1998. Frost, D., The upper mantle and transition zone, Elements, 4, 171–276, doi:10.2113/GSELEMENTS.4.3.171, 2008. Huang, X. G., Y. S. Xu, and S. I. Karato, Water content in the transition zone from electrical conductivity of wadsleyite and ringwoodite, Nature, 434 (7034), 746–749, 2005. Kohlstedt, D., H. Keppler, and D. C. Rubie, Solubility of water in the α, β, and γ phases of (Mg,Fe)2SiO4, Contributions to Mineral Petrology, 123, 345–357, 1996. Litasov, K., E. Ohtani, F. Langenhorst, H. Yurimoto, T. Kubo, and T. Kando, Water solubility in Mg-perovskites and water storage capacity in the lower mantle, Earth and Planetary Science Letters, 211, 189–203, 2003. Revenaugh, J., and S. A. Sipkin, Seismic Evidence For Silicate Melt Atop The 410 Km Mantle Discontinuity, Nature, 369 (6480), 474–476, 1994. 4 Chapter 2 Nature of a low-velocity zone atop the transition zone in Northwestern Canada 2.1 Introduction Recent studies on water solubility of mantle rocks indicate that transition zone minerals can incorporate considerably more water (and perhaps other volatiles) within their crystal structure than the primary upper and lower mantle minerals [Bell and Rossman, 1992; Kohlstedt et al., 1996; Bolfan- Casanova et al., 2005; Litasov et al., 2003]. Experimental and theoretical studies predict formation of partial melts due to the presence of H2O at pressure-temperature conditions comparable to those of the deep upper man- tle and transition zone [Inoue, 1994]. In the deep upper mantle at ∼8 GPa, it is postulated that a melt-density crossover exists, such that a dry silicate melt is denser than the solid upper mantle, but less dense than the transi- tion zone [Stolper et al., 1981; Agee and Walker , 1988; Ohtani et al., 1995; Agee, 2007], permitting the formation and accumulation of a melt layer atop the transition zone (see figure 2.1). Further studies have demonstrated that the addition of H2O into the melt increases the depth of the density cross- over through reduction of melt compressibility [Sakamaki et al., 2006, for example]. Melts with less than 4-6 wt % H2O may remain stable above the 410 km discontinuity. A version of this chapter will be submitted. A.J. Schaeffer and M. G. Bostock. Nature of a low-velocity zone atop the transition zone in Northwestern Canada. 5 Chapter 2. Slave low-velocity zone Figure 2.1: Dry silicate melt density crossover. Solid curve is density of an ultrabasic silicate melt, while diamond line represents the equilibrium olivine crystals within the melt. At pressures less than 8 GPa, the solid crystals are heavier than the liquid. At pressures greater than 8 GPa, however, the melt compressibility is greater than the solids, and as a result the melt becomes more dense. Figure modified from Agee [2008]. 6 Chapter 2. Slave low-velocity zone The recently proposed “transition zone water filter” (TZWF) model [Bercovici and Karato, 2003] embodies these concepts. Assuming that min- erals in the deep upper mantle possess lower water solubility than transition zone minerals, upwelling material that transits into the upper mantle will be oversaturated in water, leading to the possibility of dehydration induced melting and ponding atop the 410 km discontinuity due to negative buoy- ancy. The TZWF model is attractive in that it provides a means of resolving the fundamental contradiction between images of whole mantle convection afforded by seismic tomography and the geochemical requirement of isolated mantle reservoirs. There exists considerable uncertainty, however, concern- ing the density [Hirschmann et al., 2006], stability [Sakamaki et al., 2006], and steady-state thickness of a melt layer [Leahy and Bercovici , 2007], as numerical simulations predict thicknesses that vary over several orders of magnitude, from tens to thousands of meters. Furthermore, some studies [Hirschmann et al., 2005; Litasov et al., 2007] suggest that the water storage capacity of the upper mantle (olivine) is greater than previously thought, and, consequently, the degree of decompression melting of upwelling man- tle material across the 410 discontinuity may be less significant than was originally proposed for the TZWF model. Over the past decade and a half, considerable evidence for a seismic low- velocity zone (LVZ) in the deep upper mantle immediately above the man- tle transition zone has emerged. Regional studies have documented LVZ’s above the 410 km discontinuity in South America [Sacks and Snoke, 1977], easternmost Asia [Revenaugh and Sipkin, 1994], northwestern Canada [Bo- stock , 1998], the Arabian Plate [Vinnik et al., 2003], northern Mexico [Gao et al., 2006], northwestern United States [Song et al., 2004], southern Rocky Mountains [Jasbinsek and Dueker , 2007], eastern United States and Gulf of Mexico [Courtier and Revenaugh, 2006], and the eastern Australian Plate [Courtier and Revenaugh, 2007]. One of the earliest detections by Revenaugh and Sipkin [1994] made use of ScS reverberations that sample a low-velocity layer at ∼330 km depth representing a ∼5.8% impedance decrease beneath easternmost Asia. The authors explained the LVZ by citing volatile-induced melting within the upper mantle and melt migration down to the 410 km 7 Chapter 2. Slave low-velocity zone discontinuity through some combination of slab entrainment and negative buoyancy. Later studies have exploited conversions identified on P and S receiver functions (RFs). Analysis of RFs affords higher resolution of inter- face structure immediately below a station [Lawrence and Shearer , 2006], and in studies cited above, characterize the LVZ as a 3-7% shear wave ve- locity decrease with thickness ranging from ∼20 km to ∼100 km, with an average of ∼60 km. These seismically determined thicknesses conflict with the expectation of thinner layers from geodynamic models. Hence it is desir- able to extract as much information as possible on physical properties from seismic observations so as to better constrain geodynamic models. In this study we return to the LVZ in NW Canada reported by Bostock [1998] ten years ago. He observed a clearly isolated negative arrival preceding the direct conversion from the 410 km discontinuity that was visible across a full range of slowness for P -receiver functions from the Yellowknife Seismic Array (YKA). The timing of this signal suggests it is a direct conversion from near 340 km depth. After a brief review of upper mantle structure within the study region, we proceed to investigate in more detail the physical properties of the LVZ observed at YKA. We then document the geographical extent and variability of the LVZ in NW Canada using data from two portable arrays. We conclude with a discussion of the implications of our findings. 2.2 Mantle Setting The area of interest comprises a broad E-W swath across the Northwest Territories extending from the Cordilleran deformation front in the west, through the Proterozoic Wopmay orogen, to the Archean Slave province in the east (Figure 2.2). The region has been host to several portable seismic experiments, and is situated favourably with respect to the distribution of global seismicity. Upper mantle structure throughout the region has been extensively mapped through a combination of seismic reflection/wide angle reflection, refraction, receiver functions, body-wave and surface wave tomog- raphy, SKS splitting, geochemical analyses, and natural-source magnetotel- lurics (MT). Large-scale mantle velocity structure is dominated by the tran- 8 Chapter 2. Slave low-velocity zone sition from low velocities beneath the Cordillera to high velocities below the Precambrian tectosphere, which is readily apparent in tomographic images produced from surface waves [Frederiksen et al., 2001; van der Lee and Fred- eriksen, 2005] and body waves [Grand , 1994; Bank et al., 2000; Grand et al., 1997]. Recent body wave studies indicate that the Cordillera-craton transi- tion takes place over a relatively narrow interval (∼50 km) and underlies the Cordilleran deformation front [Mercier et al., 2009], as opposed to previous evidence suggesting an extention of ancestral North American mantle west- ward to the Tintina fault [Frederiksen et al., 1998]. Rayleigh wave studies at long periods (>60 s) require that velocities beneath the Slave craton exceed the global cratonic average by ∼2% [Chen et al., 2007]. SKS splitting [Sny- der et al., 2004; Snyder and Lockhart , 2005] and MT studies [Jones et al., 2003] identify several prominent layers within the Slave tectosphere and re- veal the lithosphere-asthenosphere boundary at a depth of ∼260 km in the south of the craton. Numerous geochemical studies [e.g., Kopylova and Caro, 2004; Kopylova et al., 2004] of mantle xenoliths from abundant kim- berlite pipes predict a cooler than average lithosphere with a depth in the southern Slave in agreement with MT studies and a thinner lithosphere to the north, ∼190-200 km [Kopylova et al., 1997]. Crustal and shallow man- tle structures beneath the Wopmay orogen and Slave province associated with paleo-subduction episodes are revealed by Lithoprobe crustal profiles [Cook et al., 1998, 1999; Clowes et al., 2005; Fernández-Viejo et al., 2005] and receiver function studies [Bostock , 1998; Snyder , 2008; Mercier et al., 2008]. The Slave province can be subdivided into distinct western and east- ern portions based on Pb and Nd isotopic lines [Davis and Hegner , 1992; Davis et al., 2003]. The eastern Slave is isotopically juvenile, whereas the west is dominated by the >2.9 Ga Central Slave Basement Complex (CSBC) [Bleeker et al., 1999; Jones et al., 2003]. 9 Chapter 2. Slave low-velocity zone Figure 2.2: Map of the study area showing locations of stations with respect to regional scale tectonics. Inset map illustrating both the POLARIS Slave array and all three legs of the CANOE array. Main study region map only includes Leg A of CANOE array. Yellow dashed line indicates approximate RF profile for the CANOE dataset. Green dashed line indicates RF profile for POLARIS-Slave dataset. 10 Chapter 2. Slave low-velocity zone 2.3 Physical Properties of the LVZ at YKA In this section we examine the characteristics of deep upper mantle and transition zone structure beneath YKA. The section is broken into four parts. In the first, we introduce the YKA dataset. We then present the receiver functions and mantle reflectivity profiles, and subsequently proceed to investigate interval properties of the LVZ through analysis of traveltimes. Finally, we use the amplitudes to establish material property contrasts at the boundaries of the LVZ. 2.3.1 Data The Yellowknife Array (YKA) consists of five three-component broadband seismic stations deployed within a cross-shaped array of 15 km × 14 km near Yellowknife, Northwest Territories, and is operated by the Geological Survey of Canada within the Canadian National Seismograph Network (CNSN). Digital recording commenced in January 1989 for YKW2-4, whereas YKW1 and YKW5 were installed in January 1995 and November 2001, respectively. Of the five stations, only YKW5 has been decommissioned (February 2006) while the remainder continue to collect data. Given their proximity, RFs from all stations will be combined to improve signal-to-noise ratio (SNR), under the assumption that they represent identical underlying structure. During the period 1989-2008, ∼1000 earthquakes of magnitude≥5.9 were recorded. Events in our data set were limited to epicentral distances of 30◦ to 95◦, and seismograms were selected based on a SNR threshold and subsequent visual inspection, resulting in a dataset of 295 earthquakes and yielding a total of ∼1900 PRFs. The global distrubution of events used is illustrated in Figure 2.3a. 2.3.2 Receiver Functions at YKA Receiver function analysis involves the removal of the earthquake source- time function from scattered waves in the teleseismic P or S coda that are produced through interaction of the near-vertically propagating incident 11 Chapter 2. Slave low-velocity zone (a) 4.5 5 5.5 6 6.5 7 7.5 8 8.5 0 20 40 60 80 Slowness (s/°) N W 4.5 5 5.5 6 6.5 7 7.5 8 8.5 0 20 40 60 80 100 120 Slowness (s/°) SE (b) Figure 2.3: (a) Global earthquake source distributions for the events recorded and used at the Yellowknife array. Solid inner ring indicates epicen- tral distance minimum cutoff of 30◦, whereas the bounding circle indicates the maximum range of 95◦. (b) Slowness distribution for the NW (top) and SE (bottom) quadrants recorded at YKA. Although there are more events included in the SE corridor in total, they tend to be less evenly distributed than those for the NW corridor. 12 Chapter 2. Slave low-velocity zone wave with sub-horizontal discontinutities [Vinnik , 1977; Langston, 1979]. Particle motions of the recorded wavefield are transformed into upgoing P and S modes [e.g., Kennett , 1991] which, upon damped least-squares spec- tral deconvolution of one component by the other, produce an approxima- tion to the true Earth’s P - or S-Green’s function. The timing of the P -to-S or S-to-P direct conversions and reverberations in the resulting time series delivers constraints on large-scale velocity structure, whereas relative ampli- tudes and polarities provide insight into shorter wavelength S-velocity and S-impedance contrasts. PRF sections are plotted against slowness, p, to identify the move-out of different phases with respect to the direct wave arrival. Red represents positive conversions, where the incident wavefield converts at a ‘normal’ discontinuity (ie, velocity increase with depth), whereas blue indicates neg- ative conversions from a LVZ. The data are mapped from time to depth by shifting and stacking all PRFs along move-out curves for converted phases originating at a regularly sampled series of depths in a 1-D Earth model. The resulting mantle “reflectivity” profile is generated by interpolating along the expected time-depth trajectory for the IASP91 model [Kennett and Eng- dahl , 1991]. The stacking is augmented utilizing phase weights as introduced by Schimmel and Paulssen [1997]. The data analyzed at YKA were separated into two regions of back azimuthal coverage, a western Pacific source region (NW corridor) between 274-313◦ and a South American source region (SE corridor) between 131- 181◦ (Figure 2.3a). Previous work has indicated a strong signal dependence on azimuth, with arrivals from the NW tending to be most clear [Bostock , 1998; Cassidy , 1995]. Note that the extended duration of recording at YKA has provided a nearly continuous sampling in slowness for both the NW and SE quadrants, as illustrated in Figure 2.3b. Both NW and SE corridor datasets (Figures 2.4 and 2.5) display well defined upper mantle arrivals on the PRFs in agreement with prior results [Bostock , 1998]. Direct conversions (see figure 2.6 for illustration of phases) from both the 410 and 660 km disontinuities, P410s and P660s, are apparent on both the NW and SE corridors (figures 2.4a and 2.5a), with the expected 13 Chapter 2. Slave low-velocity zone negative differential slowness. These phases are also identified on the associ- ated mantle reflectivity plots in figures 2.4a and 2.5a. For both corridors, the 410 discontinuity is mapped to the (slightly shallower than nominal) depth of 405 km, likely due to intra-cratonic velocities that are faster than those in the background 1-D reference model IASP91. A comparable shallowing of the 660 discontinuity is observed, to a depth of 650 km. Preceeding the P410s arrival by ∼7 s within the NW corridor dataset is a negative polarity arrival, PW s, from the top of the postulated LVZ. The PW s phase exhibits the negative differential slowness of a direct conversion, and maps to an IASP91 depth of 335 km. The signal is partially obscured at greater slowness (0.0791≥ p ≥0.0751 s/km and 0.0666≥ p ≥0.0634 s/km), but remains clearly visible over the remainder of the slowness range. A similar arrival can be observed in the SE corridor dataset (figure 2.5a), although it does not appear with the same negative differential moveout. At large slowness, interference results perhaps through reverberation from a shallower structure. At lesser slowness, the arrival exhibits almost no moveout. In the mantle reflectivity plot (side panel 2.5a), the signal does not emerge above the noise level, as a result of this behaviour. As we will discuss in the following section, the reverberatory phases gen- erated through reflection at the Earth’s surface deliver important additional constraints on physical properties of the upper mantle above the transition zone. Figures 2.4b and 2.5b illustrate the reverberatory window for both the NW and SE corridors. Note the effect of contamination from the PP arrival when observing signals from these greater depths. It is important to exclude PP from the analysis because it causes degradation in the deconvolution. As a result, the PRFs are largely featureless after the time corresponding to the PP arrival. This effect limits the number of useful traces as the time of the conversion increases and slowness decreases. For Pp410s, only traces with p < 0.0569 s/km are available for analysis, whereas the effect is still more marked for Pp660s, where ∼90% of the data are unusable, p < 0.043 s/km. The longer paths travelled by the reverberations (including three tran- sits through the attenuative upper mantle and scattering crust) result in significantly higher noise levels than for direct conversions. Nevertheless, 14 Chapter 2. Slave low-velocity zone (a) Direct Conversion Window (b) P Reverberation Window Figure 2.4: Radial PRFs computed for NW corridor of YKA. Results were filtered between 0.02 and 0.4 Hz. (a) Windowed for direct conversions. IASP91 predicted moveout curves for the PW s, P410s, and P660s phases are superimposed. Mantle reflectivity profile at right. Amplitudes at depths greater than 250 km are magnified for clarity. (b) Windowed for the rever- berations. Moveout curves for Pp410s and Pp660s are superimposed, along with the predicted PP arrival time. Reflectivity profile at right is zoomed on region from 250 km to 500 km only. 15 Chapter 2. Slave low-velocity zone (a) Direct Conversion Window (b) P Reverberation Window Figure 2.5: Radial PRFs computed for SE corridor of YKA. Results were filtered between 0.02 and 0.4 Hz. (a) Windowed for direct conversions. IASP91 predicted move-out curves for the PW s, P410s, and P660s phases are superimposed. Mantle reflectivity profile at right. Amplitudes at depths greater than 250 km are magnified for clarity. (b) Windowed for reverbera- tions. Moveout curves for Pp410s and Pp660s are superimposed, along with the predicted PP arrival time. Note that in the mantle reflectivity, the Pp410s is of significantly lower magnitude. 16 Chapter 2. Slave low-velocity zone Figure 2.6: The incident P phase and its coda. Conversions or reflections from a discontinuity at depth are denoted using the subscript d. Illustrated are the direct Pds conversion and the P (Ppds) and S (Psds) reverberations relative to the incident P wave. Solid lines denote P transits while dashed lines denote those for S. 17 Chapter 2. Slave low-velocity zone the Pp410s phase can still be identified in figure 2.4b near the predicted IASP91 time as a positive polarity pulse characterized by positive differ- ential slowness. This signal coherently stacks along the IASP91 predicted move-out curve, and an isolated positive pulse at ∼412 km is marked in the Ppds mantle reflectivity (figure 2.4b). Structures below 550 km cannot be reliably imaged due to the significant restriction on slowness caused by PP interference. The same effect contributes to the diminished amplitude of Pp410s with respect to arrivals from shallower depths. Within the first 40 s preceeding Pp410s, there is only a single well-defined arrival exhibiting the negative polarity and positive differential slowness expected of PpW s. This feature has a move-out of ∼12 to 15 s with respect to Pp410s at ∼115 s after direct P . It maps to a depth of ∼360 km in the mantle reflectivity (side panel figure 2.4b), almost 30 km deeper than that predicted for the direct conversion, suggesting that mantle velocities deviate significantly from IASP91 within this depth range. The reverberation window for the SE corridor data is plotted in figure 2.5b, with the accompanying mantle reflectivity. In comparison to the NW corridor results, the Pp410s phase appears to have undergone a greater de- gree of attenuation, as the mantle reflectivity does not show a distinct peak in the vicinity of 410 km depth. Moreover, there is no obvious PpW s ar- rival near depths of ∼360 km. As clearly identified reverberation phases are required for computation of physical properties within the LVZ, we will restrict further analysis to data from the NW corridor only. 2.3.3 Kinematic Constraints It is commonly understood that, at small slowness, the traveltimes (rela- tive to the incident P -wave arrival) of the direct and reverberated scattered modes, i.e., Pds, Ppds, and Psds, (see figure 2.6) from a horizontal disconti- nuity constrain primarily the average VP /VS (or, equivalently, the Poisson’s ratio σ) and H/VP ratios of the overlying column of material [Zandt and Ammon, 1995; Zandt et al., 1995; Zhu and Kanamori , 2000]. Somewhat less widely appreciated is the fact that differential times of the various scat- 18 Chapter 2. Slave low-velocity zone tered arrivals from two separate discontinuities can be used to constrain the corresponding properties of the intervening layer [Audet et al., 2009]. Our interest lies in the layer defined by the W and 410 km discontinuities. An estimate of the Poisson’s ratio for this interval would provide important information concerning the nature of the LVZ since fluid infiltrated media exhibit higher values than their dry counterparts. We note that for plane waves in horizontally stratified media, we may derive analytic formulae to solve for H and VP /VS or σ (and, in principle, VP as well, see Kumar and Bostock [2008]). In studies of transition zone structure, however, we must be attentive to the effect of non-planar wavefront curvature in properly ac- counting for wavefield kinematics [Lawrence and Shearer , 2006]. To constrain LVZ properties, we window both direct conversions and reverberations to isolate the arrivals of interest, and time-normalize to the 410 arrival in both cases (P410s and Pp410s, respectively). The time nor- malization is performed using multi-channel-cross-correlation [VanDecar and Crosson, 1990] within a 5 or 7 s window (Pds and Ppds, respectively) about the predicted IASP91 410 arrival time. We then compute travel times for each of the arrivals of interest (PW s, PpW s, P410s, and Pp410s) for a range of 1-D models that parameterize the layer between W and 410 as a region of variable σ and thickness H perched atop the 410 km discontinuity (see figure 2.7 for example). We note that it is more convenient to parameterize our models in σ than VP /VS since σ has a finite upper limit of 0.5, whereas VP /VS increases without bound. We allow σ to vary between 0.235 and 0.48 (with δσ = 0.012) and H to vary between 10 km and 160 km (with δH =7.5 km). In any particular model, the interval thickness H above 410 retains the VP profile of IASP91, and maintains a constant σ through variation of VS . Following Zhu and Kanamori [2000], we conduct a model grid search where windowed seismograms for each phase are shifted and summed for many models to produce a plot of model fit in σ vs H space. In the ideal case (i.e., noise free and infinite frequency), the trajectory traced out by a given phase (i.e., Pds or Ppds) defines a tradeoff curve. The intersection of these two curves marks the model that, simultaneously, best describes the 19 Chapter 2. Slave low-velocity zone Figure 2.7: VS model based on IASP91, where the interval W is simulated by inserting a layer of thickness H and constant σ (denoted as VP /VS). At left, example for five different layer thickness at a single VP/VS value. At right, example for five different VP /VS values for a single layer thickness. 20 Chapter 2. Slave low-velocity zone move-out of the direct and reverberated phases. In the case of real data with finite bandwidth and for which the assumpution of perfectly radial structure may not apply, these lines are smeared into broader zones. Automated bootstrap re-sampling is employed to compute the standard error distribution for σ and H [Efron and Tibshirani , 1991]. The grid search is repeated numerous times where, in each instance, the windowed receiver functions are re-sampled (with replacement) to a random subset of the orig- inal number of traces, and the best fit solution is determined. The standard deviation of this ensemble of solutions yields an estimate of the errors in the best fit σ and H. We performed an initial grid search using a low-resolution sampling of model space. The resulting plot of model fit for each phase is illustrated in figure 2.8a. Blue bands indicate contributions from negative arrivals and red bands correspond to positive arrivals. Since the top of the LVZ creates a negatively polarized phase, we focus only on blue signals. The right panel of figure 2.8a is the sum of the plots for the two phases. The intersection of the two blue trade-off curves marks the best fitting model. As indicated, the layer is less than 60 km thick with a Poisson’s ratio >0.35. Our low-resolution grid results in a blocky distribution of maxima. Ac- cordingly, we produce two successively finer grids to resolve the parame- ters more precisely. First, a medium resolution of 0.3 ≤ σ ≤ 0.48 with δσ = 0.005 and 10 ≤ H ≤ 121 km with δH = 3.83 km (figure 2.8b), followed by a high resolution grid employing 0.35 ≤ σ ≤ 0.46 with δσ = 0.0024 and 20 ≤ H ≤55 km with δH =0.7 km. Results shown in figure 2.9 indicate a best-fit model of σ=0.416, H=36 km. Bootstrap re-sampling (680 itera- tions) provides a 2σ error of ±3 km in H, and ±0.011 in σ. We will return to the interpretation of this result in a later section, but for now, compare the predicted times for the best fit model with those of the data. In figure 2.10 we have superimposed the predicted differential times over the original raw data. As our chosen model parameterization is sensitive only to the VP /VS ratio of the interval and not to the absolute velocities, we require a base shift (4 s for Pds and 3 s for Ppds) to correct for the discrepancy in absolute timing. For the direct conversions (figures 2.10a), 21 Chapter 2. Slave low-velocity zone (a) Low-resolution gridding (b) Medium-resolution gridding Figure 2.8: Model matching results. Left and middle panels illustrate results from direct conversions and P reverberations, respectively. Right panels are sum of plots for the two phases. Blue regions correspond to conversions from a LVZ, while red corresponds to normal discontinuities. We are concerned only with the blue regions of these figures. (a) the low-resolution gridding of model space and (b) the medium-resolution gridding of model space. In both cases, the best solution lies at interval thicknesses <60 km and Poisson’s ratios >0.35. 22 Chapter 2. Slave low-velocity zone Figure 2.9: Resulting trade-offs from model matching using high resolution model. Again, left panel corresponds to direct conversions, middle panel to the P reverberations, and the right panel is the sum of the previous two plots. Plus symbol denotes the best-fit solution of σ=0.416 and H=36 km. the P410s phase is clearly defined. The negative polarity signal from the W interval arrives between 5 and 8 s prior to the positive 410 signal and demonstrates a good fit with the predicted move-out. The response in the reverberations is less apparent than that for the direct conversions. The predicted PpW s move-out curve for the best fit model falls between 11 and 14 s prior to the Pp410s arrival, and overlies the only candidate coherent negative polarity arrival within the 30 s preceding Pp410s. 2.3.4 Dynamic Constraints Whereas the timing of the arrivals provides controls on the velocity struc- ture within the LVZ, the relative amplitudes of the different phases can be used to constrain material property contrasts at the layer boundaries. The transmission and reflection coefficients for a boundary possess differing sen- sitivities to contrasts in the three independent parameters that describe an isotropic elastic medium. To leading order, scattering coefficients involving mode conversion are independent of compressional moduli. Transmission coefficients involving mode conversion at small slownesses are almost exclu- sively controlled by shear velocity contrast, with minor contribution from density contrast. The reflection coefficients are sensitive instead to shear 23 Chapter 2. Slave low-velocity zone (a) Direct Conversion Window (b) P Reverberation Window Figure 2.10: Raw PRFs from the NW Corridor YKA dataset, filtered be- tween 0.02 to 0.4 Hz. (a) Windowed about the direct conversion of P410s. The superimposed curves are the differential times predicted by the best fitting model. (b) Windowed about the P reverberations for the Pp410s. The superimposed curves are the predicted differential times, as in (a). 24 Chapter 2. Slave low-velocity zone impedance contrast. In principle, therefore, amplitudes of direct conver- sions and reverberations could be analyzed simultaneously to recover both the shear velocity and density jump across the boundingW and 410 discon- tinuities. In practice, however, reverberation amplitudes suffer from addi- tional degradation due to geometrical spreading, scattering, and absorption through additional propagation within the crust and upper mantle which is difficult to accurately account for. Consequently, we will confine our atten- tion to the estimation of shear velocity contrast at theW and 410 interfaces using transmission coefficients recovered from PW s and P410s. Under the assumption that properties across a discontinuity exhibit only modest contrasts (i.e., ∆VS VS ≪ 1), we may linearize the expressions for the transmission coefficient (T SPU ) to a simple overdetermined system in the single unknown ∆VS VS [Aki and Richards, 2002], T SPU (pk) = − pkVP 2ηkVS ( −4V 2S ( p2k + ζkηk )) ∆VS VS , (2.1) where pk is horizontal slowness, and ζk and ηk are the vertical slownesses for P - and S- waves, respectively. The subscript k indexes the data in slowness, 1 ≤ k ≤ N (where N is the number of PRFs). The transmission coefficients T SPU (pk) are extracted directly from PRF data using arrival times predicted by the best fitting model computed as described above (and where PRFs have been amplitude normalized by the maximum value of the deconvolution resolution kernel). These data are in- verted for shear velocity contrast at the interfaces bounding the LVZ using equation 2.1. The results indicate a -7.8% decrease in shear velocity at W , the top of the LVZ, and a 7.3% increase at the base of the LVZ, the 410 discontinuity. The contrast in shear velocity in the IASP91 model at the 410 km discontinuity is ∼4%, so the result we obtain is almost double the standard value. The magnitude of the velocity change observed is quantita- tively consistent with the high VP/VS ratio obtained from the grid search, as an interval with elevated Poisson’s ratio would imply significantly reduced VS . We note that for a LVZ comprising dense silicate melt ponding atop the 25 Chapter 2. Slave low-velocity zone 410 km discontinuity, the relative changes in amplitude of the transmission and reflection coefficients for the W and 410 interfaces can be predicted based on the coefficients’ sensitivity to density and shear velocity. As we have noted, T SPU is sensitive to VS alone (to first order), whereas the reflec- tion coefficient is sensitive to shear impedance (or roughly the sum of relative perturbations in density and shear velocity). At the W interface, shear ve- locity decreases and density increases, whereas at the 410 km discontinuity, both shear velocity and density increase. Consequently, we would expect the amplitude ratio of PpW s to Pp410s to be smaller than that of PW s to P410s. However, we observe the opposite behaviour, as illustrated in figure 2.4b. We do not have a fully satisfactory explanation for this behaviour, though it may simply be a result of the both the increased noise and PP contamination at these greater times, as previously discussed. 2.4 Geographic Controls Our analysis of the YKA PRF data from different azimuthal corridors sug- gests a degree of lateral variability in the properties and geometry of the LVZ. We now use data from two portable experiments (CANOE and PO- LARIS Slave arrays) to examine the geographic variation in this feature across a larger region in NW Canada. To this end, we employ both PRF and SRFs to leverage additional information on the structure. PRFs and SRFs are computed using similar approaches, although closer attention is paid to source characterization and removal for SRFs. The computation of SRFs suffers from several complications which limit their utility [Bostock , 2007], although they have proven useful in detecting low- velocity zones within the mid-upper mantle, where the corresponding time interval is typically masked by crustal reverberations in PRFs [Vinnik et al., 2003; Kumar et al., 2007; Li et al., 2004; Zhou et al., 2000]. For both CANOE and POLARIS-Slave data, the raw RFs for each station are migrated to produce a 1-D reflectivity section, similar to those shown at right of the slowness panels in figures 2.4a and 2.5a. The migration procedure improves coherency, suppresses noise, and allows variation across multiple adjacent 26 Chapter 2. Slave low-velocity zone stations to be better assessed. 1-D migrations were performed using phase- weighted stacking [Schimmel and Paulssen, 1997], to emphasize signals with coherent phase across all contributing RFs. 2.4.1 Data The Canadian Northwest Experiment (CANOE) comprised a large array of ∼60 three-component broadband seismic stations deployed in a three- leg pattern radiating outward from Fort Nelson, British Columbia. Leg A extended northeast to Yellowknife, crossing the Wopmay Orogen. Leg B spanned a large portion of the Canadian Cordillera, including the Cordilleran deformation front, terminating at Whitehorse, Yukon Territory. Leg C headed south roughly paralleling the deformation front, to Edmonton, Al- berta. The stations were deployed in stages during two separate campaigns, commencing in June of 2003 and 2004. They were decomissioned in Septem- ber of 2005, resulting in 27 and 15 months operation time, respectively. The station spacing was an average of 35 km across most of the array, with the exception of a stretch along the Mackenzie Highway on Leg A where station spacing dropped to ∼15 km. The present study employs a subset of the data from the 27 CANOE sta- tions (A01-A17 and UBC01-UBC10) comprising Leg A. Over ∼2200 earth- quakes of M ≥5.5 were recorded within the epicentral distance range of 30◦ − 95◦ utilized in the PRF study. The back-azimuthal coverage is good from 225◦ to 40◦ and from 125◦ to 175◦, and poorly sampled otherwise (fig- ure 2.11a). Quality control reduced the dataset to 4600 RFs for 365 events, with an average of ∼175 per station. Events employed in the SRF study were limited to earthquakes larger than magnitude 6.1 within the epicen- tral distance range of 55◦–85◦ for S waves, to avoid the S triplication at ∼90◦ (figure 2.11b). Back-azimuthal coverage is similar to that for the PRF study. After selection of seismograms based on a SNR threshold and visual inspection, the dataset was reduced to 30 earthquakes, resulting in ∼10-15 fair-to-good quality RFs per station. Twenty three-component broadband seismic stations operated by the 27 Chapter 2. Slave low-velocity zone Portable Observatories for Lithospheric Analysis and Research Investigating Seismicity (POLARIS) consortium constitute the POLARIS-Slave Array. The array geometry is primarily linear (2-D) and oriented perpendicular to the strike of the major trends in mantle xenolith geochemistry [Snyder , 2008]; however, several stations lie off-axis and so provide some degree of 3-D control (figure 2.2). Stations were deployed during July-September in three consecutive years, 2001-2003, and were decomissioned after a three to five year period, depending on the station. Over the deployment, ∼320 earthquakes of magnitude≥5.9 were recorded, at epicentral distances between 30◦ to 95◦ for PRFs and 55◦ to 115◦ (not in- cluding 82◦ to 95◦ due to S wave triplication) for SRFs, with back-azimuthal sampling comparable to CANOE and YKA. A total of 8900 PRFs and 2380 SRFs of fair-to-good quality (based on SNR and visual inspection) were computed and used in this study, with on average 400 and 110 per station, respectively (figure 2.11c, d). 2.4.2 CANOE Figures 2.12 and 2.13 present the migrated SRFs and PRFs for each station along CANOE leg A arranged from west to east. Stations were deployed on Phanerozoic platform sediments, consequently the strong velocity contrast between the sediments and basement results in amplitude saturation of the first 2-3 s in raw PRFs and 10-20 km in PRF reflectivity. SRF profile The SRF reflectivity profile in figure 2.12 are lower frequency than those of the PRFs. The depth of investigation was limited to 350 km since SKS seismograms for most stations possessed insufficient SNR to be incorporated. The top 50 km of the profile is dominated by the Moho. A coherent negative arrival is evident along the eastern edge of the transect, beginning beneath station A12 at ∼280 km depth and dipping eastward to ∼310-320 km depth beneath station A16. This feature would appear to lie too deep to represent the lithosphere-asthenosphere boundary, but it is approaching the depths 28 Chapter 2. Slave low-velocity zone (a) CANOE P (b) CANOE S (c) SLAVE P (d) SLAVE S Figure 2.11: Global earthquake source distributions for the CANOE and POLARIS-Slave experiments. Solid inner ring indicates the minimum epi- central cutoff, whereas the bounding circle indicates the maximum range. (a) CANOE PRF study with sources between 30◦–95◦. (b) CANOE SRF study with sources bewteen 55◦–85◦, (c) Slave PRF study with sources be- tween 30◦–95◦, and (d) Slave SRF study with sources between 55◦–115◦. 29 Chapter 2. Slave low-velocity zone Figure 2.12: SRF profile for the CANOE dataset. Traces are reflectivity profiles generated for all the data used at each station. Depth of investigation was limited due to the limited range in slowness and a poorer SNR of both S and SKS. Raw RFs were filtered between 0.02 to 0.15 Hz prior to migration. LVZ marked by arrow is most coherent at eastern extent of transect. observed for the LVZ at YKA. There are indications of a similar structure extending westward from A12, although considerably less pronounced. PRF profile Immediately below the zone saturated by sediment multiples in the PRF section (figure 2.13), the Moho conversion (PMs) is clear at depths of 30 to 40 km across the transect, and dips eastward. The strong velocity contrasts between 100 and 200 km depth are a combination of true lithospheric man- tle structure and crustal multiples mapped incorrectly to greater depths. Mercier et al. [2008] provide a detailed explanation of crust and shallow mantle structure along the transect. 30 Chapter 2. Slave low-velocity zone (a) (b) Figure 2.13: PRFs for the CANOE dataset. Individual RFs at each sta- tion were filtered from 0.02 to 0.4Hz prior to 1D migration. (a) Section illustrating the mantle reflectivity beneath each of the included CANOE stations ordered west to east along Leg A. (b) Average 1-D sections gen- erated through stacking of 4 different back-azimuth bins of RFs from each station in order to generate regional representative estimates of mantle re- flectivity. Blue curve (data from all back azimuths), green curve (265◦ to 320◦), red curve (125◦ to 185◦), and black curve (320◦ to 80◦). Amplitudes beneath 250 km are magnified 5×. 31 Chapter 2. Slave low-velocity zone At deeper levels, the 410 and 660 km discontinuities can be observed most clearly beneath the stations A09-A14 in the middle of the profile. The stations to either end exhibit possible, weaker signals from both discontinu- ities, but suffer from a lower SNR due to a shorter deployment period. The depths to both discontinuities are consistently shallower than IASP91 across the study area, due presumably to higher average velocities in the overlying mantle column in this region. Variations in transition zone thickness across the transect are minor, ranging from 240-250 km, and so are consistent with the global average of 242 km reported by Lawrence and Shearer [2006]. The slightly broadened transition zone at the eastern end of the profile is consis- tent with a cooler than average mantle temperature. In figure 2.13b we plot a composite 1-D mantle reflectivity profile for the entire CANOE PRF data set. In addition, we present composite profiles for data from 4 separate back azimuth bins: 125-185◦, 265-320◦, 320-80◦, and 0-360◦. The southwest quadrant is not included as this back-azimuthal range is poorly sampled. Amplitudes at depths greater than 250 km are magnified by a factor of 5 to enhance their visibility. Of the two transition zone discontinuities, the 410 is the dominant signal in each of the 4 averages, whereas the 660 appears to exhibit some azimuthal variation. All four PRF profiles are consistent in providing little indication of a W discontinuity. 2.4.3 POLARIS Slave Array The SRFs and PRFs computed for the POLARIS-Slave dataset are plotted in figures 2.14 and 2.15. The data are of higher SNR than those for the CANOE array, and, consequently, provide significantly clearer signals from transition zone depths. Furthermore, the stations of the POLARIS-Slave array are deployed east of the Phanerozoic platform sediments, limiting near- surface signal-generated noise. Stations on these profiles are ordered in a line running dominantly north to south through the center of the array. 32 Chapter 2. Slave low-velocity zone SRF profile The SRF section for the Slave stations is presented in figure 2.14a. The direct Moho conversion, PMs is clearly evident across all stations at a depth of ∼ 35 km. The transition zone discontinuities are observed at most sta- tions, but appear weaker to the north, with the 410 more strongly affected. There is minor variation in the relative depths of the 410 and 660 disconti- nuities, although no systematic N-S trend is present. The 410 is shallowest at ∼400 km beneath stations SNPN and MGTN, near the southern end of the array. The discontinuity deepens to its average value of ∼405 km to- ward the Yellowknife area. The 660 fluctuates about the average value of 650 km across the profile with little correlation to variations in 410. Al- though several stations across the array do exhibit weak negative polarities in the expected depth range for the W layer arrival, there is little spatial coherence between stations. The 1-D back-azimuthal averages for the SRF data are shown in figure 2.14b. Both the 410 and 660 discontinuties stack coherently, where the latter demonstrates a stronger coherence between stations. The pulse shape for both discontinuities has a double-lobed nature, with the 410 as positive- negative whereas the 660 is negative-positive. PRF profile Figure 2.15a presents the PRF section for POLARIS-Slave. Both discon- tinuities are well defined beneath all stations in the profile and appear at consistent depths of ∼405 km and ∼650 km, respectively, in contrast to ob- servations from the SRF image where the 410 discontinuity signature weak- ens to the north. Since both PRFs and SRFs are dominantly sensitive to shear velocity contrasts, the discrepancy may be due to the difference in slowness sampling and SNR for the two datasets. The higher SNR of the POLARIS-Slave PRF data yields a clearer image at depths of ∼100 km above the 410 km discontinuity. A negative polarity reflection that is coherent across many stations varies in depth from∼330 km in the north, and deepens marginally to the south between stations MLON to 33 Chapter 2. Slave low-velocity zone (a) (b) Figure 2.14: SRFs for the POLARIS Slave dataset. Individual RFs at each station were filtered from 0.02 to 0.3Hz prior to 1D migration. (a) Section illustrating the mantle reflectivity beneath each of the stations ordered north to south. (b) Average 1D sections generated through stacking together 4 different back-azimuth bins of RFs from each station in order to generate regional representative estimate of mantle reflectivity. Blue curve (data from all back azimuths), green curve (265◦ to 320◦), red curve (125◦ to 185◦), and black curve (320◦ to 80◦). Amplitudes beneath 250 km are magnified 2×. 34 Chapter 2. Slave low-velocity zone (a) (b) Figure 2.15: PRFs for the POLARIS Slave dataset. Individual RFs at each station were filtered from 0.02 to 0.4Hz prior to 1-D migration. (a) Section illustrating the mantle reflectivity beneath each of the stations ordered north to south. (b) Average 1-D sections generated through stacking together 4 different back-azimuth bins of RFs from each station in order to generate regional representative estimate of mantle reflectivity. Blue curve (data from all back azimuths), green curve (265◦ to 320◦), red curve (125◦ to 185◦), and black curve (320◦ to 80◦). Amplitudes beneath 250 km are magnified 2×. Black arrow indicates location of potential LVZ. 35 Chapter 2. Slave low-velocity zone MGTN. This feature reappears at DSMN, and at YKW3 (a station at YKA) at a depth of 340 km, determined independently from the YKA NWCorridor 1-D reflectivity. It is of sufficient amplitude and sufficiently consistent timing to stack coherently into the average 1-D reflectivity presented in figure 2.15b. Both the 410 and 660 discontinuities are clearly defined and map to depths of 405 km and 650 km, respectively. 1-D averages for both the complete dataset and that for the NW corridor data (265◦ to 320◦) show a corresponding negative peak at depth of ∼340 km. Data from the SE or NNE quadrants show this feature less clearly. 2.5 Discussion Using differential travel times of direct conversions and reverberations from the W and 410 km discontinuities, we estimate the LVZ beneath YKA to be 36±3 km thick with a Poisson’s ratio of 0.42± 0.01 (VP /VS ≈ 2.69). The thickness estimate falls within the lower end of the range for LVZs reported previously. The IASP91 Poisson’s ratio immediately above the 410 km dis- continuity is ∼0.29, rendering our estimate substantially greater than typical values. Shear velocity contrasts determined through linearized inversion of transmission coefficients are −7.8% for W , and +7.3% for the 410 disconti- nuity. The shear velocity decrease at W is equivalent to higher estimates of previous studies, which range between 3% and 7%. The velocity increase of 7.3% at the 410 km discontinuity is almost double the shear velocity contrast represented in IASP91; the magnitude of this increase is qualitatively con- sistent with the high VP/VS ratio for the LVZ extracted from travel times. Moreover, given that the amplitudes of direct conversions are normalized by those of the direct P -wave to produce estimates of transmission coefficients, our recorded velocity contrasts may actually be biased to smaller values due to the higher degree of attenuation experienced by S- versus P -waves en route to the surface. The magnitude of the observed Poisson’s ratio implies either extreme physical conditions or a bias due to some form of error. We consider each possibility in turn. 36 Chapter 2. Slave low-velocity zone 2.5.1 Implications for Physical Conditions Recent results from Yoshino et al. [2007] conclude that, at pressure and temperature conditions corresponding to the base of the upper mantle, the dihedral angle approachs 0◦ for the olivine + H2O system. With completely wetted grain boundaries, the presence of even a small degree of partial melt has dramatic effects on the seismic wave velocity and attenuation. Atten- uation is increased, and for melt fractions of 1%, shear wave velocity de- creases by as much as 20-30% [Stocker and Gordon, 1975]. At first glance, these numbers appear, at least qualitatively, consistent with our estimates of properties within the LVZ below YKA. We can perform a more quanti- tative appraisal by combining the estimates of δVS and σ to determine the quantity δ log(VS) δ log(VP ) . As argued by Takei [2002], this quantity is controlled by pore geometry and is largely independent of pore volume. We can cast it in terms of our observations, specifically the shear velocity contrast determined from dynamic modelling and the change in VP /VS (=R) ratio inferred from the kinematic modelling: δ log VS δ log VP = RδVS RδVS+VSδR , (2.2) where background values for R and VS are assumed from the IASP91 velocity profile. Accordingly, for ∆VS/VS ≈ −0.08 and ∆R/R ≈ 0.36 we recover δ log(VS) δ log(VP ) = −0.27. The negative sign implies that the substantial increase in VP/VS ratio across W requires not only a decrease in VS , but also an increase in VP . The requirement for an increase in VP has important implications for our understanding of the elastic properties between W and the 410 km discon- tinuity. The P -wave velocity within an isotropic solid depends upon bulk modulus (κ), shear modulus (µ), and density (ρ), whereas the P -wave veloc- ity for a fluid is dependent only on the bulk modulus and density. Seismic waves traversing the LVZ will be sensitive to the bulk properties of the fluid- infiltrated material; consequently, the measured values reflect propagation through both the solid matrix and the melt. Our results indicate a simul- taneous increase in VP and decrease in VS , thereby requiring an increase in 37 Chapter 2. Slave low-velocity zone average κ that significantly outweighs the decrease in average µ. We recall that the stability of a hydrated melt layer atop the 410 km discontinuity is predicated upon a density cross-over where κ of the melt is lower than that of the solid matrix. Thus, to account for the increase in VP implied by our model requires a mechanism that permits an increase in average κ without affecting the stability of the LVZ. Let us consider the solid matrix, and, in particular, κ or bulk sound velocity of anhydrous versus hydrous solids. In a recent review by Jacobsen [2006], observations of elastic properties of olivine, wadsleyite, and ringwoodite were fit to equations of state to examine varia- tion with pressure and hydration. These equations suggest that a cross-over in the bulk sound velocities of hydrous and anhydrous solids may occur at higher pressures, and for olivine, in particular, near ∼300 km depth (figure 2.16). Simultaneous modelling for µ does not predict a cross over above the transition zone, and, consequently, the shear velocity for hydrous olivine re- mains below that of anhydrous olivine. A more recent study has reproduced these general results for the transition zone mineral wadsleyite, to pressures up to 10 GPa [Holl et al., 2008]. The presence of anhydrous olivine imme- diately above a hydrous LVZ might therefore afford a means of increasing VP , provided the effect outweighs the reduction in velocity caused by the melt phase. Jacobsen [2006] emphasizes the need for further experiments at greater pressure to better constrain properties at transition zone pressures, in particular for hydrous olivine, as past studies have focused on the anhy- drous form. Moreover, the magnitude of the velocity jump and a mechanism for producing a rapid transition from anhydrous to hydrous olivine at the top of the LVZ, remain to be explained. Another potential explanation for an increase in the compressional ve- locity may be related directly to the melting process. The two olivine end- members, forstertite (Mg2SiO4) and fayalite (Fe2SiO4), have significantly different elastic wave velocities; pure forsterite (Fo) has compressional and shear velocites of 8.76 km/s and 5.06 km/s at ∼3.1 GPa [Zha et al., 1996] re- spectively, whereas the corresponding values for fayalite (Fa) are 6.67 km/s and 3.39 km/s at 3.0 GPa [Speziale and Duffy , 2004]. For any given starting mantle composition, the extraction of melt results in a more magnesium- 38 Chapter 2. Slave low-velocity zone Figure 2.16: Computed bulk sound velocities for the various transition zone (Fe,Mg)2SiO4 polymorphs (sold lines). The dash and dash-circle lines depict the velocities predicted by various 1-D seismic velocity models. Note that for dry vs hydrous olivine, a crossover in the bulk sound velocity occurs within the depth range 300 to 350 km. Similar crossovers for the other phases can be observed, though at greater depths than applicable for the LVZ atop the transition zone. Figure modified from Jacobsen [2006]. 39 Chapter 2. Slave low-velocity zone rich residue and an iron-rich melt. As a consequence, the residue velocity is increased with respect to the starting composition, as iron preferentially enters the melt phase. This provides an additional mechanism that could contribute to our inference of an increase in VP into the low velocity zone. Moreover, this mechanism would not place any specific requirement on the material immediately overlying the LVZ. 2.5.2 Possible Sources of Error Given the difficulties in explaining our estimates of VP/VS and δ log(VS) δ log(VP ) , we must also consider potential sources of error that might bias or invalidate our results. We discuss three possibilities: 1) a bias in traveltime measurements used to estimate VP/VS , 2) misidentification of phases scattered from the LVZ, and 3) modelling error in the parameterization of this layer. Audet et al. [2009, supplementary materials] report a positive bias in Poisson’s ratio estimates derived in the general manner of section 2.3.3 that results from low-pass filtering. This effect will arise if the signals PW s, P410s or PpW s, Pp410s arrive too closely to one another on the PRFs, relative to their dominant periods. The identification of discrete and well separated signals in YKA data from the NW corridor preclude this mechanism as a source of bias. A second potential source of error concerns the possible misidentification of either or both of PW s or PpW s. The timing and move-out of the PW s arrival on individual PRFs match those predicted by the IASP91 model for a horizontal interface at 335 km. If the signal has been misidentified, the most likely candidate is a reverberation from a shallower discontinuity. There are, however, a range of difficulties with this explanation. First, it is difficult to produce a move-out with negative differential slowness for a reverberation, even when appealing to higher levels of interface dip. Second, of the can- didate discontinuities identified by Bostock [1998] at shallower levels in the Slave lithosphere, only the PsHs reverberation from a layer at 70-80 km ar- rives within a time window that coincides with PW s. Reverberations from deeper interfaces, X and L, arrive well after P410s. The PsHs phase would 40 Chapter 2. Slave low-velocity zone also be expected to be weaker than PpHs, for which there is little evidence, and, moreover, it should exhibit a distinct dipole signature that is not ob- vious. Finally, given the anisotropic nature of H, one would expect to see a comparable PsHs arrival on transverse component PRFs, but no such phase is evident with negative differential slowness. Thus it appears difficult to argue for an origin of the PW s signal as anything but a direct conversion. Let us now consider our analysis of reverberations from the LVZ. Al- though individual arrivals at these late times are not so easy to identify as the direct conversions, upon migration, the reflectivity profile (side panel figure 2.4b) reveals clearly defined signals. There is only one positive po- larity signal present within a 50 km window about the predicted IASP91 Pp410s, and it maps to 411 km. This peak rises significantly above noise levels, and can be observed as a faint signal within the PRF section (fig- ure 2.4b) along the predicted IASP91 move-out curve. Consequently, there can be little doubt as to its correct identification as Pp410s. Within the 100 km depth interval above 410 km, there exists a single well defined pulse in the reflectivity profile of figure 2.4b. It is negative polarity, and maps to a IASP91 depth of 360 km. Thus, if we accept, from arguments made above, that the signal identified as PW s must be a direct conversion, then the only likely candidate for the corresponding reverberation is the feature we have identified with PpW s. An alternative interpretation for PpW s as a higher order reverberation from a shallower discontinuity [e.g., H, X, or L, from Bostock , 1998] is readily dismissed. Whereas the traveltime and move-out of such a phase could mimic that of PpW s, its amplitude would be an order of magnitude or more smaller than Pp410s, which is clearly not the case. The third source of error involves modelling assumptions and here, the observations from the SE corridor YKA data set are suggestive. A PW s sig- nal is present in the SE corridor PRFs, but it is characterized by an absence of move-out with respect to direct P , and, consequently, a lack of signature on the mantle reflectivity profile in figure 2.5a. This behaviour cannot be explained for either a direct conversion or reverberation in a radially sym- metric Earth, and thus indicates the presence of lateral heterogeneity in the LVZ. This case is supported both by the depth variations documented across 41 Chapter 2. Slave low-velocity zone CANOE and POLARIS-Slave profiles in sections, and by a previous study that has argued for rapid lateral variations in LVZ thickness above the 410 km discontinuity in the western US [Song et al., 2004]. We proceed to characterize the effects of lateral heterogeneity by per- forming forward modelling through a simple LVZ that posesses dipping pla- nar boundaries, using the approach of Frederiksen and Bostock [2000]. The LVZ is characterized by a constant σ=0.333 (VP=7.47 km/s, VS=3.74 km/s) and is overlain by a 350 km thick upper mantle of σ=0.27 (VP=8.3 km/s, VS=4.6 km/s). An infinite halfspace beginning at 400 km depth underlies the LVZ, with σ=0.293 (VP=9.03 km/s, VS=4.87 km/s). The LVZ upper boundary (i.e., W ) dip varies from 0◦ to 30◦. Synthetic receiver functions are generated for a range of slowness from 0.04 to 0.08 s/km at different back azimuths. Our first task is to investigate what circumstances might lead to an absence of move-out for a direct PW s conversion. The only model for which we could substantively remove the move-out involved a dipping upper boundary (i.e., W ) sampled from back azimuths in the updip direction. At interface dips beyond 20◦ to 30◦, the signals begin to deviate significantly from the laterally homogenous case, as a result of polarity reverals. We can produce a reduction in the move-out of reverberations (e.g., PpW s) by sampling from back azimuths in the downdip direction, although, again, be- yond dips of 10◦ to 15◦, the behaviour begins to differ strongly from that expected for a laterally homogenous model. For no geometry were we able to generate a move-out with strictly negative differential slowness (such as that observed in the NW corridor YKA section for PW s (figure 2.4a)), for reverberations. We now examine the effect that dipping interfaces have on recovery of Poisson’s ratio and thickness when we assume a laterally homogeneous LVZ in the inverse modelling, as in section 2.3.3. As before, we generate synthetic receiver functions for the slowness range 0.04 s/km to 0.08 s/km. Using the method of section 2.3.3, we determine the best-fitting horizontal layer for a range of dips and back azimuths. We present results for four representa- tive models in figure 2.17. Figure 2.17a corresponds to a horizontal LVZ for which Poisson’s ratio and thickness are correctly determined. For all models 42 Chapter 2. Slave low-velocity zone where the upper surface of the LVZ is dipping (but the 410 km discontinu- ity remains horizontal), solutions were biased to smaller Poisson’s ratios, regardless of back azimuth (for example, figure 2.17b). Where large dips produce polarity reversals, no valid solutions can be found within the model space (figure 2.17c). The only circumstance in which a positive bias in Pois- son’s ratio could be generated was through the introduction of dip on the 410 km discontinuity (figure 2.17d). We regard it unlikely that this mech- anism could be responsible for the large Poisson’s ratio observed at YKA because the 410 km discontinuity represents a phase transition and there is little evidence for significant structure on this interface either locally or globally (that is, away from subduction zones). 2.5.3 Geographical Variations Results from PRF and SRF studies of the portable array datasets docu- ment a well defined transition zone beneath a large portion of northwestern Canada. Average depths to the 410 and 660 km discontinuities are close to global averages at 405 and 650 km. The two datasets provide evidence for LVZs to the west and northeast of YKA. SRF data from CANOE leg A reveals the LVZ to dip at ∼10◦ to the east, in the direction of YKA over a distance of ∼ 230 km. It deepens from a nominal (IASP91) depth of 280 km at its western edge to 310 km at the eastern extent, consistent in extrapolation with the depth estimated at YKA. A similar feature can be traced along the central to southern half of the POLARIS-Slave array for the PRF dataset, and exhibits an average nominal depth of 340 km as determined from all azimuths. The feature is best defined for NW back az- imuths, consistent with its behaviour at YKA. Between stations ACKN and NODN (∼150 km separation), there is little variation in the nominal depth of W , whereas from MLON to MGTN the feature dips to ∼360 km over ∼120 km. Stations DSMN and YKW3, which are ∼300 km offset westward from the main profile, again reveal the LVZ at a depth of 340 km. Figure 2.18 presents a summary cartoon of the results from both the CANOE and POLARIS-Slave array datasets. The Moho, 410, and 660 km discontinuities 43 Chapter 2. Slave low-velocity zone (a) (b) (c) (d) Figure 2.17: Results from synthetic modelling for dipping geometries. Each model consists of a 350 km thick homogeneous upper mantle overlying a 50 km thick LVZ (σ =0.333), underlain by a homogeneous transition zone. (a) Flat layer model results. Direct conversions arrive between 40 and 60 s, while reverberations arrive after 100 s. (b) Simulated 10◦ dip of the W interface (with horizontal 410) sampled along strike. Solution is biased to- wards decreased Poisson’s ratio and increased layer thickness. (c) W layer dipping 20◦ (again, horizontal 410) sampled from the updip direction. (d) Simulated 10◦ dip of the 410 km discontinuity (while W interface held hori- zontal). This produces a bias towards higher Poisson’s ratios and decreased layer thickness. 44 Chapter 2. Slave low-velocity zone are indicated in red, along with the top of the LVZ (W ) in blue. The de- gree of spatial variability across both CANOE and POLARIS Slave datasets is consistent with the work of Song et al. [2004], who characterize the LVZ along the Pacific Northwest margin with still larger thickness variations of 20 to 90 km over horizontal scale-lengths of 100 to 200 km. We note that these thickness variations are defined relative to normal mantle velocity models. If elastic properties are as extreme as our inverse modelling would suggest, the variations must be taken as upper limits. Nonetheless, thickness variations are predicted by geodynamic modelling as a result of horizontal gradients in efficiency of viscous re-entrainment of wet material back into the transition zone [Leahy and Bercovici , 2007]. Finally we comment on the apparent discrepancies between results from PRF and SRF data sets along the CANOE and Slave arrays, notably the fact that only one and not both datasets reveal the LVZ in both transects. We speculate that inherent differences in frequency, SNR, geographical and slowness sampling between PRFs and SRFs result in somewhat intermittent detection of a laterally variable LVZ. 2.6 Concluding Remarks We have presented evidence for a LVZ in shear velocity atop the transition zone in northwestern Canada that is laterally variable in its apparent thick- ness. The wealth of broadband data accumulated from YKA has afforded us the opportunity to constrain the local elastic properties of the LVZ through the identification of direct conversions and free-surface reverberations gen- erated at the W and 410 km discontinuities. The relative timing of these phases leads to an estimate (0.42±0.01) of Poisson’s ratio that significantly exceeds that of typical mantle materials. In combination with a shear ve- locity contrast of −7.8% at the top of the layer, as determined from forward conversion amplitudes, our modelling implies an increase in P-velocity into the layer. This result is difficult to reconcile with the current understanding of mantle LVZ’s and, in particular, with the stability of a dense melt layer. We have examined possible sources of error in our approach and find none 45 Chapter 2. Slave low-velocity zone Figure 2.18: Cartoon representation of study region with two representative cross sections for both the CANOE dataset and the POLARIS-Slave dataset. Red layers represent approximate Moho, 410, and 660 km discontinuities. Blue layer represents the W interface, the top of the LVZ. Dashed blue curve indicates regions with no strong evidence for the LVZ. Not that depths depicted are approximate. 46 Chapter 2. Slave low-velocity zone to be overly compelling. In particular, lateral heterogeneity in the form of a dipping, planar upper boundary (W ) does not appear to be capable of pro- ducing positive bias in Poisson’s ratio estimates. We hope that the results reported here will lead to new efforts to characterize this interesting region of Earth’s mantle. 47 Bibliography Agee, C. B., Static compression of hydrous silicate melt and the effect of water on planetary differentiation, Earth And Planetary Science Letters, 265, 641–654, 2007. Agee, C. B., Compressibilty of water in magma and the prediction of den- sity crossovers in mantle differentiation, Phil. Trans. R. Soc. A, 366, 4239– 4252, doi:10.1098/rsta.2008.0071, 2008. Agee, C. B., and D. Walker, Static compresssion and olivine flotation in ultrabasic silicate liquid, Journal of Geophysical Research, 93, 3437–3449, 1988. Aki, K., and P. G. Richards, Quantitative Seismology, 2 ed., University Science Books, Sausilito, California, 2002. Audet, P., M. Bostock, N. I. Christensen, and S. M. Peacock, Seismic evidence for overpressured subducted oceanic crust and megathrust fault sealing, Nature, 457, 76–78, doi:10.1038/nature07650, 2009. Bank, C. G., M. G. Bostock, R. M. Ellis, and J. F. Cassidy, A reconnais- sance teleseismic study of the upper mantle and transition zone beneath the Archean Slave craton in NW Canada, Tectonophysics, 319 (3), 151–166, 2000. Bell, D. R., and G. R. Rossman, Water in Earth’s mantle: The role of nominally anhydrous minerals, Science, 255, 1391–1397, 1992. Bercovici, D., and S. Karato, Whole-mantle convection and the transition- zone water filter, Nature, 425 (6953), 39–44, 2003. 48 Bibliography Bleeker, W., J. Ketchum, V. Jackson, and M. Villeneuve, The Central Slave Basement Complex: Part I. Its structural topology and autochthonous cover, Canadian Journal of Earth Science, 36, 1083–1109, 1999. Bolfan-Casanova, N., N. Kleppe, M. Welch, and K. Wright, Water in the Earth’s mantle: protons in minerals, Mineralogical Magazine, 69, 229–257, 2005. Bostock, M., Teleseismic Body-Wave Scattering and Receiver-Side Struc- ture, in Seismology and the Structure of the Earth, Treatise on Geophysics, vol. 1, edited by G. Schubert, pp. 219–246, Elsevier, 2007. Bostock, M. G., Mantle stratigraphy and evolution of the Slave province, Journal Of Geophysical Research, 103 (B9), 21,183–21,200, 1998. Cassidy, J. F., A comparison of the receiver structure beneath the sta- tions of the Canadian National Seismograph Network, Canadian Journal of Earth Science, 32, 938–951, 1995. Chen, C. W., S. Rondenay, D. Weeraratne, and D. Snyder, New constraint on upper mantle structure of the Slave craton from Rayleigh wave inversion, Geophysical Research Letters, 34, 2007. Clowes, R., P. Hammer, G. Fernández-Viejo, and J. Welford, Lithospheric structure in northwestern Canada from Lithoprobe seismic refraction and related studies: a synthesis, Canadian Journal Of Earth Sciences, 42, 1277– 1293, 2005. Cook, F., A. van der Velden, K. Hall, and B. Roberts, Frozen subduction in Canada’s Northwest Territories: Lithoprobe deep lithosphereic relflection profiling of the western Canadian Shield, Tectonics, 18, 1999. Cook, F. A., A. Van der Velden, K. Hall, and B. Roberts, Tectonic de- lamination and subcrustal imbrication of the Precambrian lithosphere in northwestern Canada, mapped by lithoprobe, Geology, 26, 839–842, 1998. 49 Bibliography Courtier, A. M., and J. Revenaugh, A Water-Rich Transition Zone Be- neath the Eastern United States and Gulf of Mexico from Multiple ScS Reverberations, in Earth’s Deep Water Cycle, Geophysical Monograph Se- ries, vol. 168, edited by S. Jacobsen and S. van der Lee, pp. 181–194, AGU, Washington, D.C., 2006. Courtier, A. M., and J. Revenaugh, Deep upper-mantle melting beneath the Tasman and Coral Seas detected with multiple scs reverberations, Earth and Planetary Science Letters, 259, 66–76, doi:10.1016/j.epsl.2007.04.027, 2007. Davis, W., and E. Hegner, Neodymium isotopic evidence for the tectonic assembly of Late Archean crust in the Slave province, northwest Canada, Contributions to Mineral Petrology, 111, 1992. Davis, W., A. G. Jones, W. Bleeker, and H. Grutter, Lithospheric devlop- ment in the Slave craton: A linked crustal and mantle perspective, Lithos, 71, 575–589, doi:10.1016/S0024-4937(03)00131-2, 2003. Efron, B., and R. Tibshirani, Statistical data analysis in the computer age, Science, 253, 390–395, 1991. Fernández-Viejo, G., R. Clowes, and J. Welford, Constraints on the compo- sition of the crust and uppermost mantle in northwestern Canada: VP /VS variations along Lithoprobe’s SNorCLE transect, Canadian Journal Of Earth Sciences, 42, 1205–1222, 2005. Frederiksen, A. W., and M. G. Bostock, Modelling teleseismic waves in dipping anisotropic structures, Geophysical Journal International, 141 (2), 401–412, 2000. Frederiksen, A. W., M. G. Bostock, J. VanDecar, and J. F. Cassidy, Seismic structure of the upper mantle beneath the northern Canadian Cordillera from teleseismic travel-time inversion, Tectonophysics, 294, 43–55, 1998. 50 Bibliography Frederiksen, A. W., M. G. Bostock, and J. F. Cassidy, S-wave velocity structure of the Canadian upper mantle, Physics Of The Earth And Plan- etary Interiors, 124 (3-4), 175–191, 2001. Gao, W., E. Matzel, and S. P. Grand, Upper mantle seismic structure beneath eastern Mexico determined from P and S waveform inversion and its implications, Journal Of Geophysical Research-Solid Earth, 111 (B8), B08,307, 2006. Grand, S. P., Mantle shear structure beneath the Americas and surrounding oceans, Journal Of Geophyiscal Research, 99 (B6), 11,591–11,621, 1994. Grand, S. P., R. van der Hilst, and S. Widiyantoro, High resolution global tomography: a snapshot of convection in the Earth, GSA Today, 7, 1–7, 1997. Hirschmann, M., C. Aubaud, and A. Withers, Storage capacity of H2O in nominally anhydrous minerals in the upper mantle, Earth and Planetary Science Letters, 236, 167–181, 2005. Hirschmann, M., A. Withers, and C. Aubaud, Petrologic structure of a hydrous 410 km discontinuity, in Earth’s Deep Water Cycle, Geophysical Monograph Series, vol. 168, edited by S. Jacobsen and S. van der Lee, pp. 227–287, AGU, Washington, D.C., 2006. Holl, C., J. Smyth, S. Jacobsen, and D. Frost, Effects of hydration on the structure and compressibility of wadsleyite, β-(Mg2SiO4), American Mineralogist, 93, 598–607, 2008. Inoue, T., Effect of water on melting phase relations and melt composition in the system Mg2SiO3-H2O up to 15 GPa., Physicas of the Earth And Planetary Interiors, 85, 237–263, 1994. Jacobsen, S., Effect of Water on the Equation of State of Nominally An- hydrous Minerals, Reviews in Mineralogy and Geochemistry, 62, 321–342, 2006. 51 Bibliography Jasbinsek, J., and K. Dueker, Ubiquitous low-velocity layer atop the 410- km discontinuity in the northern Rocky Mountains, Geochemistry Geo- physics Geosystems, 8, Q10,004, 2007. Jones, A. G., P. Lezaeta, I. J. Ferguson, A. D. Chave, R. L. Evans, X. Gar- cia, and J. Spratt, The electrical structure of the Slave craton, Lithos, 71 (2-4), 505–527, 2003. Kennett, B., The removal of free surface interaction from three-component seismograms, Geophysical Journal International, 104, 153–163, 1991. Kennett, B., and E. R. Engdahl, Traveltimes for global earthquake location and phase identification, Geophysical Journal International, 105 (2), 429– 465, 1991. Kohlstedt, D., H. Keppler, and D. C. Rubie, Solubility of water in the α, β, and γ phases of (Mg,Fe)2SiO4, Contributions to Mineral Petrology, 123, 345–357, 1996. Kopylova, M., and G. Caro, Mantle xenoliths from the Southeastern Slave craton: Evidence for chemical zonation in a thick, cold lithosphere, Journal of Petrology, 45 (5), 1045–1067, doi:{10.1093/petrology/egh003}, 2004. Kopylova, M., J. Russel, and H. Cookenboo, Petrology of peridotite and pyroxenite xenoliths from Jericho Kimberlite; implications for the thermal state of the mantle beneath the Slave Craton, northern Canada, Journal of Petrology, 40, 79–104, 1997. Kopylova, M., J. Lo, and N. Christensen, Petrological constraints on seis- mic properties of the Slave upper mantle (Northern Canada), Lithos, 77 (1- 4), 493–510, doi:{10.1016/j.lithos.2004.03.012}, 2004. Kumar, M., andM. Bostock, Extraction of absolute P velocity from receiver functions, Geophysical Journal International, 175, 515–519, doi:10.1111/j. 1365-246X.2008.03963.x, 2008. 52 Bibliography Kumar, P., X. H. Yuan, M. R. Kumar, R. Kind, X. Q. Li, and R. K. Chadha, The rapid drift of the Indian tectonic plate, Nature, 449 (7164), 894–897, 2007. Langston, C., Structure under Mount Rainier, Washington, inferred from teleseismic body waves, Journal of Geophysical Research, 84 (B9), 4749– 4762, 1979. Lawrence, J. F., and P. M. Shearer, A global study of transition zone thickness using receiver functions, Journal of Geophysical Research, 111, doi:10.1029/2005JB003973, 2006. Leahy, G. M., and D. Bercovici, On the dynamics of a hydrous melt layer above the transition zone, Journal Of Geophysical Research, 112 (B7), B07,401, 2007. Li, X. Q., R. Kind, X. H. Yuan, I. Wolbern, and W. Hanka, Rejuvenation of the lithosphere by the Hawaiian plume, Nature, 427 (6977), 827–829, 2004. Litasov, K., E. Ohtani, F. Langenhorst, H. Yurimoto, T. Kubo, and T. Kando, Water solubility in Mg-perovskites and water storage capacity in the lower mantle, Earth and Planetary Science Letters, 211, 189–203, 2003. Litasov, K. D., E. Ohtani, H. Kagi, S. D. Jacobsen, and S. Ghosh, Temperature dependence and mechanism of hydrogen incorporation in olivine at 12.5-14.0 GPa, Geophysical Research Letters, 34, doi:10.1029/ 2007GL030737, 2007. Mercier, J. P., M. G. Bostock, P. Audet, J. B. Gaherty, E. J. Gar- nero, and J. Revenaugh, The Teleseismic Signature of Fossil Subduction: Northwestern Canada, Journal of Geophysical Research, 113, B04308, doi: 10.1029/2007JB005127, 2008. Mercier, J. P., M. G. Bostock, J. F. Cassidy, K. Dueker, J. B. Gaherty, E. J. Garnero, J. Revenaugh, and G. Zandt, Body-Wave tomography of western Canada, Tectonophysics, in press, accepted, 2009. 53 Bibliography Ohtani, E., Y. Nagata, A. Suzuki, and T. Kato, Melting Relations Of Peridotite And The Density Crossover In Planetary Mantles, Chemical Geology, 120 (3-4), 207–221, 1995. Revenaugh, J., and S. A. Sipkin, Seismic Evidence For Silicate Melt Atop The 410 Km Mantle Discontinuity, Nature, 369 (6480), 474–476, 1994. Sacks, I., and J. Snoke, The use of converted phases to infer the depth of the lithosphere-asthenosphere boundary beneath South America, Journal of Geophysical Research, 82 (B14), 1977. Sakamaki, T., A. Suzuki, and E. Ohtani, Stability of hydrous melt at the base of the Earth’s upper mantle, Nature, 439, 192–194, 2006. Schimmel, M., and H. Paulssen, Noise reduction and detection of weak, coherent signals through phase-weighted stacks, Geophysical Journal In- ternational, 130 (2), 497–505, 1997. Snyder, D. B., Stacked uppermost mantle layers within the Slave craton of NW Canada as defined by anisotropic seismic discontinuities, Tectonics, 27, doi:10.1029/2007TC002132, 2008. Snyder, D. B., and G. D. Lockhart, Kimberlite trends in NW Canada, Journal Of The Geological Society, 162, 737–740, 2005. Snyder, D. B., S. Rondenay, M. G. Bostock, and G. D. Lockhart, Mapping the mantle lithosphere for diamond potential using teleseismic methods, Lithos, 77 (1-4), 859–872, 2004. Song, T. R. A., D. V. Helmberger, and S. P. Grand, Low-velocity zone atop the 410-km seismic discontinuity in the northwestern United States, Nature, 427 (6974), 530–533, 2004. Speziale, S., and T. S. Duffy, Single-crystal elasticity of fayalite to 12 GPa, Journal of Geophysical Research, 109, doi:10.1029/2004JB003162, 2004. Stocker, R., and R. Gordon, Velocity and internal friction in partial melts, Journal of Geophysical Research, 80, 4828–4836, 1975. 54 Bibliography Stolper, E., D. Walker, B. H. Hager, and J. F. Hays, Melt Segregation From Partially Molten Source Regions - The Importance Of Melt Density And Source Region Size, Journal Of Geophysical Research, 86 (NB7), 6261– 6271, 1981. Takei, Y., Effect of pore geometry on VP/VS: From equilibrium geometry to crack, Journal of Geophysical Research, 107, doi:10.1029/2001JB000522, 2002. van der Lee, S., and A. W. Frederiksen, Surface-wave tomography applied to the north American upper mantle, Geophysical Monograph, 157, 67–80, 2005. VanDecar, J., and R. Crosson, Determination of telseismic relative phase arrival times using multi-channel cross-correlation and least squares, Bul- letin of the Seismological Society of America, 80 (1), 150–169, 1990. Vinnik, L., Detection of waves converted from P to SV in the mantle, Physics of the Earth and Planetary Interiors, 67, 39–45, 1977. Vinnik, L., M. R. Kumar, R. Kind, and V. Farra, Super-deep low-velocity layer beneath the Arabian plate, Geophysical Research Letters, 30 (7), 1415, 2003. Yoshino, T., Y. Nishimura, and S. Karato, Complete wetting of olivine grain boundaries by a hydrous melt near the mantle transition zone, Earth and Planetary Science Letters, 256, 466–472, 2007. Zandt, G., and C. Ammon, Continental crust composition constrained by measurements of crustal Poissons ratio, Nature, 375, 152–154, 1995. Zandt, G., S. Myers, and T. Wallace, Crust and mantle structure across the Basin and Range—Colorado plateau at 37d north latitude and implications for Cenozoic extensional mechanism, Journal of Geophysical Research, 100, 10,592–10,548, 1995. 55 Bibliography Zha, C., T. S. Duffy, R. Downs, H. Mao, and R. Hemley, Sound velocity and elasticity of single-crystal forsterite to 16 GPa, Journal of Geophysical Research, 101 (B8), 17,535–17,545, 1996. Zhou, L. M., W. P. Chen, and S. Ozalaybey, Seismic properties of the central Indian shield, Bulletin Of The Seismological Society Of America, 90 (5), 1295–1304, 2000. Zhu, L. P., and H. Kanamori, Moho depth variation in southern Califor- nia from teleseismic receiver functions, Journal Of Geophysical Research, 105 (B2), 2969–2980, 2000. 56 Chapter 3 Conclusion 3.1 Summary of Results At the beginning of this thesis, we outlined two main objectives. First, we set out to characterize interval properties of the LVZ documented in the YKA dataset. We accomplished this through the analysis of kinematic and dy- namic contraints provided by the computed PRFs. Through incorporation of the direct conversions and free-surface reverberations from the bounding interfaces (i.e., W and 410), we developed a kinematic modelling technique to constrain the Poisson’s ratio and thickness of the interval, with minimal sensitivity to the absolute velocity within the LVZ and the properties of the overlying column of material. A similar technqiue was used successfully to document overpressured fluids within the Cascadia subduction zone [Audet et al., 2009]. By correctly accounting for wavefront curvature, we have ex- tended the method to structures at transition zone depths. In addition, we estimated the shear velocity contrasts at the top and bottom of the LVZ through inversion of transmission coefficients extracted from the PRFs. The best-fitting model determined from kinematic modelling estimated a Poisson’s ratio of 0.42 and a thickness of 36 km for the LVZ, with a shear velocity decrease of 7.8% at W and a corresponding 7.3% increase at the 410 km discontinuity. As we have discussed, the magnitude of the Poisson’s ratio is higher than that predicted from global velocity models for transition zone depths, and indicates not only a decrease in VS intoW , but an increase in VP also. The changes in elastic moduli requried by these velocity con- trasts are difficult to justify, although a potential explanation involves a bulk sound velocity cross-over at high pressures for hydrous vs anhydrous min- eral phases. We reviewed the potential sources of error that could result in a 57 Chapter 3. Conclusion bias in our Poisson’s ratio estimate. These included a bias in the traveltime measurements used to estimate VP/VS , misidentification of the PW s and PpW s phases, and a modelling error in the parameterization of the layer. We excluded a modelling bias based on the frequency content and phase separation of the data, and argued for the correct identification of W based on the distinct negative moveout and the lack of a clear alternate phase. The effect of unmodelled dipping structure was demonstrated to effect the Poisson’s ratio, although reproduction of an upward biased Poisson’s ratio was difficult to achieve without the additional requirement for topography on the 410 km discontinuity. The second objective outlined at the outset of the thesis was to examine the geographic extent of the LVZ across northwestern Canada, from the Cordilleran deformation front in the west to the Archean Slave craton in the east. We computed total RFs of almost 3000 S and 14000 P from both the CANOE and POLARIS-Slave portable arrays. Profiles of mantle reflectivity were created for lines running from west to east along CANOE leg A, and north to south for POLARIS-Slave. From the latter, we documented a well defined transition zone across the entire profile, with coherent 410 km and 660 km signals. Evident in the PRFs is a LVZ at a depth of ∼340 km. The signal demonstrates a degree of variability along the profile, with a nominal interface depth between ∼320-360 km. Results from the CANOE profile tended to be noisier, due in part to the shorter deployment period. PRF results image the 410 and 660 km discontinuities, while an emergent anomaly is visible on the CANOE leg A SRFs, beginning in the west at ∼280 km and dipping eastwards towards YKA to a depth of ∼310 km. The proximity in depth and lateral position between this anomaly and that observed in the YKA NW corridor PRFs suggests they represent the same feature. The appearance of the LVZ beneath station A12 provides a westward limit of the LVZ within the middle of the Proterozoic Wopmay orogen, well east of the Cordilleran deformation front. In conclusion, we have demonstrated that the modelling of the LVZ observed at YKA using the differential times from direct conversions and reverberations from the bounding interfaces has potential utility in charac- 58 Chapter 3. Conclusion terizing the interval properties independent from the overlying column. The evidence presented suggests the existence of a melt layer situated above the 410 km discontinuity, with a minimum thickness of 36 km and an elevated Poisson’s ratio of 0.42. Geographic profiling using the portable array data indicates the presence of a potential LVZ at depths between 280 to 310 km beneath the eastern extent of CANOE leg A, and at 320 to 360 km across many stations of the POLARIS-Slave deployment. 3.2 Future Directions In order to improve the characterization of physical properties of this LVZ, both at Yellowknife and in extension to other locations, we suggest sev- eral modifications that have potential to improve the efficacy of the current technique. First, we propose the incorporation of additional reverberation phases into the modelling technique. Currently, we utilize only the direct conver- sions and the P reverberations (Pds and Ppds) from PRFs to constrain the differential times associated with the observed layer. Further refinement of the best fitting solution may be leveraged through incorporation of ad- ditional reverberation arrivals, including the Psds and Ppdp phases. The Ppdp differential times would be most advantageous, as they demonstrate very limited sensitivity with respect to the Poisson’s ratio and depend dom- inantly on the interval thickness. Analysis of global stacks of long period radial and vertical component seismograms show indications of such arrivals, including Pp410p and Pp660p [Shearer , 1991]. Promising new approaches to the determination of a high frequency source estimate for recovery of the P component Green’s function may enable the incorporation of this phase into the current technique [Bostock , 2008]. The Psds phase could also pro- vide further constraints, however, its longer traveltime limits the amount of usable data due to the effect of PP contamination in the deconvolution, in particular for conversions from the transition zone. The addition of the Ppdp to the existing Pds and Ppds phases would help to further constrain the trade-off curve intersection, and best-fit model solution. 59 Chapter 3. Conclusion Secondly, we consider the benefits of a more rigorous, three dimensional model parameterization. As we have demonstrated, the model parameteri- zation can play a role in the results of the best-fit model solution. In this study, we parameterized the LVZ as a horizontal layer of thickness H with a variable VP /VS ratio perched on the 410 km discontinuity. This simple interpretation may result in a Poisson’s ratio bias if the LVZ is laterally heterogeneous. By increasing the complexity of the model through the in- corporation of strike and dip as additional parameters, the robustness of the kinematic model technique could be improved, particularily when incorpo- rating one or both of the previously mentioned reverberations. In theory the additional model complexity is straightforward to incorporate into the current grid search scheme. In practice, modifications to the model algo- rithm would be required to compensate for the computational cost involved in accounting for non-planar wavefront curvature. This could be offset, for example, through parallelization of model iterations and traveltime calcula- tions. The proposed modifications provide an opportunity to further develop the methods discussed in this thesis. Through more complex three dimen- sional modelling and the incorporation of additional reverberation phases, we may continue to refine the estimates of physical properties of LVZs such as that imaged beneath YKA. 60 Bibliography Audet, P., M. Bostock, N. I. Christensen, and S. M. Peacock, Seismic evidence for overpressured subducted oceanic crust and megathrust fault sealing, Nature, 457, 76–78, doi:10.1038/nature07650, 2009. Bostock, M., Toward Recovery of the Earth’s High Frequency Elastic Wave Green’s Function, in EOS Trans. AGU, Fall Meet. Suppl., Abstract DI11A- 05, 2008. Shearer, P. M., Imaging Global Body Wave Phases by Stacking Long- Period Seismograms, Journal of Geophysical Research, 96 (B12), 1991. 61


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items