Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Crustal anisotropy in a subduction zone forearc : Northern Cascadia Matharu, Gian 2014

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_2014_spring_matharu_gian.pdf [ 1.87MB ]
JSON: 24-1.0167305.json
JSON-LD: 24-1.0167305-ld.json
RDF/XML (Pretty): 24-1.0167305-rdf.xml
RDF/JSON: 24-1.0167305-rdf.json
Turtle: 24-1.0167305-turtle.txt
N-Triples: 24-1.0167305-rdf-ntriples.txt
Original Record: 24-1.0167305-source.json
Full Text

Full Text

Crustal anisotropy in a subductionzone forearc: Northern CascadiabyGian MatharuB.Sc. (Hons), The University of British Columbia, 2011A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinThe Faculty of Graduate and Postdoctoral Studies(Geophysics)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)April 2014c© Gian Matharu 2014AbstractS -wave splitting analyses using high signal-to-noise ratio low frequency earth-quake (LFE) templates at 3-component stations across southern VancouverIsland (SVI) and northern Washington indicate the presence of a heteroge-neous distribution of crustal anisotropy in the North American plate. ForSVI, we investigate the contribution to anisotropy from the Leech RiverComplex (LRC), an allochthonous terrane comprised of strongly foliatedgreenschist facies phyllites and amphibolite facies schists with steeply dip-ping foliations striking E-W. On SVI, estimates of initial S -wave polarizationdirection are consistent with predictions from radiation patterns generatedby LFE focal mechanisms, providing corroboration for thrust mechanisms atthe plate boundary. Fast directions across mainland SVI are subparallel tothe dominant foliation direction in the LRC. Increases in depth normalizeddelay times from east to west, combined with small-scale azimuthal varia-tions in fast directions suggest a heterogeneous distribution of anisotropy.We test azimuthally anisotropic LRC models based upon analyses of geo-logical fabric and geometrically constrained by reflection studies, throughforward modeling using 3D spectral element method (SEM) simulations.The preferred model of a north/northeast shallowly dipping wedge of LRCmaterial with varying orientations of anisotropy terminating at mid crustaliiAbstractlevels is able to recreate mean and azimuthal variations in fast directionsalong with variations in delay times, thereby supporting the hypothesis ofthe LRC as a primary contributor to crustal anisotropy beneath SVI. Forselect stations where anisotropic LRC models do not recreate observations,fast directions are subparallel to local estimates of maximal compressive hor-izontal stress, suggesting fluid-filled cracks could be a source of anisotropy.We refute the idea that anisotropy along mainland SVI is primarily dueto stress related cracks as has been suggested by prior studies. Fast direc-tions at stations on northern Washington exhibit variations with azimuthand incidence angle suggesting complex anisotropy interpreted as due to acombination of cracks and preferred mineral orientation of metamorphosedslates of the Olympic core rocks. These slates may also underlay stationson SVI and represent another source of anisotropy.iiiPrefaceThe content of this thesis is original and does not include text from priorpublications. This thesis is based on the analysis of process data obtainedby Bostock et al. [10], Royer and Bostock [41] as detailed in chapter 2. Theanalyses in chapter 3 and 5 follow well established techniques in seismol-ogy but present original results. I conducted model testing in chapter 4with guidance from collaborator J. Tromp regarding proper model designimplementation. Chapter 4 uses open source software SPECFEM3D to per-form the modeling. Chapters 6 and 7 are based on my original analysesthat were guided through discussions with supervisors M. Bostock and N.I.Christensen.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Elastic anisotropy . . . . . . . . . . . . . . . . . . . . . . . . 32 LFE Templates . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Splitting Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 113.1 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . 113.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143.2.1 Initial polarization direction . . . . . . . . . . . . . . 163.2.2 Fast polarization direction . . . . . . . . . . . . . . . 17vTable of Contents3.2.3 Delay times . . . . . . . . . . . . . . . . . . . . . . . 183.3 Comparison with a prior forearc study . . . . . . . . . . . . . 194 Forward Modeling . . . . . . . . . . . . . . . . . . . . . . . . . 264.1 Initial model and LFE source . . . . . . . . . . . . . . . . . . 264.2 LRC anisotropy model . . . . . . . . . . . . . . . . . . . . . 294.3 Preferred model and results . . . . . . . . . . . . . . . . . . . 305 Northern Washington . . . . . . . . . . . . . . . . . . . . . . . 346 Sources of Anisotropy . . . . . . . . . . . . . . . . . . . . . . . 376.1 Extensive dilatancy anisotropy . . . . . . . . . . . . . . . . . 376.2 Mineral preferred orientation of the Leech River Complex . . 396.3 Mineral preferred orientation of the Olympic Peninsula . . . 427 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48viList of Tables3.1 Splitting parameters at stations employed in this study. Meanvalues and standard deviations for fast direction (φ) and splittime (δt). Azimuthal distribution of φ uses directional statis-tics to compute means and standard deviations [5]. . . . . . . 15viiList of Figures1.1 S -wave splitting schematic diagram. Two main sources ofcrustal anisotropy, cracks and preferred mineral orientation,are illustrated with fast directions. . . . . . . . . . . . . . . . 42.1 Map of SVI stations employed in this study. POLARIS-BCstations are represented by white triangles with remainingstations displayed as black triangles. Light grey contour illus-trates the LRC boundary, with the bounding faults labelledand divided by double zig-zag lines. Dashed lines represent20, 30 and 40 km depth contours for the subducting Juan deFuca plate [2]. . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2 Horizontal component sections of two SVI LFE templates dis-playing north (blue) and east (red) components. Clear split-ting can be seen at certain stations (e.g. TWKB, MGCB,LZB) whereas others show strong waveform distortions (e.g.TWBB, TWGB). . . . . . . . . . . . . . . . . . . . . . . . . . 9viiiList of Figures2.3 Map of local SVI earthquake epicenters and stations. LFEsare shown as red diamonds and crustal/instraslab earthquakes(Mw > 1.0, event depth < 50 km) are shown as blue dots.LRC boundary is shown in light grey. Depth profile of seis-micity along line AA’ is shown in (b). . . . . . . . . . . . . . 103.1 Stages of splitting analysis for LFE template 002. (a) Win-dowed and tapered SV -SH waveforms. (b) Elliptical par-ticle motion indicative of anisotropic wave propagation. (c)Grid search to minimize second eigenvalue of the covariancematrix. Star represents minima, thick black lines representconfidence levels. Vertical black lines mark φ0 and φ0 + pi/2.(d) Fast and slow waves (e) Linearized particle motion of cor-rected seismograms. (f) Reconstruction of original isotropicwaveform. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133.2 Rose histograms of original polarization direction φ0 plottedat station locations. Cumulative rose histogram for all sta-tions is shown at the lower left along with the local platemotion vector of the Juan de Fuca plate relative to the NorthAmerican plate. . . . . . . . . . . . . . . . . . . . . . . . . . . 22ixList of Figures3.3 Single event comparison of observed (black lines) and pre-dicted (dashed red lines) φ0 plotted at station locations forLFE template 101. The representative LFE focal mechanismis plotted at the LFE epicenter. The corresponding radia-tion pattern is shown in the inset. Polarization vectors forpredicted φ0 are displayed as red arrows in the inset. . . . . . 223.4 Single station comparisons for a range of POLARIS-BC sta-tions displaying observed (black lines) and predicted (redlines) φ0. φ0 measurements are plotted at corresponding LFEepicenters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.5 Rose histograms of fast polarization direction φ at stationlocations. Dashed cyan lines represent φmode for tremor split-ting measurements [9]. Red arrows represent estimates ofmaximum horizontal compressive stress (σHmax) [3]. . . . . . 233.6 Equal area plots displaying δt and φ for all stations. Borderrepresents 45◦ incidence angle and dashed line represents 35◦incidence angle. φ are plotted as blue bars with lengths pro-portional to delay times. Mean φ from this study (red lines)and Balfour et al. [4] (green lines) are included for compari-son. Mean delay times from this study (δt) and Balfour et al.[4] (δtb) are also included. Shaded regions indicate backaz-imuths where LFEs are thought to share similar raypaths tocrustal/instraslab events from Balfour et al. [4]. . . . . . . . . 24xList of Figures3.7 Histograms of depth normalized delay times (δt′) for E-Wmainland stations. Grey histograms represent δt′ for LFEtemplate measurements with mean values signified by a dashedblack line. Mean value is listed in black in the top right. His-tograms for synthetic δt′ computed with a subset of 50 LFEs,are shown as transparent bars with red outlines. Dashed redlines identify mean δt′ with numerical value listed in red. . . 254.1 Images of regional mesh. Top: Mesh doubling structure isdisplayed with stations as green dots on the surface. Bottom:3D tomographic Vp model used for modeling [40]. . . . . . . . 284.2 (a) Schematic diagram of LRC anisotropy model. Faultsare labelled with dip directions indicated by black arrows.Anisotropic domains are labelled, light grey lines indicate ori-entation of vertically dipping foliation. Dashed lines representprojections of LRC at depth. (b) Rose histograms of syntheticfast polarizations, φsyn. (c) Equal area plots displaying φsynat select stations, red line represents mean φsyn. (d) Equalarea plots showing azimuthal distribution of delay time resid-uals (r = δtobs-δtsyn). Red and blue circles represent fast(r < 0) and slow (r > 0) residuals respectively, clear circlesrepresent r=0. (e) Histograms of delay time residuals. . . . . 33xiList of Figures5.1 North Washington LFEs: Rose histograms of φ plotted atstation locations. Green diamonds represent northern Wash-ington LFE epicenters. Red arrows represent estimates ofσHmax from [3]. Equal area plots displaying φ are includedfor select stations. Red bars indicate φmode, φmode and meanδt are listed in boxes. . . . . . . . . . . . . . . . . . . . . . . . 366.1 Schematic diagrams illustrating geological terrane and inter-preted depth sections. (a) Major geological terranes of north-ern Cascadia with rose histograms of φ included. (b) Geo-logic interpretation of depth profile BB’ based upon Cloweset al. [20]. Illustrates typical LFE paths and lists associatedsources of anisotropy. (c) Geologic interpretation of depthprofile CC’ based upon Ramachandran et al. [40]. Illustratesunderthrusting of Olympic core rocks beneath Eocene basalts. 45xiiAcknowledgementsForemost I would like to thank my supervisor Michael Bostock for his gen-erosity, patience and unwavering support. His expertise and experience inthe field have been invaluable for the completion of this thesis along withmy growth as an aspiring researcher. I am also grateful to Nik Christensenfor his multifaceted insight into the interpretation of results along with hisinspiring enthusiasm. I would also like to thank Nik, Ron Clowes and FelixHerrmann for having served on my supervisory committee. I am thankfulto Jeroen Tromp and his group at Princeton University for accommodatingme and exposing me to the field of computational seismology, their insightinto modeling techniques was essential. I express my gratitude to KumarRamachandran for providing us with his tomographic model of VancouverIsland allowing us to improve our modeling efforts.xiiiChapter 1IntroductionThe discovery of low frequency earthquakes (LFEs) as a component of tec-tonic tremor presents a novel and unexplored means of studying local sub-duction zone structure. At the time that LFEs were originally discoveredin SW Japan as discrete events on seismograms, their relation to episodictremor and slip (ETS) had not yet been ascertained [6, 38]. Later workusing network correlation methods showed that tectonic tremor can be con-sidered as a superposition of many LFEs undergoing repeated ruptures ata range of different earthquake locations [44]. Correlation methods havebeen used to identify LFE events among tremor signals in both SW Japan[14, 43] and Cascadia [10, 14]. LFEs are small earthquakes (Mw ≈ 2.0) withcharacteristic frequencies of 1-10 Hz that result from shear slip at the plateboundary [31, 41, 44]. The predominantly horizontal motion of the ruptureleads to strong S -wave arrivals on horizontal component seismograms atsmall epicentral distances. The consistent mechanism and local nature ofLFEs suggest their use as a new seismic source to image regional subductionzone structure. In the Cascadia subduction zone where regular seismicity isrelatively scarce, LFEs present a potentially valuable imaging tool.The North American plate overrides the subducting Juan de Fuca plate1Chapter 1. Introductionin Cascadia and exhibits crustal anisotropy that has been documented inearlier S -wave splitting analyses of southern Vancouver Island (SVI) [4, 9,15, 25]. The origins of crustal anisotropy are often disputed but are generallyascribed to either stress-aligned fluid-saturated fractures [e.g. 23, 24] or thepreferred mineral orientation of rocks [e.g. 16, 17]; both have been cited asexplanations for anisotropy observed on SVI. Schematic diagrams illustrat-ing the two types of anisotropy are shown in figure 1.1. We seek to providefurther insight into the cause of anisotropy beneath SVI using LFEs. InCascadia, LFEs generate waves that propagate upwards through the forearccrust of the North American plate. Consequently, any anisotropy observedat nearby stations can be attributed directly to local crustal anisotropy asthe observed waveforms have not encountered any contribution from mantlestructure.Chapters 2-4 of this study focus solely on SVI for which the analysisis divided into two components. In the first part (sections 2-3), we out-line and conduct a splitting analysis of LFE data, present observations andcompare them with previous studies of crustal anisotropy in the region.We then expand on the work of Bostock and Christensen [9] by developinganisotropic models for SVI and computing synthetic seismograms using aspectral element method (section 4). We continue to develop the hypothe-sis that crustal anisotropy in the region is primarily influenced by mineralorientation of metamorphic rocks from the Leech River Complex (LRC), asopposed to stress-aligned fluid-saturated cracks. In section 5 we include aless extensive splitting analysis for LFE data from northern Washington, anextension that permits a more comprehensive interpretation of anisotropy21.1. Elastic anisotropyin the northern Cascadia forearc crust.1.1 Elastic anisotropyElastic anisotropy refers to the property of a medium whereby the velocityof a seismic wave depends on its direction of propagation and polarization.As a linearly polarized S -wave enters an anisotropic medium it is usuallysplit into two quasi S -waves, qS1, qS2 that are orthogonal, and have po-larization directions determined by the elastic properties of the medium.A commonly assumed form of anisotropy is transverse isotropy, which hasone axis of symmetry. If the propagation direction of an incoming S -wavelies along the axis of symmetry, the S -wave is not split nor is the vibrationdirection altered. For other directions of propagation the S -wave is splitinto two waves with different velocities (figure 1.1). The velocity differenceresults in a delay time δt between the qS1 and qS2 waves, the magnitudeof which depends on the strength of anisotropy. More specifically, the in-tegrated time delay is dependent on the difference between fast and slowwavespeeds along with the extent of the anisotropic material. In order toobserve and characterize shear wave splitting, the delay time δt and the fastpolarization direction φ, are typically measured. Splitting measurementscan be used to map anisotropy and subsequently can be related to past orpresent deformation processes.31.1.ElasticanisotropyCrack anisotropySlowFast Fast direction    to cracksFigure 1.1: S -wave splitting schematic diagram. Two main sources of crustal anisotropy, cracks and preferredmineral orientation, are illustrated with fast directions.4Chapter 2LFE TemplatesThe LFE templates [10, 41] used in this study were generated using networkcorrelation methods [13, 43] on ETS episodes occurring in 2003 to 2013 alongSVI over a set of 7 anchor stations. A group of initial LFE templates weresubject to a network cross correlation over an expanded range of stations(figure 2.1) in an iterative manner that allowed for improvement to thesignal-to-noise ratio level (SNR) of LFE template waveforms. For moredetails on the LFE template acquisition in Cascadia, the reader is referredto Bostock et al. [10] and Royer and Bostock [41].LFE template waveforms, like tremor, are bandlimited and show charac-teristic frequencies of 1-10 Hz on SVI. When an entire set of templates at anindividual station is plotted in a seismogram section, there is considerableuniformity across the section, indicative of a simple and consistent sourcemechanism. These observations were interpreted as an indication that LFEtemplate waveforms on SVI could be considered as empirical Green’s func-tions [10]. The dipolar pulse on far-field particle velocity seismograms isrepresentative of a point source characterized by a step function time de-pendence in displacement (figure 2.2).LFE epicenters on SVI are distributed in a band approximately con-5Chapter 2. LFE Templatesstrained by the 25 and 37 km plate boundary depth contours (figure 2.3)[2], a region where regular seismicity is scarce. The LFE locations showsome degree of segregation with local crustal and intraslab earthquakes (fig-ure 2.3) [10]. Similar to LFEs in Japan [43], LFEs in Cascadia have beenlocated near the top of a high Vp/Vs, low S velocity zone [10], which in Cas-cadia is inferred to be the upper oceanic crust [8, 30]. Waveform modelingof several LFE templates requires LFEs to be located within the upper 1km of the low velocity zone [39].As preprocessing steps for the splitting analysis, LFE template wave-forms are bandpass filtered between 1 and 8 Hz after which windows aremanually selected around S -wave arrivals and subjected to a cosine taper.In the event no clear arrival is apparent across any component, the windowis removed from further analysis. Given the station and event distribution,most LFEs represent small incidence angles due to steep raypaths; how-ever, some source receiver geometries are wide enough that effects of thefree surface may become important. To mitigate such effects we apply a 1-D wavefield decomposition that takes radial-transverse-vertical component(uR, uT , uZ) seismograms and transforms them into upgoing P-SV-SH wave-forms [32]. This transformation requires estimates of the horizontal slowness(p), near surface velocities (α, β) and density (ρ).PSVSH=pβ2/α 0 β2p2− 12αqα12−β2p2βqβ 0 pβ0 12 0URUTUZ(2.1)where qα =√α−2 − p2 and qβ =√β−2 − p2. We exclude seismograms6Chapter 2. LFE Templatesthat have incidence angles greater than a critical angle, θc = sin−1(β/α). Forsurface velocities in northern Cascadia this corresponds to θc ≈ 35◦. Beyondθc, surface reflection coefficients become complex, leading to phase rotationsthat preclude a standard splitting analysis. This criterion coincides with the“S -wave window” desirable for S -wave splitting analysis [7].7Chapter2.LFETemplates20’ 124oW 40’ 20’ 123oW18’24’48 oN30.00’36’42’48’SSIBGOWBSILBKELBMGCBTWKBTSJBTWBBTWGBSNBPGCLZBPFBSHVBKHVBVGZLCBCJRBC GLBCSOKBLeech River FaultSan Juan FaultSurvey Mountain Fault20km 30km40kmFigure 2.1: Map of SVI stations employed in this study. POLARIS-BC stations are represented by white triangleswith remaining stations displayed as black triangles. Light grey contour illustrates the LRC boundary, with thebounding faults labelled and divided by double zig-zag lines. Dashed lines represent 20, 30 and 40 km depthcontours for the subducting Juan de Fuca plate [2].8Chapter 2. LFE TemplatesTime [s] Template 05820 22 24 26 28 30 32SSIBSNB GOWBSILBPGC KELBMGCBTWKBLZB TSJBTWBBTWGBTime [s] Template 12520 22 24 26 28 30 32SSIBSNB GOWBSILBPGC KELBMGCBTWKBLZB TSJBTWBBTWGBFigure 2.2: Horizontal component sections of two SVI LFE templates dis-playing north (blue) and east (red) components. Clear splitting can be seenat certain stations (e.g. TWKB, MGCB, LZB) whereas others show strongwaveform distortions (e.g. TWBB, TWGB).9Chapter 2. LFE Templates125oW 30’ 124oW 30’ 123oW48oN12’24’36’48’49oN20km 30km40kmAA’0 20 40 60 80 100 120 140 1600204060A A’Distance (km)Depth(km)LFE 125LFE 058(a)(b)Figure 2.3: Map of local SVI earthquake epicenters and stations. LFEs areshown as red diamonds and crustal/instraslab earthquakes (Mw > 1.0, eventdepth < 50 km) are shown as blue dots. LRC boundary is shown in lightgrey. Depth profile of seismicity along line AA’ is shown in (b).10Chapter 3Splitting Analysis3.1 MethodologyWe perform a standard, two-parameter S -wave splitting analysis that searchesfor δt and φ that minimize the second eigenvalue λ2 of the covariance ma-trix for the shifted and rotated horizontal component seismograms [45]. Thisprocedure serves to maximize the linearity of the particle motion therebyidentifying the original, unsplit S -waveform, assuming that the medium isadequately modeled by constant anisotropy. The eigenvector correspondingto the largest eigenvalue of the corrected covariance matrix is used to obtainan estimate for the initial polarization direction φ0. Figure 3.1 shows stagesof the splitting analysis at station TWKB for LFE template 002.The standard splitting analysis is susceptible to null measurements thatoccur when there is an inherent lack of anisotropy along the raypath, whena wave propagates along the axis of symmetry in a transversely isotropicmedium or when the initial polarization direction is coincident with the fastor slow direction. For a parameter combination (φm, δtm) that minimizesλ2(φ, δt) of the splitting analysis (figure 3.1c), there are three scenariosunder which the measurement is deemed null:1) If φ0, φ0+ pi2 or δt = 0 lie within the 95% confidence level (τ) of λ2(φm,113.1. Methodologyδtm)2) δtm > δtmax, where δtmax is an upper limit imposed based uponrealistic expectations for measurements in the region (set at 0.3 s) [e.g.4, 9, 15, 25].3) Both φ0 or φ0 + pi2 and δt = 0 lie within 2τ of λ2(φm, δtm).The final criterion was chosen due to empirical evidence that showedmeasurements falling into this category consistently failed to follow expectedsplitting behaviour. The 95% confidence levels are computed assuming anF-distribution. Hereafter, usage of φ and δt refers exclusively to solutionsof the splitting analysis, φm and δtm respectively.123.1. Methodology8 10 12 14−0.0100.010.02Time (s)Input seismograms (SV/SH) − TWKB  SVSH−0.04 −0.02 0 0.02 0.04−0.0100.010.02Input particle motionMinimum eigenvalue splitting parameter estimateTime shift (s)Fast axis (deg)LFE Template 00250 100 1500.10.20.38 10 12 14−0.0200.02Time (s)Fast and slow waves  FastSlow−0.04 −0.02 0 0.02 0.04−0.02−0.0100.010.02Reconstructed particle motion8 10 12 14−0.0200.02Reconstructed seismograms in polarization coordinatesTime (s)Figure 3.1: Stages of splitting analysis for LFE template 002. (a) Windowedand tapered SV -SH waveforms. (b) Elliptical particle motion indicative ofanisotropic wave propagation. (c) Grid search to minimize second eigenvalueof the covariance matrix. Star represents minima, thick black lines representconfidence levels. Vertical black lines mark φ0 and φ0 + pi/2. (d) Fast andslow waves (e) Linearized particle motion of corrected seismograms. (f)Reconstruction of original isotropic waveform.133.2. Results3.2 ResultsA suite of 90 LFE templates representing a relatively uniform sampling ofthe tremor prone region are used to perform a splitting analysis over a groupof 20 stations located on SVI (figure 2.3). The complete analysis leads to 517valid splitting measurements out of a potential 1800 source-receiver pairs.Table 1 shows a summary of the resulting measurements at all stations. Dueto the requirement of steeply incident arrivals, the use of certain stationsbecomes limited by the event distribution. Stations SNB, GOWB, SSIBand PFB are located outside of the region most densely populated withLFEs, which naturally leads to shallower incidence angles. Other stationswith limited measurements, e.g. TWGB, are characterized by poor SNR orstrong distortions in the original S -waveforms that make isolating arrivalsdifficult (e.g. figure 2.2). Station TWGB lies in close proximity to thesurface trace of the San Juan Fault and atop potentially complex lithology,thus it is conceivable that this waveform distortion is a consequence of stronglocal anisotropy/heterogeneity. Stations TWBB and TSJB are located nearTWGB and often display similar but less severe behaviour.143.2. ResultsTable 3.1: Splitting parameters at stations employed in this study. Meanvalues and standard deviations for fast direction (φ) and split time (δt).Azimuthal distribution of φ uses directional statistics to compute meansand standard deviations [5].Station Latitude Longitude δt(s) ±δt(s) φ(◦) ±φ(◦) N. Meas.SSIB 48.7558 -123.3875 0.19 0.09 109 27 14SNB 48.7751 -123.1723 0.21 0.08 95 37 3GOWB 48.7369 -123.1848 0.16 0.06 121 34 5SILB 48.6020 -123.2815 0.11 0.08 87 32 23PGC 48.6500 -123.4500 0.10 0.04 111 32 30KELB 48.6611 -123.5701 0.10 0.04 103 26 37MGCB 48.6317 -123.6808 0.11 0.06 97 30 37TWKB 48.6449 -123.7332 0.13 0.06 99 15 44LZB 48.6117 -123.8236 0.13 0.05 99 23 46TSJB 48.6013 -123.9885 0.21 0.06 79 22 38TWBB 48.5846 -124.0920 0.20 0.07 91 26 26TWGB 48.6076 -124.2559 0.16 0.07 93 35 14PFB 48.5750 -124.4444 - - - - 0LCBC 48.4834 -124.2619 0.14 0.06 60 47 10JRBC 48.3957 -123.9601 0.10 0.06 80 44 18SOKB 48.3947 -123.6731 0.12 0.06 121 35 45GLBC 48.3960 -123.6363 0.11 0.06 115 32 37SHVB 48.4723 -123.3737 0.18 0.08 122 33 31KHVB 48.5688 -123.4663 0.12 0.04 112 30 37VGZ 48.4139 -123.3244 0.15 0.08 116 31 22153.2. Results3.2.1 Initial polarization directionWe first analyze the distribution of initial polarization directions φ0 andcompare with source mechanism predictions. Bostock and Christensen [9]acquired φ0 estimates using raw tremor waveforms and found φ0 to be scat-tered but to average close to the NE directed plate motion direction, con-sistent with the interpretation of LFEs as resulting from shear failure atthe plate boundary. Figure 3.2 shows a composite plot of φ0 represented inrose histograms at individual station locations. Each histogram contains 50bins of width 7.2◦. In comparison to φ0 from Bostock and Christensen [9],φ0 from LFE template splitting measurements exhibit a similar but moretightly defined tendency for original polarization direction to coincide withplate motion direction [36].We expand on this analysis by comparing measured φ0 with predictedpolarization directions for an isotropic 1D velocity model. We take an aver-age representative moment tensor solution for LFEs on SVI [41] and computethe predicted far-field radiation pattern (inset figure 3.3). We find strongagreement between predicted and observed initial polarization directions.Figure 3.3 displays a single event example where both observed and pre-dicted φ0 at stations TWBB and TSJB display considerable deviation fromthe plate motion direction, but are directly explained by the expected radia-tion pattern. Figure 3.4 displays single station comparisons of φ0 plotted atthe LFE epicenters. Station TWKB demonstrates an ideal case with strongconsistency in predicted and measured φ0 over the majority of LFE loca-tions. Agreement at TWBB is less consistent; whereas the main cluster of163.2. Resultsevents originating from under the LRC still exhibit reasonable agreement,there is considerably more variability for distant sources. Recovery of φ0is less reliable for noisy template waveforms or where strong waveform dis-tortions are observed. The overall agreement aids in verifying the methodsused in this study.3.2.2 Fast polarization directionWhereas original polarization directions provide corroboration for a commonthrust mechanism, fast polarization directions yield information related toanisotropy. Figure 3.5 displays rose histograms of fast polarization directionsat station locations. For stations in close proximity to the LRC we notesome similarity between the dominant fast direction and the attitude of thenearest bounding fault. Western stations on the POLARIS-BC line crossingmainland SVI show a general E-W fast direction. Stations to the southeast(KHVB, SHVB, VGZ, GLBC and SOKB) exhibit fast directions that aresubparallel to the Survey Mountain Fault, which defines the eastern extentof the surface expression of the LRC.Two recent crustal anisotropy studies in SVI also compiled measurementsof fast polarization direction using alternative datasets. Balfour et al. [4]employ local crustal and intraslab earthquakes (distribution is similar butnot identical to that for crustal/instraslab earthquakes displayed in in figure2.3), whereas the study by Bostock and Christensen [9] measures splittingon raw tremor waveforms from ETS episodes in Cascadia. Since LFE tem-plates are acquired through network correlation of tremor data, we expectsimilarity between LFE and tremor splitting measurements, with improved173.2. Resultsconsistency due to precise source locations. An earlier crustal anisotropystudy [15] measured average φ = 113◦ at PGC for shallow crustal events,exhibiting good correspondence with our result (Table 1).As expected, fast directions in this study reveal good agreement withthose derived from raw tremor at common stations [9], with a general E-W trend for stations in close proximity to the San Juan Fault (figure 3.5).Moreover, the LFE measurements extend spatial coverage to stations alongthe south-east coast of SVI. A comparison between our fast polarizationdirections and those of Balfour et al. [4] are displayed in figure 3.6; agreementis mixed and will be addressed in section 3.3.Figure 3.6 presents splitting measurements as a function of azimuth andincidence angle. We note some degree of azimuthal variation in φ at selectstations. Where present, azimuthal variations manifest as spatially coherentclusters of similar φ. Stations PGC, LZB and TWKB exhibit approximatelyE-W fast directions for south-western azimuths yet trend closer to NW-SEfor south-eastern azimuths. At SOKB, φ is rotated counter-clockwise forevents originating from eastern azimuths when compared to events to fromthe north-west. Similarity in azimuthal variations of φ between neighbour-ing stations suggest that these small scale variations are genuine indica-tions of heterogeneous anisotropy. Two such examples include station pairsTWKB/LZB and KELB/PGC (figure 3.6).3.2.3 Delay timesDelay times across all stations and events vary between 0.05 and 0.3 s,whereas mean values over all stations vary between 0.1 and 0.21 s. We do183.3. Comparison with a prior forearc studynot observe any systematic variation in delay time with backazimuth (figure3.6), with the possible exception of station TWKB. Bostock and Christensen[9] reported larger delay times corresponding to northerly azimuths withan average δt = 0.22 s, as opposed to an average δt = 0.07 s for eventsfrom south-western azimuths; we observe a similar but less definitive trend.Whereas we lack measurements from northerly azimuths, the northern mostevents display larger split times (> 0.2 s) and decrease for south-westernazimuths, with delay times ranging between 0-0.2 s. Figure 3.7 displayshistograms of depth normalized delay times (δt′) for an E-W line of stations.We observe an increase in average δt′ from east to west, with maximumvalues occurring at TSJB. The large variance within the histograms hint ata strongly heterogeneous strength of anisotropy. Measured delay times aregenerally larger than those of Balfour et al. [4] and will be addressed in laterdiscussion.Schistose rocks are known to cause splitting due to preferred mineralorientation [12, 17, 28], thus we chose to investigate the potential influence oflithology on splitting measurements by combining laboratory measurementsof LRC rock samples with structural information available on the geometryof the LRC (section 4).3.3 Comparison with a prior forearc studyWe herein refer to Balfour et al. [4] as BCDM for brevity. We credit thesomewhat complementary event distributions as the primary contributor tothe discrepancies in splitting measurements between the two studies. LFE193.3. Comparison with a prior forearc studyepicenters occur predominantly on mainland SVI and arrive with subverti-cal incidence angles, whereas BCDM epicenters tend to lie along the coastsand arrive with shallower incidence angles. Figure 2.3 is a representativedisplay of the respective distributions. Shared raypaths between LFE andcrustal/instraslab events exist, but are limited. The variable sampling ofthe crust could lead to inherent sampling bias in the average splitting mea-surements. In the presence of heterogeneous anisotropy, these differencesbecome more difficult to verify and interpret. Sampling bias may explainwhy average φ from BCDM are consistently rotated clockwise relative to av-erage φ determined in this study (figure 3.6). For example, stations LZB andTWKB show some azimuthal variation in φ but the abundance in E-W di-rected measurements means that certain measurements are not representedin the mean, an issue that is due to the sample distribution.Where azimuthal variations in φ are observed, agreement with BCDMfast directions are improved for azimuths with similar LFE/BCDM earth-quake epicenters (figure 3.6). Examples include south-western azimuths forTWKB and LZB, along with eastern azimuths at SOKB. Without a morecomprehensive event-by-event comparison it is unclear whether the improvedagreement is genuine and reflects shared sources anisotropy. Mean fast di-rections at stations SSIB, SNB, GOWB and SOKB show good agreementwith BCDM, although the limited LFE template measurements at SNB andGOWB suggest that these particular agreements should be taken with cau-tion.Average split times at TWGB, TSJB and LZB are larger than thosefrom BCDM (listed in figure 3.6), this is potentially explained by propa-203.3. Comparison with a prior forearc studygation through a medium with HTI symmetry. Maximal splitting is pro-duced when the propagation direction is perpendicular to the symmetryaxis [e.g. 9, 12, 17]. For N-S propagation, the subvertical raypaths of LFEswould produce near maximal splitting, whereas the shallower incidence ofinstraslab/crustal earthquakes could produce anisotropy of a smaller mag-nitude [9]. This difference can be significant and could explain the smallersplit times recorded by BCDM despite the generally longer raypaths forcrustal/instraslab events at these stations.Differences in delay times for stations along the east coast are difficultto reconcile due to the variable behaviour between stations. We focus on asubset of LFE events east of a line of constant longitude -123.6731◦ (longi-tude of station SOKB), with comparable ray geometries to crustal/intraslabevents used by BCDM. For this subset, PGC and KHVB have average delaytimes of 0.1 and 0.13 s respectively, both smaller than those observed byBCDM (PGC: 0.12 s, KHVB: 0.18 s). Conversely, stations SOKB, SHVBand VGZ have mean delay times greater than BCDM, retaining mean valueslisted in figure 3.6. Without more information on the contributing event dis-tribution, it is difficult to ascertain what may cause the discrepancies. Thevariable length of raypaths for intraslab and shallower crustal earthquakesmay cause variations, however, BCDM reported no systematic increase indelay time with depth. Measurements for the remaining stations are tooscattered and few to draw meaningful comparisons. In general the splittingmeasurements for LFE templates and BCDM appear to reflect differentsampling of a heterogeneous, anisotropic crust.213.3. Comparison with a prior forearc study 125oW  30’  124oW  30’  123oW  20’   48oN  30.00’  40’  50’ N56°EFigure 3.2: Rose histograms of original polarization direction φ0 plotted atstation locations. Cumulative rose histogram for all stations is shown at thelower left along with the local plate motion vector of the Juan de Fuca platerelative to the North American plate.30’ 15’ 124oW 45’ 30’ 15’20’25’48oN30.00’35’40’Figure 3.3: Single event comparison of observed (black lines) and predicted(dashed red lines) φ0 plotted at station locations for LFE template 101. Therepresentative LFE focal mechanism is plotted at the LFE epicenter. Thecorresponding radiation pattern is shown in the inset. Polarization vectorsfor predicted φ0 are displayed as red arrows in the inset.223.3. Comparison with a prior forearc study 40’  20’  124oW  40’  20’  123oW  12’  18’  24’   48oN  30.00’  36’  42’  48’ TWBB 40’  20’  124oW  40’  20’  123oW  12’  18’  24’   48oN  30.00’  36’  42’  48’ TSJB 40’  20’  124oW  40’  20’  123oW  12’  18’  24’   48oN  30.00’  36’  42’  48’ TWKB 40’  20’  124oW  40’  20’  123oW  12’  18’  24’   48oN  30.00’  36’  42’  48’ KELBFigure 3.4: Single station comparisons for a range of POLARIS-BC stationsdisplaying observed (black lines) and predicted (red lines) φ0. φ0 measure-ments are plotted at corresponding LFE epicenters.20’ 124oW 40’ 20’ 123oW18’24’48oN30.00’36’42’48’Figure 3.5: Rose histograms of fast polarization direction φ at station loca-tions. Dashed cyan lines represent φmode for tremor splitting measurements[9]. Red arrows represent estimates of maximum horizontal compressivestress (σHmax) [3].233.3. Comparison with a prior forearc studyδt: 0.19 sδtb: 0.12 sSSIBδt: 0.21 sδtb: 0.08 sSNBδt: 0.16 sδtb: 0.12 sGOWBδt: 0.11 sδtb: 0.11 sSILBδt: 0.09 sδtb: 0.12 sPGCδt: 0.10 sδtb: −KELBδt: 0.11 sδtb: −MGCBδt: 0.13 sδtb: 0.14 sTWKBδt: 0.13 sδtb: 0.09 sLZBδt: 0.21 sδtb: 0.13 sTSJBδt: 0.20 sδtb: −TWBBδt: 0.16 sδtb: 0.13 sTWGBδt: −δtb: 0.19 sPFBδt: 0.14 sδtb: −LCBCδt: 0.10 sδtb: 0.09 sJRBCδt: 0.12 sδtb: 0.08 sSOKBδt: 0.11 sδtb: −GLBCδt: 0.18 sδtb: 0.09 sSHVBδt: 0.12 sδtb: 0.18 sKHVBδt: 0.15 sδtb: 0.11 sVGZδt < 0.1 s0.1 ≥ δt ≤ 0.2 sδt > 0.2 sFigure 3.6: Equal area plots displaying δt and φ for all stations. Borderrepresents 45◦ incidence angle and dashed line represents 35◦ incidence angle.φ are plotted as blue bars with lengths proportional to delay times. Meanφ from this study (red lines) and Balfour et al. [4] (green lines) are includedfor comparison. Mean delay times from this study (δt) and Balfour et al.[4] (δtb) are also included. Shaded regions indicate backazimuths whereLFEs are thought to share similar raypaths to crustal/instraslab events fromBalfour et al. [4]. 243.3. Comparison with a prior forearc study0 1 2 3 4 5 6 7 8 91005101520PGCδt′ (ms/km)Count2.472.120 1 2 3 4 5 6 7 8 91005101520KELBδt′ (ms/km)Count2.632.610 1 2 3 4 5 6 7 8 91005101520MGCBδt′ (ms/km)Count3.013.980 1 2 3 4 5 6 7 8 91005101520TWKBδt′ (ms/km)Count3.453.550 1 2 3 4 5 6 7 8 91005101520LZBδt′ (ms/km)Count3.545.040 1 2 3 4 5 6 7 8 91005101520TSJBδt′ (ms/km)Count6.105.930 1 2 3 4 5 6 7 8 91005101520TWBBδt′ (ms/km)Count5.605.470 1 2 3 4 5 6 7 8 91005101520TWGBδt′ (ms/km)Count4.585.81Figure 3.7: Histograms of depth normalized delay times (δt′) for E-W main-land stations. Grey histograms represent δt′ for LFE template measurementswith mean values signified by a dashed black line. Mean value is listed inblack in the top right. Histograms for synthetic δt′ computed with a subsetof 50 LFEs, are shown as transparent bars with red outlines. Dashed redlines identify mean δt′ with numerical value listed in red.25Chapter 4Forward ModelingWe seek to test the validity of the assertion that crustal anisotropy in SVI isprimarily a result of mineral orientation in metamorphic rocks. We employthe spectral element method (SEM) to compute synthetic seismograms in3D anisotropic models [33, 34]. SEM allows for accurate solutions to the fullelastic wave equation with general anisotropy represented by 21 independentelastic constants. We develop a conceptually simple and geologically justifi-able model that captures the systematic trends in φ, with less attention paidto precise matching of split times. Split times are dictated by both the ex-tent and strength of anisotropic material, thus making detailed assumptionsabout either is problematic given the lack of information on their respectivedistributions.4.1 Initial model and LFE sourceWe develop a regional mesh that encompasses the SVI region lying between48.2◦-48.8◦ latitude and −123.0◦-−124.75◦ longitude and translates to a re-gion of approximately 130 km x 65 km in the E-W and N-S directions,respectively. The mesh extends to a depth of 50 km, enclosing the LFEdistribution while maintaining a sufficient distance from the domain bound-264.1. Initial model and LFE sourcearies to mitigate the influence of boundary reflections. Absorbing boundaryconditions are applied and a realistic topography and bathymetry model isused. We use a 3D P -wave velocity model [40] and compute S -wavespeedsusing a constant Vp/Vs ratio of 1.76, an average value for continental crust[18]. Density is computed using a wavespeed-based empirical relation forcrustal rocks [11]. We employ a purely isotropic background model uponwhich we superpose various anisotropic perturbation models. Two meshdoubling layers are implemented to maintain a similar number of grid pointsper wavelength throughout the mesh. The minimum period resolvable bythe mesh is ∼0.6 s. Regional mesh is displayed in figure 4.1. Although LFEtemplate waveforms display characteristic frequencies of 1-10 Hz, modelingwas restricted to frequencies between 0.5-1.67 Hz, a decision that was dic-tated by considerations of numerical cost. To allow for comparisons betweendata and synthetics, we bandpass filter the data at frequencies resolvable bythe mesh. Even within this bandpass LFE template waveforms retain aclearly defined arrival, thereby justifying our choice for a lower frequencyband. It should be noted that we chose not to include the explicit signatureof the (3-4 km) low velocity, high Vp/Vs zone that sits at the top of thesubducting plate [30]. Any reflections/conversions therefrom should arriveas S -waves within the S -wave window with geometries similar to direct S.Hence they should not alter estimates of δt or φ. Due to the considerablesimilarity of LFE focal mechanisms over the entire suite of LFE events [41]we use a single representative moment tensor for all LFEs when conductingnumerical simulations (figure 3.3).274.1. Initial model and LFE sourceFigure 4.1: Images of regional mesh. Top: Mesh doubling structure is dis-played with stations as green dots on the surface. Bottom: 3D tomographicVp model used for modeling [40].284.2. LRC anisotropy model4.2 LRC anisotropy modelWe expand upon the premise discussed in Bostock and Christensen [9] thatcrustal anisotropy in SVI is primarily a result of mineral orientation in theLRC. P -wave and S -wave velocities for a range of Leech River schist andphyllite samples demonstrate transverse isotropy with a symmetry axis per-pendicular to their foliation and fast directions parallel to the foliation [9].We first review prior constraints on geometry of the LRC. The surfaceexpression of the LRC is bounded by three faults; the Leech River fault(LRF), San Juan fault (SJF) and the Survey Mountain fault (SMF) (figure2.1). Seismic reflection surveys reveal the LRF and SMF to be thrust faults,dipping to the northeast at 35-45◦ to a depth of ∼10km [20, 29]. Subsequentreprocessing of select data [29] reveals an undulating LRF that begins as asteeply dipping (60◦) fault near the surface and becomes shallowly dippingpast 3 km depth. A lack of a reflection signature for the SJF was interpretedas indication of a steep northward dip of 60-70◦. Reflection imaging off thewest coast of SVI suggests that the LRF and SJF are both steeply dippingand merge at ∼13 km depth [21].We implement azimuthal anisotropy by assuming that anisotropy canbe represented as a transversely isotropic medium with the symmetry axisoriented in the horizontal plane (HTI). The appropriate elastic tensor iscomputed using velocity measurements for an LRC schist sample (L-4 schistsample from Bostock and Christensen [9]). The L-4 sample exhibits trans-verse isotropy with S -wave anisotropy of 11%, a relatively conservative valuein comparison to other schistose rocks from the LRC that can exhibit S -wave294.3. Preferred model and resultsanisotropy up to 30%. Orienting the symmetry axis within the horizontalplane equates to modeling a vertically dipping foliation as is prevalent insurface exposures of the LRC [27]. The two main adjustable parametersare the spatial extent of anisotropy and the orientation of symmetry axis inthe horizontal plane. We test a range of models and present the preferredmodel.4.3 Preferred model and resultsA basic anisotropic model of a homogeneous block of HTI material witha symmetry axis oriented due north (E-W oriented foliation), is improvedby iteratively increasing model complexity to reduce discrepancies betweenobserved and synthetic φ. At the first iteration we alter the LRC spatialgeometry following the constraints established in section 4.2. In later itera-tions we include differing anisotropic orientations.In 4 iterations we reach a viable anisotropic model (figure 4.2 (a)). TheLRF and SMF both dip shallowly at 30◦ to N35◦E respectively; the preferreddip of these faults lie slightly below the lower estimates from reflection sur-veys [20]. The SJF dips steeply (70◦) to the north and becomes shallowerpast 6 km to accommodate the interpretation that the SJF represents alistric thrust fault [20]. Projections of the dipping LRC bounding faultsextend to 12 km depth and terminate forming a wedge of north/north-easttrending material. Anisotropy in the model is divided into two distinct re-gions. In region 1, the symmetry axis is oriented due north and simulatesa region of material with E-W trending foliation. Region 2 has foliation304.3. Preferred model and resultsthat is oriented subparallel to the strike of the SMF, with a symmetry axisoriented at ∼N35◦E. The inferred rock type in region 1 and 2 varies only inthe orientation of the foliation.We select a subset of 50 LFEs with ray geometries that encounter theanisotropic wedge model of the LRC to perform forward modeling. Thesynthetic data vary more systematically and the signature of the two dis-tinct anisotropic regions is clearly observed (figure 4.2). We capture thegeneral E-W fast direction trend for western POLARIS-BC stations, alongwith the transition to NW-SE fast directions for more easterly stations (e.g.TWKB, KHVB )(figure 4.2 (b)). While the aggregate behaviour is reason-ably recovered, comparing the results for individual events/stations identifiesdeficiencies in the model.Figure 3.7 includes the depth normalized delay times for the syntheticdata in comparison to LFE template measurements, displaying fair overallagreement. The correspondence between trends of increasing average depthnormalized delay times from east to west is notable and could further supportLRC based anisotropy. Instances where split times are improperly matchedcan be observed through residuals, r = δtobs − δtsyn (figure 4.2(d-e)). Geo-graphical clusters of slow (r > 0) or fast (r < 0) residuals are indicative ofregions where anisotropy is too weak/strong along the corresponding ray-paths. Fast residuals are prevalent at station LZB, which is reflected in thegreater mean δt′ for synthetics (figure 3.7). TSJB displays both fast andslow residuals, overall agreement is strongest at TWKB. The deficiencies inthe current model can be attributed to an imperfect model geometry thatfails to capture complex lithology, particularly near the fault junction of the314.3. Preferred model and resultsSJF and SMF. Fairchild and Cowan [27] documented an increase in meta-morphic grade from north to south from greenschist to amphibolite facies.Variable metamorphic grade (and consequently strength of anisotropy) isnot reproduced in the current model, thereby presenting another potentialsource of error. With the current methods it is not possible to distinguishbetween the two sources of error.Stations SOKB, GLBC and VGZ all show significant splitting with strongpreferred directions for LFE template measurements, yet the current anisotropicmodel fails to reproduce splitting for synthetic seismograms (a single mea-surement at GLBC is the exception). The absence of a splitting signatureis anticipated, as the corresponding LFE raypaths for these stations do notencounter the synthetic LRC anisotropy. Further adjustments, that hon-our prior geological constraints, could likely be made to improve agreementat stations VGZ and SHVB. This is not the case for other stations, wheresignificant, implausible alterations would have to be made to account foranisotropy under an LRC model. The limited influence of the LRC suggestsan alternate source of anisotropy exists along raypaths observed at SOKB,GLBC along with the trio of SSIB, SNB and GOWB.324.3.Preferredmodelandresults−0.2 −0.1 0 0.1 0.20246810−0.2 −0.1 0 0.1 0.20246810−0.2 −0.1 0 0.1 0.20246810−0.2 −0.1 0 0.1 0.2024681020’ 124 oW 40’ 20’ 123 oW25’30’48 oN35.00’40’45’LRFSMFSJF9 km11 km7 kmkm53 km20’ 124 oW 40’ 20’ 123 oW20’25’48 oN30.00’35’40’DOMAIN1DOMAIN2abc d eTSJBLZBTWKBSHVBr (s)r (s)r (s)r (s)CountCountCountCountFigure 4.2: (a) Schematic diagram of LRC anisotropy model. Faults are labelled with dip directions indicated byblack arrows. Anisotropic domains are labelled, light grey lines indicate orientation of vertically dipping foliation.Dashed lines represent projections of LRC at depth. (b) Rose histograms of synthetic fast polarizations, φsyn. (c)Equal area plots displaying φsyn at select stations, red line represents mean φsyn. (d) Equal area plots showingazimuthal distribution of delay time residuals (r = δtobs-δtsyn). Red and blue circles represent fast (r < 0) andslow (r > 0) residuals respectively, clear circles represent r=0. (e) Histograms of delay time residuals.33Chapter 5Northern WashingtonAlthough our focus in this thesis is to ascertain the nature of anisotropybelow southern Vancouver Island, it will prove insightful in this respect toconsider splitting measurements of LFE templates determined for stationsimmediately to the south below northern Washington state. LFE data fornorthern Washington were obtained using the same correlation methodsused for SVI in section 2. For further details the reader is referred to [41].We perform a splitting analysis, as outlined in section 3, for 100 LFE tem-plates at 14 stations to produce 236 valid splitting measurements out of1400 potential source-receiver pairs. Figure 5.1 displays rose histograms ofφ at station locations along with equal area plots of φ for select stations.Splitting measurements are significantly less ordered and more complex thanthose on SVI. Mean delay times range from 0.1-0.15 s and are comparablein magnitude to delay times observed on SVI. Fast directions are scatteredand exhibit significant variability in behaviour between neigbouring stations.Observations of φ for single stations exhibit variations with both azimuthand incidence angle (figure 5.1), stations W020 and GNW are two such ex-amples. Station W020 displays near N-S φ at small incidence angles buttrends closer to NE-SW for larger incidence angles. Azimuthal variations34Chapter 5. Northern Washingtonare somewhat consistent between neighbouring stations suggesting the vari-ability is a genuine result of complex anisotropy. Complex anisotropy couldbe a consequence of highly deformed metamorphosed core rocks from theOlympic peninsula.35Chapter5.NorthernWashington 30’  124 oW  30’  123 oW  30’  30’  45’   48 oN  15’ SQMδt: 0.13 sφ: 31 ºBS11δt: 0.15 sφ: 59 ºW020δt: 0.13 sφ: 81 ºW030δt: 0.12 sφ: 63 ºGNWδt: 0.13 sφ: 59 ºFigure 5.1: North Washington LFEs: Rose histograms of φ plotted at station locations. Green diamonds representnorthern Washington LFE epicenters. Red arrows represent estimates of σHmax from [3]. Equal area plotsdisplaying φ are included for select stations. Red bars indicate φmode, φmode and mean δt are listed in boxes.36Chapter 6Sources of Anisotropy6.1 Extensive dilatancy anisotropyExtensive dilatancy anisotropy is a concept where vertically-aligned fluid-filled cracks are oriented parallel or subparallel to the direction of maximumhorizontal compressive stress (σHmax), thus generating azimuthal anisotropyin crustal rocks. Prior studies [4, 15, 25] have interpreted anisotropy ob-served on SVI as due to extensive dilatancy anisotropy [23]. A stress in-version for northern Cascadia [3] provides stress estimates at two locationsin close proximity to eastern stations on SVI (figure 3.5). Stations SHVB,VGZ, SOKB along with SSIB, SNB and GOWB to the north-east, have fastdirections similar to the local σHmax. The current anisotropic LRC modeleither lacks significant anisotropy along LFE raypaths to these stations ordoes not adequately reproduce observations of φ (figure 4.2). The similarityto σHmax combined with inability for a wedge model of the LRC to accountfor anisotropy at these stations suggests that crack induced anisotropy couldbe the primary source of anisotropy observed at these stations.σHmax directions are generally considered to be margin-normal near thetrench due to compression at the locked portion of the plate. To the eastthe subducting plate becomes weakly coupled and σHmax becomes margin-376.1. Extensive dilatancy anisotropyparallel, in Cascadia this is thought to be due to the northward push ofthe Oregon block [36, 48, 49]. Under this stress model, we do not inter-pret anisotropy observed at POLARIS-BC stations on SVI to be related tocrack anisotropy as dominant fast directions are almost perpendicular to themargin-parallel direction.With increasing depth cracks are closed due to increasing lithostaticpressure; for anisotropy to persist at depth pore fluid pressures must be nearlithostatic so that cracks remain open. Vp/Vs ratios of the continental crustoverlying the subducting Juan de Fuca plate have been obtained throughreceiver functions of teleseismic body waves [2]. At SVI stations employedin this study, Vp/Vs varies between 1.59 and 1.83 with a mean of 1.73,slightly below the global average of 1.76 for continental crust [18]. The belowaverage Vp/Vs ratios do not suggest elevated fluid pressures that would berequired to prop open cracks. Electrical conductivity studies on SVI [46]have observed elevated conductivities below SVI at depths coincident witha seismically reflective layer [20] corresponding to the low S - velocity zone[1, 40]. The high conductivity region below SVI is interpreted as a region oftrapped fluids within or above the subducting Juan de Fuca plate. However,electrical conductivities in the overlying crust of the North American plateare not elevated and do not suggest an abundance of fluid that would berequired to elevate fluid pressures. A lack of fluids would be consistentwith the interpretation of a sealed or low permeability plate boundary [1]that coincides with the ETS zone. Without elevated fluid pressures crackanisotropy would have a limited depth extent that is dependent on rock typeand porosity [18].386.2. Mineral preferred orientation of the Leech River ComplexσHmax directions on northern Washington [3] are displayed in figure 5.1and are generally N-S, trending near margin parallel. While some of theobserved fast directions on northern Washington coincide with the localσHmax directions (e.g. select azimuths at stations W020, W030 and GNW),the majority do not. The complex splitting patterns observed on northernWashington could be a result of multiple sources of anisotropy, that includecrack induced anisotropy.6.2 Mineral preferred orientation of the LeechRiver ComplexWe assert that the anisotropy we observe on mainland SVI is primarily aresult of the preferred mineral orientation of highly anisotropic rocks fromthe LRC. The geology of the LRC suggests a HTI model that adequately ex-plains the observed splitting measurements. The LRC is mostly comprisedof metamorphosed phyllites and schists that exhibit pervasive foliations withstrong phyllosilcate lattice preferred orientations [22, 27]. Extensive struc-tural analyses determined that two major deformational events have oc-curred in the tectonic history of the LRC, with the latter generating slatycleavage and schistosity [27, 42]. Orientations of foliated planar folds pro-duced by the second deformation are consistently steeply dipping and strikeapproximately east-west. The transversely isotropic LRC schists and phyl-lites [9] and observations of structural geology imply a source of azimuthalanisotropy generated by a HTI medium with a slow axis normal to the planeof foliation. The implementation of such a model in the forward modeling396.2. Mineral preferred orientation of the Leech River Complexconfirms the validity of an anisotropy model based upon preferred mineralorientation of the LRC.E-W trending φ observed at mainland POLARIS-BC stations are con-sistent with the east-west striking foliation in the main body of the LRC.Geologic mapping suggests that the foliation trend rotates to the east, be-coming subparallel to the strike of the SMF yet remains steeply/verticallydipping [35, 37]. By incorporating this change in dominant foliation direc-tion in our forward model, we can recreate the NW-SE φ observed for someeastern stations using a north/north-east dipping wedge model that extendsto 12 km. While the assumption that the foliation is vertically dippingthroughout the LRC may not be valid, highly anisotropic LRC schists andphyllites (up to 30 % S -wave anisotropy) can generate up to 0.3 s splittingwith a uniform thickness of as little as 2-3 km [9]. The trade-off betweenextent and strength of anisotropy allows for a shallower extent of verticallydipping foliation provided anisotropy is stronger; thereby supplying a moreplausible alternative.Assuming a constant strength of anisotropy in the LRC, the heteroge-neous magnitude of anisotropy indicated by the increase in normalized delaytimes from east to west could be due to the lateral variations in the extentof the LRC. Normalized delay times from forward modeling display a sim-ilar trend but generally have more scattered delay times (figure 3.7). Thesynthetic wedge model is thickest below TSJB/TWBB/TWGB whereas tothe east the dipping arm of the LRC is significantly thinner. An alternativeexplanation relates to fault zone mylonites as observed in the Brevard faultzone, a continental strike-slip/thrust fault. Above 200 MPa mylonitic meta-406.2. Mineral preferred orientation of the Leech River Complexmorphic rocks of the fault zone were found to exhibit transverse isotropy dueto preferred mineral orientation, which at lower pressures is enhanced by ori-ented cracks [19]. The presence of similar fault zone rocks in the boundingfaults of the LRC may also explain why normalized delay times are increasedfor stations in close proximity to the faults. We also acknowledge that in thevicinity of fault zones crack anisotropy can be anomalously high, owing tolarge surface fractures [23]. Increased proximity of eastern stations to LRCfault zones presents another potential explanation for the E-W increase innormalized delay times.The azimuthal variations observed at LZB and TWKB are reproducedby the two domain synthetic LRC model. Western azimuths record φ pref-erentially oriented E-W, whereas eastern azimuths tend towards NW-SE(figure 4.2). It is conceivable that anisotropy related to maximum hori-zontal compressive stress is a more slowly varying function of position incomparison to anisotropy due to mineral orientation, thus given the stressmodel established earlier we do not attribute the azimuthal variations in φ tolocal variations in σHmax. We interpret spatial clusters of φ as an indicationfor heterogeneous anisotropy on SVI resulting from complex lithology withpreferred mineral orientation that leads to small-scale azimuthal variationsin φ.416.3. Mineral preferred orientation of the Olympic Peninsula6.3 Mineral preferred orientation of the OlympicPeninsulaThe core rocks of the Olympic peninsula (figure 6.1) are comprised of twomajor accretionary terranes; western core rocks are non-slaty and locallycoherent whereas the slaty eastern core rocks are pervasively sheared andexhibit well developed slaty-cleavage. The eastern core is predominantlycomposed of shale, sandstone and siltstone that have been variously meta-morphosed to slate, semischist and phyllites. Core units form long irregu-lar packets that vary between disrupted formations of sandstone/semischistwith slate/phyllite to relatively intact interbedded sandstone and slate [47].Whereas sandstone is seismically isotropic, velocity measurements on Olympiccore slates indicate that they exhibit transverse isotropy. At 400 MPa, asample of slate from Hurricane Ridge has Vqs1 = 3.79 kms −1 and Vqs2 = 2.94kms −1 (∼ 7% S -wave anisotropy) for propagation perpendicular to thesymmetry axis. The interbedding of slate with sandstone acts to effectivelydilute the anisotropy of the aggregate medium.Structural analyses of the Olympic core rocks provide important cluesabout the origin and nature of crustal seismic anisotropy in SVI and north-ern Washington. Based on field observations of lineations, Tabor and Cady[47] have divided the Olympic Eastern core rocks into two major structuraldomains, with a boundary running approximately north-south. Slaty cleav-ages in the Western domain dip steeply to the east and northeast and thelineations plunge eastward. In the Eastern domain cleavages dip steeplyo the west and southwest and lineations plunge westward. The orienta-426.3. Mineral preferred orientation of the Olympic Peninsulations of slaty cleavage, bedding and axial planes of folds are nearly coplanarthroughout both domains. Although structural details of the core rocksshow considerable complexity, the provide a coherent model of slates with afan of cleavage extending asymmetrically to the east with dips increasing tonear vertical from west to east [47]. Of importance is the overall continuityof the slaty cleavage over distances of several kilometers, a condition that isfavourable to generate observable seismic anisotropy.Tomography [40] and reflection [20] studies have postulated that theOlympic core rocks are underthrust beneath the Eocene basalts of northernWashington and SVI as illustrated in the schematic diagrams in figure 6.1.Figure 6.1b presents an interpreted geologic profile along BB’ based uponClowes et al. [20] and includes reflectors imaged in the study. A dippingreflector beneath SVI was interpreted as the northward continuation of theHurricane Ridge fault that is underlain by core rocks from the Olympicpeninsula. The potential presence of anisotropic slate from the Olympiccore beneath SVI presents a third source of anisotropy that could explainanisotropy for stations such as SOKB, GLBC, SSIB and others from SVIthat we have not associated with the LRC. In addition, potential field studiesbeneath SVI identified a region of higher density (3140 kg/m3) material un-derlying the Eocene basalts and LRC with a lower boundary defined by theE-layer [26]. This layer has been interpreted as either underplated metamor-phosed fragments of the LRC/Eocene basalts or high density rocks resultingfrom metamorphosis of accreted sedimentary rocks. If the metamorphicrocks exhibit strong preferred mineral orientation they could represent asource of deep crustal anisotropy that has not been illustrated in figure 6.1.436.3. Mineral preferred orientation of the Olympic PeninsulaSimilarly, splitting observed on northern Washington may also be re-lated to anisotropy from underlain Olympic slates as illustrated in figure6.1c. Figure 6.1c is based upon P -wave tomographic profiles across north-ern Washington [40]. The variability of the terrane along with the mixeddegrees of deformation may be the cause of the complex anisotropy displayedby variations of φ with azimuth and incidence angle. The Eocene basaltsthat overlay the core rocks on SVI and northern Washington are isotropicand do not influence observations of anisotropy [17]. Whereas the HurricaneRidge slate exhibits relatively weak anisotropy in comparison to LRC phyl-lites and schists, the extent of the core rocks is believed to be significantlygreater than that of LRC rocks. We believe the core rocks may extendas far as the plate boundary. Using the velocities stated earlier, Olympicslates could produce the 0.1-0.15 s of splitting typically observed in northernWashington with as little as 1-2 km for propagation perpendicular to thesymmetry axis.446.3. Mineral preferred orientation of the Olympic Peninsula 125 oW  30’  124 oW  30’  123 oW  30’  20’  40’   48 oN  20’  40’   49 oN WRANGELLIA TERRANEEOCENE BASALTSOLYMPIC CORE ROCKSLEECH RIVER COMPLEXHURRICANE RIDGE FAULTVANCOUVER ISLANDOLYMPIC PENINSULAB’BUNSPECIFIED PERIPHERAL SEDIMENTARYROCKSC C’C (West)Depth (km)OLYMPIC MOUNTAINS EASTERN WASHINGTON010403020CORE ROCKSC’ (East)EOCENE BASALTS? ? ?HURRICANE RIDGE FAULTJUAN DE FUCA PLATEB (South)Depth (km)OLYMPIC MOUNTAINS STRAIT OF JUAN DE FUCA VANCOUVER ISLANDLEECH RIVER FAULTSURVEY MOUNTAIN FAULTWRANGELLIA HURRICANE RIDGE FAULT010403020CORE ROCKSCORE ROCKS?B’ (North)EOCENE BASALTSLRCHURRICANE RIDGE FAULT ?? ? ?E - LAYERJUAN DE FUCA PLATEacb 2. CRACK ANISOTROPY ? 3. MINERAL ORIENTATION OF OLYMPIC PENINSULA SLATES ?1. LRC MINERAL ORIENTATION2. CRACK ANISOTROPY ? 3. MINERAL ORIENTATION OF OLYMPIC PENINSULA SLATES ?Figure 6.1: Schematic diagrams illustrating geological terrane and inter-preted depth sections. (a) Major geological terranes of northern Cascadiawith rose histograms of φ included. (b) Geologic interpretation of depthprofile BB’ based upon Clowes et al. [20]. Illustrates typical LFE paths andlists associated sources of anisotropy. (c) Geologic interpretation of depthprofile CC’ based upon Ramachandran et al. [40]. Illustrates underthrustingof Olympic core rocks beneath Eocene basalts.45Chapter 7SummaryWe have conducted a splitting analysis using LFE templates distributedon southern Vancouver island and northern Washington and presented theassociated fast directions and delay times. Pervasively foliated and ver-tically dipping schists and phyllites in the Leech River Complex lead toanisotropy generated by a transversely isotropic medium with a symmetryaxis in the horizontal plane. Anisotropy observed at the majority of main-land stations can be readily explained by anisotropy resulting from preferredmineral orientation of rocks within the Leech River Complex. Anisotropicforward modeling using a north/north-east dipping LRC wedge model thatexhibits HTI symmetry was able to reproduce trends in dominant fast di-rections, normalized delay times and in select cases, azimuthal variationsin fast direction. Crack anisotropy may provide an alternative explanationfor anisotropy observed at stations that showed little relation to LRC prop-erties; agreement with local estimates of maximum horizontal compressivestress was typically improved at these stations. Prior studies suggested thatanisotropy observed at stations along SVI results from stress aligned cracks,thus making them candidates to monitor local stress patterns. We havedemonstrated that anisotropy at the same stations is equally well explained46Chapter 7. Summaryby preferred mineral orientation of the Leech River Complex thereby castingdoubt on prior conclusions. Splitting observations on northern Washingtonappear to be due to a mixture of crack anisotropy and the preferred mineralorientation of anisotropic slates of the Olympic core rocks that underthrustlocal terranes. We suggest that the anisotropic slates of the Olympic coremay extend beneath southern Vancouver Island and present an additionalsource of anisotropy. If so, anisotropy due to the preferred mineral orienta-tion of metamorphosed, accreted sedimentary rocks may be a global featurein forearc crusts of subduction zones.47Bibliography[1] Audet, P., M. Bostock, N. Christensen, and S. Peacock (2009), Seis-mic evidence for overpressured subducted oceanic crust and megathrustfault sealing, Nature, 457, 76–78.[2] Audet, P., M. G. Bostock, D. C. Boyarko, M. R. Brudzinski, and R. M.Allen (2010), Slab morphology in the Cascadia fore arc and its relationto episodic tremor and slip, J. Geophys. Res., 115, B00A16, doi:10.1029/2008JB006053.[3] Balfour, N. J., J. F. Cassidy, S. E. Dosso, and S. Mazzotti (2011),Mapping crustal stress and strain in southwest British Columbia, J.Geophys. Res., 116.[4] Balfour, N. J., J. F. Cassidy, and S. E. Dosso (2012), Crustal anisotropyin the forearc of the northern Cascadia Subduction Zone, BritishColumbia, Geophys. J. Int., 188, 165–176, doi:10.1111/j.1365-246X.2011.05231.x.[5] Berens, P. (2009), CircStat: A MATLAB Toolbox for Circular Statis-tics, J. Stat. Soft., 31, 1–21.[6] Beroza, G. C., and S. Ide (2011), Slow earthquakes and nonvolcanic48Bibliographytremor, Annu. Rev. Earth Planet. Sci., 39, 271–296, doi:10.1146/annurev-earth-040809-152531.[7] Booth, D. C., and S. Crampin (1985), Shear-wave polarizations on acurved wavefront at an isotropic free surface, Geophys. J. R. Astron.Soc., 83, 31–45.[8] Bostock, M. G. (2013), The Moho in subduction zones , Tectonophysics,609, 547 – 557, doi:[9] Bostock, M. G., and N. I. Christensen (2012), Split from slip and schist:Crustal anisotropy beneath northern Cascadia from non-volcanictremor, J. Geophys. Res., 117, B08303, doi:10.1029/2011JB009095.[10] Bostock, M. G., A. A. Royer, E. H. Hearn, and S. M. Peacock(2012), Low frequency earthquakes below southern Vancouver Island,Geochem., Geophys., Geosyst., 13, doi:10.1029/2012GC004391.[11] Brocher, T. M. (2005), Empirical relations between elastic wavespeedsand density in the Earth’s crust, Bull. Seis. Soc. Am., 95, 2081–2092.[12] Brocher, T. M., M. A. Fisher, E. L. Geist, and N. I. Christensen (1989),A high-resolution seismic reflection/refraction study of the Chugach-Peninsular Terrane Boundary, southern Alaska, J. Geophys. Res., 94,4441–4455.[13] Brown, J. R., G. C. Beroza, and D. R. Shelly (2008), An autocorrelationmethod to detect low frequency earthquakes within tremor, Geophys.Res. Lett., 35, doi:10.1029/2008GL034560.49Bibliography[14] Brown, J. R., G. C. Beroza, S. Ide, K. Ohta, D. R. Shelly, S. Y.Schwartz, W. Rabbel, M. Thorwart, and H. Kao (2009), Deep low-frequency earthquakes in tremor localize to the plate interface inmultiple subduction zones, Geophys. Res. Lett., 36, doi:10.1029/2009GL040027.[15] Cassidy, J. F., and M. G. Bostock (1996), Shear-wave splitting abovethe subducting Juan de Fuca Plate, Geophys. Res. Lett., 23, 941–944.[16] Christensen, N. I. (1965), Compressional wave velocities in metamor-phic rocks at pressures to 10 kilobars, J. Geophys. Res., 70, 6147–6164.[17] Christensen, N. I. (1966), Shear wave velocities in metamorphic rocksat pressures to 10 kilobars, J. Geophys. Res., 71, 3549–3556.[18] Christensen, N. I. (1996), Poisson’s ratio and crustal seismology, J.Geophys. Res., 101, 3139–3156.[19] Christensen, N. I., and D. Szymanski (1988), Origin of reflections fromthe Brevard Fault zone, J. Geophys. Res., 93, 1087–1102.[20] Clowes, R. M., M. T. Brandon, A. G. Green, C. J. Yorath, A. S. Brown,E. R. Kanasewich, and C. Spencer (1987a), LITHOPROBE southernVancouver Island: Cenozoic subduction complex imaged by deep seis-mic reflections, Can. J. Earth Sci., 24, 31–51.[21] Clowes, R. M., C. Yorath, and R. Hyndman (1987b), Reflection map-ping across the convergent margin of western Canada, Geophys. J. ofthe R. Astron. Soc., 89, 79–84.50Bibliography[22] Cowan, D. S. (2003), Revisiting the Baranof Leech River hypothesisfor early Tertiary coastwise transport of the Chugach Prince Williamterrane, Earth and Planet. Sci. Lett., 213, 463 – 475.[23] Crampin, S. (1994), The fracture criticality of crustal rocks, Geophys.J. Int., 118, 428–438.[24] Crampin, S., and S. Peacock (2005), A review of shear-wave splittingin the compliant crack-critical anisotropic earth, Wav. Motion, 41, 59– 77.[25] Currie, C. A., J. F. Cassidy, and R. D. Hyndman (2001), A regionalstudy of shear wave splitting above the Cascadia Subduction Zone:Margin-parallel crustal stress, Geophys. Res. Lett., 28, 659–662.[26] Dehler, S., and R. Clowes (1992), Integrated geophysical modelling ofterranes and other structural features along the western Canadian mar-gin, Can. J. Earth Sci., 29, 1492–1508.[27] Fairchild, L., and D. Cowan (1982), Structure, petrology, and tectonichistory of the Leech River complex northwest of Victoria, Vancouver,Can. J. Earth Sci., 19, 1817–1835.[28] Godfrey, N. J., N. I. Christensen, and D. A. Okaya (2000), Anisotropyof schists: Contribution of crustal anisotropy to active source seismicexperiments and shear wave splitting observations, J. Geophys. Res.,105, 27,991–28,007.[29] Green, A., B. Milkereit, L. Mayrand, C. Spencer, R. Kurtz, and R. M.51BibliographyClowes (1987), Lithoprobe seismic reflection profiling across vancouverisland: results from reprocessing, Geophys. J. Int., 89, 85–90, doi:10.1111/j.1365-246X.1987.tb04392.x.[30] Hansen, R. T., M. G. Bostock, and N. I. Christensen (2012), Natureof the low velocity zone in Cascadia from receiver function waveforminversion, Earth and Planet. Sci. Lett., 337338, 25 – 38, doi:[31] Ide, S., D. R. Shelly, and G. C. Beroza (2007), Mechanism of deep lowfrequency earthquakes: Further evidence that deep non-volcanic tremoris generated by shear slip on the plate interface, Geophys. Res. Lett.,34, doi:10.1029/2006GL028890.[32] Kennett, B. L. N. (1991), The removal of free surface interactions fromthree-component seismograms, Geophys. J. Int., 104, 153–154, doi:10.1111/j.1365-246X.1991.tb02501.x.[33] Komatitsch, D., and J. Tromp (1999), Introduction to the spectral ele-ment method for three-dimensional seismic wave propagation, Geophys.J. Int., 139, 806–822, doi:10.1046/j.1365-246x.1999.00967.x.[34] Komatitsch, D., C. Barnes, and J. Tromp (2000), Simulation ofanisotropic wave propagation based upon a spectral element method,Geophys., 65, 1251–1260, doi:10.1190/1.1444816.[35] Mayrand, L. J., A. G. Green, and B. Milkereit (1987), A quantitativeapproach to bedrock velocity resolution and precision: The LithoprobeVancouver Island Experiment, J. Geophys. Res., 92, 4837–4845.52Bibliography[36] McCaffrey, R., R. W. King, S. J. Payne, and M. Lancaster (2013),Active tectonics of northwestern U.S. inferred from GPS-derived surfacevelocities, J. Geophys. Res., 118, doi:1029/2012JB009473.[37] Muller, J. (1983), Geology, Victoria, Map 1553A, scale 1:100000 , geol.Surv. of Can.,Ottawa, Ont., Can.[38] Nishide, N., T. Hashimoto, J. Funasaki, H. Nakazawa, M. Oka, H. Ueno,N. Yamada, I. Sasakawa, K. Maeda, K. Sugimoto, and T. Takashima(2000), Nationwide activity of low-frequency earthquakes in the lowercrust in Japan, abstract Sk-P002 presented at Japan Earth and Plane-tary Science Joint Meeting, Geod. Soc. of Jpn., Tokyo.[39] Nowack, R. L., and M. G. Bostock (2013), Scattered waves from low-frequency earthquakes and plate boundary structure in northern Cas-cadia, Geophys. Res. Lett., 40, 4238–4243, doi:10.1002/grl.50826.[40] Ramachandran, K., R. D. Hyndman, and T. M. Brocher (2006), Re-gional p wave velocity structure of the northern Cascadia subductionzone, J. Geophys. Res., 111, B12301, doi:10.1029/2005JB004108.[41] Royer, A., and M. Bostock (2013), A comparative study of low fre-quency earthquake templates in northern Cascadia, Earth and Plane-tary Science Lett., doi:, inpress.[42] Rusmore, M. E., and D. S. Cowan (1985), JurassicCretaceous rock unitsalong the southern edge of the Wrangellia terrane on Vancouver Island,Can. J. Earth Sci., 22, 1223–1232.53Bibliography[43] Shelly, D. R., G. C. Beroza, S. Ide, and S. Nakamura (2006), Low-frequency earthquakes in Shikoku, Japan, and their relationship toepisodic tremor and slip, Nature, 442, 188–191.[44] Shelly, D. R., G. C. Beroza, and S. Ide (2007), Non-volcanic tremor andlow-frequency earthquake swarms, Nature, 446, 305–307.[45] Silver, P. G., and W. W. Chan (1991), Shear wave splitting and sub-continental mantle deformation, J. Geophys. Res., 96, 16,429–16,454.[46] Soyer, W., and M. Unsworth (2006), Deep electrical structure ofthe northern Cascadia (British Columbia, Canada) subduction zone:Implications for the distribution of fluids, Geology, 34, 53–56, doi:10.1130/G21951.1.[47] Tabor, R. W., and W. M. Cady (1978), The structure of the olympicmountains, washington; analysis of a subduction zone, USGS Profes-sional Paper, 1033.[48] Wang, K., T. Mulder, G. C. Rogers, and R. D. Hyndman (1995), Casefor very low coupling stress on the cascadia ssubduction fault, J. Geo-phys. Res., 100, 12,907–12,918.[49] Wells, R. E., and R. W. Simpson (2001), Northward migration of thecascadia forearc in the northwestern u.s. and implications for subduc-tion deformation, Earth Planets Space, 53, 275 – 283.54


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