Michael Bostock

British Columbia and Washington state encompass the northern extent of the Cascadia subductionzone and the southern termination of the Queen Charlotte Transform fault. Convergencebetween the North American plate and oceanic Juan de Fuca (JdF), Explorer, and Pacificplates is accommodated by subduction, underthrusting, and transform faulting and producesabundant seismicity that is unevenly distributed along the margin. This thesis probes severalearthquake-dense regions in Cascadia that include a unique cluster of seismicity south of TexadaIsland in BC, the highly active Puget Lowlands in Washington state, and the offshore NootkaFault zone (NFZ) region that separates the JdF from the Explorer microplate. High-resolutionrelative earthquake locations and new seismic velocity models strongly suggest that these regionsof dense seismicity are influenced by fluids. Fluids enter the subduction system throughthe oceanic plate and are expelled by increasing pressure and temperature conditions. Subsequentfluid redistribution likely influences the distribution of seismicity. The Texada cluster ofseismicity is revealed to manifest ∼20 km long faults within the JdF slab, which are inferred tobe structures inherited from a propagator wake that exhibits increased levels of hydration. Inthe Puget Lowlands, robust, tomographically-derived measurements of Poisson’s ratio suggestthat the Washington forearc is dominated by the accreted Siletzia terrane, and that this >10km thick unit of flood basalts inhibits infiltration of slab-derived fluids into the overriding crust,thereby concentrating earthquakes beneath it. A complex configuration of oceanic lithosphereat the NFZ is revealed by newly determined seismic velocity models, and the association oflarge earthquakes and abundant microseismicity with plate bending faults suggests that fluidingress lowers effective normal stresses and promotes brittle failure producing Vancouver Island’slargest earthquakes in recent decades. The penultimate chapter of this thesis presentsa new methodology for determining changes in seismic velocity from seismic coda. Repeatingearthquakes and ambient noise correlations from Haida Gwaii are utilized to demonstrate thatthe velocity of the shallow crust decreased following the 2012 Mw 7.8 earthquake, which weinterpret as due to earthquake-induced changes in fluid-filled porosity.

View record

I investigate earthquakes and slow-slip phenomena in subduction zones and their overlying forearcs, with a particular focus on how metamorphism and fluid-migration affect the locations and character of earthquakes. I first focus on low-frequency earthquakes (LFEs) associated with episodic tremor-and-slip (ETS) down-dip of the megathrust zone on the subduction interface. I search for LFEs in southern Cascadia and find 34 distinct families. As in other regions of Cascadia, LFE hypocenters suggest that the plate interface is several kilometers shallower than reported by a recent model. South of Cascadia, I discover LFEs on the Maacama and Bucknell Creek faults, and compare them to shallow repeating earthquakes with distinct spectral and recurrence characteristics.After LFEs I turn to regular earthquakes. I develop new methodologies to estimate earthquake source properties that can better inform us of the physical conditions leading to brittle failure. In particular, I develop a multichannel deconvolution algorithm for recovering apparent source-time functions and apply it to a set of earthquakes offshore Japan. The method relates to empirical Green's function deconvolution but makes fewer assumptions of the target earthquakes. I then introduce a new relative moment tensor inversion method that relies on principal component analysis on aligned P-waves of nearby earthquakes to measure relative P amplitudes, and on aligned S-waves to produce coefficients that linearly relate three earthquakes. I apply the method to intraslab earthquakes in the subducted portion of the Yakutat terrane, in eastern Alaska, that appear to be related to eclogitization of oceanic plateau crust. Nearby earthquakes with antiparallel slip imply extremely weak faults and possibly a heterogeneous stress regime.Finally I consider multiple potential causes of enigmatic low Vp/Vs anomalies in the Cascadia forearc and subducting mantle, and explore effective medium models that consider both mineralogy and porosity. Based on their correlation with dense seismicity, high attenuation, and rapid ETS recurrence, I suggest that low Vp/Vs anomalies relate to upwelling slab fluids above regions of high permeability in the subducting plate, which are caused by high slab-curvature.

View record

Inverse problems in the imaging sciences encompass a variety ofapplications. The primary problem of interest is the identification ofphysical parameters from observed data that come from experimentsgoverned by partial-differential-equations. The secondary type ofimaging problems attempts to reconstruct images and video that arecorrupted by, for example, noise, subsampling, blur, or saturation.The quality of the solution of an inverse problem is sensitive to issues such as noise and missing entries in the data. The non-convex seismic full-waveform inversion problem suffers from parasitic local minima that lead to wrong solutions that may look realistic even for noiseless data. To meet some of these challenges, I propose solution strategies that constrain the model parameters at every iteration to help guide the inversion.To arrive at this goal, I present new practical workflows, algorithms, and software, that avoid manual tuning-parameters and that allow us to incorporate multiple pieces of prior knowledge. Opposed to penalty methods, I avoid balancing the influence of multiple pieces of prior knowledge by working with intersections of constraint sets. I explore and present advantages of constraints for imaging. Because the resulting problems are often non-trivial to solve, especially on large 3D grids, I introduce faster algorithms, dedicated to computing projections onto intersections of multiple sets.To connect prior knowledge more directly to problem formulations, I also combine ideas from additive models, such as cartoon-texture decomposition and robust principal component analysis, with intersections of multiple constraint sets for the regularization of inverse problems. The result is an extension of the concept of a Minkowski set.Examples from non-unique physical parameter estimation problems show that constraints in combination with projection methods provide control over the model properties at every iteration. This can lead to improved results when the constraints are carefully relaxed.

View record

The Cascadia subduction zone is a warm subduction zone where Episodic Tremor and Slow Slip(ETS), a phenomenon with debated governing physical mechanisms, is observed. The eventualdevelopment of an accurate physical model of ETS relies on a comprehensive set of seismic andgeodetic observations. Low Frequency Earthquakes (LFE) are one of the seismic signaturesthat accompany slow slip and possess discernible phase arrivals that can be used for traveltimeinversion methods. I first address the issue of detecting LFEs by developing an automatedalgorithm that exploits cross-station waveform similarity. The algorithm detects thousands ofevents per ETS episode in southern Vancouver Island that reveal pronounced spatiotemporalclustering and complex propagation patterns that compare favourably with independent ob-servations in Cascadia. This method allows us to build a significantly improved catalogue ofLFE templates that we next employed for local double-difference earthquake tomography withthe aim of resolving structures in the vicinity of ETS and metamorphic controls on its generation. In southern Vancouver Island, tomographic images reveal high Poisson’s ratios associatedwith a dipping low‐velocity zone (LVZ) inferred to be overpressured, upper oceanic crust ofthe Juan de Fuca plate where LFEs and other slow‐slip phenomena occur. We also observe alow Poisson’s ratio anomaly ( ∼0.225) in the forearc continental crust above the mantle wedgecoinciding with high Vp and high levels of clustered microseismicity. We develop a conceptualmodel where quartz concentration is produced by metasomatism in the forearc crust and catalyzed by a focussed ingress of slab-derived fluids at high pore pressure. In northern VancouverIsland, we also observe the LVZ and a low Poisson’s ratio anomaly in a similar position, but noassociated microseismic activity. Distinct, parallel lineaments that trend slightly oblique to thestrike of the megathrust are resolved within seismicity concentrations of the Nootka Fault Zoneimmediately offshore Nootka Island. We speculate that these features may represent faultingdeveloped as a result of high strains associated with slab bending, amplified in the vicinity ofthe newly formed plate boundary. We also find possible evidence for a deep-seated, seismogenicfault near the Explorer Plate edge.

View record

Railway operators mitigate the risk of derailments caused by hazardous rocks falling onto the track by installing slide detector fences (SDF). These consist of electrical sensing wires strung on poles located uphill of the track; falling rocks snap these wires and trigger an alarm. Rocks of non-threatening size and migrating animals frequently break the wires causing prolonged false alarms and delaying rail traffic until the SDF is manually repaired, often in a hazardous environment. This thesis is concerned with the development of a prototype of the autonomous Seismic Rockfall Detection System (SRFDS) as a potential replacement for the SDF. Analysis and classification of natural and anthropogenic seismic signals which have been observed at the SRFDS field installations, is presented. A method for identification of hazardous rocks (>0.028 m³) using an empirical peak ground velocity attenuation model is outlined. Pattern recognition techniques which are based on cross-correlation and on variations in the short-term / long term averages of the ground vibrations are introduced for rail traffic identification and rockfall detection. The techniques allow the SRFDS to eliminate false activations by rail traffic, report hazardous rocks with minimum (
View record

Sedimentary rocks and shales in particular are known to be anisotropic, sometimes strongly so, and hydraulic fracturing is now common practice in shale plays to enhance the extraction of hydrocarbons. One of the ways to understand the hydraulic fracturing process is through the micro--earthquakes that it generates; it is therefore of interest to study the impact that anisotropy may have on hydraulically induced seismicity. This thesis is concerned with the inclusion of anisotropy into the geophysical forward and inverse problems of microseismic sources.Ray theory is used for the forward problem --- dynamic ray tracing in a medium composed of homogeneous layers with vertical transverse isotropy (VTI) is used and the possibility of qSv triplication is considered. Novel approaches to the inverse problem are introduced, including waveform fitting in the frequency domain to recover the source function as well as moment tensor. Uncertainties in estimated moment tensor components are quantified with multi--variate normal sampling utilizing the full covariance matrix from the linear inversion. A new decomposition of the moment tensor is described that removes the distortions due to anisotropy local to the source and a new way to visualize an earthquake source is also introduced. The impact of anisotropy on moment tensor inversion and decomposition is shown to be significant.Field data recorded by two downhole arrays of multicomponent receivers are analyzed using the new techniques. The event collection suggests a source mechanism dominated by cylindrical dilatation. This is unexpected but is supported by results from another code. Future directions include extensions to lower symmetry forms of anisotropy and application to surface arrays. Preliminary analysis of downhole recordings of aftershocks of the 2008 Mw=7.9 Wenchuan earthquake is shown.

View record

Deconvolution of seismic data is an important component of signal processing that aims to remove the seismic source from seismograms, thereby isolating the Green’s function. By considering seismograms of multiple earthquakes from similar locations recorded at a given station and that therefore share the same Green’s function, we investigate a system of equations where the unknowns are the sources and source durations. Our solution is derived using direct linear inversion to recover the sources and Newton’s method to recover source durations. For the short seismogram durations considered, we are able to recover source time functions for noise levels at 1% of the direct P -wave amplitude. However, the nonlinearity of the problem renders the system expensive to solve and sensitive to noise; therefore consideration is limited to short seismograms with high signal-to-noise ratio (SNR). When SNR levels are low, but a large multiplicity of seismograms representing a common source-receiver path are available, we can apply a different deconvolution approach to recover the Green’s function. In an application to tectonic tremor in northern Cascadia, we implement an iterative blind deconvolution method that involves correlation, threshold detection and stacking of 1000’s of low frequency earthquakes (LFEs) that form part of tremor to generate templates that can be considered as empirical Green’s functions. We exploit this identification to compute hypocentres and moment tensors. LFE hypocentres follow the general epicentral distribution of tremor and occur along tightly defined surfaces in depth. The majority of mechanisms are consistent with shallow thrusting in the direction of plate motion. We analyze the influence of ocean tides on the triggering of LFEs and find a spatially variable sensitivity to tidally induced up-dip shear stress (UDSS), suggesting that tidal sensitivity must partially depend on laterally heterogeneous physical properties. The majority of LFEs fail during positive and increasing UDSS, consistent with combined contributions from background slow slip and from tides acting directly on LFEs. We identified rapid tremor reversals in southern Vancouver Island with higher sensitivity to UDSS than the main front and which at least partially explains an observed increase in LFE sensitivity to UDSS with time.

View record

The present thesis is divided into three distinct parts and focuses both on the improvement of deconvolution technique in a teleseismic context for crustal and upper-mantle studies and on the understanding of western Canada structure and evolution through seismic imaging. The first part presents estimates of the P-component of the teleseismic-P Green's functions for three stations of the Canadian National Seismic Network (CNSN) obtained using a new deconvolution technique. Our results show evidence of the principal, first-order scattered Moho phases and, in particular, the Pp_Mp. The second part presents teleseismic receiver functions from 20 broadband three-component seismometers deployed along the MacKenzie-Liard Highway in Canada's Northwest Territories as part of the joint Lithoprobe-IRIS CAnadian NOrthwest Experiment (CANOE). These stations traverse a Paleoproterozoic suture and subduction zone that has been previously documented in detail to mantle depths using seismic reflection profiling. Our results reveal the response of the ~1.8 Ga subduction zone on both the radial and transverse component. The identification of this structure and its comparison with fine-scale mantle layering below the adjacent Slave province and from a range of Precambrian terranes provides an unambiguous connection between fossil subduction and fine-scale, anisotropic mantle layering found beneath cratons. Previous documentation of similar layering below the adjacent Slave province provides strong support for the thesis that early cratonic blocks were stabilized through processes of shallow subduction. The last part presents P- and S wave velocity models for western Canada. In this part, we focus our attention on two distinct features: 1) the transition from Phanerozoic to cratonic mantle in northwestern Canada and 2) the complex tectonic environment at the northern terminus of the Cascadia subduction zone where the plate boundary changes from convergent to transform. We find that the main transition from Phanerozoic to cratonic mantle in northwestern Canada occurs at the Cordilleran deformation front. In northern Cascadia, we have imaged and characterized the signature of the subducting Juan de Fuca plate and observed evidence of subduction beyond the northern edge of the slab. Our result show that the Anahim hotspot track is underlain by a -2% low-velocity zone.

View record

As seismic waves travel through the earth, the visco-elasticity of the earth's medium will cause energy dissipation and waveform distortion. This phenomenon is referred to as seismic absorption or attenuation. The absorptive property of a medium can be described by a quality factor Q, which determines the energy decay and a velocity dispersion relationship. Four new ideas have been developed in this thesis to deal with the estimation and application of seismic absorption. By assuming that the amplitude spectrum of a seismic wavelet may be modeled by that of a Ricker wavelet, an analytical relation has been derived to estimate a quality factor from the seismic data peak frequency variation with time. This relation plays a central role in quality factor estimation problems. To estimate interval Q for reservoir description, a method called reflectivity guided seismic attenuation analysis is proposed. This method first estimates peak frequencies at a common midpoint location, then correlates the peak frequency with sparsely-distributed reflectivities, and finally calculates Q values from the peak frequencies at the reflectivity locations. The peak frequency is estimated from the prestack CMP gather using peak frequency variation with offset analysis which is similar to amplitude variation with offset analysis in implementation. The estimated Q section has the same layer boundaries of the acoustic impedance or other layer properties. Therefore, the seismic attenuation property obtained with the guide of reflectivity is easy to interpret for the purpose of reservoir description. To overcome the instability problem of conventional inverse Q filtering, Q compensation is formulated as a least-squares (LS) inverse problem based on statistical theory. The matrix of forward modeling is composed of time-variant wavelets. The LS de-absorption is solved by an iterative non-parametric approach. To compensate for absorption in migrated seismic sections, a refocusing technique is developed using non-stationary multi-dimensional deconvolution. A numerical method is introduced to calculate the blurring function in layered media, and a least squares inverse scheme is used to remove the blurring effect in order to refocus the migrated image. This refocusing process can be used as an alternative to regular migration with absorption compensation.

View record

Convergent continental margins are regions of intense deformation caused by the interaction of oceanic plates with continents. The spatial extent of deformation is broadly commensurate with the specific time scale of the causative phenomenon. For example, subduction-related short-term deformation is limited to
View record