Our knowledge of Earth's structure and dynamics is owed, in large part, to techniques that allow us to infer the physical properties of rocks based on observations made near the surface. In particular, passive seismic imaging relies on natural sources of elastic wave energy (typically earthquakes) to illuminate Earth's interior. The frequency-dependent dispersive characteristics of surface waves from earthquakes provides a valuable constraint on the depth heterogeneity beneath the surface, which can be used to infer structure of the lithosphere (i.e., tectonics). The first part of this thesis develops novel tools for passive seismic imaging considering surface-wave dispersion. Specifically, Bayesian (probabilistic) methods are developed that provide rigorous uncertainty quantification. The ability to estimate the directional dependence of surface-wave speeds (i.e., seismic anisotropy) is demonstrated. Furthermore, a general approach for considering circular (wrapped) data, such as surface-wave phase measurements, is developed and applied to estimate the average dispersion between pairs of seismic stations. These ideas are applied to data recorded at seismic stations over British Columbia, Canada, to produce a large volume of data products that will help improve our understanding of the tectonics throughout the region. The second part of this thesis investigates the structural and mechanical conditions in subduction zones, where tectonic plates collide and one plate is thrust beneath the other. Specifically, a type of passive seismic imaging based on recordings of body waves from distant earthquakes (known as receiver functions) is used to infer subduction zone structure in relation to the coupling between tectonic plates. Receiver function data calculated for stations over the Cascadia subduction zone (southwestern Canada) suggest that episodes of slow slip on the plate interface occur in tandem with changes in mechanical conditions. Receiver function data calculated for stations over the Hikurangi subduction zone (New Zealand) suggest that stress and deformation along the margin are spatially linked to plate coupling. These results improve our understanding of the dynamics of these tectonic systems, including the seismic hazards that they pose.

Ce travail vise à améliorer la compréhension des signaux sismiques dérivés des fonctions de corrélation inter-récepteur du bruit sismique, ce qui est critique pour une imagerie fiable de la Terre profonde basée sur le bruit. La thèse comprend sept chapitres. Le chapitre 1 introduit les connaissances de base sur le bruit sismique, de la terminologie à ses origines diverses. Le chapitre 2 fournit une vue d'ensemble de la littérature sur l'historique et le développement de la méthode récente de corrélation de bruit, et passe en revue diverses techniques pour le prétraitement des données de bruit sismique et le post-traitement des fonctions de corrélation de bruit. Des méthodes de traitement du bruit basées sur les statistiques et un schéma modifié pour calculer la fonction de corrélation sont développés dans ce chapitre. Le chapitre 3 propose plusieurs techniques basées sur la transformée de Radon pour mesurer les lenteurs des champs d'ondes corrélés et analyser en termes de phases sismiques les signaux dérivés du bruit. Le chapitre 4 montre que les ondes de volume sondant la Terre profonde peuvent être extraites des corrélations de bruit à des distances télésismiques, avec des enregistrements de bruit provenant de deux réseaux sismiques régionaux. Le chapitre 5 applique les techniques proposées au chapitre 3 aux corrélations de bruit entre deux réseaux calculées au chapitre 4, et permet de comprendre l’origine de la phase précoce non-physique observée dans les données. Le chapitre 6 discute des conditions dans lesquelles apparaissent des phases sans correspondance dans la réponse physique de la Terre qui peuvent fausser les analyses des structures profondes basées sur le bruit.. Le dernier chapitre fournit un résumé sur les contributions de cette thèse et une perspective de plusieurs travaux soit en cours soit envisagés pour le futur. / This work aims toward an improved understanding of the seismic signals derived from the inter-receiver correlation functions of seismic noise, which is valuable and critical for a reliable noise-based deep Earth imaging. The thesis consists of seven chapters. Chapter 1 introduces background knowledge on seismic noise, from its classifications to various origins. Chapter 2 provides a literature overview on the history and development of the emerging noise correlation method, and reviews various techniques for the pre-processing of seismic noise data and post-processing of noise correlation functions. Statistics-based noise processing methods and a modified scheme for computing correlation function are developed in this chapter. Chapter 3 proposes several Radon-based techniques to analyze the slownesses of correlated wavefields and to unveil the origin of noise-derived seismic signals. Chapter 4 shows that body waves penetrating into deep Earth can be extracted from noise correlations at teleseismic distances, with noise records from two regional seismic networks. Chapter 5 applies the techniques proposed in chapter 3 to the double-array nose correlations computed in chapter 4, and accordingly reveals the origin of an early spurious phase observed in chapter 4. Chapter 6 discusses several situations that bring ambiguities into the noise-derived seismic signals and can potentially bias the noise-based imaging of subsurface structure. The last chapter provides a summarization over the contributions of this thesis and an outlook of several ongoing and prospected works.