Advancing Correlation Methods of Earthquake Coda in Seismic Body Wave Studies

The discovery of long-range spatial correlation in earthquake coda and ambient noise records has had far-reaching impacts on modern seismology. This approach provides a conceptual framework to extract information about the structure of the Earth from stacked cross-correlations as a function of inter-receiver distance (i.e., correlograms) using noisy seismic records. Early work concentrated on the retrieval of surface waves travelling between receiver pairs from ambient noise and used measurements of their dispersion in surface wave tomography applications. Later it was found that body-wave-like signals also emerge in correlograms. Intriguingly, many feeble signals that are sensitive to deep Earth structures can also be 'extracted' efficiently from late coda of large earthquakes. Yet, the nature of correlation features in coda correlograms has remained a puzzle. Disentangling that puzzle is a key to open new avenues for deep Earth studies. This thesis presents a compilation of published works featuring methodological and theoretical advances for exploiting correlation methods for earthquake coda. The fundamental objectives include (i) improving data processing, (ii ) understanding of the nature and generation mechanism of body-wave-like correlation features, (iii) applying new understanding to study Earth structures. In particular, the thesis contributes to local, regional and global scale problems. At the local and regional scales, contributions are made to the development of the teleseismic P wave coda autocorrelation method for imaging near-surface structures. At the global scale, a new concept the global correlation wavefield, which is relevant to studies of the deep Earth, is introduced. Seismic waves from distant earthquakes illuminate stratified structures beneath a receiver in a nearly vertical fashion. Stacked autocorrelations of the portion of ground motion records immediately after the P wave arrivals (i.e., the P wave coda) are constructed to extract reflection signals from shallow subsurface discontinuities beneath individual stations. The processing incorporates a spectral whitening operation to enhance the quality of autocorrelation results. The feasibility of the method is demonstrated through synthetic and field data. The improved procedure is then applied to a large number of seismic receivers across Antarctica to obtain estimates of the thickness and the ratio of P to S wave speed for the ice cover. These estimates are in a good agreement with prior measurements mainly obtained from active seismic and radar methods. This successful application for the Antarctic ice cover proves the potential of the method for studying other shallow structures including sediment, regolith or ice covers in other continents and on other planets in future space missions. Using the insights gained from the autocorrelation case study, the cross correlation of the late coda of large earthquakes recorded at seismic receivers around the globe is exploited. The global correlograms reveal a wealth of correlation features, including correlation phases that have timing properties similar to regular seismic phases from a surface source, and others that do not have counterparts in the direct wavefield. All features in the correlation wavefield are produced from the similarity of the waveforms of two regular seismic phases sharing a subset of propagation legs. These novel insights are then used to identify the presence of J (shear) waves in the Earth's inner core. These shear wave signals are direct evidence to confirm the solidity of the Earth deepest shell. To match the character of the observed correlograms a 2.5% reduction of inner core shear wave speeds relative to spherically symmetric models of the Earth is inferred. With further refinements, the correlation wavefield approach could lead to improved structural constraints on the Earth interior, especially for poorly sampled regions such as the lowermost mantle and core