Towards Late-Coda Correlation Tomography

The global correlation wavefield is a mathematical manifestation of the seismic wavefield, and we have recently shown that all prominent features (including non-causal ones) are formed due to the similarity between multiple pairs of seismic phases. This new paradigm sets a stage for extracting new and valuable information about Earth's internal structure by means of seismic imaging. Therefore, we are motivated to devise a new and accurate method for the late-coda correlation tomography. Ambient-noise correlation tomography has been rigorously studied for the theory and methods of inter-receiver Earth-structure reconstruction. However, earthquake-coda correlation has a fundamentally different physical mechanism from ambient noise in that it is dominated by the similarity of various pairs of body waves that sample the Earth's interior in a complex way. Here, we propose a framework for the late-coda correlation tomography. The relevant signals in the correlograms are first decomposed into discrete constituents, and we identify a unique origin of each constituent formed by the similarity between specific seismic phases. They are then selected and related to Earth structure to build sensitivity kernels for tomography. Our theoretical framework is verified via a toy-problem recovery experiment. We compare our new travel-time tomography method with a conventional tomography based on inter-receiver reconstruction. This illustrates how significant differences can arise in the interpretation of results. We present a way forward to further detailed and application-oriented method improvements and exploitation of the late-coda correlation tomography.