Final Report Summary - WAVETOMO (Imaging earth's internal structure using full waveform tomography) Seismic tomography of the earth's mantle traditionally relies on measurements of travel times of a small number of seismic waves, those that are well separated from others in seismic records. This limits the sharpness of the resulting images, since the distribution of natural earthquakes and recording stations is far from uniform on the earth's globe. Seismic waveform tomography holds the potential of utilizing all of the information contained in a seismogram within a specified frequency band, including scattered waves bouncing off major structural boundaries, but also those that illuminate smaller features, important for improving our understanding of mantle dynamics. It is now possible to compute the theoretical seismic wavefield in spherical geometry and in any given, three-dimensional elastic model, very accurately, using numerical techniques. Applying this to global seismic tomography is promising but extremely heavy computationally, which hinders progress in resolution, because the latter depends on the highest frequency of the computation, while computational time increases as the cube of the frequency.Our approach has been to make progress in our understanding of mantle dynamics through the construction of increasingly sharp tomographic images, while minimizing computational cost. During the WAVETOMO project, we developed a methodology, which we call "box tomography", which allows us to construct high resolution seismic models of a remote target region through an iterative non-linear optimization process, by using appropriately high frequencies, but computing the complete wavefield (from source to receivers) only once for every earthquake considered. During subsequent iterations, the wavefield needs only to be computed within the much smaller target region. Through various tests, we demonstrated that this method works for the case of a remote target near the earth's core mantle boundary. Meanwhile, as a first application to real data, we developed a continental scale tomographic model of the upper mantle under the North American continent using a combination of teleseismic (distant earthquakes) and regional waveform data. During the course of this project, we also developed the first whole mantle elastic tomographic model of the earth, at the global scale, developed using the spectral element method (numerical integration of the equations of motion) for wavefield computations, and the first global upper mantle model of anelastic attenuation based on the same approach. A major finding in our elastic images is the presence of ~25 broad low shear velocity conduits (representing upwelling flow) anchored at the core-mantle boundary and rising quasi-vertically in the geographical vicinity of major hotspot volcanoes. These conduits are wider than expected from conventional mantle flow calculations, suggesting, together with recent findings in mineral physics, that our views of deep mantle rheology may need to be revised. In an independent study, through forward modeling of Sdiff waveforms (shear waves that diffract along the core-mantle boundary)we determined that at least some of the roots of these broad plumes contain large, quasi axi-symmetric ultra-low-velocity zones, zones of extreme seismic properties that indicate the presence of partially molten material.We also collected and processed short period (~1s) waveforms sampling (1) the deep mantle (core reflected waves), for which we adapted a methodology first developed in exploration geophysics to enhance signal-to-noise ratio of these weak phases and separate them from other unrelated energy; (2) the vicinity of the earth's inner-core boundary (ICB, where we showed the presence of a thin zone (<50 km) of reduced velocity and higher attenuation right above the ICB.This project involved 3 PhDs, 2 Masters' students, 5 post-docs and one more senior contract researcher. It resulted in 21 publications, plus 4 that are submitted or in preparation.
