Global Centroid Moment Tensor Inversion in a Heterogeneous Earth

Abstract

Earthquakes have fascinated humans since the dawn of humanity because they remind us of Earth’s dynamic nature. They reshape Earth’s landscape and pose a risk of significant destruction in tectonically active regions. During the last century, seismology, the scientific study of earthquakes, has helped us to delineate a critical component of our dynamic planet, the tectonic plates, and partially demystify the origin of natural earthquakes, which is a stress-release mechanism of interlocked, moving tectonic plates. In this thesis, I investigate whether we can improve our current understanding of global seismicity by improving an earthquake parameter representation known as the centroid moment tensor. The dissertation starts with a gentle introduction to the concept of a centroid moment tensor and why we might be interested in studying it. We continue by introducing the historical cataloguing of moment tensors as part of the Global Centroid Moment Tensor Project, key concerns in the catalogued parameters, and how the three-dimensional modelling of earthquake wave propagation may help us remove concerns. After the introduction, we present our systematic approach to improving global earthquake parameters by optimising the parameters of 9,382 globally distributed earthquakes. While the results crystallize the need for three-dimensionally modelled seismograms in the source inversion process, it remains a computationally costly problem. We continue by delineating how to overcome this computational challenge by implementing the first global database of seismograms modelled in a heterogeneous Earth. To demonstrate the power of this database, we repeat the above optimization, however, without a limitation on the number of iterations or model parameters. We find that a large number of earthquakes have a larger double-couple component after inversion using three-dimensionally-computed seismograms, meaning that approximate forward modelling methods introduce anomalous components to the focal mechanism. The database allows us to continue working on another part of the source, the rupture history or source time function. In this last part of the dissertation, we introduce a new method to invert the optimal source time function of major and great earthquakes. The results show an overall reduction in the scalar moment and better waveform fits, particularly for earthquakes with complex ruptures.