Publication Date


Type of Culminating Activity


Degree Title

Doctor of Philosophy in Geophysics



Major Advisor

Paul Michaels


Ambient seismic noise techniques are an excellent choice for imaging the subsurface in areas that are seismically quiet or otherwise unsuitable for active source experiments due to geographic isolation or environmental sensitivity. Recently, decades-long time series were made available for download from the Incorporated Research Institutions for Seismology (IRIS) from permanent network installations, allowing access to long, uninterrupted recordings from seismometers around the world. This has spurred the development of an entire field of applications for passive seismic noise analysis. Over the continental United States, the USArray project has advanced to provide station coverage in relatively dense and regularly spaced arrays, but along the Aleutian Island arc in Alaska and other geographically isolated but seismically active locations, the hazards associated with volcanic eruptions and the difficulty of accessing stations for repair or replacement throughout most of the year has allowed only for sparse coverage.

The analysis of ambient seismic recordings generally suits one of two purposes. The first involves the parameterization of the source of each component of the ambient seismic noise spectrum and focuses on both the spatial locations and mechanisms of generation. The second purpose of looking at ambient seismic noise is to create velocity models of the subsurface below the array. The assumptions required for the traditional approach to analysis of ambient seismic noise, namely beamforming and the spatial autocorrelation coefficient method, involve specific requirements for array geometry and density that often are not met when only limited permanent network coverage is available. Furthermore, most studies focused on mapping the subsurface generally assume that the source of ambient noise is diffuse and azimuthally homogeneous. Here, principal component analysis (PCA) is used to show that the source of ambient noise is often highly directive. Methods of incorporating the direction of approach are introduced to allow one to correct the apparent velocity calculated by the consideration of an omnidirectional source when the energy is instead strongly directive. To ensure that the energy measured at a remote network is indeed pervasive energy that travels through the ground rather than local noise, a quality control algorithm based on the results of PCA is also introduced. The benefits of improving the reliability of velocity measurements for the subsurface and reducing the size of the dataset by the introduction of the quality control parameter will greatly enhance the accuracy and ease with which scientists may determine subsurface velocity profiles at especially sparse networks with particularly long recording times.