as one of three explanations of a prominenttine delay in the 6.5 km/s branch of their seismic refraction survey in the Rocky Mountain Trench, Eennett et al (1975) suggested a high-angle crustal fault crossing the trench near Radium. If the density contrast between basement and cover rocks is 0.1 g/cm3, a gravity anomaly of approximately 18 mgal should be observed. To test the fault hypothesis, a gravity survey has been carried out in and adjacent to the trench in the Radium area. The resultant data are not consistent with the proposed fault model. The principal feature of the data is a pronounced low which coincides with the trench throughout the survey area. The low is due to Cenozoic fill and interpretation by two-dimensional modeling indicates the thickness of fill is about 550 m to the north and 420 m to the south of Radium. An analysis has also been performed of the shear-wave data recorded during the seismic survey of Bennett et al (1975). Although the quality of the S save data is poor, they show consistent behavior with the P save data. There is weak evidence suggesting a basement refractor velocity of 3.5 km/s and a Moho refractor velocity of 4.2-4.5 km/s. The corresponding Poisscn's ratios are 0.30 and 0.28-0.32. To determine maximum and minimum depth limits to the Hcho allowed by the seismic data, an extremal analysis was performed on both the P and S wave record sections. From the P wave data, the limits on crustal thickness beneath the Rocky Mountain Trench are 52 km and 60 km; from the S wave data, the limits are 47 km and 59 km. Is a result of these additional studies, the tao alternative hypotheses of Bennett et al (1975) to explain the seismic data must be reconsidered. These are (1) the existence of a crustal low velocity zone and (2) a major deformation of the basement and overlying rocks due to the trench being an ancient zone of weakness which coincides with the western limit of the continental Precambrian craton. As reflections from the top of the low velocity zone are not observed by Bennett et al (1975), the second alternative is preferred. / Science, Faculty of / Earth, Ocean and Atmospheric Sciences, Department of / Graduate

Earthquake detection is an important task, focusing on detecting seismic events in past data or in real time from seismic time series. In the past few decades, due to the increasing amount of available seismic data, research in seismic event detection shows remarkable success using neural networks and other machine learning techniques. However, creating high quality labeled data sets is still a manual process that demands tremendous amount of time and expert knowledge, and is stifling big data innovation. When compiling a data set, it is unclear how many earthquakes and noise are mislabeled. Another challenge is how to promote the general applicability of the machine learning based models to different geographical regions. The models trained by data sets from one location should be applicable to the detection at other locations. This thesis explores the most popular deep learning model, convolutional neural networks (CNN), to build a single location detection model. In addition, we build more robust generalized earthquake detection models using transfer learning and meta learning. We also introduce a process for generating high quality labeled datasets. Our technique achieves high detection accuracy even on low signal to noise ratio events. The AI techniques explored in this research have potential to be transferred to other domains that utilize signal processing. There are a myriad of potential applications, with audio processing probably being one of the most directly relevant. Any field that deals with waveforms (e.g. seismic, audio, light) can utilize the developed techniques.

In this thesis I combine data from 29 volcanic earthquake swarms that follow the pattern predicted by the Generic Volcanic Earthquake Swarm Model (GVESM; Benoit and McNutt, 1996) to investigate whether the relative timing of various parameters of pre-eruptive volcanic earthquake swarms could be used to forecast the time of an impending eruption. Based on the analysis of seismic unrest preceding many eruptions, the GVESM suggests that it is common to see an increase first in high-frequency earthquakes, then low-frequency earthquakes, then the onset of volcanic tremor. While this pattern is useful to volcano-seismologists, the relative timing and durations of these three different types of volcanic seismicity, is explored here for the first time. The parameters investigated are the onset times of (i) low-frequency (LF) events and of (ii) tremor, and the time at which (iii) the peak rate (PR) of volcano-tectonic (VT) events and (iv) the maximum magnitude (MM) earthquake occur in relation to normalized time defined by swarm onset and end (i.e., eruption). The normalized time starts at the swarm onset (0%) and ends with the eruption (100%) allowing a comparison and joint consideration of parameter occurrences across swarms of different actual duration. We identify the normalized onset time of for each parameter (LF, tremor, PR, MM) with respect to the duration of each swarm. Each swarm has onset time uncertainties of the swarm itself and of its parameters. A swarm with large onset uncertainty could bias the normalized onset time of each parameter and we use weighted means to decrease the influence of swarms with large uncertainties on overall results. The weighted means of LF onset, tremor onset, MM and PR occurrence are 79% ± 23%, 96% ± 10%, 78% ± 29% and 75% ± 34%, respectively. Errors are the standard deviation of each parameter. The uncertainties for LF, MM and PR are large because their normalized onset times have the characteristics of a uniform distribution and therefore seem to have no predictive value. In contrast, tremor onset has a narrow distribution towards the end of swarms. A possible tremor mechanism consistent with this observation could be boiling of groundwater as magma nears the surface. LF onset always seems to precede tremor onset. LF and tremor start early (at less than 80% of normalized time) at five volcanoes with high SiO2 content possibly related to lower density and higher gas content of the resulting magma.

Neural networks are extremely versatile tools, as evidenced by their widespread adoption into many fields in the sciences and beyond, including the geosciences. In seismology neural networks have been primarily used to automatically detect and discriminate seismic signals within time-series data, as well as provide location estimates for their sources. However, as neural network research has significantly progressed over the past three decades, so too have its applications in seismology. Such applications now include earthquake early warning systems based on smartphone data collected from large numbers of users, the prediction of peak ground acceleration from earthquake source parameters, the efficient computation of synthetic seismograms, providing probabilistic estimates of solutions to geophysical inverse problems, and many others. This thesis contains three components, each of which explore novel uses of neural networks in seismology. In the first component, a previously established earthquake detection and location method is supplemented with a neural network in order to automate the detection process. The detection procedure is then applied to a large volume of seismic data. In addition to automating the detection process, the neural network removes the need for several user-defined thresholds, subjective criteria otherwise necessary for the method. In the second component, a novel approach is developed for inverting seafloor compliance data recorded by ocean-bottom seismometers for the shallow shear-wave velocity structure of oceanic tectonic plates. The approach makes use of mixture density networks, a type of neural network designed to provide probabilistic estimates of solutions to inverse problems, something that standard neural networks are incapable of. In the final component of this thesis, the mixture density network approach to compliance inversion is applied to a group of ocean-bottom seismometers deployed along the continental shelf of the Cascadia Subduction Zone in order to investigate shelf sediment properties.

Four geophysical methods, along with well logs and outcrop data, were used in determining the location of a fault situated on the campus of Oregon Institute of Technology, just north of Klamath Falls, Oregon. The fault displaces rocks of the Yonna Formation, of Pliocene age. Wells located northeast of the fault (on the upthrown side) produce cold water, and wells located southwest of the fault (on the downthrown side) produce hot water. The purpose of this investigation was to define the characteristics of the fault exposed behind a large water tank southeast of the OIT campus.

The geologic setting of the Portland Basin is ideal for gravity surveys because of the large density contrasts between geologic units. The Portland Basin consists of a north-northwest-trending syncline in the Columbia River basalt overlain by Pliocene to Recent alluvium. This study was undertaken to define structures in the Portland Basin which are obscured by the alluvium. An areal gravity survey of the Portland Basin covering approximately 450 square kilometers was conducted for this study.

A dissertation submitted to the Geophysics Department, School of Geosciences, Faculty of Science, University of the Witwatersrand, Johannesburg, in fulfilment of the requirements for the degree of Masters of Science. Johannesburg, February 2012 / Impala Platinum Mine (Impala), situated north of the town of Rustenburg in the North West Province of South Africa, has experienced an increase in seismicity from ~841 seismic events in the year 2005 to ~1588 seismic events in 2008. The seismologists and rock engineers need to understand the underlying mechanisms and driving forces responsible for seismicity to develop and design mining layouts and support strategies to lessen the risks posed by rockburts. However, most previous studies of seismicity conducted on Impala and other Bushveld Complex mines in the Rustenburg area provided limited information regarding the source parameters and mechanism due to insufficient data. The study is designed to investigate the seismic hazard on Impala Platinum Mine by means of two approaches: an investigation of seismic source parameters and the mechanism of potentially damaging seismic events, and mapping of the weathered layer of the near surface within the Impala mine lease area. A number of detailed investigations of rockbursts were conducted whereby damage was mapped and photographed. The investigations includes reviews of the seismic history, short-, medium- and long-term seismic hazard assessment methods, and an analysis of the source parameters of the seismic event and associated ground motions. The study has revealed that most of the seismic events occur close to the reef plane, and are the result of the failure of a volume of rock that includes the pillar and the host rock that forms the foundation of the pillar.

p.p1 {margin: 0.0px 0.0px 0.0px 0.0px; font: 12.0px 'Times New Roman'} A methodology is presented for evaluating the potential surface fault displacement on reverse faults in a probabilistic manner. This methodology follows the procedures put forth for Probabilistic Fault Displacement Hazard Analysis (PFDHA). Empirical probability distributions that are central to performing a PFDHA are derived from field investigations of reverse faulting events. Statistical analyses are used to test previously assumed properties of scale invariance with respect to magnitude for normalized displacement. It is found that normalized displacement is statistically invariant with respect to magnitude and focal mechanism, allowing for the combination of a large number of events into a single dataset for regression purposes. An empirical relationship is developed using this single dataset to be used as a fault displacement prediction equation. A PFDHA is conducted on the Los Osos fault zone in central California and a hazard curve for fault displacement is produced. A full sensitivity analysis is done using this fault as a reference, to test for the sources of variability in the PFDHA methodology. The influence of the major primary variables is quantified to provide a future direction for PFDHA.

Catalogs of seismic source parameters (hypocenter locations, origin times, and magnitudes) are vital for studying various Earth processes, greatly enhancing our understanding of the nature of seismic events, the structure of the Earth, and the dynamics of fault systems. Modern seismic analyses utilize supervised machine learning (ML) to build enhanced catalogs based on millions of examples of analyst-picked phase-arrivals in waveforms, yet the ability to characterize the time-varying spectral content of the waveforms underlying those catalogs remains lacking. Unsupervised machine learning (UML) methods provide powerful tools for inferring patterns from musical spectrograms with little a priori information, yet has been relatively underutilized in the field of seismology. In this thesis, I leverage advanced tools from UML to analyze the temporal spectral content of large sets of spectrograms generated by different mechanisms in two distinct geologic settings: icequakes and tremors at Gornergletscher (a Swiss temperate glacier) and repeating earthquakes from a 10-km-long creeping segment of the San Andreas Fault. The core algorithm in this work, now known as Spectral Unsupervised Feature Extraction, or SpecUFEx, extracts time-varying frequency patterns from spectrograms and reduces them into low-dimensionality fingerprints via a combination of non-negative matrix factorization and hidden Markov Modeling (Holtzman et al. 2018), optimized for large data sets via stochastic variational inference. This work describes the SpecUFEx algorithm and the suite of preprocessing, clustering, and visualization tools developed to create an UML workflow, SpecUFEx+, that is widely-accessible and applicable for many seismic settings. I apply theSpecUFEx+ workflow to single- and multi-station seismic data from Gornergletscher, and demonstrate how some fingerprint-clusters track diurnal tremor related to subglacial water flow, while others correspond to the onset of the subglacial and englacial components of a glacial lake outburst flood. I also discover periods of harmonic tremor localized near the ice-bed interface that may be related to glacial stick-slip sliding. I additionally apply the SpecUFEx+ workflow to earthquakes on the San Andreas Fault to unveil far more repeating earthquake sequences than previously inferred, leading to enhanced slip-rate estimates at seismogenic depths and providing a more detailed image of seismic gaps along the fault interface. Unsupervised feature extraction is a novel tool to the field of seismology. This work demonstrates how scientific insight can be gained through the characterization of the spectral-temporal patterns of large seismic datasets within an UML-framework.

In this thesis, I analyze seismic signals collected during two passive-source broadband seismic deployments that instrumented tectonic boundaries with opposing plate motion—the heavily sedimented forearc of the obliquely convergent Indo-Burman subduction zone and the Malawi rift of the divergent East African rift system—as part of the BIMA and SEGMeNT experiments. These two settings provide unprecedented opportunities to broaden the extent of our understanding of tectonic processes and linkages between atmosphere-to-solid earth seismic coupling, respectively. The Indo-Burman forearc represents an extreme endmember system for sedimentary accretion underneath Earth’s largest delta, while the Malawi rift contains one of the widest and deepest freshwater bodies and one of the first to be instrumented by a seismic array from lake bottom to lake shore. Collectively, this work represents a diverse set of seismic observations that improve our understanding of environmental and tectonic systems across a range of scales, from oblique convergence under heavy sedimentation to energy transfer between the atmosphere and the solid earth. Using the BIMA dataset, we investigate the seismic shear-velocity structure across the extensive sediment blanket, crust, and uppermost mantle of the Indo-Burman forearc margin to robustly constrain subsurface structure and lithology. We construct a comprehensive three- dimensional survey of seismic shear velocity across the region using a joint-inversion of surface- and scattered-wave constraints that explicitly parameterizes key boundary layers. We extract measurements of Rayleigh-wave phase velocities from (1) interstation Rayleigh wavefields produced from the cross-correlation and spectral waveform fitting of ambient seismic noise between 12-25 s period and (2) intra-array Rayleigh-wave phase variations form regional and teleseismic earthquakes propagating across the array between 20-80 s period, in order to constrain absolute shear velocities throughout the model. To constrain the depths to and amplitudes of significant velocity interfaces, we also develop a generalized-Radon-transform migration image across the array and incorporate the resulting scattered-wave measurements into the joint inversion. Together, these measurements complement each other’s individual limitations and allow for a comprehensive modeling analysis. Overall, the Bengal basin appears markedly slower than other heavily sedimented basins observed globally. East-west dispersion variations highlight a deepening slow structure to the east, which suggests a basin geometry primarily controlled by a down-dipping slab interface as opposed to central basin loading. Scattered-wave imaging captures three important interfaces in the velocity architecture underlying the region. Within the joint-inversion modeling, we observe two model classes that emblemize the evolution of consolidation and stress state within the uppermost sediments and metasediments along a predominantly northeast-southwest trend. We interpret variations in deeper seismic structure under two proposed scenarios: (1) a Moho at ~21-26 km underlying a package of metasediments and a thin oceanic crust, with a slow mantle lithosphere that may contain retained melt from the onset of India-Antarctica seafloor spreading; or (2) a Moho at ~50-59 km underlying a package of metasediments and a thick slug of mafic material, which may correspond to significant underplating from the Kerguelen hotspot at the time of creation of the subducting crust. These findings improve our understanding of sediment evolution and tectonic architecture across the Indo-Burman forearc margin. Using the amphibious SEGMeNT data at Lake Malawi, we explore variations in the spectral character of lake-generated microseisms to investigate the dominant parameters controlling seismic coupling between water and the solid earth. We document clear evidence for two spectral peaks in the lake microseism band, and relate variations in spectral behavior as a function of recording depth and proximity to steep lake-floor slopes and shorelines to suggest that these spectral bands may correspond to single- and double-frequency generation processes, akin to primary and secondary ocean microseisms. Some observations are otherwise complex and inconsistent with traditional microseism theory, indicating that signals may alternatively reflect interactions between differing source regions within separate basins of the lake under exclusively double-frequency generation processes, an ambiguity that might have been resolved with the availability of colocated wind and wave-state data sets. This dissertation work highlights the value of array-based seismic deployments and the incorporation of complementary data types for exploring the detailed structure and evolution of systems, especially in high-noise settings.