Doctor of Philosophy, Miami University, 2020, Geology and Environmental Earth Science

Recent studies have suggested that slow slip processes and high pore fluid pressures may have a role in promoting seismogenic events, particularly earthquake swarms. Swarms of seismicity have also been observed to be induced by the injection of fluid, either via hydraulic fracturing or wastewater disposal, such that this can provide a complimentary comparison. Here, we present three chapters that seek to detect and investigate the origin of earthquake swarms. In Chapter 1, we generate a catalog of earthquake swarms in Oaxaca, Mexico and find most events outline a steeply dipping fault in the overriding plate. Examination of GPS data reveals many of these swarms occur during slow slip events, although they occur during episodes of strike slip motion as opposed to thrust motion. This appears to be the first evidence for slow slip behavior on a sliver fault that helps to partition the oblique convergence. Conductivity studies indicate fluids released along the subduction interface may be channeled up this steep sliver fault, leaving the megathrust with drier conditions that could promote traditional fast slip behavior. In Chapter 2, we investigate the increase in seismicity in the Eagle Ford oil and gas field of south Texas and how hydraulic fracturing (HF) contributed. We compare times and locations of HF wells with a catalog of seismicity we enhanced through template matching (2014-2018). Several HF wells have seismicity nearby during operation, indicating seismicity from HF is more common in this area than previously thought. We find that HF strategy affects the probability of earthquakes. A MW 4.0 earthquake is the largest HF‐induced earthquake in the U.S. Thus, this study demonstrates that faults in this area are capable of producing felt and potentially damaging earthquakes due to ongoing HF. In Chapter 3, we seek to perform a deeper exploration of how HF has contributed to recent seismicity using template matching with newly deployed stations and a repeating sig (open full item for complete abstract)

Committee: Michael Brudzinski PhD (Advisor); Brian Currie PhD (Committee Member); Elizabeth Widom PhD (Committee Member); Jonathan Levy PhD (Committee Member); Aaron Velasco PhD (Committee Member) Subjects: Geology; Geophysics

MS, University of Cincinnati, 2017, Engineering and Applied Science: Computer Science

Supervised document classification technique, proposes a model that is trained with a training set with fixed number of labeled classes and can be effectively used to classify documents under one of these labels in the test set. The major objective of our research was to identify text documents from labels or topics which are not present in the training set, yet appeared in the test set. We devised a method to identify and eliminate documents from such labels/topics that do not occur in the training set. This technique brings together the idea of template matching and document classification by creating characteristic signatures that are unique to each label in the training set. Using these signatures any unknown label could be detected and ignored in the test data-set. Our results clearly show that, these techniques are useful in classification of documents into known categories or labels, as well as identifying labels that don't match with the predefined labels in training set.

Committee: Raj Bhatnagar Ph.D. (Committee Chair); Ali Minai Ph.D. (Committee Member); Shomir Wilson Ph.D. (Committee Member) Subjects: Computer Science

Master of Science, Miami University, 2017, Geology and Environmental Earth Science

North Texas has seen an increase in seismic activity around the Dallas/Fort Worth area since the early 2000's, with activity in Johnson County in particular culminating in magnitude 3 and 4 events in 2011 and 2015 respectively. Previous analysis of the Johnson County sequence between 2009 and 2011 concluded that many of the events were induced by wastewater injection (Frohlich, 2012), however the earthquake database was small during this time period, and the differences between inducing and non-inducing injection wells were not clearly identified. This study addresses the causes of recent seismicity in Johnson County through an in depth characterization of the seismicity, industry operations, and regional and local geology in North Texas from 2009 to 2011. Seismic template matching using 3 USArray Transportable Array station recordings of all previously cataloged earthquakes in the study area provide a more complete temporal history of seismicity, identifying 977 additional events. Earthquakes from the largest burst in activity, in June 2011, were relocated using hypoDD and seem to align along NNE-SSW trends consistent with regional stress orientations and pre-existing structures related to the adjacent Ouachita thrust front. Relocated seismicity outlines a fault plane in the Precambrian basement that extends approximately 4 km in vertical extent, and is consistent with the hypothesis that seismicity is occurring on reactivated, pre-existing, critically stressed faults. Monthly injected volumes from 9 wastewater disposal wells suggest a correlation with background levels of seismicity throughout the study timeframe, however they do not correlate with distinct spikes in seismic activity. Temporal patterns of seismicity during the June 2011 sequence resemble patterns seen in previously documented cases of hydraulically fractured induced seismicity in Ohio. While a complete stimulation database is not available from this time frame, the vast number of active hydraulic (open full item for complete abstract)

Committee: Mike Brudzinski (Advisor); Brian Currie (Committee Member); Jonathan Levy (Committee Member) Subjects: Geology; Geophysics

Doctor of Philosophy, Miami University, 2016, Geology and Environmental Earth Science

Relationships between earthquakes are observed by the clustering of seismic events in space and time. This clustering commonly occurs as mainshock-aftershock sequences, which are generally interpreted to contain the initial rupture of a fault (the mainshock) and a decaying cascade of smaller ruptures on or very near to the initial rupture plane (aftershocks). Clustering of earthquakes in space and time can also occur as earthquake swarms, which are empirically defined as an increase in seismicity rate above the background rate without a clear triggering mainshock earthquake. Earthquake swarms are often associated with volcanic regions and are studied because of their relationship to eruptions. Earthquake swarms have also been correlated with subduction zone slow slip events, including a case that led into the 2011 Tohoku earthquake. Earthquake swarms are also well associated with many induced (“human influenced”) earthquake sequences. Understanding the mechanisms that lead to earthquake swarms and the rapid detection of these events are key factor in reducing the hazard posed by these events. Here, we present four chapters that seek to detect and better characterize earthquake swarms with an emphasis on induced seismicity. We develop an efficient template matching algorithm that can be used to improve an earthquake catalog completeness by more than an order of magnitude and apply it throughout the state of Ohio. We also develop a new method, referred to as a Repeating Signal Detector (RSD), that uses agglomerative clustering to group signals of interest according to their temporal and frequency domain characteristics. Resulting signal families can be stacked, improving the signal-to-noise ratio of the recorded signals, and then the signal stack can the used in template matching. We apply the technique to detect earthquake swarms in volcanic, subduction, and induced seismicity settings throughout North America. In each case, RSD duplicates or improves upon existing c (open full item for complete abstract)

Committee: Michael Brudzinski PhD (Advisor); Brian Currie PhD (Committee Member); Jens Mueller PhD (Committee Member); Jonathan Levy PhD (Committee Member); Jacob Walter PhD (Committee Member) Subjects: Geology; Geophysics

Doctor of Philosophy, Case Western Reserve University, 2016, EECS - Computer and Information Sciences

Providing an efficient and expressive querying technique for graph structured RDF data is an emergent problem as large amounts of RDF data are available from applications in many areas. Current techniques do not fully satisfy this goal due to the nature of the RDF model which requires highly flexible use of keywords, and a structure expression in query language. Viewing RDF as graphs requires additional graph-based functionalities, such as querying a path or a tree connection. We propose a querying framework, called RDF-h, which uses the query template as a basic query unit, and supports both partially entered keywords and query conditions based on graph-structure. In order to provide efficient query evaluation, signature-based index is utilized. Though most existing techniques which utilize signature-based index claim its benefits on all datasets and queries. The effectiveness of signature-based pruning varies greatly among different RDF datasets and highly related with their dataset characteristics. The performance benefits from signature-based pruning depend not only on the size of the RDF graphs, but also the underlying graph structure and the complexity of queries. We propose several dataset evaluation metrics, namely, coverage and coherence, relationship specialty and literal diversity to understand the query performance differences among real and synthetic RDF datasets. Based on these results, we further propose an application-specific framework, called RBench, to generate RDF benchmarks. By evaluating the characteristics of RDF datasets and the complexity of query templates, RDF-h selectively utilizes signature-based pruning when it is considered to be beneficial. Two aspects of RDF-h framework are evaluated in experiments: 1. extensive query performance evaluation based on randomly generated queries for different datasets; 2. utilization of RDF-h for biomedical applications. For random query evaluation, the RDF-h algorithm can automatically capture freque (open full item for complete abstract)

Committee: Meral Özsoyoglu (Advisor); Gultekin Özsoyoglu (Committee Member); Mehmet Koyutürk (Committee Member); Marc Buchner (Committee Member); Soumya Ray (Committee Member); Andy Podgurski (Committee Member); Xiang Zhang (Committee Member) Subjects: Computer Science

Master of Sciences, Case Western Reserve University, 2016, EECS - Computer and Information Sciences

Query interface is an important tool in accessing graph databases. Traditional text-based query interfaces only provide access to databases through text queries and results, without presenting the internal graph structure. In this thesis we present a graphical query interface for exploratory querying of graph structured data, called Graph by Example (GBE). Our interface introduces a Draw and Play feature, which enables users to query graph structured data intuitively using graph templates. GBE interface also provides exploratory querying features facilitating the editing and reuse of previous queries so that users can reformulate their queries and resubmit based on the results of the previous queries. These features makes the interface simpler and easier to use compared to traditional text based query interfaces. We implement this query interface utilizing an RDF querying framework in a previous research. We also demonstrate the interface's functions and features through examples in real world databases and benchmarks.

Committee: Meral Ozsoyoglu (Advisor); Mehmet Koyuturk (Committee Member); Michael Lewicki (Committee Member) Subjects: Computer Science

Master of Science, Miami University, 2014, Geology and Environmental Earth Science

A series of earthquakes in 2011 near Youngstown, OH has been a focal point for discussions of seismicity induced by nearby wastewater disposal wells. Utilizing an efficient waveform template matching procedure, the optimal correlation template to study the Youngstown sequence was identified by varying parameters such as the stations utilized, frequency passband, and seismogram length. A catalog composed of 566 events was identified between January 2011 and February 2014. Double-difference relocation refines seismicity to a ~800 m linear streak from the Northstar 1 injection well to the WSW along the same strike as the fault plane of the largest event. Our catalog suggests triggering caused by the 2011 M 9.0 Tohoku earthquake indicating that fluid injection brought the Precambrian basement to near-critical stress. Calculated Gutenberg-Richter b-values are consistent with trends observed in other regions with seismicity induced by fluid injection.

Committee: Michael Brudzinski PhD (Advisor); Brian Currie PhD (Committee Member); Jonathan Levy PhD (Committee Member) Subjects: Geological; Geology; Geophysical; Geophysics

Master of Science in Engineering (MSEgr), Wright State University, 2008, Biomedical Engineering

Our challenge was to develop a semi-automatic target detection algorithm to aid human operators in locating potential targets within images. In contrast to currently available methods, our approach is relatively insensitive to image brightness, image contrast and object orientation. Working on overlapping image blocks, we used a sliding difference method of histogram matching. Incrementally sliding the histograms of the known object template and the image region of interest (ROI) together, the sum of absolute histogram differences was calculated. The minimum of the resultant array was stored in the corresponding spatial position of a response surface matrix. Local minima of the response surface suggest possible target locations. Because the template contrast will rarely perfectly match the contrast of the actual image contrast, which can be compromised by illumination conditions, background features, cloud cover, etc., we perform a random contrast manipulation, which we term ‘wobble', on the template histogram. Our results have shown improved object detection with the combination of the sliding histogram difference and wobble.

Committee: Julie Skipper PhD (Advisor); Daniel Repperger PhD (Committee Member); Thomas Hangartner PhD (Committee Member); S. Narayanan PhD (Other); Joseph F. Thomas, Jr. PhD (Other) Subjects: Biomedical Research; Engineering; Scientific Imaging

MS, University of Cincinnati, 2001, Engineering : Computer Science and Engineering

This paper proposes an enhancement to the image matching component of an automatic target recognition system that improves the ability to handle variations and articulations within a given class of targets. This method, called chunking, can be applied to an A.T.R. system, or any image matching system, that uses templates to match against a given input image. Using information theoretical measures, the templates are divided into sub-parts, called chunks. These sub-parts are scored individually against an input image. This adds the ability to distinguish poor scoring areas of the target from those that score well. Given this ability, it can be hypothesized that a certain input image is a variation of the template class if it scores well in all but a few chunks. If a small set of chunks score significantly worse (determined by a measure of the variance) than the others, the scores of the poor-scoring chunks are discarded. The effect of this is to increase the scores of an input image that is of the same class but a different variation than that of the template, but will have little or no effect on the score of an input image that is of another class. The results given will illustrate the effects of this approach. The chunking and matching systems are written entirely in Java 2 (v1.3).

Committee: Dr. Raj Bhatnagara (Advisor) Subjects: Computer Science

Master of Science (MS), Ohio University, 1989, Electrical Engineering & Computer Science (Engineering and Technology)

Parallel implementation of template matching on hypercube array processors

Committee: M. Celenk (Advisor) Subjects:

Master of Sciences (Engineering), Case Western Reserve University, 2009, EECS - Computer and Information Sciences

Accurate localization is one of the core requirements for autonomous vehicle navigation. This thesis presents a localization algorithm that operates on an a priori map represented as a collection of line segments. This method treats the entire map as a solid body template and performs a fit of this template to the available sensory data. Combining an analytic translation-optimization technique with an iterative heading search technique enabled an accurate and computationally efficient method for finding the localization parameters. This algorithm is further extended to deal with singularities of situations of partial observability. Statistical analysis of simulations with synthetic data indicated that the algorithm was able to handle ideal as well as noisy data with good accuracy. Its performance was further evaluated using actual sensory data from LIDAR. LIDAR-based localization proved to be an effective technique for localization a vehicle in complex indoor environments. The thesis also explores the capabilities of stereo vision and describes the initial assessment of replacing LIDAR with stereo vision for localization.

Committee: Wyatt Newman (Committee Chair); Frank Merat (Committee Member); Cenk Cavusoglu (Committee Member) Subjects: Computer Science; Electrical Engineering; Robots

Master of Sciences, Case Western Reserve University, 2009, EECS - Computer Engineering

This thesis presents a novel vision-based localization method that uses fast template-matching techniques with respect to regularly-spaced floor tiles to provide pose information. To accomplish the template matching, an edge map is created, transformed into Hough space and interpreted modulo the periodicity of the template. In this space, offsets relative to the periodicity of the floor tiles can be found. By separately tracking accumulation of periods, a global pose estimate can be determined that is immune to accumulation of incremental errors. The method is shown to be robust with respect to noise and distracter lines, and it also recognizes when a scene analysis is untrustworthy. The method is suitable for integration within a Kalman filter to contribute to improved localization.

Committee: Wyatt Newman (Advisor); M. Cenk Cavusoglu (Committee Member); Francis Merat (Other) Subjects: Artificial Intelligence; Robots