8413 research outputs found
Sort by
A Unified Unsupervised Anomaly Detection Framework with Score-based Generative Modeling for Multivariate Time Series
No full textThe challenge in unsupervised anomaly detection is the unknown nature of anomalous data points. This task requires the identification of abnormal patterns within the data, even when we lack prior knowledge about what those anomalous patterns might explicitly suggest. Existing unsupervised anomaly detection methods have attempted to address this issue by focusing on limited aspects of the overall problem. These methods can be broadly categorized into four main approaches: reconstruction-based, density estimation-based, boundary description-based, and explicit data characteristic modeling-based. Although each of these methodologies has its own advantages, they are also limited by inherent weaknesses that restrict their effectiveness yielding to sub-optimal results. In this research, I present a novel methodological framework, Unified Unsupervised Anomaly Detection (U2AD), that comprehensively addresses the problem of anomaly detection in multivariate time series. This approach provides a deeper understanding of anomalies within the data distribution space while elucidating the dynamics of non-anomalous data. The framework integrates previous techniques while offering a fresh, holistic perspective. This allows for the creation of customized solutions for various applications by increasing adaptability in selecting appropriate components needed for accurate, robust, and efficient anomaly detection in multivariate time series. Utilizing score-based generative modeling in conjunction with reengineered time-dependent score network and novel training objectives further enhances comprehension of anomalies. Additionally, reconstruction is achieved through the sampling method with deterministic numerical ordinary differential equation solver. Extensive experiments demonstrate that this methodology not only improves anomaly detection precision but also identifies anomalies at earlier stages than current state-of-the-art methods
Lab- and Pilot-Scale Sulfate-Reducing Bioreactors Treating Acid Mine Drainage from an Abandoned Nevada Gold Mine
No full textAcid mine drainage (AMD) is a global environmental hazard that is produced when water flows over exposed rocks containing sulfur-bearing minerals. AMD has many sources and is produced by active and abandoned mines, leading to a need for effective, low maintenance, and low-cost remediation solutions to prevent the contamination of groundwater, the corrosion of infrastructure, and the disruption of reproduction and growth cycles in plants and animals. Sulfate-reducing bioreactors (SRBR) are currently one of the most cost-effective and low-maintenance solutions for treating AMD. Anaerobic sulfate reducing bacteria (SRB) use sulfate as a terminal electron acceptor for their metabolism to create H2S, provide energy for the bacteria, and lower the pH of the water, resulting in the precipitation of metal sulfides. In this work, four pilot-scale SRBR were installed as in-field bioreactors in Perry Canyon, NV near the abandoned Jones-Kincaid adit. In parallel, eight lab-scale SRBRs were operated at the University of Nevada, Reno. Each SRBR contained organic substrate (corn stover, pine shavings, and dairy manure), pea gravel to maintain porosity, and a microbial inoculum. The inoculum was obtained from either the anoxic soil of a nearby lake environment (the Sparks Marina) or from the AMD-impacted ephemeral stream of Perry Canyon, with both demonstrating high sulfate-reducing performance in past work. The pilot-scale SRBR were fabricated as 115-L upflow drums and were installed below ground to modulate environmental conditions and were fed AMD directly from the adit. The lab-scale SRBR were 2-L upflow columns fed synthetic AMD mimicking the Perry Canyon AMD composition. Sulfate and metals concentrations of feed and effluent from each SRBR were monitored temporally. Although results to date indicate limited sulfate reduction occurring in the field-scale SRBR during the first six months of the in-field operation, corresponding to October through April, the reactors inoculated with the marina soil had slightly greater sulfate reduction than those inoculated with AMD-impacted soil. The colder temperature averages (0 to 10 °C) in these months were hypothesized to be responsible for the lower performance and preventing the microbial community from becoming established in the bioreactors, which led to minimal or no AMD treatment even after the temperatures increased during the summer months. The lab-scale SRBR operated in parallel were found to have high sulfate removal over the first 5 months, but then started to decrease as the acidity of the effluent also decreased. This change can be linked to the consumption of limestone sand leading to a decrease in pH and a corresponding inhibiting the SRB that gradually decreases sulfate reduction and bacterial pH neutralization. The concurrent operation of field- and lab-scale SRBR in this work provide valuable knowledge of the scale-up process of the treatment technology, as well as insight into how environmental operating conditions may impact the SRBR performance. If designed and operated properly, SRBR have the potential to be a cost-effective option for AMD remediation at locations around the world
Advancing Observational and Modeling Capabilities of Meltwater – Firn Interactions in Ice Sheet Percolation Zones
No full textApproximately 90% of the Greenland Ice Sheet and 99% of the Antarctic Ice Sheet are covered with firn, the intermediate material between fresh snow and glacial ice. The firn layer consists of interconnected, air-filled pore spaces and ranges from several to hundreds of meters thick. Many glaciological research applications require an understanding of firn structure, including monitoring changes in ice sheet mass balance, interpreting ice core paleoclimate records, and understanding ice sheet hydrology. Constraining how firn structure affects englacial hydrology is critical to quantifying firn’s meltwater storage capacity. This is especially crucial given that the Greenland and Antarctic ice sheets are the largest potential contributors to future sea level rise, but the rate and timing of their contribution are highly uncertain. Meltwater retention in porous firn delays runoff from ice sheets and mitigates mass loss. Therefore, understanding how liquid water originating from surface melt interacts with firn structure as it flows through open pore space is necessary for determining the fate of meltwater on ice sheets. While macrostructural properties such as firn density or porosity are relatively easy to measure in the field or laboratory settings, grain-scale properties that greatly influence firn’s hydraulic properties are morechallenging to describe. In this dissertation, I use a combination of laboratory observations, field data, climate reanalyses, and modeling to understand the interaction between macroscale and microscale firn properties and meltwater flow. I statistically demonstrate that firn macrostructure (density) is a poor indicator of firn microstructure (Chapter 2). I use grain size data from firn cores collected in Greenland and modeling to reveal dynamic feedback mechanisms between meltwater percolation and firn microstructure (Chapter 3). Because measuring firn grain size is tedious, time-consuming, and possibly subjective, I adapt a near-infrared hyperspectral imaging system in a cold laboratory to systematically measure grain size and develop ice layer distributions from firn cores (Chapter 4). Lastly, I develop a premelting parameterization in a one-dimensional firn hydrology model that can be used to explain observations of subfreezing liquid water in firn (Chapter 5). Together, these studies improve our understanding of meltwater fate and transport in firn on ice sheets, which is integral to assess the potential future sea level contribution from ice sheets
Dietary Impact on Potential Renal Acid Load in Chronic Kidney Disease
No full textChronic kidney disease (CKD) stands out as one of the notable chronic health conditions globally, having an overall profound impact on public health. There are several established methods for mitigating the burden of this disease, one of them being through dietary intake. Traditional renal diets have remained the same for decades, with variable updates to adjust for advancements in research and modern approaches. An emerging area of particular interest is the acid-alkaline balance of the diet and how it impacts metabolic acidosis. This has been linked to a variety of health outcomes including hyperkalemia, hyperphosphatemia, osteoporosis, malnutrition and even worsening of CKD. Preliminary research suggests that a low-acid diet may be beneficial, but this dietary pattern is inconsistent with many of the traditional renal diet recommendations. This thesis investigates the traditional renal diet, how it came to fruition, and how it related to diet acid load. Of particular interest, people with kidney failure undergoing dialysis treatment have been generally advised to consume higher amounts of animal-based protein foods, which are major contributors of diet acid load, and limit high-potassium fruit and vegetables, which are major sources of base-forming compounds in the diet. To gain a better understanding of adherence to the traditional renal diet, and how it may contribute to metabolic acidosis in the dialysis population, people undergoing maintenance hemodialysis (HD) treatment were asked to complete 3-day food records. These were analyzed for dietary intake in regards to specific nutrients and food groups. Overall the participants were found to have acidic diets, with protein being the main contributor. However, suggesting a diet that lowers acid load by reducing protein contradicts the conventional renal recommendations, opening a discussion for revision on dietary management among CKD patients
Study of Pelvic Growth and Development Through Examination of Interlandmark Distances and Geometric Morphometric Analyses: Implications for Subadult Skeletal Sex and Age Estimation
No full textThe human pelvis is extensively relied upon by forensic anthropologists for age and sex estimations of unidentified skeletal remains. However, because subadult individuals are an underrepresented demographic in skeletal collections there is a lack of ontogenetic knowledge of the sub-elements (ilium, ischium, and pubis) and the overall pelvic complex. Subsequently, there is a concomitant lack of understanding of the onset of sexually dimorphism and/or age-related size and shape changes. The purpose of this research is to provide a detailed analysis of modern human pelvic growth using in situ pelvic landmarks, which can yield size (interlandmark distances) and shape (geometric morphometric) variables to understand broad ontogenetic patterns of the pelvis and more specific patterns associated with age and sex estimation of the pelvis. It is expected that the size and shape of the pelvis stabilizes at a younger age than previously assumed and subsequently the pelvic phenotype has an increased utility in the subadult biological profile, specifically age and sex estimation. The current sample was queried from the United States sample of the Subadult Virtual Anthropology Database (SVAD) and the full body CT scans were generated prior to autopsy at the University of New Mexico Health Sciences Center, Office of the Medical Investigator between the years of 2010 to 2017 and are available via the New Mexico Decedent Image Database (NMDID). The final sample size was 807 individuals (females=324; males=483) with ages between birth and 21 years. Both chronological age and developmentally derived life history stages (LHS) were used in the analyses, which facilitates an ontogenetic framework to be used when evaluating growth patterns. The developmental LHS included infancy, childhood, juvenile, adolescent, and adult and were determined by dental and skeletal markers. Thirty-four pelvic landmarks on the left and right segmented surface models lead to 68 landmarks per individual. Unilateral and bilateral interlandmark distances (ILDs) were used in sex-specific and pooled-sex nonlinear regression models, including Loess regression and Multivariate Adaptive Regression Splines (MARS). Geometric morphometric analyses were performed to understand shape variation through ontogeny. The unilateral ILDs were tested for the utility of pelvic metrics in skeletal sex estimation using two distinct methodologies—linear discriminant function analysis (DFA) and random forest modeling (RF). Lastly, the utility of pelvic metrics for skeletal age estimation were assessed using the Mixed Cumulative Probit (MCP).
Both unilateral and bilateral measurements demonstrated sexually dimorphic growth in the dimensions associated with the true pelvis. The lack of dimorphic size relating to the false pelvis region and anterior-posterior measurements suggest these should be the focus of age estimation methods, particularly prior to fusion of the sub-elements or juvenile LHS. The vertical acetabular diameter is the earliest variable to stabilize in size and to show sexual dimorphism (~8 years of age). This suggests the biomechanical requirements for locomotion through stabilization of the hip joint is prioritized during early pelvic growth, followed by changes to the pelvic midplane and outlet relating to needs for sexual reproduction—particularly transverse expansion of the true pelvis (birth canal).
Although DFA models produced higher accuracies for skeletal sex estimation than RF models, the results are potentially invalid due to the violation of assumptions relating to normality and homogeneity of variance. The results of 85-87% accuracy achieved by the RF models are valid and still meet the 85% criteria proposed in forensic anthropology. Pelvic metrics can yield accurate and reliable skeletal sex estimations for individuals within the juvenile, adolescent, or adult LHS such that individuals potentially as young as 5 to 7 years of age can be included for accurate sex estimation.
Measurements of length across the unfused ilium, pubis, and ischium, and iliac breadth produce the most accurate and precise age estimations using the MCP. Age estimation using univariate pelvic metrics should be assessed prior to acetabular fusion, which is prior to size stabilization and the onset of sexual dimorphism both of which impact the precision of age estimates. If forensic practitioners are presented with an unfused pelvis, they can proceed with an age estimation using pelvic metrics and applying univariate MCP models. If there is active or complete fusion of the acetabulum, skeletal sex estimation using random forest model can be accurately performed