79717 research outputs found
Sort by
Drivers and Impacts of a Recent Annual Grass Invasion: Ventenata dubia and Fire in the Inland Northwest
Get PDFBiological invasions threaten native biodiversity, alter ecosystem function, and are a major cause of economic losses across the planet. The most impactful invaders alter disturbance regimes and initiate state shifts to outside the historical range of variability of the ecosystem. Concern for ecological and economic losses has prompted a rapid expansion of invasion ecology research. However, the continual arrival of new invaders with unknown ecological impacts demands further research to help close the ever-growing knowledge gap. In the Pacific Northwest, a recently introduced, rapidly spreading Eurasian annual grass, Ventenata dubia (ventenata) is poised to alter fire behavior and ecosystem function across forest-mosaic landscapes of the Inland Northwest, USA. This dissertation aims to: 1) determine the biotic and abiotic factors associated with the V. dubia invasion, 2) characterize the relationship between invasion and plant community diversity in burned and unburned areas, 3) examine how biotic and environmental factors interact to influence community invasion resistance, and 4) evaluate the influence of V. dubia on fuel characteristics and fire behavior at multiple scales.
I used field data, statistical analyses, and landscape fire simulations to determine the drivers and impacts of the V. dubia invasion at community and landscape-scales in the Blue Mountains Ecoregion of the Inland Northwest. In Chapter 2, I identified V. dubia’s unique niche in forested ecosystems of the region, including historically invasion and fire-resistant dwarf shrublands imbedded within the larger forested landscape. I demonstrated that V. dubia expands invasion impacts in these ecosystems rather than occurring in areas already impacted by other invasive annual grasses (Bromus tectorum and Taeniatherum caput-medusae), increasing the overall invasion footprint. Chapter 2 also examined the relationship between V. dubia and plant community diversity with and without fire. I found that V. dubia was weakly related to community diversity in unburned areas but was strongly negatively related to diversity and abundance of functionally similar species in burned areas. These results suggest that V. dubia may fill an otherwise seemingly unoccupied niche in unburned areas but may outcompete functionally similar species for post-fire resources.
In Chapter 3, I explored interacting drivers of community invasion resistance using an in-situ manipulation experiment across three vegetation types. I found that community biomass and some traits (specific leaf area, fine-to-total root volume, and height) may confer invasion resistance of existing communities to V. dubia. However, this was only the case in the most productive wet meadow vegetation types. I found no evidence that biomass or community trait composition contributed to invasion resistance in less productive and more stressful low sage-steppe or scab-flat vegetation types, indicating that environmental and biotic factors interact to influence invasion resistance. To assess the potential influence of V. dubia invasion on fire behavior across the region, I evaluated the influence of V. dubia on fuels and fire in Chapter 4 using a novel application of the landscape-scale Large Fire Simulator, FSim. I show that invasion increased fire spread, burn probabilities, and fire intensity across forest-mosaic landscapes by increasing fuels and fire occurrence in invaded non-forested areas adjacent to fuel rich forests.
Overall, this dissertation provides some of the first documentation of V. dubia’s niche and invasion dynamics in forested landscapes, and characterizes how this invasion differs from other problematic species in this region. My work demonstrates that V. dubia may initiate a grass-fire cycle in historically fire- and invasion-resistant scabland ecosystems and that annual grass invasion can have substantial impacts on fire behavior in uninvaded forests – ecosystems thought to be resistant to annual grass impacts. Together, these chapters provide valuable information from the invasion front to aid the management of this rapidly spreading species
Machine Learning for Architectural Design Space Exploration and Resource Control
Get PDFMachine learning has enabled significant advancements in diverse fields, yet, with a few exceptions, has had limited impact on computer architecture. Recent work, however, has begun to explore broader application to design, optimization, and simulation. Notably, machine-learning-based strategies often surpass prior state-of-the-art analytical, heuristic, and human-expert approaches. This thesis first reviews existing work applying machine learning to architecture, ranging from simulation and run-time optimization, to individual component design involving the memory system, branch predictors, networks-on-chip, and GPUs. Next, the thesis presents a novel deep-reinforcement-learning framework for design space exploration. Finally, the thesis introduces an innovative strategy for resource optimization with multiple co-scheduled workloads. Taken together, these works present a promising future for machine-learning-based architectural design
Care Team Burnout in Human and Veterinary Medicine
Get PDFMedical professionals experience higher rates of burnout than other professions due to the highly stressful nature of their environment. In 2018, 94 percent of physicians reported experiencing burnout – characterized by emotional fatigue, and feelings of depersonalization and low personal accomplishment – at some point in their careers. As emotional resources diminish, individuals become callous towards others and themselves. In the United States, this means that physicians are at least 1.4 times more likely to attempt suicide and self-harm than the general population.
Veterinarians are at an even higher risk, with rates of at least 2.1 times higher than the general population. Despite these high reported rates, current burnout mitigation methods focus on symptoms rather than organizational root-cause and do not provide long-term solutions.
Burnout mitigation methods that address root-cause and promote continuous improvement mindset often require structural changes alongside individual coping mechanisms and are therefore rare. Current burnout research suggests cognitive-behavioral techniques (CBT) that restructure cognition and motivate positive behavior as the most effective short-term method for stress management and burnout reduction. In this research, we explore the possibility of CBT mitigation methods as a long-term approach to burnout when implemented alongside structural changes.
The Conceptual Change Model (CCM) is a culture-driven framework that focuses on cognitive restructuring and behavioral activation. The CCM promotes engaging individuals to drive improvement through educating individuals, altering their behavior, and promoting their engagement with the workplace. In the CCM, individuals are classified as “single-” or “double-loop” actors. Single-loop actors engage in reactive and workaround behavior – they address symptoms or avoid problems. Double-loop actors engage in a proactive, reflective behavior – they seek to resolve root cause. The Literature suggests organizations function better when employees are double-loop actors. This research examines these single- and double-loop learning behaviors as potential factors to mitigate burnout.
A survey is used to explore the relationship between the behaviors (single and double loop) and burnout experiences in human and veterinary medicine. Organizations in the human medicine survey were located in the United States, the United Kingdom, France, Kuwait, United Arab Emirates, Bahrain, Indonesia, and Australia. Organizations in the veterinary medicine survey were located in the United States. Single- and double-loop questions were adapted from Mazur, McCrery, and Chen’s (2012) framework of behavior in healthcare to classify participant’s behaviors. The Maslach Burnout Inventory was used to measure burnout.
Findings suggest that proactive care team members in human and veterinary medicine tend to report lower rates of burnout. There also exist some similarities in burnout experiences between care team members in human medicine and veterinary medicine. Further research is needed to validate the findings observed in this study
On the Structure of the Orbit Spaces of Almost Torus Manifolds with Non-negative Curvature
Get PDFAn almost torus manifold is a closed -dimensional orientable Riemannian manifold with an effective, isometric -torus action such that the fixed point set is non-empty. Almost torus manifolds are analogues of torus manifolds in odd dimension and share many of the characteristics of torus manifolds. For example, both almost torus manifolds and torus manifolds are -fixed point homogenous. Just as torus manifolds are important examples of manifolds admitting so called isotropy-maximal actions, almost torus manifolds are important if one hopes to understand manifolds admitting almost isotropy-maximal actions. Recently Wiemeler classified simply connected torus manifolds with non-negative sectional curvature. To obtain this result, he proved a structure theorem for the quotient spaces of torus manifolds of non-negative curvature. In particular, he showed that non-negatively curved torus manifolds are locally standard, with orbit spaces homeomorphic to convex polytopes with acyclic lower dimensional faces. In this thesis, we obtain an analogous structure theorem for the orbit spaces of almost torus manifolds. Namely, we analyze the orbit spaces of simply connected almost torus manifolds with non-negative sectional curvature. The main result we obtain is that the action of is locally standard and although is homeomorphic to a disk and shares the combinatorial structure of a polytope, lower dimensional faces are in general not acyclic. Unlike locally standard torus manifolds, orbit spaces of locally standard almost torus manifolds need not be manifolds with corners. Nevertheless, the analysis of the orbit space structure of shows that almost torus manifolds are similar enough to torus manifolds, so that the result of this thesis can then be combined with results about extending isometric almost isotropy maximal actions to smooth actions to obtain a classification of non-negatively curved, simply connected almost torus manifolds
The Evolution of Scleractinian Microbial Symbionts
Get PDFThe modern world has presented many threats to the health and stability of ecosystems worldwide. One of the most biodiverse ecosystems, coral reefs, faces particularly strong pressures, and is already declining rapidly in complexity and area. Although the stressors that affect reefs are diverse, ranging from nutrient pollution to overfishing, invasive species to climate change, the impact of many of these stressors is ultimately mediated through interactions between the coral animal and its microbial associates, or microbiome. Some such interactions are readily apparent and have been studied for decades. For instance, coral bleaching, which is caused in part by increases in water temperature due to climate change, has devastated large swaths of reefs in recent years. The visual ‘bleaching’ that characterizes this phenomenon is the result of a breakdown in the symbiosis of the coral with photosynthetic algae of the family Symbiodiniaceae that normally live within its tissue. These algae provide the coral with essential energy, nutrients, and other services, but under temperature stress, they are expelled from the transparent tissue, leaving the white underlying coral skeleton visible and the animal without its food source.
However, other interactions between the coral and its microbiome are less well-defined. Many coral diseases, for instance, may be caused by the opportunistic overgrowth of fungi and bacteria in nutrient-rich water or under conditions of general stress. Even less clear is how bacteria may act as mutualists in the coral ‘holobiont’. Other cnidarians have been shown to require developmental stimulation from particular bacterial species, and non-Symbiodiniaceae microbes have also been hypothesized to act as nutritional symbionts or as defense against other, pathogenic microbes.
Scleractinian corals are diverse; having been evolving for more than 450 million years and including over 1,500 species. Because of this, they are likely to have many different modes of interaction with their microbiomes. To begin to better understand the similarities and differences among the microbiomes of corals, I conducted during my PhD the Global Coral Microbiome Project (GCMP), which sampled thousands of coral colonies from dozens of phylogenetically diverse species. Through the course of my work, I identified the similarities and differences between the microbiomes of these many species, showed the importance of considering shared evolutionary history in the analysis of such datasets, and developed new, rigorous methods of microbiome analysis that separate the effects of distinct evolutionary and ecological processes
Strategic Application of Computations to Diverse Chemical Systems
Get PDFMy work has focused on streamlining the study of organic compound structure, properties, and reactions by 1) understanding and using available computational software, 2) developing software packages to facilitate data manipulation as well as augment public programs, and 3) applying public and in-house developed software to further the understanding of organic chemistry phenomena.
An evaluation of the strengths and weaknesses of parameters while modeling complex, flexible compounds with molecular mechanics led to several suggestions to help harness the computer-driven portion of conformational searching. Several case studies were used to show the practical considerations for modeling both ground and transition state structures.
A comprehensive joint experimental and computational investigation into a supramolecular host/guest binding event led to the observation of an emergent equilibrium isotope effect (EEIE). Computations were used to fragment the detector into key components to identify the origin of the H/D isotope effect and to explore the non-classical C-H/D hydrogen bonding. However, the fragments were unable to reproduce the isotope effect observed in the full detector, leading to the hypothesis that the whole detector is more than the sum of its parts.
A continuation of the supramolecular detector work was carried out to understand the unusual binding behavior when conformationally flexible hosts bind disparate oxoanion guests. Experiments showed all three hosts exhibit similar affinities for all three oxoanion guests when in a mixed, less polar solvent system (10% DMSO/90% H2O-saturated CHCl3). However, computations predict the binding affinity between all host/guest combinations to closely resemble the trends in pKa of the conjugate acid of the guest in a range of density functional theory (DFT) methods and solvents. The trend in experimental association constant better matches the computational predictions when a simple, more polar solvent system is used (acetonitrile). Entropic and solvation effects likely explain the discrepancy between computational and experimental results in the mixed solvent system.
A study on the stereocontrol of the first highly selective hydroboration of alkyl-substituted aldimines to provide medicinally-relevant a-amidoboronates illustrates the importance of subtle hydrogen contacts. A ferrocenyl group on the planar-chiral N-heterocyclic carbene (NHC) ligand stabilizes the copper-BPin species and is responsible for the bulk of the reaction selectivity.
A joint experimental and computational investigation into the mechanism for the formation of perfluoroalkyl-substituted b-lactones using an isothiourea catalyst emphasizes the importance of superimposing experimental and computational results. Computations provided both a stepwise aldol/lactonization process and a concerted [2+2] process depending on the DFT method used. Natural abundance kinetic isotope effect (KIE) studies were used to disseminate the operative mechanism in which computations and experiments corroborate a concerted [2+2] process
Automated Synthesis of Metal-Organic Frameworks using Graph Grammars and Monte Carlo Tree Search
Get PDFIn this work, a fully automated approach to synthesis Metal-Organic Frameworks is presented. We use graph to represent the structure of Metal-organic Frameworks and use graph grammars which are backbone rules and functional group rules to generate metal-organic framework. For a given parameter value of the user defined metal-organic framework, the design space is searched to find the candidate metal-organic framework which has very close parameter value by using Monte Carlo Tree Search.
To test the effectiveness of Monte Carlo Tree Search, we choose random search as baseline to compare with Monte Carlo Tree Search. The results from using three different evaluation function prove the superior performance of Monte Carlo Tree Search
Meeting End-to-End Deadlines in Real-Time Networks Using Software Defined Networking (SDN)
Get PDFNetwork flows in Real-Time (RT) systems need to meet stringent end-to-end deadlines in order for such systems to operate safely and reliably. Today, such systems use custom or domain specific network system designs to meet end-to-end deadlines and other constraints of real-time flows. In this work we explore the design of real-time networks using common-off-the-shelf (COTS) components by leveraging Software-Defined Networking (SDN) paradigm. In particular, we explore the effectiveness of using i) spatially varying but locally static flow priorities and ii) the impact of using Least-Slack Prioritization on the performance of network path layout and provisioning algorithms. Specifically, we propose different heuristics for spatial variation of static flow priorities in a real-time network and empirically show that spatial variation of priorities can accommodate more real-time flows than simple static priorities. Further, we show that least-slack based flow priority assignment performs better than deadline monotonic priority assignment for multiple path layout algorithms considered in this work
Digital Solutions for Analog Shortcomings in Delta-Sigma Analog-to-Digital Converters
Get PDFPortable, high power efficiency communication devices is a growing market in the semiconductor industry. Analog-to-digital converters (ADC) are key interface that are used to digitize the sensed information. Recently, digital techniques have been proposed to improve analog building block power efficiency in sub-micron technologies. This research focuses on mixed signal approaches to improve the power efficiency of the noise shaping ADCs and mitigate analog inaccuracies such as non-linearity and mismatch. First, a novel continuous-time filtering delta-sigma ADC is proposed to save power and area. Digital techniques have been proposed to make the architecture more robust to out-of-band unwanted signals. A prototype was fabricated in a 65 nm CMOS technology achieving an SNDR of 72.4 dB operating at 250 MHz sampling frequency over 7 MHz bandwidth, with a power consumption of 16.3 mW. Next, A novel digital circuitry is proposed to improve the tolerance of a discrete-time delta sigma ADC to mismatch and enhance the resolution of an ADC in the presence of mismatch. A custom IC was fabricated in a 65 nm CMOS technology consuming 40.4 μA from a 1 V supply. It achieves 76.18 dB SNDR operating at 1.2 MHz sampling frequency and 25 kHz signal bandwidth
A Framework to Evaluate the Risk of Human- and Component-related Vulnerability Interactions
Get PDFMost accidents and malfunctions in complex engineered systems are attributed to human error. However, a closer inspection would reveal that such mishaps often emerge as a result of poor design and human- and component-related vulnerabilities acting together. To fully understand and mitigate potential risks, the effects of such interactions between component and human fallibilities (in addition to their independent effects) need to be considered early in the design process. Existing risk assessment methods either quantify the risk of component failures or human errors in isolation or are only applicable during later design stages. This work takes the view that the combined effects of human errors and component failures are better understood when they are studied together. To this effect, this research introduces an early design stage computational framework to model the system level effects of component failures and human errors. Then, an automated fault scenario generation technique and a severity quantification model are introduced to help designers generate a wide range of potential fault scenarios (involving both humans and components) and prioritize them based on severity. Next, the applicability of the framework to complex engineered systems and the accuracy of scenario generation and severity quantification are explored. Finally, this research demonstrates an application of the framework to promote risk-informed ergonomic assessments with the use of digital human modeling simulations. The ultimate goal of this research is to help designers detect the combined effects of human- and component-related vulnerabilities (in addition to their effects in isolation) in complex engineered systems during early design stages to improve performance and safety while minimizing the potentially costly design changes and rework later in the design stages