Electronic Theses and Dissertations
Permanent URI for this collection
Browse
Recent Submissions
Item Unraveling an uncertain future for the Great Basin: how climate, grass invasion, and wildfire interact to drive ecosystem function(2024) Strain, Maxwell Kay; Hanan, Erin; Blaszczak, Joanna; Leger, Elizabeth; Gustin, Mae; Verburg, PaulDrylands are the earth's largest terrestrial biome and are rapidly experiencing change. In drylands of the western United States, three of the most influential global change processes are shifting precipitation regimes, annual grass invasion, and wildfire. Understanding how these drivers interact to influence ecosystem function, and how such interactions may change in the future, can provide critical insight for building ecological resilience. This dissertation explores how changing moisture regimes interact with grass invasion and wildfire to modify three key cycles in the Great Basin: carbon (C), nitrogen (N), and water. This work bridges multiple spatial and temporal scales using empirical studies conducted in the field, greenhouse, and laboratory, along with simulation modeling. The work in chapter one used a greenhouse and laboratory experiment to identify how soil aridity influences root exudation by cheatgrass (Bromus tectorum), an invasive annual grass, alters the cycling of carbon (C) and nitrogen (N) in the soil. We observed that cheatgrass was able to uniquely access N in dry soils, and that changes in soil condition that implicate increased C and N cycling were caused when cheatgrass root exudates were experimentally added to soils. This work shows that root exudation by cheatgrass accelerates soil C cycling and suggests that it modifies N cycling and provides cheatgrass with unique access to N under dry conditions. When combined with seasonal dynamics and annual grass phenology, these mechanisms may provide cheatgrass with greater access to soil N compared to native species. In chapter two, I used simulation modeling to mechanistically test the effects of changing precipitation regimes and grass invasion on water and N dynamics at larger, watershed scales. Using the Regional Hydro-Ecologic Simulation System (RHESSys)�"a process-based ecohydrologic model that couples the cycling of water, C, and N�"I examined how precipitation variability and invasion influence stream discharge and N export. I simulated four climate scenarios with three levels of cheatgrass invasion in a factorial design. Streamflow decreased with both climate change and grass invasion, both of which increased rates of evapotranspiration. Watershed N export increased with climate change and invasion. This work can help us understand how change and annual grass invasion in the Great Basin may interact to influence ecosystem water supply and water quality. Chapter three investigated the role of aridity and burn severity on N retention after wildfire using a field and laboratory approach. I quantified how burn severity and aridity influence some of the key factors driving N uptake by plants and soil microbes after fire. Plant N uptake exhibited a non-monotonic relationship with burn severity, where sites that burned severely had suppressed N uptake for up to two years following fire. Dry conditions muted plant and microbial N uptake for all burn severities. N uptake was lowest after severe fire and when soils were relatively dry. Soil N that exceeded plant and microbial demand when N uptake is suppressed is likely more vulnerable to export. Using a conceptual framework, I hypothesize that postfire N loss is likely to occur at a lower burn severity when conditions are dry after fire. This hypothesis expands existing N saturation theory to better capture the role of fire and its interactions with soil aridity in drylands.Item Data-Driven Voltage Control of DERs Integrated Distribution Grids Using Deep Reinforcement Learning(2024) Hossain, Rakib; Livani, HanifThe proliferation of Distributed Energy Resources (DERs) in modern power distribution grids presents a significant challenge for voltage control and power loss due to their intermittent nature and diverse operational characteristics. Although different control strategies have been developed to improve the voltage profile of the distribution networks, Volt-VAR control (VVC) is one of the efficient tools that can be used to control the reactive power set-points of legacy reactive power resources or newly added inverters to regulate the voltage of distribution networks. VVC methods regulate the voltage by exploiting the reactive power absorption/injection capabilities of capacitor banks, and smart inverters. Although existing classical VVC control/ optimization based approaches can control the set-point of the slow responding legacy devices and fast responding smart inverters to control the the voltage of the distribution network, machine Learning (ML) based VVC/VVO approaches especially deep reinforcement learning (DRL) based VVC approaches has become popular for real time control of the voltage. DRL can optimally control the reactive power set-point of the inverter to minimize the voltage fluctuation and avoid the voltage violation. This dissertation investigates the application of Deep Reinforcement Learning (DRL) techniques for voltage control in distribution grids with high penetration of DERs. Leveraging the capabilities of artificial intelligence, specifically deep neural networks and reinforcement learning algorithms, our approach addresses the complexity of managing voltage levels in the presence of decentralized energy sources. It presents a comprehensive analysis of the challenges posed by high DER penetration and demonstrate the effectiveness of DRL-based control strategies in optimizing grid voltage profiles. By training deep reinforcement learning agents on real-time data from distribution networks with a diverse array of DERs, we achieve precise and adaptive voltage control. The proposed methodology not only ensures the stability and reliability of the distribution grid under varying operating conditions but also maximizes the utilization of renewable energy sources. Existing DRL based approaches do not consider the topology information of the network and they don't consider the consequence of missing/unobservable data. Therefore, incorporating graph convolution layers and generative adversarial network with DRL algorithm, this research address these problems. The implementation of the existing DRL based approaches in real power grids has been impeded by the absence of clear assurances regarding stability and safety. Therefore, this research utilizes a monotonically decreased neural network that can guarantee the stability of the policy during training and execution processes. The capability of the DRL based approaches in real-time implementation in tested using real-time digital simulator (RTDS). Through extensive simulations and comparative analyses, we showcase the superior performance of our DRL-based voltage control approach in comparison to traditional methods. The results underscore the potential of advanced machine learning techniques in enhancing the reliability and efficiency of distribution grids amidst the rapid integration of DERs. This research contributes valuable insights and practical solutions for utilities and researchers aiming to optimize voltage control strategies in distribution systems with a high penetration of renewable energy resources.Item Applying U-Th Disequilibrium to Dating Siliceous Sinter(2024) Sankovitch, Lauren; Munoz-Saez, Carolina; Hudson, Adam; Stillings, Lisa; Arienzo, MonicaContinental hydrothermal systems are critical avenues for the crustal transport of heat and mass captured for geothermal energy and mineral exploration. Thus, understanding their temporal evolution and longevity is important for resource characterization. In particular, high temperature reservoirs (> 170°C), which are commonly marked at the surface by deposits of microlaminated siliceous sinter, have the potential to trace hydrothermal histories. Previous attempts to use radiocarbon (14C) dating in active modern systems have encountered problems matching ages with stratigraphy, possibly due to contamination with old carbon leading to older apparent ages. Geothermal reservoirs are often positioned within uranium (U)-rich silicic volcanic rock where subsurface fluid-rock interactions extract U into hydrothermal fluids. As U is precipitated in the surface sinter deposit, assuming closed-system behavior, it begins to produce Th as an intermediate daughter product within the U-Th-Pb decay chain. If this assumption is correct, the U-Th disequilibrium geochronologic method becomes an additional dating option. While U-Th dating has been broadly utilized in carbonates, a systematic methodology has not been established for siliceous sinter deposits. As U-Th ages require only milligrams of material, this would potentially allow detailed chronologies of microlaminated material. For this work, we collected samples from El Tatio geyser field in the altiplano of northern Chile, the largest geothermal system in the Andes. Radiocarbon dating has been conducted at El Tatio and its arid climate offers exceptional preservation of deposits. Our resulting U-Th ages, along with the water and deposit chemistry, suggest trends of U and Th accumulation along the sinter apron. While distal facies containing the highest U concentrations (> 50 μg/g) show the least effect from detrital Th, they can display suspected open-system behavior. In contrast, more medial facies, where bacterial mats and other porous textures are commonly concentrated, have only trace amounts of U (< 0.1 μg/g), which leads to unreliable or geologically improbable ages. More proximal facies tend to date most consistently, with repeat age results on one sample even producing a statistically significant isochron. By comparing existing 14C ages with U-Th results, El Tatio serves as an independent constraint, with U-Th ages trending younger than the 14C ages, possibly supporting the presence of an old carbon influence, and pushing forward the late Pleistocene onset of geothermal activity at El Tatio by more than 10 ka.Item Data-Driven Analysis and Topology-Aware Learning of Phasor and Waveform Measurements for Enhanced Situational Awareness in Power Systems(2024) Mansourlakouraj, Mohammad; Livani, HanifPower grids are evolving with the integration of more renewable generation resources and different types of loads. This shift introduces new types of challenging events, oscillations, and controller responses, in addition to typical faults and outages. Additionally, phasor and waveform measurement devices are being increasingly used, measuring system variables such as voltage and current at a high reporting rate across the grid. This provides opportunities to enhance modern power grid monitoring during events and grid responses. Therefore, a fundamental question is how to analyze this valuable recorded data for practical power system monitoring and achieve better situational awareness. This dissertation is concerned with data-driven analysis of events and operational changes of assets by using statistical analysis, signal processing, and topology-aware learning methods. Events, abrupt changes and oscillations in power systems create specific signatures on the measurement signals, so effectively analyzing them helps enhance event detection, clustering, classification, and localization of their sources. These attempts, relying on the proposed methods in this dissertation, such as graph-based learning, short-time modal analysis, and statistical wavelet-based studies, can provide energy utilities with insight into the ongoing conditions in the power system. This can enable them to take proper actions, predict grid responses, and prevent failures at early stages, ensuring reliable and sustainable energy for people.Item Exploring the Neural Foundations and Causal Mechanisms of Real-World Cognition(2024) Fairchild, Grant; Walsh-Snow, Jacqueline; Strother, Lars; Rudd, Michael; Lescroart, Mark; Jiang, Fang; Kidd, ThomasThe brain evolved to perceive and to facilitate interaction with an environment consisting of real, three-dimensional objects, an environment in which humans and animals still live most of their lives. However, our knowledge of visual cognition and visual neuroscience is based predominantly on studies that have used 2-D images as visual stimuli. If behavioral and brain responses to real objects and 2-D stimuli differ, then many of the traditional frameworks of cognitive neuroscience may need to be reevaluated. Recent studies have in fact shown that real objects elicit different behavioral responses than images in many cognitive domains, including perception, action, memory, and attention. However, neither the neural correlates underlying these differences, nor the causal mechanisms responsible for these differences, have been extensively investigated. Across three studies, this dissertation explores the neural and causal bases of real-object effects. In the first study, I use fMRI to compare the effects of distance and size on brain representations of real objects vs. printed pictures. In the second study, I use fMRI to compare the neural correlates of recognition memory for real objects vs. computer images. In the third study, I use eye tracking to compare gaze patterns for real objects, 2-D images, and 3-D images. My behavioral data shows that real objects are better remembered than images, and that gaze towards real objects, compared to both 2-D and 3-D images, gravitates more towards object parts relevant to the current task. My neuroimaging data show that the functions of the dorsal and ventral visual pathways may be more integrated during viewing of real objects than of pictures, and that regions supporting memory retrieval and object-action associations respond differently during recognition memory of real objects vs. images. My results suggest that the different responses to real objects vs. images depend largely on the potential of real objects for genuine physical interaction. These findings have major translational implications in an era of increased screen-time dominated by artificial, digital objects. These findings also reveal the limitations of 2-D experimental stimuli and highlight the value of real-world experimental stimuli in providing a comprehensive characterization of naturalistic visual cognition.