22962 research outputs found
Sort by
Carbon Sequestration in Riparian Buffer Systems as a Nature-based Solution for Climate Change Mitigation
No full textRiparian buffer systems (RBS) are increasingly recognized as effective nature-based solutions (NbS) that enhance carbon sequestration in agricultural landscapes, thereby contributing to climate change mitigation while providing environmental, economic, and social co-benefits. The literature suggests that different types of RBS and variables, including landowners’ management decisions, influence RBS as NbS. RBS are adopted on public and private lands in the County of Wellington and Grand River Watershed, located in southern Ontario, Canada. In riparian tree buffers, deciduous species grow faster and accumulate carbon quickly, though they store it for shorter periods, while coniferous species grow more slowly but retain carbon over longer periods. Government initiatives play a crucial role in promoting the establishment, protection, and long-term maintenance of RBS. Findings revealed that RBS provide multiple ecosystem services, water quality, enhanced biodiversity, and carbon sequestration.Natural Sciences and Engineering Research Council of Canada2026-12-1
Advancing Automated Biodiversity Monitoring: A Deep Learning Framework for Bulk Arthropod Samples
No full textAs climate change threatens insects worldwide, a need for automated biodiversity monitoring has arisen. While vision-based methods are promising, they rely on individual specimen images, representing a domain shift from the unsorted “bulk images” produced by ecological surveys. We focus on the challenge of localizing small, densely packed insects from bulk images through two contributions. First, we benchmark instance segmentation models on the Mixed Arthropod Sample Segmentation and Identification (MassID45) dataset, a collection of 49 bulk images with nearly 18 000 expert-annotated insect masks. Our experiments demonstrate the importance of expert-annotated training data for ecological analyses, where foundation models underperform their supervised counterparts. Second, we significantly improve our MassID45 benchmarks using image super-resolution (SR). We discover that simple interpolation-based methods outperform deep learning-based SR methods, which points to the feasibility of low-compute solutions for large-scale biodiversity studies. Together, these contributions pave the way for advances in automated, vision-based insect monitoring.Vector Institute for Artificial IntelligenceMitacsNatural Sciences and Engineering Research Council of CanadaQueen Elizabeth II Graduate Scholarship in Science and TechnologyUniversity of Guelp
The Influence of Sex and Age on Myocardial Gene Expression in the Left Ventricle and Left Atrium of both Young and Adult Healthy Male and Female Cats
Full text linkDifferences in the development and progression of cardiac diseases between females and males have been reported in humans and cats. However, how sex and age influence variation in cardiac gene expression in healthy feline hearts has not been investigated. We analyzed left ventricle (LV) and left atrium (LA) myocardial tissue samples from young and adult cats and hypothesized that mRNA profiles in both chambers would exhibit sex and age-specific differences. The aims were to compare RNA-sequencing data between 1) young female and male hearts and 2) adult female and male hearts. Across ages, female LV and LA samples exhibited activation of pathways and genes related to immune regulation, development and repair. Conversely, male LV and LA samples displayed enrichment in pathways and genes linked to inflammation, oxidative stress and apoptosis. The findings provide a better baseline understanding of how sex and age influence healthy feline hearts from a transcriptomic perspective
Design and performance evaluation of a strawberry gripper for greenhouse elevated culture harvesting
No full textThis thesis addresses the problem of automating strawberry picking in elevated greenhouse settings. Strawberry farming is a relevant agricultural activity worldwide, as it has a broad economic impact that will continue to grow with the demand for fresh fruit products. Strawberry crops are hand-picked due to the fruit’s delicate nature, which incurs higher costs in time and logistics. Harvesting year-round in controlled greenhouse environments increases this output. Automation of strawberry harvesting could be a potential solution to increase process efficiency and reduce costs.
The design, development, and evaluation of end effectors for automated strawberry harvesting allowed for a final tilting robotic gripper capable of detaching greenhouse strawberries with minimal damage. The work was done through multiple prototype iterations that integrated findings from mechanical performance, in-lab testing, and field observations.
Overall, this work assesses the feasibility of a tilting-based harvesting mechanism and contributes insights into mechanical interactions, sensing requirements, and variety-dependent behaviour in automated strawberry picking mechanisms, thereby widening the state of the art.2026-12-1
Detecting Deceptive Behaviors in Decentralized Finance
Full text linkDecentralized Finance (DeFi) represents a major transformation of traditional finance and has become a foundational component of the broader Web3 ecosystem. DeFi’s core is a blockchain-based financial system that enables open and permissionless interactions without reliance on centralized intermediaries. While this open and permissionless structure fosters innovation and accessibility, it also creates conditions for deceptive behaviors that exploit user trust and informational asymmetries. This dissertation proposes practical frameworks to detect three underexamined forms of deception in DeFi: wash trading, honeypot token traps, and documentation--code inconsistencies. First, we analyze wash trading on decentralized exchanges (DEXs), where manipulators use self-controlled addresses to generate artificial volume and misleading market signals. We introduce an entity-recognition-based detection method that uncovers collusive activity by examining ETH transfer links and temporal transaction sequences while removing non-target transaction interference. Experiments on major exchanges reveal distinctive behavioral patterns, including the coordinated use of multiple wallets and flash loan-based manipulation strategies. Second, we study honeypot token traps that allow users to buy tokens but block subsequent selling, locking victim assets in malicious liquidity pools. We present the first systematic taxonomy of honeypot mechanisms and develop a hybrid detection framework based on transaction simulation and log analysis. Evaluation on real-world datasets demonstrates high precision and recall across diverse trap categories. Third, we define documentation--code inconsistency as deception in which project descriptions conflict with deployed contract logic. To address this problem, we propose DeFiAligner, an end-to-end framework that integrates a customized symbolic Ethereum virtual machine (SEVM) with large language models (LLMs). SEVM extracts path conditions and the logic of balance changes, while LLMs analyze their alignment with natural-language documentation. DeFiAligner enhances auditing capability in closed-source settings by exposing hidden functions and semantic mismatches. These studies demonstrate that representative deceptive behaviors in DeFi can be effectively detected under real-world constraints by relying on publicly available data. Furthermore, the proposed frameworks can be integrated into user-facing systems such as Web3 wallets, blockchain explorers, or auditing pipelines, thereby providing practical safeguards for users.Natural Sciences and Engineering Research Council of Canad
The Dark Calculus: Applying Marxist Theatre Techniques to Interactive Digital Narrative in Caves of Qud
Full text linkLiterature applying concepts from Marxist theatre to interactive digital narratives has neglected both the political character of the theory from which it draws as well as the roguelike videogame form. By revisiting the fundamentals of Epic Theatre and Theatre of the Oppressed in relation to the field of videogames today, this thesis shows how engagement with the politics of these practices leads to a deeper understanding of their potential applications in digital media. Through a case study of Caves of Qud, I show how one particular videogame exemplifies techniques parallel to those of Marxist theatre through both its position within and advancements beyond the traditional roguelike form. My research also presents further direction for theoretical and practical applications of Marxist theatre techniques to interactive digital narratives.Ontario Graduate Scholarshi
A Comparative Study of Embedding and Classification Techniques for Mapping Job Descriptions to O*NET-SOC Codes
Full text linkThe evolving job market demands continuous understanding of skills required for various occupations as new technologies reshape workforce needs. This thesis is divided into two complementary sections. The first section surveys existing literature on job skills extraction, comparing methodologies employed by economists and computer scientists. By synthesizing these approaches, the survey highlights strengths and limitations inherent to each discipline, offering insights into diverse data sources and techniques utilized. The second section employs machine learning and artificial intelligence to classify job descriptions and titles into standardized O*NET codes. By comparing various embedding types and classification algorithms, this section enhances understanding of how computational methods apply to real-world job skills classification, thereby bridging theoretical insights from the survey with practical applications. Through this dual approach, literature synthesis and empirical analysis, this paper contributes to ongoing discourse on job skills analysis and provides a foundation for future research in both economic and computational perspectives
A Physics Based Digital Twin Framework for Motor Operated Valves
Full text linkMotor Operated valves (MOVs) are essential safety components that perform an on-off isolation and injection function under operational accidents in hydraulic power systems. The difficulty of obtaining run-to-failure data and the incapacity of basic estimation models to represent complex, concurrent, and non-linear faults limit degradation analysis for these components. Additionally, pure machine learning techniques require expert interpretation to understand fault characteristics, incurring additional costs. The model presented is built with the fidelity required to allow estimates on the MOV's availability using residual-based techniques and dynamically depict its state. To more accurately determine the current operating state, simulation strokes produced by this numerical solver are utilized for parameter estimation. The presented model serves as a precursor for future experimental validation and verification. While the simulated results from the model show agreement with literature “generalized” MOV stroke signatures, these parameters approximated
Phosphoproteome modulation by zinc and iron in Klebsiella pneumoniae reveals new virulence mechanisms and antibacterial targets
Full text linkKlebsiella pneumoniae is a Gram-negative bacterium causing 10-20% mortality in immunocompromised individuals and is known to rapidly develop resistance towards existing antibiotics. This underscores a need for novel antimicrobials or strategies for resensitization. Bacterial adaptation to iron and zinc, which are key nutrients linked to growth, division, and virulence, offers promising targets for antibiotic intervention. In this thesis, mass spectrometry-based phosphoproteomics was used to profile K. pneumoniae in iron/zinc-replete and -limited media, identifying 574 phosphopeptides from 315 phosphoproteins. Two candidates, SbmA and TolQ, were selected for disruption based on hypothesized roles in iron acquisition and membrane stabilization, respectively. Through iron quantification assays, growth curves, electron microscopy, antibiotic susceptibility testing and infection assays, we identified a previously uncharacterized role for SbmA in iron acquisition and identified reduced growth, viability, increased β-lactam susceptibility, and enhanced immune clearance for ΔtolQ. High-throughput drug screening revealed two compounds that target SbmA or TolQ, reducing bacterial growth and enhancing immune clearance. Together, these findings highlight phosphoproteomics as a powerful tool for identifying nutrient-regulated targets and position TolQ and SbmA as promising candidates for antimicrobial development.Natural Sciences and Engineering Research Council of Canad
Assessing the Capacity of Municipalities in Ontario to Support Agri-food Systems Planning
Full text linkMunicipal governments in Ontario play a key role in agri-food systems planning. Through land use planning, economic development, environmental stewardship, and broader decision making, municipal staff and elected officials have the ability to encourage or hinder agri-food systems in their communities. This dissertation investigates the capacity of municipalities in Ontario related to agri-food systems planning and addresses four research questions: 1. How can municipal agri-food systems planning capacity be conceptualized? 2. Are municipal planning departments well positioned to respond to and support agri-food systems in Ontario? 3. What activities are municipal planning departments in Ontario undertaking in support of local and regional agri-food systems? 4. What best practices and resources would strengthen municipal agri-food systems planning capacity? This research is underpinned by a mixed methods methodological paradigm, grounded in pragmatism, and employed a convergent mixed-methods research approach. Data and methods involved in this research include a literature review, survey questionnaires, semi-structured interviews, and secondary data. Data was analyzed using a variety of methods including thematic analysis, data visualization, and statistical analysis. In total, 290 Ontario municipalities participated in this research through at least one of the data collection methods. Findings from this research confirm that municipal capacity to support agri-food systems planning in Ontario is variable. Factors contributing to municipal capacity include planning department structure and personnel, relationships with other municipal departments and external actors, the land use planning framework in Ontario, and a commitment to local and regional agri-food systems. This capacity positions municipalities to facilitate agri-food systems planning processes including the use of regulatory and non-regulatory tools, and leveraging partnerships with agri-food systems stakeholders. Additional resources, including training, guidance materials, and funding, are needed to support municipal staff and councils undertaking this important work. A conceptual framework for municipal agri-food systems planning capacity is presented, and includes an enabling environment (the elements that provide the context), an action environment (the elements and processes that contribute to undertaking the work), and interim effects (relationships between elements that contribute positively to municipal agri-food systems planning capacity).Ontario Agri-Food Innovation Allianc