Masters Theses

Permanent URI for this collection

https://hdl.handle.net/10919/9291

Browse

Now showing 1 - 20 of 24009

Application of Foam Fractionation Technology for the Removal of Disinfection Byproduct Precursors: Challenges and Opportunities
Haque, Md Sazzadul (Virginia Tech, 2026-07-30)
Foam fractionation is a cost-effective physicochemical process that concentrates surface-active compounds at the air-water interface and is recognized for its effectiveness in removing per- and polyfluoroalkyl substances (PFAS) from different aqueous matrices. To this date, evaluation of simultaneous control of disinfection byproduct (DBP) formation in drinking water and potable reuse systems through foam fractionation had not been investigated. This thesis evaluated the removal of DBP precursors using a bench-scale foam fractionation system assisted by cetyltrimethylammonium bromide (CTAB), analyzing surrogate solutions and water samples from the Occoquan Reservoir, an indirect potable reuse source, at pH levels between 6.5-8.5. Foam fractionation experiments demonstrated high removal efficiencies for the carbonaceous surrogate (humic acid; 80-96%, average 88.7%) and the nitrogenous surrogate (L-tryptophan; 80-99%, average 92.4%). In the real water matrix, post-fractionation decreased dissolved organic carbon by 35%, UV absorbance and specific UV absorbance at 254 nm by 78% and 65%, respectively, indicating effective removal of aromatic and hydrophobic natural organic matter. Following chlorination, regulated trihalomethane formation fell 88-97% and haloacetic acids by 60-82% at pH 6.5-7.0; haloketones, haloacetaldehydes, and trichloronitromethane were near-completely suppressed in the post-treatment. In the treated water, toxicity-weighted responses increased by 58-110% across all pH levels, attributed to a shift towards brominated species incorporation from CTAB-derived bromide. At the lowest CTAB dosing tested, toxicity-weighted response declined to approximately 1%, whereas the bromine substitution factor plateaued at 30-40% of untreated levels. Bromine substitution therefore persisted at bromide concentrations well below those associated with the full CTAB dose, indicating that reducing bromide loading alone did not remove it under the tested CTAB removal conditions. These findings show that foam fractionation can effectively reduce the downstream DBP formation potential, while emphasizing the importance of careful surfactant selection and further optimization in engineering design for achieving public health goals in drinking water treatment context.
Reconstructing vertical deformation using stratigraphy and microfossils to infer megathrust rupture history on Sitkinak Island, Alaska
Nowak, Tabitha Elizabeth (Virginia Tech, 2026-08-10)
Characterizing long-term subduction zone behavior is critical to prepare for future seismic hazards along vulnerable coastlines. The Alaska-Aleutian subduction zone (AASZ) has previously produced devastating multi-section, tsunamigenic earthquakes exhibiting variable rupture patterns that complicate hazard assessments. The incomplete historical record (~200 yr) of earthquakes along the AASZ and the fragmentary paleoseismic archive in the western portion of the megathrust require millennial-scale geologic investigations to resolve uncertainties regarding rupture scenarios and timing. Sitkinak Island, located in the southwest Kodiak Archipelago at the western extent of the 1964 rupture zone, is well-positioned to study the persistence of rupture boundaries through multiple earthquake cycles. Previous work in south Sitkinak Lagoon identified a unique mixed coseismic uplift and subsidence record, marked by sharp, laterally extensive peat-over-mud (subsidence) and mud-over-peat (uplift) contacts in coastal marsh stratigraphy. However, questions remain about possible non-seismic sources of stratigraphic change in south Sitkinak Lagoon. A multi-core analysis and quantitative approach across diverse depositional settings are needed to corroborate previous findings, establish site-wide continuity of paleoseismic evidence, and generate estimates of the amount of land-level change recorded. Here, we evaluate evidence for coseismic deformation at a new coring site in south Sitkinak Lagoon. We infer that four sharp contacts, representing two subsidence and two uplift events, record deformation associated with megathrust earthquakes along the AASZ. Age estimates obtained using 137Cs analysis, radiocarbon dating, and age modeling constrain the timing of these events to 1964 CE and 700-554 cal yrs (subsidence), and 524-482 cal yrs BP and 785-735 cal yrs BP (uplift). Across the sharp contacts, we analyzed fossil diatoms, key indicators of salinity changes, and applied a transfer function to statistically relate fossil diatom assemblages to elevations within the tidal frame. Paleoelevation reconstructions indicate 0.29-0.64 m of deformation across the four interpreted earthquake contacts. The timing and direction of land-level change at our site are consistent with those reconstructed from a previously studied core in a different geomorphic setting in south Sitkinak Lagoon, demonstrating site-wide continuity of paleoseismic evidence across the lagoon and strengthening confidence in a mixed uplift and subsidence history. These results support the interpretation that Sitkinak Island occupies a complex deformation setting near the lateral edge of Kodiak section ruptures. Our quantitative estimates of coseismic land-level change provide new model constraints on rupture behavior and slip variability in this region, improving future earthquake and tsunami hazard assessments in Alaska.
The Influence of Compost Tea on Plant Growth Performance in Corn and Winter Wheat
Flory, Erika Ivy (Virginia Tech, 2026-08-10)
Nitrogen (N) fertilizer is one of the most important inputs in modern grain production. Without N many crops, including corn (Zea mays L.) and winter wheat (Triticum aestivum L.), would lack one of the most vital nutrients used to produce proteins, amino acids, and chlorophyll. Rising fertilizer costs and increasing interest in sustainable nutrient management practices have prompted research into alternative products that may improve N use efficiency (NUE) while maintaining crop grain yield. Compost tea (CT), a liquid extract produced from composted organic materials, has been widely used in organic vegetable production systems and home gardening applications. The reported potential benefits of CT include increased soil microbial activity, enhanced nutrient cycling, and improved plant health. Limited research has evaluated the effectiveness of CT in intensive row crop production. This study investigated the effects of CT application in combination with varying N fertilizer rates on the growth, physiological status, and grain yield of winter wheat and corn under Virginia growing conditions. Field experiments were conducted during the 2024-2025 growing season at Blacksburg, Orange, and Blackstone, Virginia, using a nested randomized complete block design. Treatments included 100% recommended N with and without CT, 75% recommended N with and without CT, and 50% recommended N without CT. Nitrogen was supplied as split applications of urea ammonium nitrate, and CT was applied once at 5 gal ac⁻¹. Plant height, chlorophyll content (SPAD), and grain yield were measured. Compost tea did not significantly affect plant height, SPAD readings, or grain yield in either crop. Location significantly influenced plant height and grain yield, while SPAD readings were generally similar across treatments and locations. These results indicate that environmental conditions and soil characteristics were the primary determinants of crop performance. Under the conditions evaluated, CT provided no measurable agronomic benefit when combined with conventional N fertilizer programs. Additional multi-year research is needed to determine whether CT may improve crop performance or N use efficiency under specific production conditions.
Thromboelastography in Dogs with Suspected Primary Brain Tumors
Kahlon, Vikram (Virginia Tech, 2026-08-10)
Thromboelastography in dogs with suspected primary brain tumors Vikram Kahlon (Academic Abstract) Background: Dogs with primary brain tumors typically have chronic progressive clinical signs, but some dogs will decline suddenly and unexpectedly. It is currently unknown if dogs with primary brain tumors experience hemostatic derangements. Dogs can be screened for hypercoagulability using a whole blood test called thromboelastography (TEG). Hypothesis: Dogs with suspected primary brain tumors have TEG alterations consistent with hypercoagulability compared to healthy control dogs, characterized by one or more of the following: shortened R and K values, increased α-angle, and increased MA. Animals: Thirty-seven dogs were prospectively enrolled with suspected primary brain tumors. Following MRI evaluation, ten dogs met inclusion criteria, and twenty-seven were excluded due to the absence of intracranial neoplasia. Forty healthy adult dogs from a previously established control population were used for comparison. Prospective observational preliminary study. Thromboelastography will be performed on all dogs with the Haemonetics TEG 6s system using the global hemostasis cartridge. Dogs will be diagnosed with a primary brain tumor if they have clinical signs and exam findings suggestive of intracranial disease and brain MRI results consistent with a primary brain tumor. Dogs must have a blood chemistry, complete blood count (CBC), and urinalysis within 1 week of brain MRI to exclude concurrent disease that would affect hemostasis. Dogs receiving medications known to alter hemostasis within 2 weeks of enrollment will be excluded. Healthy adult dogs from a previously established control population were used for comparison. Power analysis indicated that nine dogs per group were required. Thromboelastography parameters were compared between groups using appropriate parametric or non-parametric statistical tests, with significance set at p < 0.05. Results: Results revealed that 30% (3/10) of dogs with primary brain tumors had a prolonged R value. Additionally, R was significantly prolonged in dogs with primary brain tumors compared to healthy dogs (p = 0.0003). RapidTEG MA was significantly increased in dogs with primary brain tumors compared to healthy dogs (p = 0.0266). No significant differences were identified between groups for K, α-angle, or MA. Conclusions: Overall, dogs with primary brain tumors have TEG features of hypocoagulability. The clinical significance of increased RapidTEG MA with normal MA in dogs with primary brain tumors is unclear.
Coupled Effects of Tailings Particle Size and Borehole Size on Cement Paste Backfill Fracture Initiation Pressure
Arnold, Brian Nicholas (Virginia Tech, 2026-07-31)
Cement paste backfill (CPB) is a mixture of mine tailings, water, and binder used to support underground openings. Its mechanical strength is a critical design parameter for safety and the current strength testing methods are highly variable and unrepresentative of in-situ conditions. A recently proposed strength testing method, fracture initiation pressure (FIP), uses a hydraulic fracturing technique for estimating the in-situ tensile strength. Previously, the Point Stress (PS) model was used to interpret the FIP values, which indicate that both the material's internal flaws and borehole size govern hydraulic fracturing behavior. Because the particle size distribution (PSD) controls packing, porosity, and flaw size, this thesis investigates the coupled effects of PSD and borehole size on FIP in CPB. Three ground-silica PSDs (Sil-Co-Sil 52, 90, and 106; with D50 of 16.1, 21.5, 32.9 μm, respectively) were initially tested at 4 and 8 wt.% binder and 7, 14, and 28 days of curing for unconfined compressive strength (UCS), Brazilian tensile strength (BTS), Mode I fracture toughness (KIC), FIP, and isothermal calorimetry. Strength generally increased with binder content, curing duration, and particle fineness with the fineness effects being attributed to the filler effect where increased surface area from fine particles leads to increased hydration product formation. Although the FIP results exceed tensile strength significantly, they correlate linearly with conventional strength measurements, confirming that FIP could be used as a strength measurement. Because the PS model indicates that the FIP will approach the true tensile strength as borehole size increases, the FIP was measured across eight borehole diameters (0.125 inch to 2.375 inch) for the same three PSDs. The results reveal a strong borehole size dependence where FIP decreases with increasing diameter size and plateaus beyond a critical size. The PS model and Lecampion mixed criteria models were used to evaluate the data and a modified mixed model incorporating an additional length factor, taken as the fracture process zone length, best fit the observed results. The plateau pressure at large boreholes is proposed as an effective tensile strength that correlates strongly with conventional strength measures, indicating that FIP measured at a significantly large borehole could provide a reliable relative in-situ CPB strength measurement.
Efficient Time-Domain Simulation of Sub-Terahertz Multipath: A Hybrid Architecture and Multi-Tiered Validation
Van Horn Jr, Charles Kente (Virginia Tech, 2026-07-29)
As wireless communication systems transition toward sub-terahertz (sub-THz) and terahertz frequencies to support 6G data rates, deterministic channel modeling becomes computation- ally bottlenecked. At sub-millimeter wavelengths, common indoor textures such as drywall roughness and carpet weave act as complex diffuse scatterers rather than simple specular re- flectors. Standard full-wave numerical methods, such as the Finite-Difference Time-Domain (FDTD) method, are computationally prohibitive at these macroscopic scales, while efficient ray-based techniques typically rely on approximations to estimate surface roughness effects. This thesis presents a novel, efficient time-domain hybrid computational electromagnetic sim- ulator designed to deterministically resolve sub-THz multipath scattering without the need for high-performance computing. The three-phase architecture integrates a two-dimensional FDTD engine with analytical ray launching and Rayleigh-Sommerfeld scalar diffraction prop- agation. A Simultaneous Grid Subtraction technique is also implemented to eliminate nu- merical dispersion during cylindrical Total-Field/Scattered-Field (TF/SF) source injection. The hybrid simulator is validated against commercial full-wave solvers (HFSS) and empir- ical measurements utilizing a 75–115 GHz Vector Network Analyzer (VNA) in a modified Naval Research Labs (NRL) Arch configuration. Time-domain Monte Carlo simulations of stochastically generated rough surfaces, modeled using Perfect Electric Conductor (PEC) boundary conditions, demonstrate excellent agreement with physical measurements, achiev- ing a 0.97 correlation in the extracted Power Delay Profiles. The simulator accurately predicts macroscopic channel statistics, matching empirical Mean Excess Delay to within 2.05 ps and Rician K-Factor to within 1.55 dB. The localized FDTD domain partitioning achieves an 86% reduction in active grid cells, shrinking the static memory footprint from 2.54 GB to 346 MB. This optimization yields a 16x reduction in execution time compared to a full-wave equivalent, executing sub-THz Monte Carlo campaigns in under 10 hours. This work provides a computationally viable, deterministic alternative to statistical models and ray-based deterministic models, bridging the gap between macroscopic simulation efficiency and microscopic electromagnetic precision.
Investigating the Process of Mammary Epithelial Cell Loss Into Milk and Its Relationship With Lactation Stage, Milk Production, and Lactation Persistency
Gilpatrick, Elizabeth Marie (Virginia Tech, 2026-07-29)
Mammary epithelial cells (MEC) are the milk-secreting cells of the mammary gland and are lost into milk throughout a cow's lactation. Because milk yield is a function of the number and productivity of MEC, understanding when and by which mechanisms cells are lost into milk can help us better understand local regulation of milk production and how it relates to milk yield and lactation persistency. The objective of this thesis was to investigate the relationship between MEC loss into milk and lactation persistency and better understand cellular mechanisms behind MEC loss into milk. This thesis consisted of two studies. The first study evaluated loss of MEC into milk longitudinally across seven stages of lactation and the relationship of MEC loss into milk with four corresponding persistency estimates for each stage. There was an inverse relationship between milk yield and MEC loss into milk but no relationship between any persistency estimate and MEC loss into milk. Mammary epithelial cell loss into milk declined in early lactation to a minimum shortly after peak and increased thereafter. Persistency parameter calculations were uncertain in early stages of lactation but converged and remained consistent in the later stages. The second study used an immortalized line of bovine mammary epithelial cells, BME-UV cells, to evaluate the relationship between apoptosis, epithelial crowding, and cellular extrusion. There was an association between apoptosis and extrusion, with increasing actin ring accumulation around cells as they progressed later into apoptosis. However, inhibition of canonical extrusion signaling pathways involving sphingosine-1-phosphate (S1P) and Rho-associated kinase (ROCK) did not disrupt extrusion. Increased epithelial crowding promoted extrusion of non-apoptotic cells, suggesting that crowding induces live cell extrusion. Taken together, the results of both studies suggest that MEC detected in milk may originate, at least in part, from apoptotic and crowding-induced extrusion processes that function to maintain tissue homeostasis. However, the signaling pathways involved in these extrusion processes in BME-UV cells may differ from canonical S1P-ROCK-mediated pathways ob-served in other epithelial models. Additionally, while MEC loss into milk is inversely related to milk yield, persistency estimates show that milk yield declines at a constant rate post-peak production. This suggests that the mechanisms responsible for MEC loss from the mammary gland, including loss into milk and other pathways, may shift across lactation. This work provides insight into local regulation of milk production in the mammary gland, an important consideration when evaluating milk production and lactation persistency.
Opt-In Audibility for Bitcoin Transactions via Cryptographic Commitments in Taproot Change Outputs
Chandra, Pratyaksh (Virginia Tech, 2026-07-27)
Bitcoin transactions carry no native mechanism for linking an on-chain payment to its real- world commercial purpose. A self-custodying user who wishes to demonstrate that a specific transaction corresponds to a specific invoice — for tax reporting, corporate audit, or reg- ulatory compliance — has no standardised, cryptographically sound way to do so without revealing their full wallet history. Existing approaches either lack on-chain binding (BIP-70, manual record-keeping), impose visible overhead and privacy costs (OP_RETURN commit- ments), or operate on different platforms entirely (Zcash viewing keys, Lightning invoices). This thesis presents an opt-in auditability protocol that embeds a cryptographic commit- ment to a structured invoice inside a standard Taproot change output using a pay-to-contract key tweak. The commitment adds zero bytes to the transaction, produces an output that is byte-for-byte indistinguishable from any other Pay-to-Taproot output, and requires no changes to Bitcoin's consensus rules. A new invoice format — the Signed Bitcoin Invoice (SBI) — organises payment metadata into a Merkle tree with per-field blinding nonces, en- abling field-level selective disclosure to auditors via inclusion proofs. An optional verifiable credential, embedded as a Merkle leaf, binds the merchant's identity to the payment address, providing pre-payment address authenticity verification. Five security properties (commitment binding, commitment hiding, disclosure soundness, disclosure privacy, and non-repudiation) are formally argued via reductions to standard cryp- tographic assumptions. A game-theoretic analysis demonstrates that the invisibility of the mechanism is a prerequisite for the opt-in property to remain voluntary under regulatory pressure.
The EGP Broker: Overcoming LMS Constraints to Support Alternative Grading Practices in Computer Science Education
van Duijl, Ilse (Virginia Tech, 2026-07-17)
Traditional points-based grading systems often fail to accurately reflect student proficiency, frequently penalizing the learning process. While Equitable Grading Practices (EGP), such as mastery-based or categorical grading, offer a more representative assessment of student learning, their adoption is hindered by a mismatch in infrastructure. Institutional Learn ing Management Systems (LMS), such as Canvas or Moodle, are architecturally rigid and built around traditional, quantitative gradebook exchanges. This thesis presents the EGP Broker, a middleware proxy built on a Nuxt 4 and Nitro server architecture backed by a PostgreSQL database. The core contribution of this work is the design and implementation of a protocol-agnostic LTI (Learning Tools Interoperability) gateway that sits between the LMS and external assessment tools. By intercepting and managing the launch and grade passback payloads, the broker provides a technical framework for implementing EGPs and enhancing LTI communication. Because the broker operates as a symmetric proxy, it can extract rich, multi-dimensional performance data that external tools possess which the in stitutional LMS is not natively built to receive. This layer handles, translates, and combines these complex datasets before forwarding back to the LMS. Ultimately, this work delivers a scalable, highly extensible institutional and research infrastructure that transforms how alternative grading models can be sustainably deployed in computer science education.
Simulation of Cloud Cavitation: Evaluation and Qualitative Assessment of Turbulence Predicted by k-omega SST and k-omega SAS Models
Sankhala, Ritika (Virginia Tech, 2026-07-15)
Cloud cavitation is a strongly unsteady two-phase flow in which vapor formation, cavity shedding, collapse, and turbulence interact over a wide range of spatial and temporal scales. In numerical simulations, these processes are commonly modeled using Reynolds-averaged or scale-adaptive turbulence closures, but agreement in mean flow or shedding frequency does not necessarily imply agreement in turbulent momentum transport. This thesis evaluates cloud-cavitating Venturi-flow simulations obtained using k-omega SST and k-omega SST-SAS models and compares them with experimental velocity and grayscale image data. The experimental dataset corresponds to a 48 L/min cloud-cavitating Venturi case. High-speed grayscale images were used to estimate an apparent vapor-fraction field and an apparent mixture-density field. These fields were combined with measured velocity fluctuations to form an apparent density-weighted Reynolds shear stress. The numerical SST and SST-SAS datasets were compared using spectral diagnostics, time-averaged velocity and density fields, density-weighted stress fields, and phase-averaged quantities over the cavitation cycle. A spatial filtering-based decomposition was also used to define experimental resolved-equivalent and residual stress contributions for diagnostic comparison with SST-SAS resolved and modelled stress terms. The spectral analysis showed that the experiment had a dominant shedding frequency near 95.21 Hz, while SST-SAS produced a close peak near 97.65 Hz. SST remained in the same low-frequency cavitation range but showed a shifted dominant response near 117.17 Hz. The mean velocity fields showed that both simulations reproduced the broad cavity-induced acceleration and flow deflection. However, the density weighted stress comparisons revealed larger differences. The experimental stress was weaker and more localized, while SST and SST-SAS produced stronger and more organized stress along the cavity shear layer and downstream closure region. In SST-SAS, this difference was present in both resolved and modelled stress contributions. The results show that matching the dominant cavity-shedding time scale is not sufficient to reproduce density-weighted turbulent momentum transport. The main discrepancies are associated with local vapor distribution, wall-normal interface motion, re-entrant-flow organization, u'-v' covariance, and the eddy-viscosity-based modelled stress response. The thesis therefore demonstrates the value of density-weighted and phase-resolved diagnostics for qualitative assessment of cavitating-flow turbulence predictions.
Oasis
Berrada, Yacine (Virginia Tech, 2026-07-15)
This thesis explores how architecture can transform a rural educational center into a place of attraction, gathering, and continuity in the dispersed mountain landscape of Al Haouz, Morocco. In this territory, villages are separated by distance, topography, climate, and limited access to shared infrastructure. Education is therefore not only a social question, but also a spatial one. The project positions the place of learning as a civic anchor between four villages along Oued Iss, creating a neutral ground where children, families, landscape, material, and light are gathered. The architectural proposal develops the idea of the oasis not as a natural condition, but as a constructed spatial strategy. In a context marked by exposure, distance, and fragmentation, the oasis becomes a place of shade, orientation, protection, and collective presence. The educational center is designed to be more than a collection of classrooms. It becomes a visible point of arrival, a shaded threshold, and a shared community space that supports learning, care, and social exchange. The project responds to the site through a careful relationship with topography, local materiality, and environmental conditions. Rammed earth walls root the building in the ground and provide thermal mass, while a lighter roof structure creates shade, filters light, and gives the educational center a recognizable civic presence in the landscape. Courtyards, controlled openings, and layered thresholds organize the transition between exterior exposure and interior comfort. Light is treated not only as an environmental element, but also as a cultural and spatial material that shapes atmosphere, memory, and orientation. Ultimately, this thesis argues that architecture can give educational challenges a spatial response. By gathering dispersed communities around a shared place of learning, the rural educational center becomes an oasis: a place where distance is reduced, arrival is dignified, and education gains a visible and lasting presence in the landscape.
Classification of Canine Primary Sensory Neurons
Jankelunas, Leanne Noelle (Virginia Tech, 2026-07-15)
Chronic pain accounts for nearly half of owner-reported canine euthanasia decisions, yet dogs remain underutilized as a large-animal model for studying pain and developing therapeutics. Here, we present a canine dorsal root ganglion (DRG) cell atlas. Our dataset comprises 3,026 neurons and 11,734 non-neuronal cells and resolves 15 neuronal subtypes that map onto A- and C-fiber classes. We further identify eight non-neuronal subtypes, including glial, vascular, and immune populations and characterize neuronal and non-neuronal expression of physiologically relevant neuropeptides, receptors, and ion channels. We identify region-specific differences in subtype composition between lumbar and sacral DRGs. Cross-species comparisons reveal that canine DRG subtypes are broadly conserved with human and mouse, while also exhibiting canine-specific and canine–human shared molecular features relevant for translation. Together, this atlas serves as a valuable resource for understanding canine sensory neurobiology, comparing DRG organization across mammals, and leveraging dogs as a translational model for pain research.
Gauss Bonnet Theorem for Singular Curves in CPn
Levy, Benjamin Philip (Virginia Tech, 2026-07-14)
This paper provides a generalization of the Gauss-Bonnet theorem to singular irreducible projective algebraic curves. For any singular curve $C$ with singularity set $Sigma$, we prove that if one integrates the Gaussian curvature provided by restricting the Fubini-Study metric to the noncompact manifold $C setminus Sigma$ that the result is well defined and equal $2pi$ times the Euler characteristic of $C$'s desingularization plus finitely many correction terms specified by the structure of $C$'s singularities.
Post-Disturbance Assessment of Bat Communities in the Southern Coastal Plain of New Jersey
King, Amanda Jane (Virginia Tech, 2026-07-14)
The Southern Coastal Plain of New Jersey (SCPNJ) is being rapidly altered by human development, altered forestry practices, and climate change. Understanding the influence of such effects on species in decline may provide insights into management efforts and conservation actions. White-nose syndrome (WNS), caused by the fungal pathogen Pseudogymnoascus destructans, was discovered in New York during winter 2006. Since then, several species of cave-hibernating bats have experienced range-wide population declines. The SCPNJ provides habitat for several bat species, including the endangered northern long-eared bat (Myotis septentrionalis). Despite this endangered status, remnant populations persist in coastal regions such as nearby Long Island, New York and on the offshore islands of Massachusetts. Additionally, the species appears to persist year-round in the coastal Carolinas and locally in portions of the Deep South, where shorter and milder winters allow individuals to avoid mortality from WNS. Although the SCPNJ experiences milder winters than nearby inland areas, the region contains a unique dynamic of increasing urbanization and habitat loss within a fire-adapted ecosystem. Prior to European settlement, the Pine Barrens within the SCPNJ were frequently burned and maintained by indigenous populations for hunting and forest management practices. After European settlement and subsequent forest harvesting, disturbance patterns shifted in response to fire suppression campaigns and land use change. To assess whether the SCPNJ functions as a refugium for the northern long-eared bat, I collected acoustic data from 2024–2026 using a spatially balanced sampling design. I processed calls using an automated call classification software and manually vetted files identified as potential northern long-eared bat calls. I used landscape and weather variables to fit a priori frequentist occupancy models. I used the top-performing model to parameterize a Bayesian occupancy model and then compared performance and predictions of these approaches. My results indicated a positive relationship between northern long-eared bat occupancy and the proportion of area burned within the past 10 years. Notably, fitting the top frequentist model within a Bayesian analysis using informative priors provided more stable and informative coefficient estimates likely due to limited detections. Based on the positive association between northern long-eared bat occupancy and fire, combined with recent wildfire events in the New Jersey Pine Barrens, I investigated the short-term effects of fire on total bat activity. Specifically, I collected acoustic data in two recent (< 1 year) wildfire areas and constructed Bayesian activity models. From April to November 2025, I deployed detectors along transects in burned interior, forest edge, and unburned sites. I processed data using automated call classification software, filtered by MLE, and aggregated total bat activity for each detector at the weekly scale. I then modeled activity within a Bayesian framework using normalized difference vegetation index (NDVI) as a proxy for fire disturbance effects on bat activity. I selected the most parsimonious model from three competing candidates. This model revealed a negative relationship between bat activity and NDVI, indicating higher activity in recently burned areas or along fire breaks. This suggests that fire reduces clutter that would otherwise obstruct flight paths whereby also likely promoting habitat diversity for both roosting and foraging. Results from this research indicate that periodic fire disturbance may benefit the bat community as a whole and a forest-dwelling endangered species such as northern long-eared bat, specifically. Land managers may use prescribed fire to promote a mosaic of productive roosting and foraging habitat to support bat populations on conserved lands that are otherwise facing increasing anthropogenic pressures elsewhere in the SCPNJ.
Expert-Driven Evaluation Platforms for Scientific Information Extraction Using Large Language Models
Gatne, Vasundhara Sagar (Virginia Tech, 2026-07-14)
Researchers are increasingly adopting Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) to extract scientific information in structured formats for training ML models. However, generative models are stochastic and prone to hallucinations. This necessitates expert-driven evaluation to verify factual accuracy and scientific reasoning. Traditional evaluation workflows rely on generic survey tools and spreadsheets. These tools impose cognitive load and may introduce psychological biases into the evaluation process. This thesis presents the design, architecture, and deployment of two visual analytics platforms to address these cognitive and operational challenges in LLM evaluation. Specifically, these platforms support the evaluation of extraction pipelines developed for computational virology. These underlying computational models are designed to parse academic literature to identify mutations enabling zoonotic host shifts and to extract quantitative viral inactivation metrics. The VILLA Evaluation Platform provides a structured environment for the qualitative assessment of LLM-generated reasoning regarding viral mutations. By implementing interface strategies such as blind evaluation, randomized presentation, and side-by-side component layouts, VILLA mitigates anchoring and central tendency biases. The Persist Extraction Interface serves as a diagnostic tool for evaluating the extraction of structured, quantitative data. The Persist interface aligns model outputs directly with human-verified ground-truth datasets to capture feedback across different models and pipeline iterations. These process-oriented interfaces translate raw probabilistic outputs into the actionable diagnostic data required by machine learning engineers. By standardizing collaborative workflows, this work provides a scalable foundation for assessing and optimizing research agents in specialized scientific domains.
Canine placental changes in normal and abnormal parturition: a transcriptomic approach from prepartum to second stage labor
Carneiro de Sousa, Gabriela (Virginia Tech, 2026-07-14)
The molecular mechanisms involved in canine parturition and the role of the placenta in this process remain unclear. This work aims to provide a thorough review of canine pregnancy, the placenta, and normal and abnormal parturition. Furthermore, in our study, we hypothesized that global gene expression profiling through RNA sequencing (RNAseq) would assist in the overall understanding of the role of the canine placenta throughout the parturition process. Twenty-two bitches were allocated into four groups: planned Cesarean section (C-section) at term pregnancy not in labor (PCS, n = 6, serum progesterone, P4 > 3.0 ng/ml); elective C-section in first stage labor (ECS, n = 6, P4 < 1.5 ng/ml); obstructive dystocia in second stage labor with spontaneous abdominal contractions (OD, n = 5); abnormal parturition progression diagnosed as primary uterine inertia (PUI, n = 5). One randomly collected placenta per dog during C-section was prepared and submitted for RNAseq. Transcripts for 13225 genes were detected and analyzed. The highest yield of differently expressed genes (DEG, n=1409) was identified in the OD vs PCS contrast (False Discovery rate [FDR]< 0.1 in edgeR and DESeq2). The OD group had 1121 upregulated genes, and the PCS had 288 upregulated genes relative to OD. The PUI vs PCS, OD vs PUI, and PUI vs ECS contrasts presented 8, 2, and 1 DEG, respectively. Functional characterization and enrichment analysis of the upregulated DEGs in the OD vs PCS contrast showed 109 pathways inferred to be activated in the OD group, which included but were not limited to the Rap1 and MAPK signaling pathways, cytokine-cytokine receptor interaction, leukocyte transendothelial migration, focal and cell adhesion molecules, complement and coagulation cascade, platelet activation, and oxytocin signaling. Collectively, functional analysis and KEGG pathways in the OD group appeared to be associated with immune-mediated processes, vascularization and VEGF signaling, hemostasis and platelet activation, tissue remodeling and oxytocin response. These findings corroborate the predominantly inflammatory nature of second-stage labor as strong uterine contractions develop and the placenta detaches for fetal expulsion. We were unable to definitively pinpoint the underlying placental cause for PUI.
Assessing the environmental, biotic, climatic, and anthropogenic factors driving patterns of space-use of jaguars and co-occurrence of three sympatric felid species through occupancy modelling in Belize, Central America
Lugo, David A. (Virginia Tech, 2026-07-14)
The jaguar is an apex predator whose conservation is critical for maintaining ecosystem structure across the Neotropics. Despite its ecological importance, country-wide assessments of jaguar population dynamics remain limited in Belize. We conducted the first large-scale, long-term evaluation of jaguar distribution across Belize by integrating over two decades of camera trap data from multiple protected areas and wildlife corridors spanning ~4,000 km². Using single-species, single-season and multi-season occupancy modeling frameworks, we assessed the influence of environmental, biotic, climatic, and anthropogenic factors on jaguar space use, detection, colonization, and extinction dynamics. Across all study areas, jaguar occupancy was consistently high (ψ > 0.60) and relatively stable through time in well-protected landscapes. In single-season models, occupancy and detection were most strongly influenced by biotic covariates, particularly prey trap rates, while environmental covariates such as enhanced vegetation index (EVI) at multiple spatial scales and influence of distance to water bodies varied across sites and years. In multi-season models, environmental drivers played a larger role in shaping long-term dynamics, with occupancy positively associated with vegetation structure and hydrological features. Detection probability was highly variable and influenced by a combination of prey availability, vegetation structure, climate (precipitation), and anthropogenic covariates. Colonization (γ) was generally higher than extinction (ε) across sites, with colonization associated with favorable habitat conditions, indicating dynamic but resilient occupancy patterns. In contrast, areas with greater anthropogenic pressure exhibited reduced space use and increased potential for human–wildlife conflict. To further contextualize felid spatial dynamics, we evaluated co-occurrence patterns among jaguars, pumas, and ocelots using single-season co-occurrence models. Felid occupancy and co-existence were structured by environmental gradients, particularly vegetation productivity (EVI) and proximity to water, in combination with prey availability. Patterns of co-occurrence suggest spatially driven niche partitioning; whereby sympatric felids adjust space use in response to one another while maintaining overlap in shared habitat and resource preferences. Overall, our findings demonstrate that jaguar populations in Belize remain broadly stable but are governed by complex, scale-dependent ecological processes. By integrating single-season, multi-season, and co-occurrence modeling approaches, this study provides a comprehensive assessment of felid spatial ecology and highlights importance of multi-use landscapes in maintaining carnivore communities. These results offer critical insights for conservation planning, emphasizing the need to maintain habitat structure, ensure water availability, and incorporate prey dynamics into landscape-level management strategies in the Selva Maya.
Determining the role of density-dependent effects in disease induced population impacts
Newell, Rose Gabrielle (Virginia Tech, 2026-07-14)
Emerging infectious diseases have contributed to the loss of biodiversity and threaten conservation efforts across the globe. Managing infectious diseases can be challenging in wildlife, often due to the complex nature of pathogen transmission in muti-host systems that can lead to varied decline and recovery patterns. Common disease management strategies focus on the reduction of pathogen transmission at low host densities to be effective. Density-dependent contact rates are generally assumed to be the mechanism driving these decline patterns. However, other density-dependent processes that influence transmission probability could also influence mortality in a density-dependent manner. Here we examine the processes by which population density contributes to the observed declines of bats infected with white-nose syndrome, a fungal disease that has decimated bat populations across North America. We evaluated the contribution of both density-dependent transmission rates and density-dependent environmental exposure on pathogen prevalence and host population declines. Our analyses indicate that the importance of population density varies depending on the phase of the epidemic, and that density-dependent declines are not observed in years when we see density-dependent transmission. Instead, we find that the amount of pathogen a host is exposed to in the environment is density-dependent and likely contributes to the observed patterns of decline. This research contributes to our general understanding of the role of density-dependent processes in disease dynamics and can help inform when certain disease mitigation efforts will be most effective.
SoundBreak: Studying vulnerabilities exposed by Audio Adversarial Attacks on Trimodal Models
Hussain, Aafiya Shamshad (Virginia Tech, 2026-07-13)
Foundation models that jointly process audio, visual, and textual inputs have demonstrated strong performance across a wide range of reasoning and generation tasks. Despite these advances, the robustness of such systems against adversarial manipulation remains insufficiently explored. This thesis presents SoundBreak, a comprehensive study of untargeted adversarial attacks that perturb only the audio modality in trimodal audio-video-language models. We introduce six base attack objectives and a combined objective that target different components of the multimodal pipeline, including audio encoder representations, cross-modal attention distributions, internal hidden-state dynamics, and output token likelihoods. Extensive experiments conducted across four state-of-the-art multimodal models and multiple benchmark datasets show that audio-only perturbations are sufficient to induce substantial performance degradation, achieving attack success rates of up to 96%. The combined objective achieves the highest attack success rate, while the encoder-space objective provides the clearest high-effectiveness, low-distortion trade-off, reaching 89.12% ASR with LPIPS 0.08 and SI-SNR -1.77 dB. The lowest-distortion objectives achieve LPIPS 0.06 and SI-SNR around 0 dB, but with lower ASR. Our analysis further reveals that prolonged optimization contributes more significantly to attack effectiveness than increased dataset scale, while transferability across models and audio encoders remains limited. Additionally, automatic speech recognition systems such as Whisper are found to respond primarily to overall distortion magnitude, reaching attack success rates exceeding 97% under high-distortion objectives. Collectively, these findings expose a previously underexplored single-modality vulnerability in multimodal systems and highlight the need for robust defense mechanisms based on cross-modal consistency and agreement verification.
The One, The Void, The Other: The Art of Connection
Blanton, Caleb Judah (Virginia Tech, 2026-07-10)
This thesis examines the nature of physical, metaphysical, and metaphorical connections within the urban landscape, exploring the spatial and philosophical relationships between edges and the void that separates them. The research begins with a comprehensive exploration of bridge inhabitation, examining historic real-world infrastructure and the multitude of fairtales, stories, and cultural narratives. Through the analysis of physical, metaphysical, and metaphoric connections, the thesis investigates how "moments of transaction", the threshold where boundaries blur, form in relation to the edges. These discoveries set a framework for site analysis and development, reinforcing the theoretical grounds for bridges to be considered more than mere infrastructure, rather, "new land" with opportunities for place-making and inhabitation. The theory is applied to the Alexandria Aqueduct abutment ruins nested between the Chesapeake and Ohio (CandO) Canal and the Potomac River in Georgetown, D.C. Established as an historic landmark, the Alexandria Aqueduct abutment ruins are situated upon a major intersection of topographical voids, urban thresholds, and a distinct metaphysical connection to history. The final project proposal is an architectural response that treats the existing structure as the foundational anchor for the "moment of transaction" between various connections, activates the surrounding edges, and establishes inhabitable space.

Browse

Recent Submissions