62880 research outputs found
Sort by
UrbanTALES: a large-eddy simulation dataset for urban canopy layer turbulence and parameterization
Get PDFThe urban canopy layer (UCL) exhibits complex, heterogeneous flow patterns shaped by urban geometry. Traditionally, research has relied on microscale simulations over limited and often idealized building arrays, leaving a need for more extensive datasets to capture the dynamics across diverse urban neighborhoods. Responding to this gap, we developed an extensive dataset, known hereafter as Urban Turbulence Analyses from Large-Eddy Simulations (UrbanTALES), based on state-of-the-art Large Eddy Simulations (LES) over 538 urban layouts (generated using over 3,000,000 CPU hours and 35 TB of storage) with both idealized and realistic configurations. Realistic urban neighborhood configurations were obtained from major cities worldwide, incorporating wide variations in building plan area densities [0.06-0.64] and height distributions [4-50m]. Idealized urban arrays, on the other hand, include two commonly studied configurations (aligned and staggered building arrays), featuring both uniform and variable height scenarios along with oblique wind directions. UrbanTALES offers canopy-averaged flow data as well as 2D and 3D flow fields tailored for different applications in urban climate research such as the development and testing of urban canopy models. The dataset provides time-averaged wind flow properties, as well as second and third-order flow moments that are critical for understanding turbulent processes in the UCL. Here, we describe the UrbanTALES dataset and its applications, noting the unique opportunity to use high-fidelity simulated flow in realistic urban neighborhoods to: a) revisit neighborhood-scale urban canopy parameterizations in various climate models; and b) inform in-canopy flow and turbulent analyses in complex urban configurations. UrbanTALES is openly available at https://urbantales.climate-resilientcities.com/ and can be extended to incorporate future LES datasets in the field
Enhancing the performance and transparency of Machine Learning (ML) using MRI-derived data: alternative approaches to ML interpretability-explainability
Get PDFThis thesis addresses the three fundamental challenges for enhancing the performance of
Machine Learning (ML) models. Despite their evolving predictive capabilities, MLsstill present
significant limitations in generalisability, particularly in high-dimensional settings,
interpretability, and high data requirements. These issues require methodologies that reduce
input data dimensionality, enhance transparency, and utilise prior knowledge to moderate
the scale of data requirements, thereby improving the performance, reliability, and efficiency
of machine learning solutions in practical applications.
Accordingly, this thesis introduces three independent methods responsive to the
above main limitations that need to be overcome to enhance the performance and
transparency of models in complex task domains. First, two filter-based feature selection
techniques—one correlation-driven and the other clustering-based—are developed to reduce
redundancy and enhance generalisability in high-dimensional data. The correlation-based
technique outperforms the state-of-the-art (as represented by ReliefF) in both internal and
external validations. Second, an ensemble explainability framework integrates Shapley
Additive Explanations (SHAP) values with Sobol indices, combining their rankings to yield
stable and interpretable attributions. Third, a multi-stage algorithm couples transfer learning
with an autoencoder to minimise labelled data requirements without adversely affecting
performance.
All proposed methods yielded quantifiable improvements. The feature selection
techniques reduced input dimensionality while enhancing accuracy and generalisability
compared to ReliefF. The ensemble explainability framework produced consistent attributions
under varying data distributions and reliably identified informative input features. The multi-stage algorithm achieved enhanced classification performance with reduced reliance on
labelled data.
Case-Study: The proposed methods were validated in the context of medical diagnosis
for early-stage prediction of dementia, utilising a structural Alzheimer’s MRI dataset. In this
application, optimising the feature selection, as described above, enhanced the cross-cohort
accuracy and decreased the data dimensionality. The explainability framework consistently
identified clinically relevant regions, such as hippocampal subfields (W. Zhao et al., 2019) and
the temporal horn (Vernooij and van Buchem, 2020), supporting the credibility of feature
relevance. The data-efficient multi-stage pipeline achieved an accuracy of 73.26%, exceeding
prior baselines (Li et al., 2015; Oh et al., 2019).
This thesis concludes that the proposed correlation and clustering-based feature
selection, ensemble explainability combining SHAP and Sobol, and transfer learning with
autoencoders have led to enhanced accuracy, robustness, and transparency of the
performance of the machine learning models. Although this was validated for the Alzheimer’s
validation task, these methods are domain-agnostic and provide scalable, reliable, and
resource-efficient approaches for high-dimensional, data-limited real-world applications
Identifying key genes for European canker resistance in apple: machine learning and gene expression profiling of quantitative disease resistance
No full textEuropean canker, caused by Neonectria ditissima, is a major disease of apple (Malus × domestica) with limited control options, making host resistance a key management strategy. Although quantitative disease resistance (QDR) has been identified, the underlying molecular basis remains poorly understood. We investigated candidate genes associated with resistance using transcriptomic profiling of a bi-parental population segregating for six QTLs linked to canker resistance. RNA sequencing combined with machine learning enabled the identification of key biomarkers predictive of disease resistance. Integration of expression and QTL data highlighted genes involved in phenylpropanoid biosynthesis, immune receptors (NLRs, RLKs, WAKs), and epigenetic regulators, implicating their roles in host defense. Expression patterns were further resolved into cis- and trans-regulatory effects, providing insight into allele-dependent regulation. Independent validation in a separate dataset confirmed the robustness of key expression patterns. These findings advance understanding of the genetic architecture underlying QDR in apple and provide a basis for marker development to support breeding of cultivars with durable resistance to European canker
Case 1: Reflecting on assessment and feedback strategies in the Foundation Year Psychology programme
No full textMolecular dynamics simulation and performance verification of γ-polyglutamic acid/cold water–soluble starch film formation and permeability
Get PDFSix types of γ-polyglutamic acid (γ-PGA)/cold water–soluble starch (St) composite-film models were constructed using molecular dynamics simulation, and their properties were investigated and compared with the corresponding experimental values. The compatibility between the composite film components was analyzed using the radial distribution function and mean square displacement (MSD). The hydrogen bond number and bond energy were used to track the film-formation process. The mechanical property data of the films were extracted, and MSD was used to analyze the permeability of the film to carbon dioxide, oxygen, water vapor, and carbon-16 saturated fatty acids. Finally, the simulated values of mechanical properties and permeability were compared with the experimental values. The results demonstrated that γ-PGA is well compatible with St. The intramolecular and intermolecular hydrogen bonds of γ-PGA and St did not change considerably during the film-formation process. The simulated values of the mechanical properties exhibited a similar trend as the experimental values; however, in terms of permeability, a difference was observed between the initial values of the simulated design and actual material parameters, as well as the complexity of the experiment
The effect of supplementary LED illumination of Romaine lettuce on midribs pinking after harvest
Get PDFPinking of midribs is a major postharvest issue in cut lettuce. Here, the effects of cultivar, light intensity and time of storage on pinking discolouration and related metabolites were elucidated. Two cultivars of Romance lettuce, Keona (fast pinking) and Icarus (slow pinking) were grown under four light intensities (L1 – L4: 1044, 578, 386 and 338 µmol.m −2.s −1 respectively); we determined their effects on pinking of leaf midribs, phenolic acids, soluble sugars and total ascorbic acid concentrations after eight days of cold storage. Differences in pinking index of the midribs of the two cultivars were only observed when the plants were grown in higher light intensities. All phenolic acids increased during storage and were highest at L1. Keona consistently contained higher concentrations of glucose, galactose and sucrose regardless of light intensity compared with Icarus after both 0 days and 8 days of storage. Principal component analysis (PCA) shows that coumaric acid, caffeic acid, and chlorogenic acid were positively correlated with the pinking index for both cultivars. The study revealed that pinking was reduced when the plants were grown at a low light intensity. Using low pinking cultivars offers a clear benefit in improving postharvest quality, especially when plants are grown under high light intensity