42337 research outputs found
Sort by
Machine learning-aided fluvial system analysis
No full textMid-resolution satellite imagery from platforms like Sentinel and Landsat has made large-scale monitoring of river systems feasible, yet challenges persist in analysing complex fluvial features, such as river morphology and sediment transport, on a global scale. These difficulties stem from the volume and complexity of remote sensing data, along with the need for fully automated methods to extract accurate and meaningful insights. Machine learning offers a powerful solution, enabling efficient processing of large datasets and uncovering patterns across spatial and temporal dimensions.
This thesis leverages machine learning to address challenges in large-scale fluvial system analysis, automating the extraction and representation of key fluvial components and advancing our understanding of river systems at global scales. Using data from Sentinel-1 and Sentinel-2, this study applies machine learning techniques—including Random Forest and U-Net models—to segment and classify features such as water bodies, sediment deposits, and surrounding land cover (e.g., vegetation, urban areas). Additionally, a data-driven approach combines spatial features that describe the geomorphology of upstream catchments with time series data to analyse flow dynamics. Geomorphic descriptors serve as inputs to Long Short-Term Memory (LSTM) models, capturing temporal variations in river behaviour and providing insights into complex interactions between rainfall, landscape characteristics, and river discharge. This integrated framework of spatial and temporal analysis offers a robust approach for studying river processes across diverse environments.
Key findings from the applications include improved river network extraction by conditioning DEMs with river distributions derived from classified river channels, resulting in better alignment with actual flow paths; effective fluvial system segmentation using incremental learning to classify fluvial sediment from existing global land cover products; and explainable river discharge predictions with LSTM models, which reveal model multiplicity (equifinality in numerical modelling). Collectively, these results underscore the potential of machine learning to provide deeper insights into the complex dynamics of river systems across diverse landscapes.
The main contribution of this thesis is the development of machine learning methodologies that automate and enhance the analysis of fluvial systems, providing scalable, data-driven insights into river dynamics, which holds significant implications for large-scale environmental monitoring and management. Future research should focus on refining machine learning models for greater precision in capturing river dynamics, integrating higher-resolution data, and exploring the adaptability of these methods across diverse hydrological and geomorphological contexts
Social problems and urban governance in Edinburgh and Canongate, 1560-1640
No full textThis thesis explores urban governance and the experience of authority in Edinburgh and Canongate, c.1560 to 1640, with a specific interest in the processes of marginality and exclusion. It analyses the relationship between local authorities and town inhabitants of non-burgess status, such as migrants, vagrants, petty criminals, and other undesirables. The period witnessed the Reformation, civil wars, plague outbreaks, rapid population expansion, and perhaps the most severe nationwide famine that Scotland has ever faced.
Analysis of the aims and objectives of both local and national government policies in response to the growing vagrancy crisis in Edinburgh will be presented along with an assessment of the efficacy of the same at street level. Adopting an interdisciplinary approach, this thesis examines the nature of the relationship between established residents and incoming migrant poor, arguing that networks of support and cooperation established between the two gave agency to the otherwise impotent migrant poor, and in doing so limited governmental authority at street level
Patient participation in nursing care in a Chinese hospital: a focused ethnographic study
No full textBACKGROUND:
Patient participation in health care refers to actively involving patients in their own care,
treatment, and decision-making processes. Supporting patients to be engaged in their health care
can lead to improved outcomes, better experience, and potentially more efficient healthcare
systems. Within professional nursing practice, patient participation is advocated and aligned with
philosophies of care such as person-centred care. Patient participation in nursing care
incorporates establishing nurse-patient relationships, information and knowledge sharing,
relinquishing power from nurses to patients, and mutual engagement in intellectual and, or
physical activities. Yet, there remains a lack of understanding of nurses’, patients’, and family
members’ experiences of patient participation in nursing care. In particular, there is a dearth of
perspectives from contexts outside the Western world. China, which has different policies and
cultural background, is an important context.
AIM:
This focused ethnographic study aimed to explore patient participation in inpatient nursing care
in China as perceived and experienced by nurses, patients with chronic illness, and their family
members.
METHODOLOGY:
Focused ethnography was adopted to explore patient participation in nursing care in a Chinese
hospital. Nurses, chronically ill patients, and family members were recruited. Data was collected
over an 8-month period of fieldwork from February to September 2021 in a Neurology
Department. Fieldnotes from 90 hours of participant observation and 30 interviews were
included (ten nurses, 13 patients, three family members, and four joint patient-family members).
Following transcription, the data were analysed in NVivo 12 with a reflexive thematic analysis
approach.
FINDINGS:
Four themes were developed: the context of the Neurology Department; patient self-care;
factors influencing patient participation; and minor decision making in nursing care. The Context
of the Neurology Department comprised organisational and interpersonal contexts of the setting,
and included nurses’ experience of role ambiguity and emotional exhaustion. Patients
participated in physical, intellectual, and emotional self-care activities and many aspects of
nursing care. Patients were interdependent with family members during their hospital stay, and
required support from nurses. Patient participation was diverse and dynamic depending on a
variety of factors, including capability, responsibility, and willingness to participate. Patients
generally desired more information and communication. Capability to participate referred to
patients’, family members’, and nurses’ ability to participate or provide accessible support, and
resources. Responsibility in patient care included patients’ responsibility for their health, family
obligations, nurses’ professional accountability, and duties in nursing care. Willingness to
participate was related to nurses’ and patients’ preferences, emotional status, and nurse-patient
relationships. Decision making in nursing care was related to minor issues of patient care, in
contrast to treatment decisions. Patient participation and autonomy were respected where
decisions were co-determined, but challenged where unilateral determination and
organisational determination operated.
CONCLUSIONS:
Patient participation was seen in self-care and decision-making processes in nursing care.
Promoting patient autonomy and creating a caring environment could help to eliminate barriers
and facilitate participation. Nurses should increase their awareness of involving patients in their
own care. Family members’ contributions to care should be valued. Meanwhile, attention should
be paid to paternalistic approaches from both nurses and family members causing potential
hinderance for patient participation. Organisational managers and policy makers should set
associated regulations and policies in place, to create a supportive environment for patient
participation
Linear bilevel optimization and applications in the grid integration of energy storage
Get PDFDespite only having been introduced in the 1930s, bilevel optimization has rapidly
gained researchers’ interest as a useful hierarchical modelling framework for numerous applications. In particular, bilevel optimization has been playing an important role for studying the effective integration of energy storage systems into
the energy grid, a pivotal step for addressing challenges in decarbonizing the
energy sector as we shift to renewable-powered networks. Within this context,
this thesis explores key theoretical aspects of bilevel optimization and practical
applications in grid-integration of energy storage, and it is composed of three
main research contributions. First, we investigated unboundedness in bilevel
and multilevel optimization, an often overlooked issue as most research assumes
boundedness. In this first contribution, we show that deciding whether an optimistic linear bilevel problem is unbounded is strongly NP-complete, and that
the hardness part of this result is valid for the pessimistic formulation. In general, we show that deciding unboundedness of an optimistic k-level problem is
Σ
p
k−1
-hard for linear problems and Σ
p
k
-hard for mixed-integer problems. We also
propose two algorithmic approaches for detecting unboundedness and compare
their performance through computational experiments. In the second part of this
thesis, we focus on an application of bilevel optimization in the energy industry.
We propose an innovative business model to harness the potential of aggregating
the behind-the-meter residential storage that arises with the emergence of pro-sumers. In this business model, a grid-scale aggregator compensates prosumers
for on-demand use of their storage systems, while allowing them to sell the electricity they generate at wholesale market price. Our computational results for
a realistic Texas case study show that the model has strong economic potential, with participants and the aggregator both achieving profitability. Lastly, we
focus on another energy application arising from the energy transition and, we
study the viability of hydrogen storage as a supplemental source of baseload support in a fully-renewable energy grid. Using a two-stage stochastic optimization
model, we analyse investment decisions in renewable plants and hydrogen storage, while accounting for the operational costs of running the hydrogen storage
systems under uncertain renewable generation. Our results indicate that green
hydrogen is particularly valuable in high-wind environments and that long-term
liquid hydrogen storage is more profitable than intraday hydrogen gas storage.
The optimization model presented provides a framework to further investigate
the potential of green hydrogen under many different energy grids
Impact of over-pressured sequences on seismic attributes and AVO analysis
No full textOverpressure mechanisms typically involve transferring some load from the rock framework to the pores. This reduces velocities and increases the porosity of overpressured rocks. Chemical diagenesis complicates the relationship between pore pressure and rock properties. Despite efforts to improve well drilling safety through understanding and predicting these mechanisms, there is little information on how they influence AVO analysis. The Foz do Amazonas and French Guiana basins are close to the Amazon River delta, hence the overpressure in the area is driven by the large volumes of sediment from the Amazon River and due to their arkosic nature, also caused by chemical alterations. This setting could explain the high bulk modulus and high-impedance seismic reflections on the cretaceous sandstones targeted by wells drilled in the area.
To explore the effects of overpressure on elastic properties and AVO analysis, I developed a pressure-dependent rock physics model that incorporates porosity as a variable. This model accounts for the different responses of facies and the mineralogical differences between shales and sands. Using datasets from multiple scales—plugs, logs, and seismic—from the Foz do Amazonas and French Guiana offshore basins, I demonstrate that high pore-pressure settings may cause amplitude variation, by increasing acoustic impedance and Poisson Ratio contrasts and and potentially induce AVO anomalies.
I have also shown how differential diagenesis influences porosity in the studied hydrocarbon-saturated reservoirs. Oil saturation may prevent quartz cementation in Well-A, resulting in higher porosity and different elastic properties than brine-saturated zones. This effect is noticed by a reduction on the acoustic impedance and angle reflectivity. However, while this effect is noticeable at the well scale, it is too subtle to be detected by the available surface seismic data.
Through synthetic seismic modelling and 3D seismic data interpretation, I found that lithological variations predominantly influence the seismic response of these reservoirs, rather than fluid content or pore pressure. For instance, substituting oil with brine in the target reservoir resulted in minimal changes to seismic attributes, underscoring the negligible impact of fluid presence due to the high compaction setting in which the reservoirs are inserted.
Future studies should consider the interconnection of complex geological processes—including sediment provenance, diagenesis, and lithological variations—on overpressure effects and how they collectively influence the elastic properties of rocks, and, consequently, seismic attributes and AVO analysis
Modelling logistics strategies for the installation of offshore wind farms
No full textThe construction of an offshore wind project is a complex process that involves many uncertainties and technical risks. Challenges include installation schedule delays and cost overruns caused by harsh weather conditions, as well as supply chain restrictions and constraints regarding port infrastructure and a shortage of specialised installation vessels. These uncertainties may impact the achievement of the global offshore wind installed capacity target of over 1,100GW by 2050. Therefore, optimising offshore logistics and installation methods is critical to strengthen the risk identification process at the planning and pre-construction phases of future offshore wind projects, including fixed-bottom and floating wind technologies.
The stages and processes involved in the project lifecycle of an offshore wind farm are well-documented, ranging from the initial project planning and permitting through to the decommissioning (or repowering) phases. Models and simulation tools for long-term logistical planning have been developed, yet great uncertainty remains in the logistics requirements for the transport and installation operations of large-scale offshore wind projects, particularly for floating offshore wind technologies. Most of the tools have focused on operations and maintenance (O&M) and operating cost (OPEX) estimation, with limited research addressing the performance of installation vessels and the likely impact of logistics strategies on project installation time and costs.
This research contributes to the knowledge by exploring the current installation methods and logistical approaches for the installation of monopiles and semi-submersible floating structures –the most common offshore wind turbine foundations to date. After analysing current and future trends in offshore wind industry development, a novel construction model was developed using an advanced forecasting and discrete-event simulation support tool. The construction model was validated through: i) real system data validation; ii) benchmarking; and iii) industrial case studies.
The model represents key offshore wind farm T&I processes, predicts campaign installation times, and assess the impact of installation methods and operational weather conditions on offshore installation duration. Validation results indicate that the construction simulation model closely aligns with existing T&I procedures and accurately represents different strategies for ports, vessels, and offshore logistics operations.
Research findings suggest that large-scale floating wind projects in the UK's North Sea waters are more sensitive to the distance between the marshalling port and the wind farm site than monopile-based projects, highlighting the need for significant capital investment in port infrastructure to meet offshore wind ambitions of large deployment by 2050. Moreover, the study underscores the importance of precise input data and assumptions in reducing uncertainty in offshore wind logistics models. Inadequate or inconsistent input information may lead to inaccurate risk assessments related to weather downtime, installation timing, and vessel cost estimation.
Ultimately, this research highlights the necessity of enhanced industry collaboration and data-sharing initiatives to refine offshore wind T&I strategies
Ageing in place: exploring perspectives and lived experiences of Chinese empty nesters and their adult children through interpretative phenomenological analysis (IPA)
No full textBACKGROUND:
China is undergoing a rapid transition into an ageing society, with empty nesters emerging as a significant demographic among the elderly population. Many empty nesters choose 'Ageing in Place (AIP)', which is defined as the ability of older people, regardless of their economic status, age, or intrinsic capacity, to live independently, safely, and comfortably in their own homes and communities. In the Chinese context, the role of adult children in supporting AIP is particularly critical due to the cultural value of filial piety. However, there is limited understanding of how AIP is experienced and interpreted by both empty nesters and their adult children within this cultural framework.
AIMS:
This study aims to investigate the lived experiences and perceptions of AIP in China from the perspectives of both empty nesters and their adult children, as well as to explore the intergenerational understanding of AIP in a Chinese sociocultural context.
DESIGN:
The study employed an Interpretative Phenomenological Analysis (IPA) methodology to examine the lived experiences of participants.
METHODS:
Purposive sampling was used to recruit empty nesters and adult children in Changsha, China. The two study groups were not related. Data were collected from December 2021 to May 2022 through photo elicitation and semi-structured interviews with empty nesters (n=9), and semi-structured interviews with adult children (n=8). Empty nesters were instructed to take photographs of their daily lives prior to the interviews. These photographs were used as prompts during the interviews to help participants recall and reflect on their experiences, thereby enriching their responses. Adult children only participated in 90-minute interviews. All interviews were audio-recorded and transcribed verbatim, and the data were analysed using IPA analysis.
RESULTS:
The findings on AIP for empty nesters revealed five group experiences themes: 1) Autonomy and freedom in ageing, with sub-themes including a) A life centred around oneself, b) Managing boundaries in intergenerational relationships, and c) Restriction of mobility on autonomy and freedom; 2) Collective, connective and reciprocal ageing, with sub-themes including a) Interdependence with spouses, b) Mutual support and reciprocal care with adult children, c) Social connectedness with old ties and neighbours, and d) State support in ageing; 3) Making sense of ‘place’, with sub-themes including a) Having a physical home to age, b) Keeping an appropriate distance for ‘a bowl of hot soup’, c) Intangible place-ageing in supportive social networks, d) Ageing in a digital world, and e) Not to be ‘Institutionalised’; 4) Losing social participation and social connectivity, with sub-themes including a) Isolated by other residents and the local community and b) Disconnected from each other; and 5) Insufficient access to healthcare resources. From the interviews with adult children, three group experience themes emerged: 1) Managing expectations and negotiating competing demands, with sub-themes involving a) Expecting autonomy and freedom, b) Addressing role transformation and transfer of power within the family, and c) Navigating the 'sandwich generation' dilemmas; 2) Making sense of living apart from ageing parents, with sub-themes involving a) Distance producing beauty, b) Distance bringing about concerns and a sense of insecurity, c) Adopting mitigation strategies to manage risks, d) Living with mutual support and reciprocal care, and e) Needing more state support and community support; and 3) Making sense of ‘filial piety’, with sub-themes involving a) Responsibility and filial obligation, b) Contrasting values on the next generation regarding filial piety, and c) A two-way process.
CONCLUSION:
Both empty nesters and adult children, shaped by the cultural tradition of filial piety, demonstrated diverse interpretations of AIP. The study's findings contribute to expanding the understanding of AIP across different cultural contexts and have important implications for developing culturally sensitive policies that support AIP in China and globally
Using submarine telecommunications cables as seismometers
No full textInterrogating optical fibres to get information about the Earth is the current
state-of-the-art method for seismic data acquisition. Until around a decade ago
seismologists relied primarily on well-developed single point measurements from
a range of instruments that can measure ground motion, pressure and strain. We
now have the capability to use optical fibre as a sensing element. This can be
actively deployed for a specific data acquisition or performed on optical links that
make up the existing networks. The latter makes previously inaccessible parts of
the world open to monitoring. Different interrogation techniques exist but the most widely used is distributed acoustic sensing (DAS) which allows a fibre to be used as a large-N linear sensing array over distances up to 100 km. The focus of this thesis is on an integrated interferometric method (IIM), discovered serendipitously during frequency metrology experiments. IIMs can overcome the range limitation of DAS at the cost of spatial resolution and they can also be used on links that concurrently carry other traffic (live fibre), whereas DAS is only able to work on dark (unused) fibre.
The IIM I have investigated has been deployed on a telecommunications cable
that crosses the North Atlantic from Southport in the UK to Halifax, Canada via
Dublin, Ireland. It is 5,860 km long and, whilst in use as a sensor, is providing
world-first data from a range of marine environments such as the continental shelf, the deep ocean, and the Mid-Atlantic Ridge. The aim of this thesis is to classify the signals we detect and build a full understanding of the IIM for seismological purposes.
The thesis is structured as follows. I begin with a brief introduction outlining how the research on interferometry-based earthquake detection started, and continue into chapter 2, giving some background and motivation for development of the sensor.
Chapter 3 introduces the particular cable on which we have been acquiring data and includes an account of the cable segmentation, that turns one cable sensor into a chain of discrete sensing spans. I will show that we are able to identify earthquake detections, tidal signals, microseisms and storm detection through wave height sensitivity. This work was published in May of 2022. Chapter 4 is based on experiments performed in a controlled setting on pieces of submarine cable in order to understand the sensitivity and response of an armoured fibre sensor and this work was published at the end of 2024. Chapter 5 shows the most recent data from the cable including earthquake detections over the magnitude range Mw 4 - Mw 8. I also propose sources of deep ocean noise and investigate the feasibility of implementing the data in applications such as seismic interferometry.
An improved understanding of the detections made by IIMs means that the data acquired can well complement existing seismic networks and further our knowledge of Earth processes such as deep ocean tectonics, ocean dynamics and deep Earth structure. Wider implementation on multiple fibre links in the global network could rapidly broaden the seismic monitoring across parts of the world otherwise inaccessible to seismology
Pixels to pitch: extending few-shot learning to audio tasks
Get PDFFew-Shot Learning has gained significant attention in recent years as a possible tool for solving tasks which have too little data for traditional machine learning pipelines, such as user adaptable AI systems or rare event detection. At the start of this study, research on few-shot learning was heavily focused on the imagery domain, with only a small handful of works considering other settings. This bias toward the imagery domain, and moreover toward a few very popular sub-tasks, has the potential to negatively impact the development of well-generalised few-shot learning, capable of performing well across a variety of domains and problem settings. Works available at the time that did try to span out from few-shot imagery were largely around few-shot audio classification and event detection. These works largely suffered from the same lack of reproducibility as one another, and as such did not effectively built from one another. To more effectively branch-out from few-shot imagery, and simultaneously advance both momentum and state-of-the-art in few-shot audio classification, this thesis investigates several important threads of research, from benchmark creation and development of general purpose self-supervision approaches, to transfer learning and performance prediction.
The first part of this thesis focuses on the creation of MetaAudio, a large-scale, fully reproducible and extendable few-shot audio classification benchmark containing 10 evaluation datasets and 4 experimental tracks. We provide a detailed description of the benchmark construction and setup, as well as a comprehensive suite of experimental results using popular Meta-Learning approaches. Alongside this, our results in MetaAudio highlight key differences between few-shot in the imagery and audio domains. In the second part, we propose MT-SLVR, a novel general-purpose and domain-agnostic self-supervised algorithm, capable of learning both an augmentation-invariant and augmentation-sensitive feature space. After training and utilising MT-SLVR models, it achieved state-of-the-art performance across almost all benchmarked few-shot audio tasks. The role of transfer learning for few-shot classification was investigated next. In particular, the effectiveness of large-scale self-supervised speech models was investigated while also determining the relatedness of few-shot audio classification to existing audio benchmarking tasks. Additionally, we investigate how effectively more popular and readily available image-based models can be leveraged. The final part of the thesis looks at downstream tasks and investigates how to utilise dataset-dissimilarity measures in order to perform few-shot performance prediction. We found that either transductive or inductive class-separability measures can be used effectively to predict both task-hardness and final performance. By providing both a fully reproducible and stable benchmark as well as transfer learning evaluations and state-of-the-art approaches, this thesis contributes to the advancement of both few-shot and self-supervised learning research. Our work here will aid in the further development of these fields for audio-related tasks, and as an exemplar for sequential data in general
Cost and profitability of DAC in Scotland
No full textThis report maps global advancement and innovation in Direct Air Carbon Capture and Storage. Based on a 0.5 Mt Direct Air Capture (DAC) plant in Scotland operating in 2040, the research models costs and potential markets to estimate profitability
Access Repository Dashboard
Do you manage Edinburgh Research Archive? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!